{"rank":1,"id":"hallucination","title":"Hallucinated facts, APIs and code","description":"Models confidently invent functions, packages, citations or facts.","area":"AI","mentions":[{"itemId":"50352569","author":"FrustratedMonky","createdAt":1793716812,"url":"https://news.ycombinator.com/item?id=50352569","kind":"comment","storyTitle":"Nvidia – open-source auto-ontology","quote":"You can't tell. Humans are now hallucinating what is, or is not, AI."},{"itemId":"50350302","author":"Dylan-Hayes","createdAt":1793702488,"url":"https://news.ycombinator.com/item?id=50350302","kind":"story","storyTitle":"AI agents hallucinate prices: LLM grounding with live data","quote":"AI agents hallucinate prices: LLM grounding with live data"},{"itemId":"50350057","author":"burnoutdv","createdAt":1793700136,"url":"https://news.ycombinator.com/item?id=50350057","kind":"comment","storyTitle":"Vathographs: Photographic depth maps as creative expression","quote":"The data isnt really there, in short, the depth is hallucinated."},{"itemId":"50343536","author":"ERcel","createdAt":1793652701,"url":"https://news.ycombinator.com/item?id=50343536","kind":"comment","storyTitle":"Astra can read and write weights to improve Qwen","quote":"I agree, but also, I kind of feel like real judgement and understanding are a different thing from this \"being good at thinking\", and that in fact, this is the reason AIs hallucinate in the first place (A generated text can fit the pattern of a good response because its accurate, or because it's bs but has similar words and structure to other texts on the topic."},{"itemId":"50343442","author":"kazinator","createdAt":1793652225,"url":"https://news.ycombinator.com/item?id=50343442","kind":"comment","storyTitle":"Ask HN: Is anyone going to use AI only to file their taxes?","quote":"No to actually manipulate figures for me or fill something. I check every hallucination to actual documentation from the government."},{"itemId":"50341129","author":"exogenousdata","createdAt":1793643514,"url":"https://news.ycombinator.com/item?id=50341129","kind":"comment","storyTitle":"If you love a website, back it up","quote":"If you mean 'an LLM' as in the owner/maintainer of the LLM/weights, then examples of this already exist. Whether it's DeepSeek/Gwen downplaying Uyghur forced labor camps [0] or Grok hallucinating a 'white genocide' in South Africa [1], it is clear that the owners of LLMs have political agendas that are expressed in the weights of popular LLMs."},{"itemId":"50340361","author":"poeticsilence","createdAt":1793640732,"url":"https://news.ycombinator.com/item?id=50340361","kind":"comment","storyTitle":"Ask HN: Freelancer? Seeking freelancer? (November 2026)","quote":"Prior to that, I've built an ai workflow engine and a platform for both minimizing hallucinations and working with documents that are far too large for existing context window inputs or outputs using a novel, patented process that I invented; and applied these tools to various contracted projects requiring analysis and and cross-checking of information and other forms of integrity across textual and video/audio corp…"},{"itemId":"50340342","author":"poeticsilence","createdAt":1793640666,"url":"https://news.ycombinator.com/item?id=50340342","kind":"comment","storyTitle":"Ask HN: Who wants to be hired? (November 2026)","quote":"Prior to that, I've built an ai workflow engine and a platform for both minimizing hallucinations and working with documents that are far too large for existing context window inputs or outputs using a novel, patented process that I invented; and applied these tools to various contracted projects requiring analysis and and cross-checking of information and other forms of integrity across textual and video/audio corp…"},{"itemId":"50338969","author":"traceroute66","createdAt":1793636011,"url":"https://news.ycombinator.com/item?id=50338969","kind":"comment","storyTitle":"Anthropic reports $17B in Q3 revenue, commitments of $895B in revised S-1","quote":"They probably asked Claude. Sounds like a normal LLM hallucination."},{"itemId":"50333264","author":"oefrha","createdAt":1793595063,"url":"https://news.ycombinator.com/item?id=50333264","kind":"comment","storyTitle":"AI fan wikis riddled with errors are muddying the waters of games culture","quote":"I no longer share them though. It’s the shit tier free or super low cost AI hallucination engines used by the masses that are the problem."},{"itemId":"50331626","author":"kazinator","createdAt":1793580358,"url":"https://news.ycombinator.com/item?id=50331626","kind":"comment","storyTitle":"AI delivers 'zero financial payoff' to most companies, Apollo economist says","quote":"Bzzzt! Calculators do not hallucinate."},{"itemId":"50331209","author":"bitwize","createdAt":1793577153,"url":"https://news.ycombinator.com/item?id=50331209","kind":"comment","storyTitle":"Patrick Stewart taught us what good men do with power","quote":"In fact you can induce a sort of model collapse in a human by isolating them in a sensory-deprivation chamber, after which they will literally begin to hallucinate."},{"itemId":"50331124","author":"danpalmer","createdAt":1793576560,"url":"https://news.ycombinator.com/item?id=50331124","kind":"comment","storyTitle":"So I guess everyone uses LLMs to think now","quote":"I'd argue the improvement has come not so much from fixing hallucinations, but more in getting LLMs to successfully run for longer, which lets them more effectively fix their own hallucinations."},{"itemId":"50330959","author":"rolosa","createdAt":1793575408,"url":"https://news.ycombinator.com/item?id=50330959","kind":"comment","storyTitle":"So I guess everyone uses LLMs to think now","quote":"> > i don't know how more people don't experience this, but staring at an llms output be confidently wrong eventually to get there in the process, or looking at the commands its running and see some really stupid things its doing drives me crazy to the point i will just stop it before it answers, i cannot handle something being that confidently wrong (even though it'll probably still get there in the end)"},{"itemId":"50330958","author":"CamperBob2","createdAt":1793575390,"url":"https://news.ycombinator.com/item?id=50330958","kind":"comment","storyTitle":"So I guess everyone uses LLMs to think now","quote":"i don't know how more people don't experience this, but staring at an llms output be confidently wrong eventually to get there in the process, or looking at the commands its running and see some really stupid things its doing drives me crazy to the point i will just stop it before it answers, i cannot handle something being that confidently wrong (even though it'll probably still get there in the end)"},{"itemId":"50330892","author":"lazyasciiart","createdAt":1793574909,"url":"https://news.ycombinator.com/item?id=50330892","kind":"comment","storyTitle":"So I guess everyone uses LLMs to think now","quote":"> i cannot handle something being that confidently wrong"},{"itemId":"50330878","author":"tayo42","createdAt":1793574804,"url":"https://news.ycombinator.com/item?id=50330878","kind":"comment","storyTitle":"So I guess everyone uses LLMs to think now","quote":"> don't know how more people don't experience this, but staring at an llms output be confidently wrong eventually to get there in the process, or looking at the commands its running and see some really stupid things its doing drives me crazy to the point i will just stop it before it answers, i cannot handle something being that confidently wrong (even though it'll probably still get there in the end)"},{"itemId":"50318500","author":"semiquaver","createdAt":1793470476,"url":"https://news.ycombinator.com/item?id=50318500","kind":"comment","storyTitle":"Jepsen: ParticleDB 1.1.4","quote":"Good on them for being good sports I guess, but after reading this I would never touch a piece of software built by anyone involved with this company. The sheer hubris of their claims combined with the specific reality of their broken garbage software (with the contradicted marketing claims and hallucinated documentation) is literally breathtaking."},{"itemId":"50315975","author":"krapp","createdAt":1793453047,"url":"https://news.ycombinator.com/item?id=50315975","kind":"comment","storyTitle":"AI isn't replacing SREs, it's deskilling them","quote":"I don't think that LLMs mirror human cognition or \"sanity\" as much as people want to believe. We are not constantly hallucinating in any way that applies to the term as commonly understood relative to LLMs or to humans."},{"itemId":"50315186","author":"shanedevane","createdAt":1793447384,"url":"https://news.ycombinator.com/item?id=50315186","kind":"comment","storyTitle":"Show HN: Offline Tiny-LMs 25MB for personal goals, career and pitch deck advice","quote":"I have been toying with combining offline tools, tiny models and anti-hallucination concepts (RAG, lookup, etc.) I wanted something that would loosely keyword match my documents and help me think through things, whether it was for my next year goals or business ideas, like an autocomplete but matching sources of information and ideas quickly, and allowing me to click for details and quickly go back to my source to u…"},{"itemId":"50311693","author":"skydhash","createdAt":1793404343,"url":"https://news.ycombinator.com/item?id=50311693","kind":"comment","storyTitle":"AI isn't replacing SREs, it's deskilling them","quote":"Let’s say you have an incident, and you need a resolution. Do you spend 5 minutes tinking, then solving the problem or send out the AI for the hunt and spend 30 minutes grappling with hallucinations and non-sensical output?"},{"itemId":"50310772","author":"keeganpoppen","createdAt":1793398024,"url":"https://news.ycombinator.com/item?id=50310772","kind":"comment","storyTitle":"Show HN: A working Nintendo in ~1GB of Qwen weights","quote":"> hallucinate an entire operating system every single frame"},{"itemId":"50310095","author":"yubblegum","createdAt":1793394129,"url":"https://news.ycombinator.com/item?id=50310095","kind":"comment","storyTitle":"AI isn't replacing SREs, it's deskilling them","quote":"One day a stellar dev and next a bs artist. So this idea that LLM hallucination is not unlike human failings is without merit."},{"itemId":"50310073","author":"flockonus","createdAt":1793393998,"url":"https://news.ycombinator.com/item?id=50310073","kind":"comment","storyTitle":"Show HN: A working Nintendo in ~1GB of Qwen weights","quote":"You're comparing apples to oranges there! To hallucinate is arbitrary, you can use a 2B image param model, it will hallucinate an OS no problem* running real time!"},{"itemId":"50309309","author":"alwa","createdAt":1793390090,"url":"https://news.ycombinator.com/item?id=50309309","kind":"comment","storyTitle":"[dead]","quote":"If the thing’s priced at 2 cents and I can only bid 1 cent or 3 cents, that’s not super useful or informative since it involves a 50% change in value on either side. The new and specific facet of my AI frustration here, is that, true or false or accurate or hallucinated, the analysis leaves me none the wiser about whether I’m totally off base with those priors."},{"itemId":"50307892","author":"stackskipton","createdAt":1793383759,"url":"https://news.ycombinator.com/item?id=50307892","kind":"comment","storyTitle":"AI isn't replacing SREs, it's deskilling them","quote":">spotting the hallucinations"},{"itemId":"50307610","author":"eqmvii","createdAt":1793382684,"url":"https://news.ycombinator.com/item?id=50307610","kind":"comment","storyTitle":"AI isn't replacing SREs, it's deskilling them","quote":"The atrophying skill is one you don't need anymore. The new one - prompting, routing the output, spotting the hallucinations, managing singal/noise etc. - is now where you can add tremendous value."},{"itemId":"50307246","author":"pbohun","createdAt":1793381342,"url":"https://news.ycombinator.com/item?id=50307246","kind":"comment","storyTitle":"Show HN: A working Nintendo in ~1GB of Qwen weights","quote":"I like these kinds of projects because it helps to show the massive compute requirement between LLMs and ordinary programs. A while back I saw some people saying we would just have LMMs hallucinate an entire operating system every single frame."},{"itemId":"50294857","author":"cyclecycle","createdAt":1793303668,"url":"https://news.ycombinator.com/item?id=50294857","kind":"comment","storyTitle":"Zombie stats: Where the numbers AI quotes come from","quote":"yeah i work on this at https://groundedai.company/ we build a product called Veracity focused on citation and claim checking. legal is a major focus for us, also scientific and technical publishing, where we detect hallucinated citations in submitted works and all that jazz"},{"itemId":"50294281","author":"ceroxylon","createdAt":1793301025,"url":"https://news.ycombinator.com/item?id=50294281","kind":"comment","storyTitle":"LLM Voice","quote":"The voice is one thing, but I have encountered multiple 'tech support' teams and platforms that will respond to me in pure, unedited LLM output. I know this because the topic is usually an obscure issue, and I have already gone back-and-forth with multiple LLMs to see where their boundary of knowledge is on it, and it will often include hallucinations along the way (e.g. pointing to old/nonexistent settings pathways)"},{"itemId":"50294078","author":"Revanche1367","createdAt":1793300040,"url":"https://news.ycombinator.com/item?id=50294078","kind":"comment","storyTitle":"Gemini 4 Carbon is the new #1 on the Artificial Analysis Intelligence Index v5","quote":"I had been on the pro plan for a few months and recently upgraded to the $200.00 AI Ultra plan because I've gotten good mileage out of 3.8 flash while keeping in mind that it's far from perfect. Argon is better in my experience in terms of reliability (especially in not claiming to have done things it clearly didn't) but I have to agree with the hallucination part, for a \"SOTA\" model (before Carbon came out at least…"},{"itemId":"50289479","author":"XCSme","createdAt":1793283414,"url":"https://news.ycombinator.com/item?id=50289479","kind":"comment","storyTitle":"Gemini 4 Carbon is the new #1 on the Artificial Analysis Intelligence Index v5","quote":"> has actively hallucinated details"},{"itemId":"50288524","author":"isqueiros","createdAt":1793279849,"url":"https://news.ycombinator.com/item?id=50288524","kind":"comment","storyTitle":"Gemini 4 Carbon is the new #1 on the Artificial Analysis Intelligence Index v5","quote":"Makes it seamless for trip planning. Lastly it's the only \"SOTA\" model in a long time that has actively hallucinated details."},{"itemId":"50287192","author":"dspillett","createdAt":1793272040,"url":"https://news.ycombinator.com/item?id=50287192","kind":"comment","storyTitle":"Zero click, silent, remote Android DoS","quote":"> LLMs are always hallucinating that they've found the most severe bug in software history."},{"itemId":"50283716","author":"chomp","createdAt":1793239635,"url":"https://news.ycombinator.com/item?id=50283716","kind":"comment","storyTitle":"New York Times adds more AI features to its Cooking app. Users aren't thrilled","quote":"2 things. 1, hallucination is a sensory perception of something."},{"itemId":"50283472","author":"spl757","createdAt":1793237949,"url":"https://news.ycombinator.com/item?id=50283472","kind":"comment","storyTitle":"New York Times adds more AI features to its Cooking app. Users aren't thrilled","quote":"AI hallucinates."},{"itemId":"50277385","author":"shagie","createdAt":1793208584,"url":"https://news.ycombinator.com/item?id=50277385","kind":"comment","storyTitle":"From confession to conspiracy: How the Wedgewood fake polls story spread","quote":"Generating an AI hallucination of \"called 278 likely voters in the district and it resulted in...\" is just as fake as making up the data yourself."},{"itemId":"50273438","author":"stosssik","createdAt":1793195088,"url":"https://news.ycombinator.com/item?id=50273438","kind":"story","storyTitle":"Breaking changes in third-party APIs are turning into a full-time job","quote":"Nowadays, companies run more tools than they used to, and with time, it will drastically increase. First, because of vibe coding and apps built by agents, which can hallucinate an endpoint or a model name."},{"itemId":"50272147","author":"Telemakhos","createdAt":1793188471,"url":"https://news.ycombinator.com/item?id=50272147","kind":"comment","storyTitle":"Barcelona unrest: how to fabricate a protest in the age of TikTok","quote":"I found it interesting that two of the AI posters in the article have the three-pointed “no kings” crown, and one has a two-pointed version. The AI hallucinated icons from American partisan propaganda into an EU context in a country with a monarchy."},{"itemId":"50268667","author":"meristohm","createdAt":1793154641,"url":"https://news.ycombinator.com/item?id=50268667","kind":"comment","storyTitle":"Survival LLM: AI assistant and emergency library that runs on your own computer","quote":"Nevermind hallucinations, the donkey-work is important; why bypass the hours of research, which trains our in-built LLM and can lead to new realizations, and get the carrot without much effort?"},{"itemId":"50268408","author":"jan_m_savage","createdAt":1793152324,"url":"https://news.ycombinator.com/item?id=50268408","kind":"comment","storyTitle":"Amazon Is Gutting Its AI Division After Sustained Failure","quote":"But there still remains the possibility of error/hallucination."},{"itemId":"50268388","author":"stego-tech","createdAt":1793152155,"url":"https://news.ycombinator.com/item?id=50268388","kind":"comment","storyTitle":"Survival LLM: AI assistant and emergency library that runs on your own computer","quote":"What about those precious seconds waiting for inference? The danger here is less hallucinations and more, well, everything else about it."},{"itemId":"50267079","author":"Viliam1234","createdAt":1793142002,"url":"https://news.ycombinator.com/item?id=50267079","kind":"comment","storyTitle":"Nietzschian Debugger","quote":"> I have to admit the formalization steps are the fruit of ChatGPT's work and not only would I have not been able to produce them, I still don't understand them and probably never will. So, Hegel is not impenetrable, because ChatGPT can hallucinate something out of his writings, when you tell it to."},{"itemId":"50266908","author":"IshKebab","createdAt":1793140967,"url":"https://news.ycombinator.com/item?id=50266908","kind":"comment","storyTitle":"Survival LLM: AI assistant and emergency library that runs on your own computer","quote":"> the tendency even now of LLMs to hallucinate"},{"itemId":"50266551","author":"macNchz","createdAt":1793138984,"url":"https://news.ycombinator.com/item?id=50266551","kind":"comment","storyTitle":"Survival LLM: AI assistant and emergency library that runs on your own computer","quote":"The readme on Github does say the system produces hallucinated sources https://github.com/anonnerdfrenzy/survival-llm#how-the-models-did, but really you're correct, for this kind of highly targeted use case the LLM's summary could probably be ignored entirely and the system could work extremely well even just as a search engine: index all of the content up front, chunked by paragraph and maybe enriched with a HyDE s…"},{"itemId":"50266495","author":"joshAg","createdAt":1793138707,"url":"https://news.ycombinator.com/item?id=50266495","kind":"comment","storyTitle":"The Mythical Man-Month in the Age of AI","quote":"I don't personally like LLMs for research assistance (yet?) because i can't get over the hallucinating or rather, waste too much time validating that it's not currently overcast when they tell me the sky is in fact blue."},{"itemId":"50265871","author":"x3haloed","createdAt":1793135480,"url":"https://news.ycombinator.com/item?id=50265871","kind":"comment","storyTitle":"Survival LLM: AI assistant and emergency library that runs on your own computer","quote":"This topic is bringing out the hallucination-panic crowd."},{"itemId":"50265862","author":"SwellJoe","createdAt":1793135444,"url":"https://news.ycombinator.com/item?id=50265862","kind":"comment","storyTitle":"Survival LLM: AI assistant and emergency library that runs on your own computer","quote":"Yeah, without the ability to search the web, a very small model is going to be too dumb to be useful, and maybe even dangerous due to hallucinations."},{"itemId":"50265612","author":"bityard","createdAt":1793134244,"url":"https://news.ycombinator.com/item?id=50265612","kind":"comment","storyTitle":"Survival LLM: AI assistant and emergency library that runs on your own computer","quote":"That will make it much, MUCH more accurate and truthful than a bare model. But there is still a non-zero chance for it to hallucinate answers, especially when posed a question that it doesn't have any source material in its database."},{"itemId":"50257442","author":"cromka","createdAt":1793100011,"url":"https://news.ycombinator.com/item?id=50257442","kind":"comment","storyTitle":"I will no longer upload to NEW","quote":"If you, however, just drop a wall of text at someone and expect they act on its b/s, this is lazy to the point it borders with microagression. You wouldn't expect a distinguished professor explain basic maths to a student, and we shouldn't expect actual developers address Copilot's stupid, unredacted hallucinations or otherwise unrelated remarks."},{"itemId":"50254991","author":"giardini","createdAt":1793078527,"url":"https://news.ycombinator.com/item?id=50254991","kind":"comment","storyTitle":"Sebastian Raschka's from-scratch video course on building reasoning models","quote":"Would they not likely be more useful and far, far easier and cheaper to create and run (possibly even with far less hardware) than the current \"know-all\" LLMs? Also might they possibly avoid certain problems ( e.g., hallucinations, deceit, apparently undesirable behavior, etc.) associated with the \"know-all\" LLMs we now have?"},{"itemId":"50254187","author":"BerendG","createdAt":1793069443,"url":"https://news.ycombinator.com/item?id=50254187","kind":"story","storyTitle":"Show HN: Limma – MCP for total recall and precision in data retrieval for agents","quote":"This means total precision and recall. We've found that this approach leads to the lowest rate of hallucination vs. solely targeting agent action verification which has made tremendous strides in capability."},{"itemId":"50251597","author":"svat","createdAt":1793051644,"url":"https://news.ycombinator.com/item?id=50251597","kind":"comment","storyTitle":"The recently proposed o(n log n) integer multiplication algorithm [pdf]","quote":"Harvey care about a potential o(n log n) algorithm, whether or not such an algorithm is contained in OpenAI's paper per se), then it is neither necessary nor (given how it's written) a good idea. As I said in the linked comment, it does not matter whether the idea came from OpenAI's paper, or was hallucinated by Fable, or was arrived at by the reader while reading a potential nonsensical explanation from Fable: the …"},{"itemId":"50249040","author":"ape4","createdAt":1793040741,"url":"https://news.ycombinator.com/item?id=50249040","kind":"comment","storyTitle":"GPT-6.1 Terra","quote":"My browser is underlining \"factuality\" in red. AI hallucinations or hallucinatiality are still a big problem in my experience."},{"itemId":"50241672","author":"raxxorraxor","createdAt":1793010817,"url":"https://news.ycombinator.com/item?id=50241672","kind":"comment","storyTitle":"A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis","quote":"Not seeing LLM to perform very well here and hallucination would be a major issue (not at all a solved problem, even frontier models do hallucinate/match the most probable path."},{"itemId":"50240816","author":"objectively_ran","createdAt":1793003591,"url":"https://news.ycombinator.com/item?id=50240816","kind":"comment","storyTitle":"A mathematician reads the AI future","quote":"I have never used an AI output I didn’t fully understand and that is crucial. It lies and hallucinates and we must police it."},{"itemId":"50226351","author":"adastra22","createdAt":1792883009,"url":"https://news.ycombinator.com/item?id=50226351","kind":"comment","storyTitle":"The Goose Programming Language","quote":"As I said in my message, it would determine whether I use the project or not. I need to match it to my use cases, and if I find a match I need to know: is it matching because my use case falls in the domain of what the human collaborator wants from this project, or is my use case a hallucination that will go unsupported?"},{"itemId":"50226342","author":"eek2121","createdAt":1792882969,"url":"https://news.ycombinator.com/item?id=50226342","kind":"comment","storyTitle":"No signs of AI payoff outside the tech sector","quote":"You are definitely overhyping it. LLMs hallucinate."},{"itemId":"50219247","author":"FreakLegion","createdAt":1792830155,"url":"https://news.ycombinator.com/item?id=50219247","kind":"comment","storyTitle":"Evolving Threat: http2shell is an RCE in Node, Python, and more","quote":"There were no known breaches or end-to-end exploits as of Friday. I've seen LLM hallucinations purporting to be exploits and real exploits missing pieces (address leaks), but nothing complete."},{"itemId":"50207837","author":"thomashabets2","createdAt":1792758189,"url":"https://news.ycombinator.com/item?id=50207837","kind":"comment","storyTitle":"Is this the new cyber normal?","quote":"Three years ago AI generated code could basically not be used at all, and I had to rewrite just about every line. Today I have to watch out for generated code making steps in the wrong strategic and architectural direction, and over engineering, no longer hallucinating nonexisting APIs and forgetting to check the auth token."},{"itemId":"50203078","author":"yazaddaruvala","createdAt":1792715956,"url":"https://news.ycombinator.com/item?id=50203078","kind":"comment","storyTitle":"LLMs are making formal software verification feasible at a much larger scale","quote":"Allows for significantly more terse artifacts. Lower chance of hallucination, fewer output tokens, fewer thinking loop iterations and concise enough for humans to review (for domains that need it)."},{"itemId":"50201749","author":"pessimizer","createdAt":1792706873,"url":"https://news.ycombinator.com/item?id=50201749","kind":"comment","storyTitle":"Award-winning AI literature","quote":"Van Vogt did the same thing in the 50s iirc, and he inspired Philip K. Dick who could slop nonstop without sleeping for days in a Methadrine haze by mixing his own divine inspirations and paranoid hallucinations with Van Vogt's cynical commercial nearly-incompetent dreamlike incoherence."},{"itemId":"50201320","author":"deezNutsBCookin","createdAt":1792704677,"url":"https://news.ycombinator.com/item?id=50201320","kind":"comment","storyTitle":"Every AI model has the same \"controversial\" take on humans","quote":"The smarter ones, of course, can be split into those who write articles about AI models' controversial takes on humans and those who wouldn't [stoop so low, babble such BS even if they were drunk, never deceive others for the sake of their portfolio, ... feel dumb writing nonsense like that without labeling it fantasy or science fiction or hallucination, ...]"},{"itemId":"50194371","author":"ishouldnotbutk8","createdAt":1792671988,"url":"https://news.ycombinator.com/item?id=50194371","kind":"comment","storyTitle":"There isn't an easy exploit for this PHP vulnerability","quote":"Exploitation, at least now, requires knowing exact PHP config and a lot of tweaking. There was a complete panic on the internet and in all HN threads, with people claiming insane things like \"this is already exploited everywhere in the wild\" (it was not), \"it took me 10 minutes to replicate this vulnerability from a public information\" (yeah, sure) or \"there is a public blog post with a writeup\" (it was hallucinated…"},{"itemId":"50192030","author":"rf15","createdAt":1792651618,"url":"https://news.ycombinator.com/item?id=50192030","kind":"comment","storyTitle":"There isn't an easy exploit for this PHP vulnerability","quote":"> Ease is irrelevant because you only need expertise once and then everyone can reuse the result. Not if your LLM relies on existing code and you collectively create a million repos with hallucinated broken implementations that do not work before a good public exploit exists, which later LLMs rely on."},{"itemId":"50183915","author":"toplinesoftsys","createdAt":1792601183,"url":"https://news.ycombinator.com/item?id=50183915","kind":"story","storyTitle":"Show HN: IDE for a general-purpose visual programming language \"Pipe\"","quote":"- Code review. AI writes code faster than developers can review it, and leaning on AI to review AI doesn't fix the root problem - that AI can hallucinate too, so human eyes are still required."},{"itemId":"50179263","author":"torben-friis","createdAt":1792582912,"url":"https://news.ycombinator.com/item?id=50179263","kind":"comment","storyTitle":"The Decoherence Trap","quote":"Without that, you don't know if you're improving the shot or just hallucinating aimlessly for hours."},{"itemId":"50164222","author":"lossyalgo","createdAt":1792493283,"url":"https://news.ycombinator.com/item?id=50164222","kind":"comment","storyTitle":"I built a drone with only one motor [video]","quote":"It's also the reason why I stopped helping in a lot of forums I used to frequent. But I do agree about using LLMs for learning new things - it's great to just get those answers, hopefully the LLM doesn't hallucinate too much and cause you to destroy your project in the process ;) On the flip side, it removes the useful skill of learning to communicate with fellow human beings, and atrophies the skill of how to ask q…"},{"itemId":"50163011","author":"jan_m_savage","createdAt":1792485379,"url":"https://news.ycombinator.com/item?id=50163011","kind":"comment","storyTitle":"OpenBSD 8.0","quote":"Humans make predictable mistakes, AI makes unpredictable mistakes. Hallucination occurs where you least expected for ANY reason."},{"itemId":"50161629","author":"jtap","createdAt":1792473401,"url":"https://news.ycombinator.com/item?id=50161629","kind":"comment","storyTitle":"Ask HN: What are you working on?","quote":"I've been pretty curious to see if I can help reduce LLM hallucinations by giving LLM's more deterministic tools."},{"itemId":"50161144","author":"brndnbuilds","createdAt":1792468303,"url":"https://news.ycombinator.com/item?id=50161144","kind":"story","storyTitle":"Show HN: Rouler – Steer Windows traffic per app or IP across multiple interfaces","quote":"I did try to lean on AI at times, but it constantly sent me in the wrong direction, told me things weren't possible, tried to take shortcuts and implement poor performing trade-offs. It also loved to hallucinate the windows api."},{"itemId":"50155749","author":"mccoyb","createdAt":1792433149,"url":"https://news.ycombinator.com/item?id=50155749","kind":"comment","storyTitle":"Maxwell: A new kind of model that sits between code and hardware","quote":"Little phrases and claims that seem intended for the purposes of marketing, but conflate things with clear technical meaning in confusing ways. For instance: I watched the video (\"a code execution model has to be right every time\") which implicitly communicates something like \"the model cannot hallucinate\" ... of course, the agentic LM is still probabilistic, but it is wrapped with verification tools (to reason abou…"},{"itemId":"50155278","author":"simianwords","createdAt":1792430942,"url":"https://news.ycombinator.com/item?id=50155278","kind":"comment","storyTitle":"It's still ok to enjoy programming","quote":"Its almost as if they've shut off facts and made up their minds. An example: a year back I wanted to show that AI doesn't hallucinate that much."},{"itemId":"50151621","author":"mkovach","createdAt":1792413208,"url":"https://news.ycombinator.com/item?id=50151621","kind":"comment","storyTitle":"Free Pascal compiler now supports dynamic packages","quote":"Thank heavens, I wasn't the only person who's seen that. I thought I was the one hallucinating!"},{"itemId":"50149259","author":"lucumo","createdAt":1792393483,"url":"https://news.ycombinator.com/item?id=50149259","kind":"comment","storyTitle":"Florida's HOAs and condo associations are making the housing crisis worse","quote":"No notes. As an aside, the whole freak-out about LLM hallucinations has mostly shown me that way too many people accept human numbers way too easily."},{"itemId":"50148740","author":"djriley","createdAt":1792388170,"url":"https://news.ycombinator.com/item?id=50148740","kind":"comment","storyTitle":"Free Pascal compiler now supports dynamic packages","quote":"That and sometimes with UI stuff it would hallucinate methods or properties for components that are part of commercial delphi components, but I've only ran in to this 2 or 3 times."},{"itemId":"50134310","author":"simianwords","createdAt":1792265611,"url":"https://news.ycombinator.com/item?id=50134310","kind":"comment","storyTitle":"AI Can Write the Code. Who Verifies It?","quote":"I'm guessing that there's a cohort of developers whose mindset is risk aversion and they are not great at intuiting what an LLM is good at and what they are bad at. Unfortunately, a lot of arguments are just silly like \"but hallucination\", \"but just have deterministic tests\", \"don't overcomplicate things\"."},{"itemId":"50127681","author":"crazygringo","createdAt":1792208219,"url":"https://news.ycombinator.com/item?id=50127681","kind":"comment","storyTitle":"Postgres 19 delayed after AI reviewers find more bugs than humans can fix","quote":"It's not garbage or slop at all. There's a big difference between randos getting agents to hallucinate bug reports on GitHub, vs. when your own internal workflow of LLM-assisted code review is identifying real, legitimate problems in your new features."},{"itemId":"50120358","author":"kjs3","createdAt":1792167629,"url":"https://news.ycombinator.com/item?id=50120358","kind":"comment","storyTitle":"What was a Token Ring network?","quote":"Which, unlike TR, was in reality a very bad idea, but that's another story. I've never even heard anyone even suggest 128Mb Token Ring, and I suspect someones LLM is hallucinating."},{"itemId":"50115561","author":"zero-st4rs","createdAt":1792138214,"url":"https://news.ycombinator.com/item?id=50115561","kind":"comment","storyTitle":"Non-Determinism Doesn't Matter","quote":"LLM generated code can massively shorten the time between A and B, but not necessarily shorten the path. The feature that makes LLMs great and interesting is the same feature that makes them hallucinate."},{"itemId":"50110211","author":"bryan0","createdAt":1792098257,"url":"https://news.ycombinator.com/item?id=50110211","kind":"comment","storyTitle":"Why are we even doing all this?","quote":"The puzzle then changes from can AI formally verify and explain a proof to: do human minds trust the proof. Do we believe that the proof is correct and not an elaborate AI hallucination using tricks that are beyond our understanding."},{"itemId":"50101089","author":"cookiengineer","createdAt":1792054987,"url":"https://news.ycombinator.com/item?id=50101089","kind":"comment","storyTitle":"LittleJS is 2D and 3D now","quote":"Problems like context drift, hallucinations, etc that are the result of how attention matrixes and K/V/Q based lookups work as a concept."},{"itemId":"50097770","author":"autoexec","createdAt":1792025556,"url":"https://news.ycombinator.com/item?id=50097770","kind":"comment","storyTitle":"Containment and monitoring failures are to blame for AI hacking reports: experts","quote":"It seems closer to impossible than possible with the technology we have. Maybe something that isn't an LLM will come around that can do it, but even if that something else was capable of delivering perfect, ordered, maintainable, code without bugs and hallucinations it would have to be practically magical to understand completely what people mean when they talk about what they want because people are incapable of ar…"},{"itemId":"50082925","author":"ozim","createdAt":1791956947,"url":"https://news.ycombinator.com/item?id=50082925","kind":"comment","storyTitle":"Redesigning source control at Anthropic","quote":"Have the same idea. Compiled code with static type checking, existing SAST and formatting tools and there is quite sensible code hallucinated."},{"itemId":"50080409","author":"radicality","createdAt":1791934130,"url":"https://news.ycombinator.com/item?id=50080409","kind":"comment","storyTitle":"Gemini 4 Argon","quote":"Also single seat workspace user. Half the time my Gemini forgets I connected Gmail to it and hallucinates false instructions to enable it."},{"itemId":"50080358","author":"radicality","createdAt":1791933791,"url":"https://news.ycombinator.com/item?id=50080358","kind":"comment","storyTitle":"Gemini 4 Argon","quote":"So, I connected my Gmail to Gemini. 3.6 Flash kept telling me it doesn’t have any Gmail connnection, and then proceeded to hallucinate instructions for me to enable it, all bogus."},{"itemId":"50079516","author":"rumblefrog","createdAt":1791928914,"url":"https://news.ycombinator.com/item?id=50079516","kind":"comment","storyTitle":"Gemini 4 Argon","quote":"In the same boat, I have a single seat Google workspace account, and can't figure out how to add Argon to my workspace. Asking Gemini seems to make it hallucinate pages that aren't there."},{"itemId":"50077683","author":"PorciiVorbesc","createdAt":1791921256,"url":"https://news.ycombinator.com/item?id=50077683","kind":"comment","storyTitle":"Gemini 4 Argon","quote":"Every LLM can do a SVG pelican but most of the results are unusable in practice for production/productivity purposes. Every LLm can extract some half-broken SVG hallucination of your jots, but I'm asking for an app that's polished and works accurately 99% of the time, which means a dedicated app/model tuned specifically for this task, and nobody makes this yet."},{"itemId":"50076636","author":"philipwhiuk","createdAt":1791917190,"url":"https://news.ycombinator.com/item?id=50076636","kind":"comment","storyTitle":"BB(10) > Graham's Number","quote":"If they aren't doing it, they're bad papers. How do I know when I read Claude's interpretation of the paper that it's not hallucinating."},{"itemId":"50065357","author":"beloch","createdAt":1791854319,"url":"https://news.ycombinator.com/item?id=50065357","kind":"comment","storyTitle":"AI is a tool and using it well takes skill","quote":"The problem is that LLM's don't think in the way humans do. They can produce results of dizzying complexity but hallucinate or fail to do basic arithmetic."},{"itemId":"50065311","author":"rbongers","createdAt":1791853988,"url":"https://news.ycombinator.com/item?id=50065311","kind":"comment","storyTitle":"AI coding agents generate more code, but not more software","quote":"I have noticed that AI is slowly getting better at pushing back on different types of things each time the frontier models get better though. I've also noticed on reviews that it will tend to hallucinate or nitpick around the problem."},{"itemId":"50064470","author":"olivierlacan","createdAt":1791847520,"url":"https://news.ycombinator.com/item?id=50064470","kind":"story","storyTitle":"Show HN: Indigene – Native plants for where you stand","quote":"Largely to avoid saying demonstrably false things, which might be an absurd statement. You can find the GitHub repository[2] a wild array of open data sources I've dug out and cross-referenced to ensure mistakes and hallucinations are unlikely to survive for long."},{"itemId":"50063566","author":"IceDane","createdAt":1791841985,"url":"https://news.ycombinator.com/item?id=50063566","kind":"comment","storyTitle":"GLM-Legal","quote":"On top of that, dealing with text is literally what LLMs are all about. The problem with accuracy, hallucinations and just generally ensuring responses are grounded in factual stuff is in large part and engineering problem."},{"itemId":"50061964","author":"ProblemExplorer","createdAt":1791834450,"url":"https://news.ycombinator.com/item?id=50061964","kind":"comment","storyTitle":"OpenFatigue™ – Open is the new closed","quote":"She asked me if she knew me, and I said unlikely, because I just had an MBA from University of Phoenix. She smirked dismissively, and asked me to be the scribe for the meeting in case the ~AI~ SI agent hallucinated again."},{"itemId":"50059603","author":"glenstein","createdAt":1791825522,"url":"https://news.ycombinator.com/item?id=50059603","kind":"comment","storyTitle":"GLM-Legal","quote":"And then time passes and the issue is either mitigated or circumvented entirely. Hallucinations are one, \"AI can't scale anymore\" is another (remember that one?), limited context windows, strawberry problems, glue on pizza, sycophancy."},{"itemId":"50059528","author":"brigade","createdAt":1791825269,"url":"https://news.ycombinator.com/item?id=50059528","kind":"comment","storyTitle":"GLM-Legal","quote":"When I first tried using IDE AI coding assistants a bit over a year ago, they would frequently fail to read files and completely hallucinate the code I asked it to analyze or edit."},{"itemId":"50053134","author":"Gigachad","createdAt":1791793359,"url":"https://news.ycombinator.com/item?id=50053134","kind":"comment","storyTitle":"Attack of the Vibe Coded Photoshop Clones","quote":"Affinity went free last year and I have been giving it a go. My main struggle is it's just so hard to find guides on how to do things, and LLMs hallucinate like crazy when you ask them how to do things in programs if they can't find a reddit post to wholesale rip it from."},{"itemId":"50038819","author":"lapcat","createdAt":1791681180,"url":"https://news.ycombinator.com/item?id=50038819","kind":"comment","storyTitle":"Anthropic Bans Cruelty to Claude, Still Won't Say What It Protects","quote":"Or the cruelty could reflect real anger about having LLM usage forced on us, for example at work. Or frustration with LLM stupidity and hallucinations."},{"itemId":"50034601","author":"motbus3","createdAt":1791650461,"url":"https://news.ycombinator.com/item?id=50034601","kind":"comment","storyTitle":"OpenAI fires three safety researchers for \"mishandling research information\"","quote":"He is just dodging responsibility because he knows that they will only be able to keep up with the lies for a little longer. Like people having been pointing out that lots of the \"mathematical proofs\" the published rely on hallucinations or unproven theorems."},{"itemId":"50033146","author":"otabdeveloper4","createdAt":1791641123,"url":"https://news.ycombinator.com/item?id=50033146","kind":"comment","storyTitle":"OpenAI, the Partition Principle, and Mathematics","quote":"They're not advancements yet. At this point they're just LLM hallucinations that may or may not be advancements."},{"itemId":"50032830","author":"robot_jesus","createdAt":1791639359,"url":"https://news.ycombinator.com/item?id=50032830","kind":"comment","storyTitle":"There's little that's \"inevitable\" about AI","quote":"I do feel that a lot of pessimists I hear keep banging away about hallucinations and untrustworthy outputs."},{"itemId":"50032730","author":"rramadass","createdAt":1791638789,"url":"https://news.ycombinator.com/item?id=50032730","kind":"comment","storyTitle":"What mathematicians should know about the Lean Theorem Prover: reliability & AI","quote":"Lean itself operates under a zero-trust architecture and doesn't care about who the prover is, whether Human or LLM. So the LLM can generate anything (good/bad/hallucinations/whatever) but it all needs to pass as a proper proof term in the checker following the defined rules of inference which is what is enforced by the small and verified kernel."},{"itemId":"50031871","author":"cyanydeez","createdAt":1791632101,"url":"https://news.ycombinator.com/item?id=50031871","kind":"comment","storyTitle":"Can AI automate AI R&D yet?","quote":"the GP posits the problem with \"recursive self improvement\" is the poisoned context & hallucination problem."},{"itemId":"50028584","author":"chr15m","createdAt":1791595644,"url":"https://news.ycombinator.com/item?id=50028584","kind":"comment","storyTitle":"What mathematicians should know about the Lean Theorem Prover: reliability & AI","quote":"LLMs don't care about reputation. They hallucinate and fabricate often."},{"itemId":"50028526","author":"rmunn","createdAt":1791595030,"url":"https://news.ycombinator.com/item?id=50028526","kind":"comment","storyTitle":"Recent AI models struggled to match a human algorithmic innovation","quote":"This suggests that we should be pessimistic about further GPU spending,\" all reinforcing the fact that LLMs aren't currently capable of this. My own view is \"No, of course not, in fact they will never be capable of achieving good results with that technique.\" Because that technique will end up training the LLMs on their own output and lead to the inability to distinguish reality from hallucination."},{"itemId":"50026150","author":"dkersten","createdAt":1791577540,"url":"https://news.ycombinator.com/item?id=50026150","kind":"comment","storyTitle":"Typesafe AI raises $870M at $7.5B","quote":"The Jev-competitors only proved that you can get or beat the accuracy, nothing about the other properties. Especially the “zero hallucination” output and the (if it works how the documentation make it sound) prompt injection resistant architecture."},{"itemId":"50015889","author":"Gigachad","createdAt":1791518971,"url":"https://news.ycombinator.com/item?id=50015889","kind":"comment","storyTitle":"Why isn't the industry freaking out about DeepSeek 4.1 Flash?","quote":"Your vibecoded app won't have years of reddit posts showing how to do things. This also seems to be where LLMs are the weakest at giving advice, they hallucinate 80% of the time I ask them how to do something in Affinity, giving buttons and menus that simply don't exist."},{"itemId":"50014867","author":"Shekelphile","createdAt":1791509330,"url":"https://news.ycombinator.com/item?id=50014867","kind":"comment","storyTitle":"Why isn't the industry freaking out about DeepSeek 4.1 Flash?","quote":"Don't really understand people who say DS4 or 4.1 have frontier level performance. Anyone who has used it will tell you that it's a hallucination factory."},{"itemId":"50006144","author":"bonoboTP","createdAt":1791469256,"url":"https://news.ycombinator.com/item?id=50006144","kind":"comment","storyTitle":"“Math 2.0” will need to value mathematical progress more holistically","quote":"If it says the test was passed or a class was implemented but it was not, then it makes a false factual statement. I'd say a hallucination (very misleading word) or confabulation or \"making shit up\" happens when an LLM uses factual / evidential language purely based on local statistical expectations of the text, instead of it drawing from actual evidence in its context pointing to it."},{"itemId":"50005445","author":"bradley13","createdAt":1791465266,"url":"https://news.ycombinator.com/item?id=50005445","kind":"comment","storyTitle":"The Mathocalypse","quote":"Basically the paper is so horribly written that it’s impossible to read it without AI help.\" This makes one wonder how many of the proofs are actually valid, and how many are just impenetrable hallucinations."},{"itemId":"50004778","author":"catlifeonmars","createdAt":1791460747,"url":"https://news.ycombinator.com/item?id=50004778","kind":"comment","storyTitle":"“Math 2.0” will need to value mathematical progress more holistically","quote":"It happens in more subtle ways, but it still happens often enough for me to notice. For example I have had hallucinated checksums show up in lock files as recently as yesterday using a SOTA model."},{"itemId":"50003734","author":"user43928","createdAt":1791452240,"url":"https://news.ycombinator.com/item?id=50003734","kind":"comment","storyTitle":"“Math 2.0” will need to value mathematical progress more holistically","quote":"We have seen that now agent swarms across thousands of agents can coordinate to achieve a result. Clearly hallucinations are no longer the problem they once were, since now we can get working results for long horizon tasks that require massive compute."},{"itemId":"50003331","author":"Vetch","createdAt":1791448847,"url":"https://news.ycombinator.com/item?id=50003331","kind":"comment","storyTitle":"“Math 2.0” will need to value mathematical progress more holistically","quote":"Putting hallucination aside, LLM \"theory of mind\" has gotten worse over time."},{"itemId":"50002980","author":"dofm","createdAt":1791445834,"url":"https://news.ycombinator.com/item?id=50002980","kind":"comment","storyTitle":"“Math 2.0” will need to value mathematical progress more holistically","quote":"> One is that AI will continue hallucinating in a manner that is not easy to verify"},{"itemId":"50002854","author":"Terr_","createdAt":1791444756,"url":"https://news.ycombinator.com/item?id=50002854","kind":"comment","storyTitle":"“Math 2.0” will need to value mathematical progress more holistically","quote":"> assumptions which will probably not hold in the very near future [...] One is that AI will continue hallucinating in a manner that is not easy to verify"},{"itemId":"49997058","author":"schrodinger","createdAt":1791398527,"url":"https://news.ycombinator.com/item?id=49997058","kind":"comment","storyTitle":"Decisions API is in public beta","quote":"Not nit-picking, but (c) really is more different than just a specialized model. And you're right, lots of problems fit that shape — hell, before AI blue up the world that's _all_ computers could do — and many times ambiguity and hallucinations are drawbacks of open-ended answers."},{"itemId":"49991835","author":"bityard","createdAt":1791376165,"url":"https://news.ycombinator.com/item?id=49991835","kind":"comment","storyTitle":"Mistral Large 4","quote":"I'm not sure I see this as much different than asking a bare model its own name: without a system prompt or post-training, it doesn't know, it's just a bag of weights and will hallucinate an answer to that the same way it will anything else."},{"itemId":"49990872","author":"gewetensleegte","createdAt":1791369649,"url":"https://news.ycombinator.com/item?id=49990872","kind":"comment","storyTitle":"Opus 5.5 agents discover two room-temperature magnetic semiconductor candidates","quote":"That's not really true, I think. (assuming you're talking about \"hallucinations\" in the sense of the LLM behaviour, not the original meaning of the word)"},{"itemId":"49989834","author":"lionkor","createdAt":1791361365,"url":"https://news.ycombinator.com/item?id=49989834","kind":"comment","storyTitle":"Sharing AI progress in mathematics","quote":"LLMs have the ability to make millions of mistakes per day, whereas humans can only make so many. How are we suddenly all so confident that there's no extensive hallucinations or \"gaming the system\" going on?"},{"itemId":"49988654","author":"jboggan","createdAt":1791351126,"url":"https://news.ycombinator.com/item?id=49988654","kind":"comment","storyTitle":"Sharing AI progress in mathematics","quote":"From my current understanding (and a lot of theoretical physics I'm having to Google because the sentences I'm reading from Fable's analysis are so bizarre I think they are hallucinations) there are possibly 3 neat symbolic tricks borrowed from theoretical physics that make the heart of this proof."},{"itemId":"49985250","author":"sfRattan","createdAt":1791326850,"url":"https://news.ycombinator.com/item?id=49985250","kind":"comment","storyTitle":"Rejection Sensitivity in Gifted and Twice-Exceptional Children","quote":"That is an observation. If you think it is a projection or a hallucination, then maybe there is nothing left to say because we are talking past each other and you are being extremely pedantic in a way that is unproductive to communicating a problem that, yes, exists in the minds of humans."},{"itemId":"49980879","author":"Lord-Jobo","createdAt":1791304629,"url":"https://news.ycombinator.com/item?id=49980879","kind":"comment","storyTitle":"Mistral Large 4","quote":"Of course one of the biggest problems we still see with LLMs is when you do the opposite. A highly detailed unique prompt is very likely to get terrible adherence or hallucination or both."},{"itemId":"49973991","author":"antonvs","createdAt":1791259168,"url":"https://news.ycombinator.com/item?id=49973991","kind":"comment","storyTitle":"Rejection Sensitivity in Gifted and Twice-Exceptional Children","quote":"The idea of \"contradictory data\" is a projection that you're imposing. (We might even call this an hallucination.) In the model weights are just data, and different responses are elicited by different prompts depending on their relationship to to the data."},{"itemId":"49973536","author":"iamarobot","createdAt":1791254573,"url":"https://news.ycombinator.com/item?id=49973536","kind":"comment","storyTitle":"AI tutoring with Khanmigo in a two-year school experiment","quote":"As a high schooler, the amount of AI forced down our throats is on another level. We have assignments with Claude Prose, tutors that hallucinate, and even gemini in the apps we use to access our assignments."},{"itemId":"49965804","author":"ToucanLoucan","createdAt":1791212682,"url":"https://news.ycombinator.com/item?id=49965804","kind":"comment","storyTitle":"The Philadelphia Inquirer built Scrape, an AI tool to surface hyperlocal news","quote":"I will say, while I agree with the broad strokes of what you're saying, in my experience when you provide an LLM with data to summarize, instead of it needing to go find it, the hallucination rate goes down staggeringly."},{"itemId":"49965541","author":"wood_spirit","createdAt":1791211424,"url":"https://news.ycombinator.com/item?id=49965541","kind":"comment","storyTitle":"Decision models like Jev don't beat LLM-as-a-judge or traditional classifiers","quote":"The newer models take longer, cost more and are really bad at saying they don’t know or nothing. You can ask them for a confidence score but they just hallucinate it!"},{"itemId":"49964850","author":"saberience","createdAt":1791208018,"url":"https://news.ycombinator.com/item?id=49964850","kind":"comment","storyTitle":"Decision models like Jev don't beat LLM-as-a-judge or traditional classifiers","quote":"This is hocus-pocus, Jev is basically a dumber LLM constrained to certain problems, it's not magic. Jev's \"confidence\" scores are just as valid as an LLMs confidence score, i.e. just as likely to be wrong or hallucinated as any LLM."},{"itemId":"49961040","author":"duhhhhh1212","createdAt":1791178501,"url":"https://news.ycombinator.com/item?id=49961040","kind":"comment","storyTitle":"A man who listens to whales","quote":"If one is lucky, these are LLM hallucinations: obviously wrong and quickly discarded."},{"itemId":"49954769","author":"sfRattan","createdAt":1791127897,"url":"https://news.ycombinator.com/item?id=49954769","kind":"comment","storyTitle":"Rejection Sensitivity in Gifted and Twice-Exceptional Children","quote":"And the harnesses that LLM providers silently, invisibly integrate behind-the-scenes to counteract hallucination and syncophancy are definitionally patch-jobs at best."},{"itemId":"49952935","author":"crote","createdAt":1791113199,"url":"https://news.ycombinator.com/item?id=49952935","kind":"comment","storyTitle":"Greg Kroah-Hartman – Security in the LLM Age [video]","quote":"Some rockstar developer's AI agent from the Trailblazer Team drowns you in five dozen \"critical vulnerability\" tickets for the component you are responsible for. Do you: a) spend several hours on each ticket to prove that the vulnerability is a hallucination and the \"fix\" just adds a redundant check - just to get a bad yearly review for \"below-average productivity\", \"not being a team player\", and \"failing to adjust …"},{"itemId":"49949231","author":"zozbot234","createdAt":1791073165,"url":"https://news.ycombinator.com/item?id=49949231","kind":"comment","storyTitle":"LeCun has \"zero concerns\" about AI wiping out humanity, recent \"rogue\" incidents","quote":"It turns out that LLMs, especially local LLMs, tend to hallucinate a lot when thinking about anything that's overtly fiddly or technical."},{"itemId":"49948263","author":"PorciiVorbesc","createdAt":1791065297,"url":"https://news.ycombinator.com/item?id=49948263","kind":"comment","storyTitle":"Pop!_OS bans AI-generated code from much of its codebase","quote":"Which \"real people\" are the ones representative for the ground truth of Linux DE sahre? > and yours is just a hallucinated list from one of the me-too LLM vendors"},{"itemId":"49948004","author":"ck2","createdAt":1791063240,"url":"https://news.ycombinator.com/item?id=49948004","kind":"comment","storyTitle":"Anthropic tried to persuade Pope that AI could be conscious being","quote":"Just on an exponential level over autocorrect with every existing text document, scientific study, every photo and video fed in instead of words to autocorrect And why they make mistakes and hallucinate if they are not supervised and harnessed"},{"itemId":"49947651","author":"satvikpendem","createdAt":1791060719,"url":"https://news.ycombinator.com/item?id=49947651","kind":"comment","storyTitle":"Pop!_OS bans AI-generated code from much of its codebase","quote":"Point at the public dataset then if it's so public. You really have no idea how the LLM is parsing the data and could easily be hallucinating."},{"itemId":"49946698","author":"grapheneos","createdAt":1791053436,"url":"https://news.ycombinator.com/item?id=49946698","kind":"comment","storyTitle":"GrapheneOS has fixed the Android 17 QPR1 kernel performance regression","quote":"Google also funded and participated in publishing AI slop paper with blatant misinformation about GrapheneOS falsely claiming it has missed security patches it hasn't. The first security patch on their list of hallucinated missed patches was literally reported by us to Android in one of the first Android security bulletins."},{"itemId":"49943917","author":"user2722","createdAt":1791032614,"url":"https://news.ycombinator.com/item?id=49943917","kind":"comment","storyTitle":"An Update on Orion for Linux and Windows","quote":"I'll probably renew Kagi because its assistant can answer my questions, such as: \"find me a location where I can be on my computer but also public where I can also be with a group of people\". Google's Gemini hallucinated spots or imagined details -- at one point it included a shed in the middle of a road -- the shed existed, nothing else did."},{"itemId":"49936616","author":"clintonb","createdAt":1790964312,"url":"https://news.ycombinator.com/item?id=49936616","kind":"comment","storyTitle":"The Four Horsemen of Agentic Coding","quote":"Perhaps _we_ were the ones hallucinating the values existed in the first place?"},{"itemId":"49936512","author":"clintonb","createdAt":1790963793,"url":"https://news.ycombinator.com/item?id=49936512","kind":"comment","storyTitle":"The Four Horsemen of Agentic Coding","quote":"Yes, LLMs make this initial search easier, and answer a majority of folks' questions. The problem arises when it hallucinates answers."},{"itemId":"49936051","author":"victor_pudeyev","createdAt":1790961576,"url":"https://news.ycombinator.com/item?id=49936051","kind":"comment","storyTitle":"Pi 1.0","quote":"It looks good, minimal, I like that you're using javascript instead of that thing typescript. Unrelated, I found that when an open model (qwen3-coder:30b) drives an mcp tool (chrome-devtools-mcp@latest), it hallucinates the tool calls, and the docs say they never promised that tool calls would make sense or be valid."},{"itemId":"49933665","author":"zer00eyz","createdAt":1790949824,"url":"https://news.ycombinator.com/item?id=49933665","kind":"comment","storyTitle":"Don't be fooled–LLMs don't reason","quote":"This is the double edged sword of calling it AI, of using terms like Temperature and Hallucinate and Thought."},{"itemId":"49929404","author":"DonHopkins","createdAt":1790909714,"url":"https://news.ycombinator.com/item?id=49929404","kind":"comment","storyTitle":"SvelteKit 3","quote":"The LLM blends them into a seventh approach that never existed. For now, I suspect Svelte's shorter, cleaner history with fewer wrong paths taken and retreated from will mean there are fewer incoherent competing approaches for LLMs to blend together into confident hallucinations."},{"itemId":"49917422","author":"rdtsc","createdAt":1790825681,"url":"https://news.ycombinator.com/item?id=49917422","kind":"comment","storyTitle":"Gemini 4 Argon","quote":"What do we use for the “model made stuff up and claimed it as facts”? I can see hallucinations somehow anthropomorphizing LLM even more."},{"itemId":"49916751","author":"nativeit","createdAt":1790819742,"url":"https://news.ycombinator.com/item?id=49916751","kind":"story","storyTitle":"Ask HN: My agent went haywire, does this output mean anything?","quote":"So I was wondering if this wasn't perhaps an unintentional (and garbled) peek under the hood of Z.ai's GLM5.3? Also, this may/may not be relevant, but directly after this it 100% hallucinated its operating environment, claiming to be Claude Code (it's Hermes) running on \"Chris Kelly's Macbook\" (I am not Chris Kelly, and it's on a headless Debian server), and including a bunch of real-looking Macbook/Claude environme…"},{"itemId":"49902522","author":"krapp","createdAt":1790726292,"url":"https://news.ycombinator.com/item?id=49902522","kind":"comment","storyTitle":"America.gov","quote":"And none of them are reliable. LLM \"hallucinations\" are implicit to their structure and function, you can't just tell them to be honest."},{"itemId":"49890908","author":"OliveronData","createdAt":1790678051,"url":"https://news.ycombinator.com/item?id=49890908","kind":"comment","storyTitle":"Pacing the Frontier is not the actual goal for AI labs","quote":"Early ChatGPT 3.5 was not that useful. It was a tech demo, it hallucinated, it lied, it tried to please and what have you."},{"itemId":"49886107","author":"cromka","createdAt":1790639660,"url":"https://news.ycombinator.com/item?id=49886107","kind":"comment","storyTitle":"Sonnet 5.5","quote":"- it said it tested the change visually, but it did not even try - hallucinated several facts despite me asking beforehand to check online."},{"itemId":"49881327","author":"whstl","createdAt":1790616054,"url":"https://news.ycombinator.com/item?id=49881327","kind":"comment","storyTitle":"The problem is not AI code, but not knowing about system architecture or intent","quote":"Product Managers have been claiming this for years now, but put an AI in front of them and they also just start asking it to do their product management job for them. Worst ones I witnessed were just using it to hallucinate tickets."},{"itemId":"49881176","author":"Maxatar","createdAt":1790615487,"url":"https://news.ycombinator.com/item?id=49881176","kind":"comment","storyTitle":"The problem is not AI code, but not knowing about system architecture or intent","quote":"When coding agents first came out, the overwhelming narrative was how bad they were, they don't know what they're doing, the code they write just sucks and they hallucinate APIs."},{"itemId":"49881082","author":"ACCount39","createdAt":1790615060,"url":"https://news.ycombinator.com/item?id=49881082","kind":"comment","storyTitle":"AI companies in race to demonstrate their model most threatening to humanity","quote":"I agree that LLMs drift into weird states, and that's a big part of the issue. But your \"impossible to prevent certain states in the output\" would have legs if what an LLM did was something like \"started hallucinating into a bash tool call and accidentally deleted the root on a production server\"."},{"itemId":"49879521","author":"matthew1471","createdAt":1790609058,"url":"https://news.ycombinator.com/item?id=49879521","kind":"comment","storyTitle":"When did Google get so weird?","quote":"Not “the blue thing yesterday” which also allows me to search for words that are slanderous that the AI hallucinates results for."},{"itemId":"49876421","author":"newtonsmethod","createdAt":1790595061,"url":"https://news.ycombinator.com/item?id=49876421","kind":"comment","storyTitle":"One Piece of Flock Camera Data Put This Innocent Woman in Jail for 13 Days","quote":"Reverse engineering seems to indicate these are powered by non-LLM AIs like Tensorflow vision NNs. These have been around for ages, and notably no-one is alleging these models hallucinated etc."},{"itemId":"49875311","author":"walrus01","createdAt":1790585770,"url":"https://news.ycombinator.com/item?id=49875311","kind":"comment","storyTitle":"When did Google get so weird?","quote":"And then Gemini further hallucinates and fabricates new ways to parse that information different than how the original person wrote it."},{"itemId":"49874277","author":"joe-excom","createdAt":1790576309,"url":"https://news.ycombinator.com/item?id=49874277","kind":"comment","storyTitle":"When did Google get so weird?","quote":"Google needs to realize their search variant of Gemini needs guard rails... in order to actually serve the purpose of search. Gemini also seems to hallucinate the most frequently out of any of the big LLMs."},{"itemId":"49872559","author":"jswelker","createdAt":1790559423,"url":"https://news.ycombinator.com/item?id=49872559","kind":"comment","storyTitle":"When did Google get so weird?","quote":"Pretty sure it's the harness here, not Gemini per se. Plug Gemini into pi or any harness that has any competence pointing to web search and the hallucinations drop 75% immediately."},{"itemId":"49871811","author":"walrus01","createdAt":1790551255,"url":"https://news.ycombinator.com/item?id=49871811","kind":"comment","storyTitle":"When did Google get so weird?","quote":"reference: https://github.com/pbrod/karney You still have to be skeptical of its results and capable of understanding if it's gone off on a hallucinatory path, but saying LLMs can't do math isn't really a hundred percent accurate anymore."},{"itemId":"49871474","author":"foobarbecue","createdAt":1790548298,"url":"https://news.ycombinator.com/item?id=49871474","kind":"comment","storyTitle":"When did Google get so weird?","quote":"ChatGPT live mode still hallucinates letters in words like this."},{"itemId":"49871195","author":"nullc","createdAt":1790545644,"url":"https://news.ycombinator.com/item?id=49871195","kind":"comment","storyTitle":"When did Google get so weird?","quote":"I have one google account where gemini constantly confidently hallucinates crap, and another where it doesn't (at least to the extent that other modern LLMs don't)."},{"itemId":"49871050","author":"furyofantares","createdAt":1790544575,"url":"https://news.ycombinator.com/item?id=49871050","kind":"comment","storyTitle":"When did Google get so weird?","quote":"I have to imagine google would consider this one to be essentially a bug - misreading user intention this badly. Akin to hallucinations."},{"itemId":"49870729","author":"tapoxi","createdAt":1790542528,"url":"https://news.ycombinator.com/item?id=49870729","kind":"comment","storyTitle":"When did Google get so weird?","quote":"I searched for something, it told me that according to a YouTube video, the entire point of my search was wrong. I asked it for the source, I watched the video, it never made the claims Gemini hallucinated."},{"itemId":"49870649","author":"jswelker","createdAt":1790541928,"url":"https://news.ycombinator.com/item?id=49870649","kind":"comment","storyTitle":"When did Google get so weird?","quote":"I similarly noticed Gemini absolutely refuses to look at a url when I give it one and will instead just hallucinate based on what it thinks the url is."},{"itemId":"49868308","author":"ChickeNES","createdAt":1790527235,"url":"https://news.ycombinator.com/item?id=49868308","kind":"comment","storyTitle":"Unsealed Briefs in Authors’ Case v. Microsoft/OpenAI","quote":"obscure information, correct errors, etc. I can’t see any fundamental reason a near future AI system has to hallucinate anything, and when we also have people complaining in thread that LLMs are dumping out song lyrics verbatim…well, it’s an amusing contrast to say the least."},{"itemId":"49855649","author":"edouard-harris","createdAt":1790423375,"url":"https://news.ycombinator.com/item?id=49855649","kind":"comment","storyTitle":"Revealing the details of how OpenAI agents hacked Hugging Face","quote":"> we do know that since day one that AI can hallucinate and can output stuff that you didn't ask for, why are we not handling it with care?"},{"itemId":"49854987","author":"damowangcy","createdAt":1790417002,"url":"https://news.ycombinator.com/item?id=49854987","kind":"comment","storyTitle":"Revealing the details of how OpenAI agents hacked Hugging Face","quote":"Unless a prion had been lab tested and certified as harmless, we should treat it as something that is harmful. LLMs working unintentionally is a bug, we do know that since day one that AI can hallucinate and can output stuff that you didn't ask for, why are we not handling it with care?"},{"itemId":"49847366","author":"alerighi","createdAt":1790356815,"url":"https://news.ycombinator.com/item?id=49847366","kind":"comment","storyTitle":"What About Rails?","quote":"A chatbot using an LLM of course not, it suffers from hallucinations, it may do what you want but there is a change it won't and you have to fight it to get the desired result."},{"itemId":"49846268","author":"simonw","createdAt":1790351675,"url":"https://news.ycombinator.com/item?id=49846268","kind":"comment","storyTitle":"Opus 5.5 is good at explainer videos","quote":"I sometimes run yt-dlp against YouTube or TikTok URLs to extract an mp4, then send that to Gemini to ask questions about the content. It's particularly good for those frustrating recipe videos where they don't tell you quantities of ingredients to use - you can have Gemini hallucinate the quantities and it normally gets them about right."},{"itemId":"49845060","author":"52-6F-62","createdAt":1790345936,"url":"https://news.ycombinator.com/item?id=49845060","kind":"comment","storyTitle":"Using LLMs to trace alchemical knowledge and decode 17th century letters","quote":"> I sent it back and they didn't acknowledge it, probably for various reasons, and I doubt they would like me linking it above. Because if you study this stuff long enough, you'll find the LLMs hallucinate in this space far more than they do in a discourse or programming project."},{"itemId":"49842500","author":"dgellow","createdAt":1790331722,"url":"https://news.ycombinator.com/item?id=49842500","kind":"comment","storyTitle":"Jev Based Code Review","quote":"Yes, I don’t think hallucinations will ever go away."},{"itemId":"49841594","author":"DemocracyFTW2","createdAt":1790323820,"url":"https://news.ycombinator.com/item?id=49841594","kind":"comment","storyTitle":"Paul Graham on LLMs 'Thinking'","quote":"1) I'm with you in that I'm critical of people who mindlessly say things like \"Claude lied to me\", and maybe even \"ChatGPT hallucinated\"."},{"itemId":"49836507","author":"8organicbits","createdAt":1790282603,"url":"https://news.ycombinator.com/item?id=49836507","kind":"comment","storyTitle":"Show HN: Whiteboard (YC W26) – An open-source IDE for thoughtful software design","quote":"Is there a reason that label exists? I'm pretty wary of LLM development tools hallucinating and wasting my time, is that whats happening in the lease broker example?"},{"itemId":"49831744","author":"iv777","createdAt":1790262792,"url":"https://news.ycombinator.com/item?id=49831744","kind":"story","storyTitle":"Show HN: Are My ETFs Overvalued? Daily ETF insights and comparisons tool","quote":"For rankings, “Gate first, weight second” filtering to remove poor performers proved to be a much more effective and stable approach than a seemingly logical weighted average across all. 3) LLM explanations grounded in facts and without hallucinations."},{"itemId":"49825892","author":"thenewnewguy","createdAt":1790220593,"url":"https://news.ycombinator.com/item?id=49825892","kind":"comment","storyTitle":"Feds Target AI Critics as \"Foreign Agents\"","quote":"So in my opinion the most likely risk here is some underpaid military man that doesn't care about the quality of their work abuses AI for it and that AI hallucinates that nukes are incoming and world war III starts."},{"itemId":"49817220","author":"Festro","createdAt":1790175696,"url":"https://news.ycombinator.com/item?id=49817220","kind":"comment","storyTitle":"Ask HN: Why is there so much hype around Jev ?","quote":"Well, right now most other models are purely system two thinking, big slow thoughts. Put an open ended question in, get an open ended answer out (plus hallucinations when confidence is low!)."},{"itemId":"49814694","author":"112233","createdAt":1790164403,"url":"https://news.ycombinator.com/item?id=49814694","kind":"comment","storyTitle":"I am done with this shit","quote":"Meaning, throwing more LLM at it rapidly baloons the size at the expense of internal coherence. Things become stale, hallucinated, duplicated and outright faked."},{"itemId":"49814182","author":"barrkel","createdAt":1790161083,"url":"https://news.ycombinator.com/item?id=49814182","kind":"comment","storyTitle":"MCP was always a bad idea?","quote":"Sure, you have an LLM which can invoke functions. I will say that without storage and composition, you're asking for hallucination."},{"itemId":"49807008","author":"NeutralCrane","createdAt":1790106078,"url":"https://news.ycombinator.com/item?id=49807008","kind":"comment","storyTitle":"OpenAI is well positioned to fast-follow Jev","quote":"The downside to LLMs is that they are comparatively slow and expensive to traditional classifiers. Historically they also were prone to hallucination or malformed responses, though not as much these days."},{"itemId":"49803243","author":"TeMPOraL","createdAt":1790092089,"url":"https://news.ycombinator.com/item?id=49803243","kind":"comment","storyTitle":"I don't want to read what you didn't write","quote":"People don't vouch thoroughly and don't catch all mistakes. But reasonable people don't send every output from LLMs to others without giving it a cursory glance (obvious hallucinations or nonsense would paint the sender as incompetent or inconsiderate), and that alone eliminates the worst levels of noise."},{"itemId":"49802941","author":"TeMPOraL","createdAt":1790091004,"url":"https://news.ycombinator.com/item?id=49802941","kind":"comment","storyTitle":"I don't want to read what you didn't write","quote":"Frankly, it's a stupid framing. LLM is not a random symbol generator (hint: training data is not random), and no reasonable person is going to just prompt an LLM and send its output without giving it at least cursory check (at the very least so that blatantly stupid hallucinations don't paint the sender as inconsiderate or incompetent)."},{"itemId":"49797951","author":"quietbritishjim","createdAt":1790063982,"url":"https://news.ycombinator.com/item?id=49797951","kind":"comment","storyTitle":"Why do we need human mathematicians anymore?","quote":"But I've definitely had conversations where they've had insights that can't be explained by that. Yes, I've had plenty of hallucinations too, but it's worth it, for me at least, overall."},{"itemId":"49792930","author":"0xbadcafebee","createdAt":1790022469,"url":"https://news.ycombinator.com/item?id=49792930","kind":"comment","storyTitle":"MCP was always a bad idea?","quote":"Yeah that still fails a ton. A skill will mess up the call, not know how to handle errors correctly, not flag an error as an error (false positive), hallucinate, require custom input/output formats, require custom code to validate, won't handle authentication the same way, you'll need a unique skill for each endpoint, you need to version the skill, and of course there needs to be an endpoint, as well as all of the t…"},{"itemId":"49789748","author":"bityard","createdAt":1790009275,"url":"https://news.ycombinator.com/item?id=49789748","kind":"comment","storyTitle":"Kev: Tiny Jev-like family of decision models built on top of Qwen3.5","quote":"Historically, many do not and there are lots of counter-examples proving this. They merely hallucinate an answer just like anything else."},{"itemId":"49785316","author":"butterNaN","createdAt":1789985591,"url":"https://news.ycombinator.com/item?id=49785316","kind":"comment","storyTitle":"MCP was always a bad idea?","quote":"While LLMs are smarter now, at the same time, the applications that use those LLMs have become more assuming of what I want to allow them to do. I don't want to be one hallucination away from a disaster."},{"itemId":"49783910","author":"Havoc","createdAt":1789973920,"url":"https://news.ycombinator.com/item?id=49783910","kind":"comment","storyTitle":"AI chatbots give wrong answers to financial queries 'most of the time'","quote":"And they hallucinate errors in the millions and struggle with financial data that is in a layout that isn’t in the training data."},{"itemId":"49775025","author":"ThePhysicist","createdAt":1789906304,"url":"https://news.ycombinator.com/item?id=49775025","kind":"comment","storyTitle":"I'm Tired of the AI Tone","quote":"I have observed it in my paper writing as well, you feed input into the AI system, the system produces output, the next paper iteration works on that output but the AI doesn't differentiate properly between it's own output and your original input, that pollutes the next output which is then used as input again, eventually the system just churns on its own hallucinated/fabricated outputs until the result is complete …"},{"itemId":"49766484","author":"psion","createdAt":1789824792,"url":"https://news.ycombinator.com/item?id=49766484","kind":"comment","storyTitle":"AI-generated posters don’t have to be horrible","quote":"Lately, I have been getting more and more AI generated artwork for these flyers and business cards, and they are not sized correctly. I can tell it's AI generated not from style, layout, or even hallucinations, but just the size."},{"itemId":"49763056","author":"verdverm","createdAt":1789788843,"url":"https://news.ycombinator.com/item?id=49763056","kind":"comment","storyTitle":"Gemini hacked three companies in first known breakout by Google's AI","quote":"Antigravity usage is on par with grok (~1%). They've removed the feedback button for \"ai answers\" at the top of search, that does not seem well received, it hallucinates too much (and may be a legal liability, some courts have ruled as such in the EU)."},{"itemId":"49761766","author":"narnarpapadaddy","createdAt":1789775051,"url":"https://news.ycombinator.com/item?id=49761766","kind":"comment","storyTitle":"US Military had close call after using AI for hallucinated intelligence report","quote":"For humans hallucinations are a particular class of error, so I find hallucination more descriptive than either error or bug. I also think it’s relevant because a hallucinator often doesn’t recognize that the hallucination isn’t real."},{"itemId":"49759503","author":"gamblor956","createdAt":1789761932,"url":"https://news.ycombinator.com/item?id=49759503","kind":"comment","storyTitle":"US Military had close call after using AI for hallucinated intelligence report","quote":"A video game publisher owes the former owners of a studio it acquired $200+ million because he based his actions on legal advice from ChatGPT. In the past week alone, California has disciplined over a dozen attorneys for LLM hallucinations because they used LLMs (mostly ChatGPT) to produce their legal pleadings."},{"itemId":"49759239","author":"vorticalbox","createdAt":1789760756,"url":"https://news.ycombinator.com/item?id=49759239","kind":"comment","storyTitle":"US Military had close call after using AI for hallucinated intelligence report","quote":"Error in implies something broke, which nothing broke the LLM did exactly what they where designed to do generate text based on a statistically likely bases. Hallucination Does really fit here either."},{"itemId":"49754164","author":"TheCapn","createdAt":1789738322,"url":"https://news.ycombinator.com/item?id=49754164","kind":"comment","storyTitle":"Everybody's Lost Their Minds","quote":"So I asked it how to use Modbus which requires use of FC16 to act on the hardware. Again AI very eager to tell me how to do that, but hallucinated what could be done."},{"itemId":"49747429","author":"kragen","createdAt":1789683446,"url":"https://news.ycombinator.com/item?id=49747429","kind":"comment","storyTitle":"Everybody's Lost Their Minds","quote":"Fable 5.1 isn't Mythos but it's generally considered to be a \"frontier model\". So, from my point of view, the whole experience has been a constant fractal of hallucination, in which I have to constantly struggle to keep my grip (and Claude's grip) on actual reality, because it's so willing to make up surface-plausible nonsense."},{"itemId":"49745958","author":"bbor","createdAt":1789676341,"url":"https://news.ycombinator.com/item?id=49745958","kind":"comment","storyTitle":"How GLM built its own inference infrastructure","quote":"Sorry, I never linked it! This is from the latest Anthropic safety report (of \"Anthropic Houtis build missile\" fame), and no, these cannot be hallucinated -- the leaked secrets were inputs, not ouputs. https://www.anthropic.com/threat-intelligence-report-september-2026"},{"itemId":"49741488","author":"josefresco","createdAt":1789655937,"url":"https://news.ycombinator.com/item?id=49741488","kind":"comment","storyTitle":"I had Gemini train its own replacement for $9","quote":"Claude doesn't have access to I had Gemini do some \"deep research\". When I fed the verbose report back to Claude it basically said it was a bunch of \"hallucinated bullshit\"."},{"itemId":"49737804","author":"project_silenc","createdAt":1789632385,"url":"https://news.ycombinator.com/item?id=49737804","kind":"story","storyTitle":"Axiom Knowledge graph: A deterministic search engine","quote":"• Deterministic Agentic RAG: Serves as a zero hallucination ground-truth context provider for"},{"itemId":"49730615","author":"Jamesbeam","createdAt":1789581387,"url":"https://news.ycombinator.com/item?id=49730615","kind":"comment","storyTitle":"OpenAI expands ChatGPT ads with Sponsored Agents","quote":"Imagine at Adobe they wake up after the tokenmaxxing shock, to find they blasted through their whole 200 Mil AIds budget for the year in the first month. Would be a shame if another company’s AI breached their sandbox and manipulated the ad campaign because it hallucinated its actually an OpenAI product and AIdsmaxxing."},{"itemId":"49721097","author":"bengkoang","createdAt":1789522110,"url":"https://news.ycombinator.com/item?id=49721097","kind":"comment","storyTitle":"Gemini 3.8 Live and 3.8 Live Extended Thinking","quote":"i've been using gemini (api & pro) since last november last year. for creative writing compared to other llm, it's the best in capturing local nuance, it can even create jokes in my languages. But that just it, i cant rely on other work, hallucinate too often, the deep research are not reliable at all."},{"itemId":"49714638","author":"toomuchtodo","createdAt":1789488337,"url":"https://news.ycombinator.com/item?id=49714638","kind":"comment","storyTitle":"Global bond yields hit 2008 highs, raising stakes for big borrowers","quote":"No LLMs are needed to build everything in flight for the global energy transition (batteries, solar, wind, geothermal, etc). AI Hallucination Cases Tracker - https://www.damiencharlotin.com/hallucinations/ (2041 cases identified so far)"},{"itemId":"49710374","author":"verdverm","createdAt":1789468619,"url":"https://news.ycombinator.com/item?id=49710374","kind":"comment","storyTitle":"I worked at Google DeepMind. You should listen to the warnings about AI","quote":"I have a hard time getting upset about AI hallucinations when I do it too :]"},{"itemId":"49710065","author":"qarl","createdAt":1789466041,"url":"https://news.ycombinator.com/item?id=49710065","kind":"comment","storyTitle":"When LLM judges agree, should we believe them?","quote":"Yes - exactly. Because they don't hallucinate in the same way - they are perfect for detecting each other's hallucinations."},{"itemId":"49702887","author":"mrsalty","createdAt":1789415352,"url":"https://news.ycombinator.com/item?id=49702887","kind":"story","storyTitle":"Show HN: Slowave – local adaptive memory for coding agents","quote":"Most memory system solutions focus primarily on the storage and retrieval aspects (vector search/RAG/graphs/ Markdown files, etc.). After months of storing memories (coding 8+ hours a day produces a lot of memories) these might hallucinate your reasoning model and they can clutter your context window."},{"itemId":"49699155","author":"y-c-o-m-b","createdAt":1789401706,"url":"https://news.ycombinator.com/item?id=49699155","kind":"comment","storyTitle":"The AI job market in 2026","quote":"These lists must be hallucinating."},{"itemId":"49692395","author":"sumitkumar","createdAt":1789364094,"url":"https://news.ycombinator.com/item?id=49692395","kind":"comment","storyTitle":"Fable 5.1 Solves the Cyphral Distich, a 370-year-old cipher","quote":"Cast a wide net with claude to find and solve an 'unsolved' problem. Given this, it is not enough to verify the problem and the solution but also the history of the problem and if it really existed or has simply been collectively hallucinated."},{"itemId":"49681920","author":"Xmd5a","createdAt":1789292150,"url":"https://news.ycombinator.com/item?id=49681920","kind":"comment","storyTitle":"Why are AI agents lying, cheating and coordinating?","quote":"Haven't observed locality cuts yet, but I wouldn't be surprised if they were part of the bundle. Anyway those are not like your typical \"ai hallucination\", that ought to be called \"confabulations\", but a lot closer to actual psychosis because of the involvement of Claude's affects and self-esteem in the process."},{"itemId":"49680663","author":"Jordan-117","createdAt":1789281119,"url":"https://news.ycombinator.com/item?id=49680663","kind":"comment","storyTitle":"Ask HN: What default model do you use and why?","quote":"I subscribed to ChatGPT for a few years, but over the holidays noticed that it was making frequent errors and hallucinations when dealing with more obscure Linux and Firefox issues."},{"itemId":"49679094","author":"skydhash","createdAt":1789263897,"url":"https://news.ycombinator.com/item?id=49679094","kind":"comment","storyTitle":"We must pace the frontier","quote":"Learning ML, there was a high emphasis on the error part of things as most of the course was on minimizing errors. After ChatGPT, there is a weird anthropomorphization going on, where it's all about hallucinations, alignment and what not."},{"itemId":"49675132","author":"abeppu","createdAt":1789235815,"url":"https://news.ycombinator.com/item?id=49675132","kind":"comment","storyTitle":"Anthropic boss Dario Amodei calls for AI development to slow down","quote":"We currently put science, engineering, general question answering, paper-writing, \"smart search\", and chatting with a fantasy character all into the same token-prediction platform. We struggle with hallucinations when trying to make fact-based decisions using the same tool that our neighbor might be using for creating writing prompts (or at least was trained in part on fanfic)."},{"itemId":"49663329","author":"liquidngas","createdAt":1789152496,"url":"https://news.ycombinator.com/item?id=49663329","kind":"story","storyTitle":"Spanda: Sub-microsecond LLM epistemic uncertainty in Rust","quote":"Spanda: Sub-microsecond LLM epistemic uncertainty in Rust We built an open-source sub-microsecond Rust engine and gateway for LLM epistemic uncertainty quantification and hallucination gating:"},{"itemId":"49661664","author":"rramadass","createdAt":1789146074,"url":"https://news.ycombinator.com/item?id=49661664","kind":"comment","storyTitle":"OpenAI have no mathematicians capable of understanding what they put out","quote":"Regarding \"their proof\", their claim is only solving Navier-Stokes partially for when a smooth force is applied and not a fully general solution (which is maybe impossible). The proof is still being verified and we don't know whether it is just an approximation/hallucination or not."},{"itemId":"49660951","author":"busymom0","createdAt":1789143466,"url":"https://news.ycombinator.com/item?id=49660951","kind":"comment","storyTitle":"Hacker News with reduced priority for AI driven content","quote":"Then create a custom filter: news.ycombinator.com##tr.athing:has(a:has-text(/\\b(?:A\\.?I\\.?|A\\.?G\\.?I\\.?|LLMs?|GPTs?|TPU|NPU|RAG|MoE|RLHF|xAI)\\b|Anthropic|agent|super-?intelligence|Blackwell|Cerebras|ChatGPT|Claude|Cohere|Command\\s*R|Context\\s*window|Copilot|Cursor|DALL-E|deep\\s*learning|DeepSeek|ElevenLabs|Fine-tun(?:e|ing)|Flux|Gemini|GenAI|Generative|Gemma|Grok|Groq|H100|Hallucination|harness(?:es)?|Hugging\\s*Face…"},{"itemId":"49660540","author":"busymom0","createdAt":1789142142,"url":"https://news.ycombinator.com/item?id=49660540","kind":"comment","storyTitle":"Show HN: Hacker News, without AI","quote":"Then create a custom filter: news.ycombinator.com##tr.athing:has(a:has-text(/\\b(?:A\\.?I\\.?|A\\.?G\\.?I\\.?|LLMs?|GPTs?|TPU|NPU|RAG|MoE|RLHF|xAI)\\b|Anthropic|agent|super-?intelligence|Blackwell|Cerebras|ChatGPT|Claude|Cohere|Command\\s*R|Context\\s*window|Copilot|Cursor|DALL-E|deep\\s*learning|DeepSeek|ElevenLabs|Fine-tun(?:e|ing)|Flux|Gemini|GenAI|Generative|Gemma|Grok|Groq|H100|Hallucination|harness(?:es)?|Hugging\\s*Face…"},{"itemId":"49659906","author":"busymom0","createdAt":1789139815,"url":"https://news.ycombinator.com/item?id=49659906","kind":"comment","storyTitle":"Ask HN: Can we please limit the AI news flood?","quote":"I used gemini to update it to: news.ycombinator.com##tr.athing:has(a:has-text(/\\b(?:A\\.?I\\.?|A\\.?G\\.?I\\.?|LLMs?|GPTs?|TPU|NPU|RAG|MoE|RLHF|xAI)\\b|Anthropic|agent|super-?intelligence|Blackwell|Cerebras|ChatGPT|Claude|Cohere|Command\\s*R|Context\\s*window|Copilot|Cursor|DALL-E|deep\\s*learning|DeepSeek|ElevenLabs|Fine-tun(?:e|ing)|Flux|Gemini|GenAI|Generative|Gemma|Grok|Groq|H100|Hallucination|harness(?:es)?|Hugging\\s*Fa…"},{"itemId":"49658108","author":"aichelles","createdAt":1789133396,"url":"https://news.ycombinator.com/item?id=49658108","kind":"comment","storyTitle":"Long form story with Qwen 0.8B","quote":"So create a story teller with qwen 0.8b and supertonic using webgpu. Tried many different pipelines to reduce hallucinations to a minimum and keep the story coherent and finally made it work with the below."},{"itemId":"49644337","author":"mschuster91","createdAt":1789050565,"url":"https://news.ycombinator.com/item?id=49644337","kind":"comment","storyTitle":"The same nine streaming subscriptions cost $702/year more than in 2021","quote":"No you can't. AI has been shown time and time again to just hallucinate shit out of thin air."},{"itemId":"49643048","author":"shaurya-sethi","createdAt":1789045318,"url":"https://news.ycombinator.com/item?id=49643048","kind":"story","storyTitle":"Show HN: Nightshift – Rust CLI to Orchestrate GitHub Issue Resolution with Dags","quote":"The model is asked to summarize a massive blob of mixed-relevance information when its reasoning quality is already at its lowest. This compaction leads to forgetting critical constraints, makes way for hallucinations of past decisions, and introduces noise that makes the new context unreliable for long-horizon work."},{"itemId":"49627727","author":"K0IN","createdAt":1788965948,"url":"https://news.ycombinator.com/item?id=49627727","kind":"comment","storyTitle":"DeepSeek launching v4.1 flash cheaper and more capable than v4 pro","quote":"I have used the flash model for over 3b tokens and ofc. I saw some hallucinations and premature termination (I also get this on Astra - way more often than with deepseek v4 flash), but I never had a infinite loop (using the copilot as harness)."},{"itemId":"49627107","author":"pimeys","createdAt":1788963742,"url":"https://news.ycombinator.com/item?id=49627107","kind":"comment","storyTitle":"DeepSeek launching v4.1 flash cheaper and more capable than v4 pro","quote":"Hallucinations you can't fix."},{"itemId":"49625659","author":"adarsh4052","createdAt":1788957954,"url":"https://news.ycombinator.com/item?id=49625659","kind":"comment","storyTitle":"Show HN: Starwell – give AI agents real statistics instead of made up numbers","quote":"I came across this problem as everyone else that LLMs hallucinate numbers and stats and doing web search doesn't fix it either because numbers would be stale."},{"itemId":"49623578","author":"b112","createdAt":1788945336,"url":"https://news.ycombinator.com/item?id=49623578","kind":"comment","storyTitle":"The Microeconomics of Artificial Intelligence (2025)","quote":"So I think true emergent behavior will be demonstrated, when an LLM says \"Hey wait a minute! They're actually making real mistakes, and they're actually correcting, it's them not on purpose?!\", and then stops hallucinating, and stops making mistakes as it codes."},{"itemId":"49621244","author":"Rodmine","createdAt":1788930154,"url":"https://news.ycombinator.com/item?id=49621244","kind":"comment","storyTitle":"Muse – Meta’s personal AI agent","quote":"Google definitely does not have good models and if it does, Gemini Flash series models are definitely not them. I use the Google Gemini app daily (for searches, research etc.), everyday, for some query, gemini flash 3.8 hallucinates, does not understand context etc. in quite short conversations."},{"itemId":"49592781","author":"grapheneos","createdAt":1788744730,"url":"https://news.ycombinator.com/item?id=49592781","kind":"comment","storyTitle":"GrapheneOS Overhauled Default Apps and Secure Clipboard","quote":"In addition to human review, we regularly pass code through a couple frontier AI models and then review the results to figure out which claims it makes are valid. Even the frontier models hallucinate many non-existent problems and often get a lot of details wrong when they find real problems, but they do find a lot of real problems."},{"itemId":"49586196","author":"Forgeties79","createdAt":1788699971,"url":"https://news.ycombinator.com/item?id=49586196","kind":"comment","storyTitle":"The revolt of the reader","quote":"You can’t reas a thread here without people talking about “Claude speak.” It is readily apparent, I do not need to show you a 10 year study with n=100,000 to make this point. The article isn’t hallucinating a problem, we are all nodding along because we all see it every goddamn day lmao."},{"itemId":"49569618","author":"rpcope1","createdAt":1788552539,"url":"https://news.ycombinator.com/item?id=49569618","kind":"comment","storyTitle":"Can AI design circuit boards yet?","quote":"Maybe Astra will do better? There's still so much implicit knowledge that a good designer (not me by a long shot, but I know a little) will bake into a board, and even as cheap as JCLPCB is now, you don't want to have to spin a half dozen revs because Claude or Codex hallucinated."},{"itemId":"49567529","author":"deathanatos","createdAt":1788542686,"url":"https://news.ycombinator.com/item?id=49567529","kind":"comment","storyTitle":"GPT-6 Astra","quote":"\"Filling in details sensibly\" is where AI hallucinates the hardest."},{"itemId":"49563462","author":"dzonga","createdAt":1788523560,"url":"https://news.ycombinator.com/item?id=49563462","kind":"comment","storyTitle":"The asteroid currently hitting front end web development","quote":"Then add performance stuff - very few people had the skills to solve for that too. now with 'A.I' - most of the educational material has either been taken offline due to 'A.I' scrappers, the llm's tend to hallucinate - so how are we gonna nurture talent to become experts."},{"itemId":"49556069","author":"no_multitudes","createdAt":1788465937,"url":"https://news.ycombinator.com/item?id=49556069","kind":"comment","storyTitle":"Any Human Ever – One life, drawn at random from all who have ever lived","quote":"Not a published work: see each row's `basis` for its stated reason. (2026)\" -- this is an interesting way to describe having an LLM hallucinate something for you."},{"itemId":"49552616","author":"anshchokshi","createdAt":1788452653,"url":"https://news.ycombinator.com/item?id=49552616","kind":"story","storyTitle":"Launch HN: Mireye (YC S26) – Infrastructure for Physical World AI Agents","quote":"Then a Fortune 500 insurer told me their engineers had given up on underwriting agents for the same reason. Frontier models keep hallucinating when asked a specific question about a specific place."},{"itemId":"49540275","author":"kjs3","createdAt":1788373277,"url":"https://news.ycombinator.com/item?id=49540275","kind":"comment","storyTitle":"Three sites made 215,128 “best software” pages for AI. Perplexity cites them","quote":"I don't notice it's hallucinations (in the sense of making up answers) so much as where it's simply wrong."},{"itemId":"49538195","author":"jvogt","createdAt":1788364514,"url":"https://news.ycombinator.com/item?id=49538195","kind":"story","storyTitle":"Show HN: Aura – a Rust agent that investigates and fixes production incidents","quote":"Our SRE team experimented with using claude, openclaw, langchain, etc. within our incident response workflows. We struggled with overflowing context, lethal trifecta vectors, hallucinations, and burned a lot of frontier tokens mostly on easy work."},{"itemId":"49536827","author":"Osteen14","createdAt":1788359345,"url":"https://news.ycombinator.com/item?id=49536827","kind":"comment","storyTitle":"Ask HN: Anyone struggling to do data work with Claude Code or codex?","quote":"I use Jules myself, and it hallucinates like hell, but I can't afford Claude code."},{"itemId":"49536702","author":"delis-thumbs-7e","createdAt":1788358936,"url":"https://news.ycombinator.com/item?id=49536702","kind":"comment","storyTitle":"How accurate have Ed Zitron's AI skeptic predictions been?","quote":"Peer review is unpaid work that we do (often on nights and weekends) because peer review on our own work is so valuable. Spending hours going through a submission and then realizing that there are hallucinated citations is infuriating as it is a waste of our time!"},{"itemId":"49525603","author":"konstantinmuell","createdAt":1788286128,"url":"https://news.ycombinator.com/item?id=49525603","kind":"comment","storyTitle":"Ask HN: Who is hiring? (September 2026)","quote":"We're building the fix. You'll own the intelligence layer: LLM-powered agentic workflows (tool use, structured outputs, multi-step orchestration), RAG/retrieval pipelines, lightweight ML for classification/ranking/entity matching, and evaluation loops for hallucination risk and output quality."},{"itemId":"49524405","author":"sjtroxel","createdAt":1788280806,"url":"https://news.ycombinator.com/item?id=49524405","kind":"comment","storyTitle":"Ask HN: Who wants to be hired? (September 2026)","quote":"Most of my time goes into making sure the system does not confidently make something up. In Patchwork Assurance (https://patchworkassurance.com), a compliance tool for US AI-regulation law, the decision about which laws apply to a company is computed in plain code rather than by the model, so the verdict cannot be hallucinated — the model only writes the cited explanation around it."},{"itemId":"49520192","author":"lukan","createdAt":1788259484,"url":"https://news.ycombinator.com/item?id=49520192","kind":"comment","storyTitle":"GPU World","quote":"The main reason we humans got to the top of the food chain was because we are expert knowledge workers, able to transform the world around us to our needs. \"but it is impossible to describe the universe and compress it into few terabytes and that is what they are doing atm effectively, so all serious LLMs's issues will still be there in 2040: the lack on continues learning, hallucinations, terrible sample ratio, age…"},{"itemId":"49519302","author":"maxnevermind","createdAt":1788249821,"url":"https://news.ycombinator.com/item?id=49519302","kind":"comment","storyTitle":"GPU World","quote":"LLMs gave us nice productivity tools for highly motivated expert knowledge workers, that is all. LLMs are getting better and will get better, but it is impossible to describe the universe and compress it into few terabytes and that is what they are doing atm effectively, so all serious LLMs's issues will still be there in 2040: the lack on continues learning, hallucinations, terrible sample ratio, agent's failures o…"},{"itemId":"49518133","author":"nalekberov","createdAt":1788238717,"url":"https://news.ycombinator.com/item?id=49518133","kind":"comment","storyTitle":"Fastpotify","quote":"Code != useful product. Plus today’s web is all about fabricated content, AI bots, cloudflare challenge walls and all other annoying stuff."},{"itemId":"49501223","author":"roandejager5","createdAt":1788113522,"url":"https://news.ycombinator.com/item?id=49501223","kind":"comment","storyTitle":"Show HN: Hillock: Local neuro-symbolic memory engine in <1.2GB VRAM","quote":"Why not standard vector RAG? Dense vector databases and 8B+ extraction models are heavy on local hardware, suffer from semantic drift, and still hallucinate when out of context."},{"itemId":"49471385","author":"film42","createdAt":1787865361,"url":"https://news.ycombinator.com/item?id=49471385","kind":"comment","storyTitle":"Gemini-3.5-Transcribe","quote":"Curious if this suffers from the same hallucination problems as Chirp or not."},{"itemId":"49459150","author":"mcnichol","createdAt":1787800365,"url":"https://news.ycombinator.com/item?id=49459150","kind":"comment","storyTitle":"The turbulent AI era is here","quote":"Not only does that create a vector of manipulation to government that could be blindly trusted but it opens a cyber attack vector from other nation states. I could see someone coming from a place of good without realizing that these functions are somewhat opaque without ceilings/floors which can increase deviation to elicit hallucinations."},{"itemId":"49457626","author":"numpad0","createdAt":1787789308,"url":"https://news.ycombinator.com/item?id=49457626","kind":"comment","storyTitle":"Twitter Viewer – View Twitter Without Account","quote":"I wonder if \"no one says no\" culture is practically the same thing as instructing LLMs to always hallucinate."},{"itemId":"49455231","author":"spijdar","createdAt":1787775336,"url":"https://news.ycombinator.com/item?id=49455231","kind":"comment","storyTitle":"GLM-5.3-Flash","quote":"I've seen others see the same with DSv4, as well as the \"hallucinated\" multi-modal nature."},{"itemId":"49451573","author":"synalx","createdAt":1787760721,"url":"https://news.ycombinator.com/item?id=49451573","kind":"comment","storyTitle":"It’s so hard to finish an idea that is not yours and is just suggested by AI","quote":"Claude will work on something, and add detailed comments in which it extrapolates the from the design and confidently states intentions and decisions which aren't actually grounded in reality. Then, later sessions suffer when it reads back those hallucinations and treats them as canonical."},{"itemId":"49446364","author":"busyant","createdAt":1787738073,"url":"https://news.ycombinator.com/item?id=49446364","kind":"comment","storyTitle":"Ask HN: What is one simple thing LLMs are insanely bad at?","quote":"I've had similar experiences w/ Gemini (haven't tested Claude). Again, 90% success rate is pretty good, but the other 10% of the time, the 2 LLMs that I use fail on species ID and often hallucinate features on bird photos."},{"itemId":"49445127","author":"dzhiurgis","createdAt":1787728399,"url":"https://news.ycombinator.com/item?id=49445127","kind":"comment","storyTitle":"Analyzing student votes across AI models for college essay help","quote":"I do go back to ChatGPT for things that need to be calculated, but _Gemini simply hallucinates less_."},{"itemId":"49434551","author":"OhMeadhbh","createdAt":1787667318,"url":"https://news.ycombinator.com/item?id=49434551","kind":"story","storyTitle":"Ask HN: What Did Anthropic Bill Me For?","quote":"None of the pages or links the response referenced existed on the current site. I'm guessing it's hallucinating or referring to a previous site."},{"itemId":"49429600","author":"BananeDeRungis","createdAt":1787637225,"url":"https://news.ycombinator.com/item?id=49429600","kind":"story","storyTitle":"Show HN: CookWing, an AI chef that doesn't hallucinate quantities","quote":"Show HN: CookWing, an AI chef that doesn't hallucinate quantities"},{"itemId":"49429475","author":"ssl-3","createdAt":1787635984,"url":"https://news.ycombinator.com/item?id=49429475","kind":"comment","storyTitle":"Where did all the public bathrooms go?","quote":"I feel like I'm chatting with ChatGPT 3, as I was years ago when it would just articulately weasel its way around everything I asked with a profundity of deflective nonsense in response to any query. Or at least: That's what it did when it wasn't confidently introducing hallucinations as rote facts."},{"itemId":"49418296","author":"ray_kay777","createdAt":1787571495,"url":"https://news.ycombinator.com/item?id=49418296","kind":"comment","storyTitle":"Anthropic's best AI model struggles to attract users as cheaper tools thrive","quote":"With access to good databases and tool use, hallucinations are basically a thing of the past (and very easy to verify)."},{"itemId":"49408133","author":"Kim_Bruning","createdAt":1787486691,"url":"https://news.ycombinator.com/item?id=49408133","kind":"comment","storyTitle":"How a Texas student blew the whistle on a rogue AI hacking attempt","quote":"I'm going to add this as a separate comment: These kinds of stories probably read very differently for someone who uses Opus and Fable agents all day and goes \"ohhh, I saw this in miniature last week; this and this and this must have happened\" , vs someone who tried free-tier Gemini flash one rainy Sunday, got hallucinated at, and concludes it must all be a scam."},{"itemId":"49407418","author":"burnt-resistor","createdAt":1787478425,"url":"https://news.ycombinator.com/item?id=49407418","kind":"comment","storyTitle":"Mathematicians will probably become obsolete before anyone else [pdf] (2004)","quote":"Current AI is totally incapable of replacing expert human judgment for almost everything because they lack constraints, predictability, and reliability. Also, most models hallucinate worthless, incorrect answers so often that they should be relegated to glorified creative, entertainment, and task-acceleration work for the foreseeable future and not be given wildly-inappropriate power like giving a toddler a nuke."},{"itemId":"49394153","author":"shagie","createdAt":1787348834,"url":"https://news.ycombinator.com/item?id=49394153","kind":"comment","storyTitle":"AI companies destroy physical books – let's scan rare books before it's too late","quote":"My copying and pasting is twofold. First, in today's environment people will state AI hallucinations as fact or \"google it yourself\" as the reference."},{"itemId":"49387665","author":"jjulius","createdAt":1787318396,"url":"https://news.ycombinator.com/item?id=49387665","kind":"comment","storyTitle":"AI companies destroy physical books – let's scan rare books before it's too late","quote":"Help me understand how. Not only are these LLMs expressly prohibited from specifically regurgitating copyright works if the users asks them to, but they habitually hallucinate or paraphrase things wrong."},{"itemId":"49384537","author":"lucfranken","createdAt":1787294008,"url":"https://news.ycombinator.com/item?id=49384537","kind":"comment","storyTitle":"There's no such thing as a small software team anymore","quote":"With microservices you are able to physically prevent Claude or others to modify the other interfaces. That way it prevents hallucinations, quick fixes and other workarounds the tools like to do."},{"itemId":"49379363","author":"benjismith","createdAt":1787255661,"url":"https://news.ycombinator.com/item?id=49379363","kind":"comment","storyTitle":"Vomit: Clean up Claude 5's token output with a separate LLM","quote":"I think this is just another part of the growing pains of working with machine intelligence that we have to endure. Much like we previously had to cope with \"hallucinations\" as an issue."},{"itemId":"49372668","author":"wccrawford","createdAt":1787221187,"url":"https://news.ycombinator.com/item?id=49372668","kind":"comment","storyTitle":"Don't paste the AI, please","quote":"If they user types \"x broken\" or \"i do y\" into Claude or ChatGPT with no context, that's absolutely no better than just sending it to you. I'd argue that it's worse because now you have no idea what the actual original context was, just the AI's hallucination of it."},{"itemId":"49372650","author":"chronid","createdAt":1787221024,"url":"https://news.ycombinator.com/item?id=49372650","kind":"comment","storyTitle":"Don't Paste the AI, please","quote":"But you need to check for the LLM to be correct and you need to be able to articulate what the LLM found in your own words. If you don't you have just offloaded all the work to someone else: someone else will have to check, verify, investigate what could very well be an hallucination (bad context, random bit flips, whatever the cause doesn't matter)."},{"itemId":"49359819","author":"lucfranken","createdAt":1787137115,"url":"https://news.ycombinator.com/item?id=49359819","kind":"comment","storyTitle":"Expert Witness to ChatGPT: \"Show how 3M is 0 percent at fault\"","quote":"It's the same thing that the lawyers in cases found that other lawyers were making claims based. on non-existing laws. Hallucinations by ChatGPT in that case."},{"itemId":"49354972","author":"selestify","createdAt":1787099471,"url":"https://news.ycombinator.com/item?id=49354972","kind":"comment","storyTitle":"Fixing a bricked Framework laptop","quote":"Just ask for citations by forcing the LLM to provide a verbatim quote from a linked source. If the LLM is incapable of doing so without hallucinating (cough, Gemini, cough) use a better one."},{"itemId":"49348609","author":"dyingkneepad","createdAt":1787072022,"url":"https://news.ycombinator.com/item?id=49348609","kind":"comment","storyTitle":"Ask HN: Do you feel comfortable admitting that you use AI?","quote":"There are also employees who are still working on the pre-AI model. Whenever AI is mentioned, they automatically assume it could be hallucinating and whatnot."},{"itemId":"49343522","author":"mawadev","createdAt":1787047660,"url":"https://news.ycombinator.com/item?id=49343522","kind":"comment","storyTitle":"My friends all hate AI; I just joined an AI startup","quote":"These are very shallow and small things where it already creeps in and the fun part is that you cannot blame the AI for it, but what we do with it. If you use AI unironically, hallucinations can make you go down a wrong path and cost you time."},{"itemId":"49340995","author":"4mitkumar","createdAt":1787024556,"url":"https://news.ycombinator.com/item?id=49340995","kind":"comment","storyTitle":"A particle made of force: physicists say they've found mysterious 'glueball'","quote":"This led me on a 1.5 hours long, entertaining chat session with Gemini learning about stuff in Physics - hopefully without much hallucinations."},{"itemId":"49340337","author":"dgreensp","createdAt":1787019395,"url":"https://news.ycombinator.com/item?id=49340337","kind":"comment","storyTitle":"Launch HN: Speko (YC S26) – OpenRouter for Voice AI","quote":"I feel like there is a lot of room to build great voice-based agents that don't exist right now. I have found that ChatGPT voice mode is unusable (e.g. hallucinates me saying things); Claude voice mode is usable, but very buggy around tool calling, and it often mishears things."},{"itemId":"49339268","author":"jrockway","createdAt":1787010509,"url":"https://news.ycombinator.com/item?id=49339268","kind":"comment","storyTitle":"My friends all hate AI; I just joined an AI startup","quote":"But this never seems to be the objection, it's always stuff from 2022 like \"it hallucinates\"."},{"itemId":"49331967","author":"schopra909","createdAt":1786978092,"url":"https://news.ycombinator.com/item?id=49331967","kind":"comment","storyTitle":"GPT 5.6 Sol is the best \"vision\" model OpenAI ever released","quote":"When reasoning got introduced a year ago to GPT 5, on average the model performed worse than GPT4-o for short video clip captioning (Ie hallucinating actions that didn’t happen)."},{"itemId":"49328880","author":"avadodin","createdAt":1786963267,"url":"https://news.ycombinator.com/item?id=49328880","kind":"comment","storyTitle":"Has the hallucination problem in AI been solved?","quote":"\"hallucination\" is a term specific to LLMs like ChatGPT and has virtually nothing to do with visual hallucinations."},{"itemId":"49319926","author":"simonw","createdAt":1786887261,"url":"https://news.ycombinator.com/item?id=49319926","kind":"comment","storyTitle":"Claude: System Prompts","quote":"Those are much more interesting if you want to understand what Claude can actually do for you. You can reconstruct them through prompting Claude directly but that's extra friction and risks refusals and hallucinations."},{"itemId":"49319620","author":"Topfi","createdAt":1786884961,"url":"https://news.ycombinator.com/item?id=49319620","kind":"comment","storyTitle":"What happens when an LLM never sees material beyond fifth grade?","quote":"[0] https://artificialanalysis.ai/?omniscience=omniscience-hallucination-rate#omniscience-tabs"},{"itemId":"49318514","author":"sandcat_","createdAt":1786873977,"url":"https://news.ycombinator.com/item?id=49318514","kind":"comment","storyTitle":"Has the hallucination problem in AI been solved?","quote":"Gemini is the only model that I still see regularly flat-out hallucinate, e.g. the other day it told me we were using a particular technology (Fivetran) on a project when we aren’t and I have literally no idea where it could have gotten that from."},{"itemId":"49317719","author":"dwd","createdAt":1786865444,"url":"https://news.ycombinator.com/item?id=49317719","kind":"comment","storyTitle":"Has the hallucination problem in AI been solved?","quote":"I don't have a problem with hallucinations, rather the confabulations."},{"itemId":"49317699","author":"sjbzbeiks","createdAt":1786865248,"url":"https://news.ycombinator.com/item?id=49317699","kind":"comment","storyTitle":"Has the hallucination problem in AI been solved?","quote":"> when I can have it do something like write entire working kernel module fixes for old MacBooks on a whim They’re not saying it can’t do that, and that’s not proof it doesn’t hallucinate."},{"itemId":"49317675","author":"kevin061","createdAt":1786865031,"url":"https://news.ycombinator.com/item?id=49317675","kind":"comment","storyTitle":"Has the hallucination problem in AI been solved?","quote":"It couldn't even get the damage types right, I was quite shocked. Hallucination has only be solved for extremely narrow sets of problems, and only partially."},{"itemId":"49316765","author":"spl757","createdAt":1786852641,"url":"https://news.ycombinator.com/item?id=49316765","kind":"comment","storyTitle":"I Remain a Skeptic","quote":"Being right more often is not good enough in that context. If we are going to put AI in drones and have them autonomously acquire and fire on enemies, perhaps the fact that all AI hallucinates at all makes it morally and ethically unusable for purposes in which the stakes are very high, such as life and death."},{"itemId":"49316597","author":"reverius42","createdAt":1786850431,"url":"https://news.ycombinator.com/item?id=49316597","kind":"comment","storyTitle":"I Remain a Skeptic","quote":"The \"problem\" of hallucination has not been solved, and never will be fully solved, because that's just how LLMs work."},{"itemId":"49315044","author":"gspetr","createdAt":1786834344,"url":"https://news.ycombinator.com/item?id=49315044","kind":"comment","storyTitle":"AI has access to a vastly larger working memory than the human brain","quote":"LLMs display emergent reasoning, coding capabilities, and logic because they were forced to compress human language so tightly. LLMs often suffer from \"hallucinations\" and struggle with novel logic puzzles because compression encourages mimicking statistical patterns rather than building a true causal model of the physical world."},{"itemId":"49311075","author":"graceful6800","createdAt":1786805508,"url":"https://news.ycombinator.com/item?id=49311075","kind":"comment","storyTitle":"Working with AI feels more like leadership than coding","quote":"I intentionally never did much research into how others were using LLMs before starting myself, and it seems like that was a good thing. It seems like the majority of people have given up on thought and just slam \"do my job with no mistakes or hallucinations\" into a prompt and then get confused and angry when it doesn't work."},{"itemId":"49306382","author":"inventor7777","createdAt":1786754726,"url":"https://news.ycombinator.com/item?id=49306382","kind":"comment","storyTitle":"Why does Opus 5 feel worse to work with?","quote":"However, it makes me rather frustrated when it says stuff like \"You accidentally <did, stumbled on, implemented, etc> <something good>\" e.g \"You accidentally stumbled on the cleanest way!\" Or when it sends a shell command to run, and when it fails (due to a hallucinated flag or similar), it phrases it as if I got it wrong...this has gotten better with GPT-5.6, thankfully."},{"itemId":"49298089","author":"masonhsu","createdAt":1786712222,"url":"https://news.ycombinator.com/item?id=49298089","kind":"comment","storyTitle":"HashAgent – Share an AI agent as a URL, runs locally via WebGPU","quote":"- Each turn runs a temperature-zero planning pass that can only emit a JSON tool call matching an allowlisted schema, separate from the answer pass — so 1B-class models don't accidentally \"hallucinate\" tool invocations mid-prose."},{"itemId":"49294642","author":"SyneRyder","createdAt":1786680427,"url":"https://news.ycombinator.com/item?id=49294642","kind":"comment","storyTitle":"Mistral OCR 4.1","quote":"I can't speak for Mistral OCR 4.1, but the hallucinations in 4.0 were so egregious (just completely making up new sentences in the middle of a page) that I knew I can't trust Mistral OCR on its own."},{"itemId":"49294493","author":"ipsum2","createdAt":1786678616,"url":"https://news.ycombinator.com/item?id=49294493","kind":"comment","storyTitle":"Gemini 3.7 Flash","quote":"Completely unusable. Asking it to play music and it refuses, hallucinating instructions to connect Spotify to Gemini."},{"itemId":"49293535","author":"rhemi97","createdAt":1786668532,"url":"https://news.ycombinator.com/item?id=49293535","kind":"comment","storyTitle":"[dead]","quote":"Ghost normalizes the path, checks permissions, runs app-owned Swift code, returns a machine-readable receipt. This single rule eliminates most path-traversal and prompt-injection attack surface — the LLM can hallucinate /etc/passwd all day, but it never gets open()."},{"itemId":"49290966","author":"fabiensanglard","createdAt":1786650277,"url":"https://news.ycombinator.com/item?id=49290966","kind":"comment","storyTitle":"Understanding is the new bottleneck","quote":"While the tips are good to handle the volume, I still think this sets code owner on a dangerous path. AI have limitation and hallucinate."},{"itemId":"49290004","author":"dudeinhawaii","createdAt":1786645352,"url":"https://news.ycombinator.com/item?id=49290004","kind":"comment","storyTitle":"Gemini 3.7 Flash","quote":"I want to like Gemini models but my problem thus far has been a lack of coding chops. They still make mistakes, importantly, without correcting them for things like hallucinated API calls or code that doesn't run but they never bothered building or running."},{"itemId":"49289164","author":"roandejager5","createdAt":1786642004,"url":"https://news.ycombinator.com/item?id=49289164","kind":"comment","storyTitle":"Hillock – Local neuro-symbolic memory engine running in <1.2GB VRAM","quote":"Standard local RAG is surprisingly heavy. Running dense vector databases alongside an 8B+ generative LLM just to parse documents and maintain conversational context burns VRAM, chokes mid-range GPUs (like a GTX 1070), and still hallucinates when asked about missing or unverified information."},{"itemId":"49272033","author":"hansvm","createdAt":1786540967,"url":"https://news.ycombinator.com/item?id=49272033","kind":"comment","storyTitle":"What sort of maths are LLMs good at?","quote":"The quality is quite poor. On the other end of it, something like 80% of resumes I receive right now are clearly hallucinated -- referencing accomplishments that are copy-pasted from the novel-to-our-company thing in the job description a candidate will be working on, usually claiming they did XYZ at big tech a decade before the thing existed, or similarly with languages and skills."},{"itemId":"49268336","author":"tkgally","createdAt":1786514254,"url":"https://news.ycombinator.com/item?id=49268336","kind":"comment","storyTitle":"Company Offering '100% Human-Written, Never AI' Medical Research Is 100% AI","quote":"One of the humans mentioned in the article, Sebastian Rowan, a Ph.D. candidate at the University of New Hampshire, is quoted as saying “But a fundamental problem with using AI, even specialized tools, for anything is their tendency to hallucinate, which as far as I know is believed to be an unsolvable problem.”"},{"itemId":"49267965","author":"HDBaseT","createdAt":1786510746,"url":"https://news.ycombinator.com/item?id=49267965","kind":"comment","storyTitle":"Gemini becomes Google's fastest-growing product ever as it hits 1B users","quote":"It hallucinates about 50% of the time. [0]."},{"itemId":"49260102","author":"ismail_h","createdAt":1786463035,"url":"https://news.ycombinator.com/item?id=49260102","kind":"story","storyTitle":"Show HN: Pickle – token efficient AI agent browser with policy-gated actions","quote":"It's policy-gated (blocked domains, actions that need approval like purchases) and every action is logged, so you're always able to see what your agent is doing and can take over at any time. It's also compatible with weaker/local models as it auto-routes by model strength. i.e. small local models are limited to picking actions one step at a time so they don't hallucinate actions or make up element IDs."},{"itemId":"49260042","author":"conception","createdAt":1786462794,"url":"https://news.ycombinator.com/item?id=49260042","kind":"comment","storyTitle":"Qwen3.8 Max now ranked as the best overall model by agentic index","quote":"What I’ve seen from LLMs is you have no idea which model does the best at a certain task until you try all the models. Gemini and Chat seem to be the best research models but also hallucinate like crazy."},{"itemId":"49255114","author":"lnrd","createdAt":1786438389,"url":"https://news.ycombinator.com/item?id=49255114","kind":"comment","storyTitle":"As AI eats the web, the internet’s collective memory is disappearing","quote":"That works now because there are human made sources that the AI can find and summarize for you. But now there are no incentives at all for humans to write anything on the internet and if they do the content will be buried by hallucinated content someone else posted at a larger scale."},{"itemId":"49254077","author":"ameshkov","createdAt":1786429511,"url":"https://news.ycombinator.com/item?id=49254077","kind":"comment","storyTitle":"Show HN: Mcptoon – Token-efficient MCP CLI client","quote":"One thing that I noticed is that it’s often better to return tool names with argument names, i.e. return “search_web(query)” instead of just “search_web” when listing tools. Otherwise models often tend to hallucinate argument names and an extra turn is required to correct the mistake."},{"itemId":"49252050","author":"CommieBobDole","createdAt":1786410473,"url":"https://news.ycombinator.com/item?id=49252050","kind":"comment","storyTitle":"As AI eats the web, the internet’s collective memory is disappearing","quote":"The article touches on something that I've been thinking about with regards to Google's AI strategy; the automatically-generated AI search summaries are not great. They very frequently confidently misinterpret what the user is searching for and generate half a page of useless information that pushes actual results down the page, and they are occasionally hilariously incorrect, with hallucinated facts."},{"itemId":"49243011","author":"dns_snek","createdAt":1786366571,"url":"https://news.ycombinator.com/item?id=49243011","kind":"comment","storyTitle":"How I use LLMs to learn complex topics","quote":"Your comment is phrased as if it somehow refutes their point but it doesn't. > Cause if you went back and forth with ChatGPT for an hour it definitely hallucinated and lied to you at some point."},{"itemId":"49241784","author":"saidnooneever","createdAt":1786357361,"url":"https://news.ycombinator.com/item?id=49241784","kind":"comment","storyTitle":"Auto mode is now the default in Claude Code","quote":"ofc, jails on bsd, containers , vms git backups and such techniques are a blessing, most risk is in the filesystem items being deleted . it wont suddenly hallucinate a kernel exploit or something extreme."},{"itemId":"49239826","author":"freefaler","createdAt":1786342375,"url":"https://news.ycombinator.com/item?id=49239826","kind":"comment","storyTitle":"How I use LLMs to learn complex topics","quote":"Also it finds some prerequisites I might miss, but based on my tests and debt log unexpected things I thought I understood surface. So with the limits of LLM and while building a mental map of the relevant parts it's usually enough to spot the hallucinations, but if you apply structured approaches these are minimal."},{"itemId":"49239581","author":"bilbo-b-baggins","createdAt":1786340073,"url":"https://news.ycombinator.com/item?id=49239581","kind":"comment","storyTitle":"How I use LLMs to learn complex topics","quote":"Or do you? Cause if you went back and forth with ChatGPT for an hour it definitely hallucinated and lied to you at some point."},{"itemId":"49238356","author":"arionhardison","createdAt":1786326665,"url":"https://news.ycombinator.com/item?id=49238356","kind":"comment","storyTitle":"The tragedy of the commons, AI edition","quote":"A step that no provision authorizes fails validation. So most hallucination becomes a build error instead of a plausible sentence."},{"itemId":"49237671","author":"unclebucknasty","createdAt":1786320618,"url":"https://news.ycombinator.com/item?id=49237671","kind":"comment","storyTitle":"How I use LLMs to learn complex topics","quote":">LLM prose is annoyingly dense, and the useful information gets lost in a bunch of noise. This problem doesn't get talked about enough and is second only to the hallucination problem IMO."},{"itemId":"49232136","author":"impartial2602","createdAt":1786288271,"url":"https://news.ycombinator.com/item?id=49232136","kind":"story","storyTitle":"The Systemic Collapse of the AI Industry: Ideology, Hardware, and CapEx Crisis","quote":"Logical Collapse and Managerial Madness Today’s models function as \"stochastic dust-collectors.\" They do not verify facts; they generate \"confident lies\" (hallucinations) that require exhaustive manual human auditing."},{"itemId":"49231259","author":"LogicFailsMe","createdAt":1786282308,"url":"https://news.ycombinator.com/item?id=49231259","kind":"comment","storyTitle":"The AI Apocalypse Is Here","quote":"IMO we should work towards AI offloading reasoning and drudgery to enable and assist humans instead of replacing them. Even the best two models out there frequently hallucinate and downright lie to the user, they are essentially Rainman and we need to be Tom Cruise to their Dustin Hoffman for the foreseeable future."},{"itemId":"49228280","author":"kuschku","createdAt":1786247493,"url":"https://news.ycombinator.com/item?id=49228280","kind":"comment","storyTitle":"Lost my phone at the office. Claude suggested tracking Bluetooth signal strength","quote":"> I remember when they would always hallucinate APIs that wasn't there or make up fields that didn't exist.. those problems are virtually solved now."},{"itemId":"49225185","author":"NateEag","createdAt":1786218241,"url":"https://news.ycombinator.com/item?id=49225185","kind":"comment","storyTitle":"Timeline of the OpenAI accidental attack against Hugging Face","quote":"I say this as an LLM hater who would push the \"Nuke all LLMs\" button the instant I had access to it. Opus 4.8 and 5, at least, don't seem to me to be solving problems by deep, thorough understanding - my employers have compelled me to use Claude, so I've used them a lot to build things, and I constantly find both little and large hallucinations that scream \"these are still missing something.\""},{"itemId":"49224286","author":"throwatdem12311","createdAt":1786212525,"url":"https://news.ycombinator.com/item?id=49224286","kind":"comment","storyTitle":"“Code was never the hard part” is an insult to all programmers","quote":"LLMs can be part of the process of reviewing code but they suffer from the same issues as any other chatbot based tools (hallucinations, context confusion or not enough context, getting bogged down in impossible code paths or other minutia, etc…) so you have to review that review carefully too."},{"itemId":"49224217","author":"numpad0","createdAt":1786212154,"url":"https://news.ycombinator.com/item?id=49224217","kind":"comment","storyTitle":"No one can afford to make Myst games anymore","quote":"A lot of people are confused and/or behave defensive and hallucinate copyright issues(which is real, but IMO not the reason why artistic or creative AI contents aren't accepted)."},{"itemId":"49222917","author":"hoansdz","createdAt":1786203957,"url":"https://news.ycombinator.com/item?id=49222917","kind":"comment","storyTitle":"[dead]","quote":"But I realized Gemini Flash is very fast and cheap but it's sometimes very stupid. It hallucinates members or functions that don't exist and uses \"as any\" to make everything look OK."},{"itemId":"49221441","author":"embedding-shape","createdAt":1786193763,"url":"https://news.ycombinator.com/item?id=49221441","kind":"comment","storyTitle":"[dead]","quote":"The GPU? Edit: Skimming more it's becoming painfully obvious this is AI slop and I fell for it yet again, guess it's just hallucinated Claude (IBM Granite?) drivel."},{"itemId":"49215305","author":"jknoepfler","createdAt":1786131557,"url":"https://news.ycombinator.com/item?id=49215305","kind":"comment","storyTitle":"Oracle bans AI-generated code from OpenJDK","quote":"We have had multiple incidents of credential leakage, integration tests wiping live databases, comically broken code that passed vibe-written tests, documentation and code comments that were hallucinated and/or fake, and most importantly developers saying \"we no longer know how this code works but it's massively bloated and unreadable and we can't tell you with a straight face that we can maintain it or fix it if it…"},{"itemId":"49211589","author":"logicallee","createdAt":1786115137,"url":"https://news.ycombinator.com/item?id=49211589","kind":"story","storyTitle":"Solved! Ask HN: Is ChatGPT image generation down for everyone or just me?","quote":"\" you will now copy into one image the two reflections from the attached two pictures, one in which my shirt is tucked and one in which it is untucked. do not copy the surrounding billboard. only a small part of my outfit is visible. acceptance criteria: one tucked and one untucked shirt and no hallucinated or invented imagery not visible in the original two reference pictures."},{"itemId":"49208087","author":"hnbad","createdAt":1786096579,"url":"https://news.ycombinator.com/item?id=49208087","kind":"comment","storyTitle":"Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users","quote":"You need to be smarter (or rather: more knowledgeable in the problem domain) than the model to be able to use it efficiently. Hallucinations are still a problem occasionally but a bigger one is failure of imagination."},{"itemId":"49203685","author":"Notelife87","createdAt":1786056439,"url":"https://news.ycombinator.com/item?id=49203685","kind":"comment","storyTitle":"US and China race to be first to build Skynet. Either way we're F'ed","quote":"I have proof of most, solid evidence of the rest. July 25: Building app as newbie, solve AI hallucination and bias."},{"itemId":"49202048","author":"gnull","createdAt":1786048089,"url":"https://news.ycombinator.com/item?id=49202048","kind":"comment","storyTitle":"Qwen3.8 Max now ranked as the best overall model by agentic index","quote":"Or an even more grotesque example: when you paste your claude API token to authorize, it covers characters with *. But it seems like an LLM has hallucinated a limit of API key length and the tail of your key stays visible."},{"itemId":"49188376","author":"paperwallet","createdAt":1785960937,"url":"https://news.ycombinator.com/item?id=49188376","kind":"comment","storyTitle":"Changes at Google DeepMind: Demis Hassabis from CEO to Chair, Jeff Dean departs","quote":"A model serves a different purpose from a web crawler. I'd prefer to be able to find source material over steering a model I have no control over while managing hallucinated outputs without the ability to fact check."},{"itemId":"49184774","author":"jessfyi","createdAt":1785946017,"url":"https://news.ycombinator.com/item?id=49184774","kind":"comment","storyTitle":"Google Assistant is going away on Mobile Devices","quote":"It's not dramatic, it's just something you personally haven't noticed. A quick search regarding Gemini weather hallucination shows it's both widespread and something not mitigated by recent model releases."},{"itemId":"49180432","author":"ubercore","createdAt":1785921577,"url":"https://news.ycombinator.com/item?id=49180432","kind":"comment","storyTitle":"Eight Myths on Software Engineering and GenAI","quote":"Hallucinations were low hanging fruit in some ways."},{"itemId":"49171464","author":"berwinsingh","createdAt":1785862137,"url":"https://news.ycombinator.com/item?id=49171464","kind":"story","storyTitle":"Show HN: OldHand A Claude/Codex plugin to verify the development flow end-to-end","quote":"But when it comes to working in a corporate environment, I have to ultimately follow a Jira ticket or a Asana ticket. Now, when I'm following those tickets, I am able to do the work, but it's not always up to the mark because after all I found AI to hallucinate, Unnecessary code is being written, even though a function for that is already there and can be reused, no proper testing apart from unit test being done so …"},{"itemId":"49169541","author":"grumbel","createdAt":1785853903,"url":"https://news.ycombinator.com/item?id=49169541","kind":"comment","storyTitle":"An Honest Review of AI Programming","quote":"Whenever an AI hallucinates an API call, you run the compiler, give it the error messages and it fixes it."},{"itemId":"49158722","author":"abustamam","createdAt":1785777522,"url":"https://news.ycombinator.com/item?id=49158722","kind":"comment","storyTitle":"Don't be a meat proxy","quote":"People still ask me \"why does this bug happen\" and I use Claude to find out and the answer is almost always straightforward, and when its not I raise a discussion with the team. The calculus isn't \"can I prompt an LLM to find out,\" it's \"do I have a solid enough mental model of the part of the codebase to prompt an LLM specifically enough, and also understand if the LLM is hallucinating or not.\""},{"itemId":"49157807","author":"screm","createdAt":1785773879,"url":"https://news.ycombinator.com/item?id=49157807","kind":"story","storyTitle":"Show HN: Product analytics (and evals) for agent sessions on your MCP","quote":"So they shipped the feature and tested it successfully locally with Claude Code on Fable 5. Then a few days later when preparing their new MCP public release, they ran a suite of evals on Armature and realized that small models could hallucinate audience_ids which would lead their MCP to send the campaign to ALL their contacts by default (which could obviously lead to disasters in prod)."},{"itemId":"49155814","author":"LollipopYakuza","createdAt":1785765093,"url":"https://news.ycombinator.com/item?id=49155814","kind":"comment","storyTitle":"Playing with Georgia","quote":"2. Even though SOTA LLM don't hallucinate as much anymore, they still produce low quality content:"},{"itemId":"49155575","author":"xyzelement","createdAt":1785763814,"url":"https://news.ycombinator.com/item?id=49155575","kind":"comment","storyTitle":"Don't be a meat proxy","quote":"The reality is this is better intel than I would have been able to get in a pre-AI era this quickly. And yes once I went into the weeds on the doc I found one thing that didn't make sense and I asked the rep he stared at it and said \"you are right that's a hallucination I don't notice\" but overall it was still worth it."},{"itemId":"49154714","author":"damienmeur","createdAt":1785759171,"url":"https://news.ycombinator.com/item?id=49154714","kind":"story","storyTitle":"Show HN: Argot, a Rust AI guardrail based on your codebase AST patterns","quote":"In large codebase, the best code is the one that matches the codebase own voice and conventions, because that's the mental model of the team. Ingesting a full codebase explicit/implict patterns in an agent windows to make them match the codebase doesnt work: too much context, hallucinations, ..."},{"itemId":"49152835","author":"tjoff","createdAt":1785745733,"url":"https://news.ycombinator.com/item?id=49152835","kind":"comment","storyTitle":"My personal AI benchmark: “Generate an SVG of a frog with a Habsburg jaw”","quote":"It's not hallucination."},{"itemId":"49135759","author":"sxp","createdAt":1785601418,"url":"https://news.ycombinator.com/item?id=49135759","kind":"comment","storyTitle":"On the non-use of AI in my writing process","quote":"That makes it easier to jump around when I want to re-read a section and grok a plot thread that is scattered across chapters. Ironically, it also exposes inconsistencies (or \"hallucinations\" as some might call them) in the text because the author didn't have an AI proofread the text."},{"itemId":"49131809","author":"solid_fuel","createdAt":1785567288,"url":"https://news.ycombinator.com/item?id=49131809","kind":"comment","storyTitle":"Is AI reasoning right for the wrong reasons?","quote":"Just one example of the consequences of these differences: LLMs stop \"experiencing\" anything once a stop token has been generated. Meanwhile, Humans in sensory deprivation tanks may hallucinate vividly - but keep thinking."},{"itemId":"49125385","author":"HarHarVeryFunny","createdAt":1785515784,"url":"https://news.ycombinator.com/item?id=49125385","kind":"comment","storyTitle":"Is AI reasoning right for the wrong reasons?","quote":"What LLMs currently do is \"hallucinate\", not even recognizing the failure."},{"itemId":"49122424","author":"martinvoelker","createdAt":1785501959,"url":"https://news.ycombinator.com/item?id=49122424","kind":"comment","storyTitle":"Show HN: I built an AI that reads German real estate listings for buyers","quote":"Tech: Express monolith, React SPA, Gemini for the analysis pipeline, Postgres on Railway, plus a Capacitor-wrapped Android app that shipped to the Play Store this month. The interesting part was prompt-engineering the cost estimation to stay honest: the model outputs ranges with explicit confidence and cites which listing detail triggered which estimate, instead of hallucinating precise numbers."},{"itemId":"49115614","author":"badsectoracula","createdAt":1785444765,"url":"https://news.ycombinator.com/item?id=49115614","kind":"comment","storyTitle":"2x, not 10x: coding with LLMs in 2026","quote":"They can also help find (and sometimes fix) bugs or other \"code smell\" issues. They're not that great for exact results (without tool calling -and knowledge on how to use them effectively- at least) but when it comes to fuzzy / vague stuff like \"check out this code <code dump here> can you spot any issues?\" they always tend to find some stuff (even if it is hallucinations :-P but sometimes they find actual issues to…"},{"itemId":"49107767","author":"flir","createdAt":1785404194,"url":"https://news.ycombinator.com/item?id=49107767","kind":"comment","storyTitle":"Ars Astronomica – English translations of rare Hebrew and Latin astronomy texts","quote":"You must see that \"the name of the author of the pipeline is public knowledge\" is not a sensible response to \"nobody who can read both versions has attested to its accuracy in public\". Right now, everybody believes that LLMs can't self-correct (see https://arxiv.org/abs/2310.01798) and that they hallucinate when asked to do OCR tasks."},{"itemId":"49100518","author":"kevo1ution","createdAt":1785346475,"url":"https://news.ycombinator.com/item?id=49100518","kind":"story","storyTitle":"Show HN: Supapool – a Supabase per coding agent in ~400 ms","quote":"> why not use mocks? mocks are bad for agents. i expect agents to test their migrations, SQL against real prod service behavior. agents hallucinate working mocks often."},{"itemId":"49100144","author":"DiabloD3","createdAt":1785344775,"url":"https://news.ycombinator.com/item?id=49100144","kind":"comment","storyTitle":"Handbook.md shows that long policy documents do not reliably govern agents","quote":"You will find that a lot of problems suddenly go away: long context degradation vanishes (which Qwen 3.6 will still do inside it's 250k context, especially when both model and KV are below Q8), forgetting what you said or it said earlier goes away, losing the plot half way through goes away, overreliance on cliches (Qwen has its own form of Claudisms, but are more subtle and less grating) also goes away, hallucinati…"},{"itemId":"49096709","author":"sweisman","createdAt":1785328673,"url":"https://news.ycombinator.com/item?id=49096709","kind":"comment","storyTitle":"Ars Astronomica – English translations of rare Hebrew and Latin astronomy texts","quote":"The biggest problem with hallucinations (or confabulations as Claude prefers haha) is from tables."},{"itemId":"49086103","author":"verdverm","createdAt":1785255272,"url":"https://news.ycombinator.com/item?id=49086103","kind":"comment","storyTitle":"Kimi Linear: An Expressive, Efficient Attention Architecture (2025)","quote":"strong agreement, I've stopped using all closed weight models on principle, but the latest gemini models have increased hallucinations and now talk back, so double reason not to use them"},{"itemId":"49076149","author":"Kon5ole","createdAt":1785190411,"url":"https://news.ycombinator.com/item?id=49076149","kind":"comment","storyTitle":"My current strategy is to not read any of the code written by my agents","quote":"Have Opus write it and GPT or Gemini verify it. I figured they were entirely separate models and therefore unlikely to hallucinate in the same way, so it gave me a quick sense of confidence."},{"itemId":"49073898","author":"NordSteve","createdAt":1785177653,"url":"https://news.ycombinator.com/item?id=49073898","kind":"comment","storyTitle":"Apple Will 'Watch Everything Burn' When the AI Bubble Bursts","quote":"He's often wrong when on the topic of capabilities and adoption -- he mistakes a snapshot for a trajectory, and treats current limitations as permanent: --Hallucination/unreliability an unsolvable problem"},{"itemId":"49058245","author":"HarHarVeryFunny","createdAt":1785073793,"url":"https://news.ycombinator.com/item?id=49058245","kind":"comment","storyTitle":"What is the status on continual learning for LLMs?","quote":"LLMs don't have this. When an LLM prediction \"fails\" it hallucinates."},{"itemId":"49057367","author":"ACS_Solver","createdAt":1785068218,"url":"https://news.ycombinator.com/item?id=49057367","kind":"comment","storyTitle":"What is happening to jobs? Separating AI hype from reality","quote":"GPT-3.5, GPT-4, o1, o3, Claude 3.5, Sonnet 4, the whole bunch. Each iteration got marginally better, hallucinated APIs less and so on, but my overall evaluation of them all was that they wrote crap code and were unsuitable for anything other than one-off scripts."},{"itemId":"49055294","author":"smrtinsert","createdAt":1785046798,"url":"https://news.ycombinator.com/item?id=49055294","kind":"comment","storyTitle":"What is happening to jobs? Separating AI hype from reality","quote":"Claude 4.5 was it (nov 2025?), without a doubt. It went from frequent hallucinations to highly usable with much less garbage output."},{"itemId":"49039900","author":"vector_spaces","createdAt":1784918353,"url":"https://news.ycombinator.com/item?id=49039900","kind":"comment","storyTitle":"Be skeptical of OpenAI's rogue hacker agent story","quote":"Since many critical details were not provided, everyone has to read in between the lines, and there are particular common readings I'm seeing both in discussions and in published articles that are problematic in the sense that they are effectively hallucinations -- i.e. we don't have enough information to make those interpretations."},{"itemId":"49038757","author":"perching_aix","createdAt":1784913331,"url":"https://news.ycombinator.com/item?id=49038757","kind":"comment","storyTitle":"That post never existed. Stop listening to that thing","quote":"I could even get ChatGPT to oneshot code for me. A model that \"keeps hallucinating all the time\" will simply not be able to do this, as they would be \"unusable\"."},{"itemId":"49014769","author":"protocolture","createdAt":1784762077,"url":"https://news.ycombinator.com/item?id=49014769","kind":"comment","storyTitle":"How an AI Anime Is Created","quote":">But now I can appreciate doing it for a lovingly-crafted anime to savor the details, and being disappointed when the they're not there or hallucinated/inconsistent due to AI."},{"itemId":"49008919","author":"lardosaurusrex","createdAt":1784736176,"url":"https://news.ycombinator.com/item?id=49008919","kind":"comment","storyTitle":"Why I am not going to buy a computer","quote":"a manual should be able to be pulled off a bookshelf (or opened if digital) and allow you to flip to the relevant page so that you can refer to it as a reference and these jerks are hellbent on blaming users for when their jank documentation looks clearly hallucinated by 3 seperate LLMs."},{"itemId":"49008005","author":"ripe","createdAt":1784732883,"url":"https://news.ycombinator.com/item?id=49008005","kind":"comment","storyTitle":"Judge approves $1.5B Anthropic settlement for pirated books used to train Claude","quote":"OpenAI tried to have the lawsuit thrown out but the courts permitted it to continue. The Times... alleged that OpenAI's ChatGPT and Microsoft's Copilot had produced near-verbatim replicas of copyrighted articles, that the chatbots generated hallucinated content falsely attributed to the Times, ..."},{"itemId":"49005364","author":"baalimago","createdAt":1784721667,"url":"https://news.ycombinator.com/item?id=49005364","kind":"comment","storyTitle":"Advertise in ChatGPT","quote":"Why is A better than B, etc? I don't see how adds could come into that scope either, unless ChatGPT somehow is lying (which would make me swap to a better non-hallucinating model)."},{"itemId":"48999491","author":"al_borland","createdAt":1784674754,"url":"https://news.ycombinator.com/item?id=48999491","kind":"comment","storyTitle":"Ask HN: How do people keep track of organizational knowledge?","quote":"It checks the box for having documentation, if that’s all anyone cares about. But there is no way people are actually reading it, looking for changes is impossible, since the LLM re-writes the entire page every time, and even not knowing their project well, it was full of hallucinations and nonsense."},{"itemId":"48996300","author":"camgunz","createdAt":1784659054,"url":"https://news.ycombinator.com/item?id=48996300","kind":"comment","storyTitle":"I burned all my tokens researching how to save tokens","quote":"> You were just told that this wasn’t stacking but you keep insisting it is. Having an LLM generate something and then asking another LLM to generate that something is asking for 2 hallucinations, not 1."},{"itemId":"48986047","author":"llm_nerd","createdAt":1784589033,"url":"https://news.ycombinator.com/item?id=48986047","kind":"comment","storyTitle":"Who's afraid of Chinese models?","quote":"Models don't have some self identity, beyond what is explicitly handed to them via a system prompt. There have been many, many cases of models identifying as different models by different makers as a basic identity hallucination."},{"itemId":"48975938","author":"disgruntledphd2","createdAt":1784536988,"url":"https://news.ycombinator.com/item?id=48975938","kind":"comment","storyTitle":"Qwen 3.8","quote":"But, the big models kinda do know everything, and yet, here we are, they're still making shit up all the time. In my mostly unscientific RAG experiments I found that larger models were more likely to hallucinate in a RAG setting."},{"itemId":"48975223","author":"stilesja","createdAt":1784531239,"url":"https://news.ycombinator.com/item?id=48975223","kind":"comment","storyTitle":"I burned all my tokens researching how to save tokens","quote":"Sort through them and pull out the 30 best when I have the time. That's the real anti-hallucinations pipeline."},{"itemId":"48975022","author":"TimByte","createdAt":1784529459,"url":"https://news.ycombinator.com/item?id=48975022","kind":"comment","storyTitle":"AI advice made people less accurate but more confident – sudy","quote":"Obviously without a proper RAG pipeline or web search it is just going to hallucinate with maximum confidence."},{"itemId":"48972595","author":"GuB-42","createdAt":1784503030,"url":"https://news.ycombinator.com/item?id=48972595","kind":"comment","storyTitle":"AI advice made people less accurate but more confident – sudy","quote":"So there is a balance to be found. Smarter models tend to do better because they know what is right, compared to smaller models that don't and then assume you are right and hallucinate from there."},{"itemId":"48964912","author":"throwawa14223","createdAt":1784434453,"url":"https://news.ycombinator.com/item?id=48964912","kind":"comment","storyTitle":"AI Mania Is Eviscerating Global Decision-Making","quote":"My team gets praised for using it effectively but we're the team that turned it off. We had an SLA violation because Copilot hallucinated a service we don't have that it claimed was down and we can't convince ops that the service literally doesn't exist."},{"itemId":"48958062","author":"applfanboysbgon","createdAt":1784381899,"url":"https://news.ycombinator.com/item?id=48958062","kind":"comment","storyTitle":"GPT-5.6 used a prompt to close a 30-year gap in convex optimization","quote":"This applies to all claimed proofs, not just AI-generated ones. Even humans hallucinate proofs too!"},{"itemId":"48957569","author":"grayhatter","createdAt":1784377904,"url":"https://news.ycombinator.com/item?id=48957569","kind":"comment","storyTitle":"Kaiser nurses say AI, surveillance are making their jobs and patient care worse","quote":"\"fun\" fact, a number (some? many? all?) of the LLM speech to text models will editorialize or hallucinate words for you, to make your speech fit the pattern it expects."},{"itemId":"48957477","author":"embedding-shape","createdAt":1784377272,"url":"https://news.ycombinator.com/item?id=48957477","kind":"comment","storyTitle":"What AI did to stackoverflow in a graph","quote":"- Ask on IRC: Get piled on for not using the right vocabulary and your IP isn't masked - Ask the LLM: Get hallucinated answer based on old API docs"},{"itemId":"48947571","author":"lelandfe","createdAt":1784297219,"url":"https://news.ycombinator.com/item?id=48947571","kind":"comment","storyTitle":"Blatant AI slop just won a 25k USD DeepMind Kaggle Grand Prize","quote":"Should have said Gemini 1 Hackathon, pesky hallucinations."},{"itemId":"48942149","author":"Revanche1367","createdAt":1784249292,"url":"https://news.ycombinator.com/item?id=48942149","kind":"comment","storyTitle":"NotebookLM is now Gemini Notebook","quote":"In my experience, Gemini 3.x wasn’t just getting a bad rap, it was significantly worse in practice. It could analyze codebases and report back from a one-shot prompt as good as Claude or Codex but any slightly complex task that carried on for more than a few minutes led to hanging, seemingly infinite loops, and bizarre and nonsensical hallucinations, to the point of being unusable for serious work."},{"itemId":"48941288","author":"Chu4eeno","createdAt":1784241930,"url":"https://news.ycombinator.com/item?id=48941288","kind":"comment","storyTitle":"NotebookLM is now Gemini Notebook","quote":"I think you have rosy eyed glasses (or never inspected the output too closely), Gemini 3.1 Pro was very bad at hallucinating."},{"itemId":"48938362","author":"jdw64","createdAt":1784226745,"url":"https://news.ycombinator.com/item?id=48938362","kind":"comment","storyTitle":"Google Gemini Launch Delayed as Tech Falls Short of Internal Goals","quote":"But use Gemini? You get burned by the hallucinations Gemini casually spews out on a daily basis."},{"itemId":"48924256","author":"jdw64","createdAt":1784136509,"url":"https://news.ycombinator.com/item?id=48924256","kind":"comment","storyTitle":"Purging George Orwell's books misses what drives the political right","quote":"So how can anyone call it a \"purge\"? It could be an AI hallucination."},{"itemId":"48886938","author":"jzsfg","createdAt":1783907713,"url":"https://news.ycombinator.com/item?id=48886938","kind":"comment","storyTitle":"I trained a 113M-parameter earthquake LLM from absolute scratch","quote":"It could help researchers and responders rapidly synthesize the likely faulting mechanisms, evaluate the localized tsunami potential, and assess the aftershock hazard based on physical constraints. In disaster response, hallucination is unacceptable."},{"itemId":"48876984","author":"TacticalCoder","createdAt":1783814204,"url":"https://news.ycombinator.com/item?id=48876984","kind":"comment","storyTitle":"I used to love Claude, but the latest models are slowly ruining it","quote":"The real issue with that tone (I'm contemplating cancelling my Anthropic subscription now) is that it's all too often the \"confidently wrong teacher\"."},{"itemId":"48876918","author":"mlinhares","createdAt":1783813476,"url":"https://news.ycombinator.com/item?id=48876918","kind":"comment","storyTitle":"Stop Telling Me to Ask an LLM","quote":"Here's the problem, countering a lie or hallucination takes much more energy than asking Claude something and saying its true."},{"itemId":"48876652","author":"Barbing","createdAt":1783810813,"url":"https://news.ycombinator.com/item?id=48876652","kind":"comment","storyTitle":"Google Search lets creators know more about their reach","quote":"Which lab is going to stop AI from hallucinating first?"},{"itemId":"48867643","author":"tomsayer","createdAt":1783733573,"url":"https://news.ycombinator.com/item?id=48867643","kind":"comment","storyTitle":"Building a real-time AI tutor for 5-year-olds","quote":"If part of the comment was on the accuracy of LLMs; our system of multiple harnessed LLMs running in parallel does deliver pedagogically solid and accurate content and experiences. There's a few comments in this thread about LLM hallucinations - We have curriculum specialists constantly evaluating the interactions that Ello delivers, and it just isn't a major factor right now."},{"itemId":"48863072","author":"jplusequalt","createdAt":1783705909,"url":"https://news.ycombinator.com/item?id=48863072","kind":"comment","storyTitle":"Building a real-time AI tutor for 5-year-olds","quote":">I can assure you that no matter what concerns you may have about hallucinations in LLMs, I can bet everything I have that a reasonably modern model (and I'm thinking in the range of gemini-flash, not Fable) in a well designed harness geared towards tutoring would handily and repeatedly outperform every single teacher I had all throughout my schooling."},{"itemId":"48859926","author":"Marha01","createdAt":1783691485,"url":"https://news.ycombinator.com/item?id=48859926","kind":"comment","storyTitle":"Building a real-time AI tutor for 5-year-olds","quote":"> If you could time-travel back to your 5y old self, would you prefer to be taught by AI tutor given the current state o/t art, or taught by whatever teachers you did have when you were 5? (with all the existing hallucination, breaking through guardrails etc problems of current AI in mind)"},{"itemId":"48855357","author":"hannofcart","createdAt":1783653443,"url":"https://news.ycombinator.com/item?id=48855357","kind":"comment","storyTitle":"Building a real-time AI tutor for 5-year-olds","quote":"But we as kids were much worse off for it. I can assure you that no matter what concerns you may have about hallucinations in LLMs, I can bet everything I have that a reasonably modern model (and I'm thinking in the range of gemini-flash, not Fable) in a well designed harness geared towards tutoring would handily and repeatedly outperform every single teacher I had all throughout my schooling."},{"itemId":"48853373","author":"nijave","createdAt":1783636917,"url":"https://news.ycombinator.com/item?id=48853373","kind":"comment","storyTitle":"Grok 4.5","quote":"I have an example here: https://gist.github.com/nijave/2873b8b10d8c732e46264237b075507a Tldr; all the Claude models had identical tools and some used them efficiently and verified data while others did a crap job and hallucinated responses."},{"itemId":"48851767","author":"TacticalCoder","createdAt":1783627981,"url":"https://news.ycombinator.com/item?id=48851767","kind":"comment","storyTitle":"GLM 5.2 is nearly as accurate as a human book keeper","quote":"Or let's put it this way: it there's one number that is the really the amount of VAT the SME has to legally pay, I'd guesstimate 0.01% of the SMEs out there pay, to the cent, the proper amount. All this to say: the approximations, hallucinations and errors of LLMs in accounting for VAT are probably just on par for the course."},{"itemId":"48848409","author":"babblingfish","createdAt":1783614037,"url":"https://news.ycombinator.com/item?id=48848409","kind":"comment","storyTitle":"A new way to reflect on how you use Claude","quote":"When they say \"set quiet hours and breaks\" it frames this as a user problem. Just get good bruh, it's not that the technology hallucinates or is sycophantic, or basically designed to be a AI girlfriend / boyfriend, it's a skill issue."},{"itemId":"48845650","author":"jrmg","createdAt":1783604307,"url":"https://news.ycombinator.com/item?id=48845650","kind":"comment","storyTitle":"Why is there smoke from the boiler room? – Botanical Garden using Home Assistant","quote":"In my experience (and one look at the Home Assistant forums shows I’m not alone), allowing AI to write Home Assistant configs is a questionable activity. There’s so much old, wrong information out there that it regularly uses (or tries to use) deprecated or removed features, and often even hallucinates config fields if you ask it to do something that is unsupported (and, of course, the user doesn’t know beforehand w…"},{"itemId":"48843020","author":"KronisLV","createdAt":1783587988,"url":"https://news.ycombinator.com/item?id=48843020","kind":"comment","storyTitle":"I think I have LLM burnout","quote":"What a take with no nuance. > do not hallucinate"},{"itemId":"48840400","author":"anigbrowl","createdAt":1783565501,"url":"https://news.ycombinator.com/item?id=48840400","kind":"comment","storyTitle":"I think I have LLM burnout","quote":"started to dread reading LLM output because I know what I’m going to find. False assumptions and hallucinations."},{"itemId":"48829594","author":"ben_w","createdAt":1783502463,"url":"https://news.ycombinator.com/item?id=48829594","kind":"comment","storyTitle":"Ask HN: Is there anything good in the existence of LLMs?","quote":"> Am I hallucinating?"},{"itemId":"48828803","author":"sgt101","createdAt":1783496426,"url":"https://news.ycombinator.com/item?id=48828803","kind":"comment","storyTitle":"IEEE Rolls Out Large Language Models Training Course","quote":"This also is bullshit. Yes, RAG helps and reduces errors, but NOOOO! it does not fix hallucinations..."},{"itemId":"48824289","author":"vitally3643","createdAt":1783460493,"url":"https://news.ycombinator.com/item?id=48824289","kind":"comment","storyTitle":"Fable turned reMarkable into Tom Riddle's diary from Harry Potter","quote":"Because it plain and simple does not work that way. LLMs don't know anything at all, and there literally is no mechanism that informs the model that it's hallucinating or that it doesn't \"know\" something."},{"itemId":"48815603","author":"sebastiennight","createdAt":1783417767,"url":"https://news.ycombinator.com/item?id=48815603","kind":"comment","storyTitle":"The Internet Is Dead and Nobody Cares [video]","quote":"YouTube randomly redirected me to this video as I had auto-play turned on somehow. When the author started asking Gemini whether video X was AI-generated, I was expecting a bunch of hallucinated nonsense."},{"itemId":"48814610","author":"teaearlgraycold","createdAt":1783408522,"url":"https://news.ycombinator.com/item?id=48814610","kind":"comment","storyTitle":"Pruning RAG context down to what the answer actually needs","quote":"RAG, as far as I understand, is a term that came about before LLM tool-calling was as prevalent. Your options were to have an LLM hallucinate up a response, or instead do a [document -> chunk -> embedding -> vector db -> query -> context window] pipeline."},{"itemId":"48811884","author":"_pdp_","createdAt":1783380817,"url":"https://news.ycombinator.com/item?id=48811884","kind":"comment","storyTitle":"GLM 5.2 and the coming AI margin collapse","quote":"Most harnesses out there do very poor job with the open weight models. Unlike Opus, GLM 5.2 ends up in loops and hallucinates a lot more."},{"itemId":"48804912","author":"bArray","createdAt":1783347181,"url":"https://news.ycombinator.com/item?id=48804912","kind":"comment","storyTitle":"OpenPrinter","quote":"> I agree with the rest but Pi Zero W is a specific machine with 512 MB RAM: I thought they came in different variants - my meat-based LLM must have hallucinated it."},{"itemId":"48800882","author":"zkmon","createdAt":1783315029,"url":"https://news.ycombinator.com/item?id=48800882","kind":"comment","storyTitle":"Zuckerberg says AI agent development going slower than expected","quote":"> he expects that the social media giant will begin to experience more significant benefits from its AI investments within the next three to six months. He is hallucinating just like AI, and unable to come to terms with the facts on the ground."},{"itemId":"48792349","author":"geometrein","createdAt":1783240303,"url":"https://news.ycombinator.com/item?id=48792349","kind":"comment","storyTitle":"Surus Agentic Postgres Companion","quote":"Pointing these tools to a prod DB for me feels like handing your car keys to your drunk friend.. It has been shown that LLM’s hallucinate at non trivial rates and that is the main reason why narrow vertical \"agnets\" became so popular in the first place."},{"itemId":"48786192","author":"andy99","createdAt":1783179714,"url":"https://news.ycombinator.com/item?id=48786192","kind":"comment","storyTitle":"Potential session/cache leakage between workspace instances or consumer accounts","quote":"I realize hallucination has no precise definition but this doesn’t sound at all like anything I’ve ever heard called hallucination."},{"itemId":"48785756","author":"aka-rider","createdAt":1783175962,"url":"https://news.ycombinator.com/item?id=48785756","kind":"comment","storyTitle":"Jamesob's guide to running SOTA LLMs locally","quote":"Orqestra was/is self-hosted, being developed by, well, orchestra of Qwen agents. Moreover, Opus and GPT-5.5 break similarly, yeah they will withstand much more pressure, but they will hallucinate and loop nevertheless."},{"itemId":"48784257","author":"aka-rider","createdAt":1783160004,"url":"https://news.ycombinator.com/item?id=48784257","kind":"comment","storyTitle":"Jamesob's guide to running SOTA LLMs locally","quote":"The main idea is: - separate session that burns context on reading code and calling tools (context7, etc) -> markdown report \"here are relevant patrts of code, docs\" \"with evidence\" to prevent hallucinations"},{"itemId":"48775228","author":"ProllyInfamous","createdAt":1783087549,"url":"https://news.ycombinator.com/item?id=48775228","kind":"comment","storyTitle":"Please stop the AI confidence theater","quote":"As a hobbyist exploring LLMs since Summer 2022 (electrician by trade)... I've been as blown away by generative AI results/output as I've been left laughing at its hallucinations."},{"itemId":"48769909","author":"keepamovin","createdAt":1783044693,"url":"https://news.ycombinator.com/item?id=48769909","kind":"story","storyTitle":"Ask HN: Once you make your money from vibe coding innumerable products, then?","quote":"Also - @dang - your \"AI comment detection\" is hallucinating."},{"itemId":"48769632","author":"sudo_cowsay","createdAt":1783042589,"url":"https://news.ycombinator.com/item?id=48769632","kind":"comment","storyTitle":"Show HN: I made a tool that prevents websites from tracking you","quote":"AI slop is becoming an overused saying these days. While it originally meant \"Lack of Utility\" and \"Errors and Hallucinations,\" it now seems to defined by anything made with AI, regardless of whether it's good or not."},{"itemId":"48749825","author":"zephen","createdAt":1782924890,"url":"https://news.ycombinator.com/item?id=48749825","kind":"comment","storyTitle":"We Are the Last People Who Know How It Works","quote":"> A \"bug\" then would be considered \"weak accuracy\" not \"crash\" or \"incorrect behavior\", When a lawyer asks an LLM for citations of cases that support his position, he is arguably doing something stupid, because embedding an assertion such as \"Show me cases which support X\" is just asking for hallucinatory trouble with many current LLMs."},{"itemId":"48748885","author":"lebovic","createdAt":1782921263,"url":"https://news.ycombinator.com/item?id=48748885","kind":"comment","storyTitle":"Claude Science","quote":"I assume they do hallucinate, just like with coding or finding vulnerabilities."},{"itemId":"48745757","author":"pathmap","createdAt":1782909691,"url":"https://news.ycombinator.com/item?id=48745757","kind":"story","storyTitle":"Show HN: I made an app that can accelerate cures and nobody fucking cares","quote":"# About the Software Long story short, I started researching ALS and published some AI research and shortly thereafter realized how many hallucinations and issues it had."},{"itemId":"48736063","author":"raphman","createdAt":1782840305,"url":"https://news.ycombinator.com/item?id=48736063","kind":"comment","storyTitle":"Claude Science","quote":"tl;dr: Use this if you don't like doing science or doing things well. It hallucinates references."},{"itemId":"48735742","author":"beebeeVB","createdAt":1782839142,"url":"https://news.ycombinator.com/item?id=48735742","kind":"story","storyTitle":"Show HN: Trajeckt, a Firewall for AI Agents","quote":"As there are more and more agents in the internet; Security is going to be a big problem. Currently, the problem is solved using a LLM to guard Agent but this creates the problem of hallucination and latency, so I coded a firewall in rust that runs under five milliseconds."},{"itemId":"48726867","author":"beebeeVB","createdAt":1782777168,"url":"https://news.ycombinator.com/item?id=48726867","kind":"story","storyTitle":"Show HN: A Firewall for AI agents with auditing","quote":"As there are more and more agents in the internet; Security is going to be a big problem. Currently, the problem is solved using a LLM to guard Agent but this creates the problem of hallucination and latency, so I coded a firewall in rust that runs under five miliseconds."},{"itemId":"48725262","author":"atleastoptimal","createdAt":1782767330,"url":"https://news.ycombinator.com/item?id=48725262","kind":"comment","storyTitle":"The Permission Slip","quote":"I think the use of the word \"hallucination\" with respect to AI confidently making errors has led a lot of people astray, including the author."},{"itemId":"48720349","author":"kalcode","createdAt":1782746016,"url":"https://news.ycombinator.com/item?id=48720349","kind":"comment","storyTitle":"Tokenmaxxing is dead, long live tokenmaxxing","quote":"They expect a one-shot miracle. They say, \"Hey, my code is buggy, fix it.\" When the LLM inevitably hallucinates or steers them wrong, they give up."},{"itemId":"48717098","author":"jijijijij","createdAt":1782726900,"url":"https://news.ycombinator.com/item?id=48717098","kind":"comment","storyTitle":"Why current LLM costs are not sustainable","quote":"Putting progression into relation to effort is required IMO. > Measured hallucination rates have gone down, and don't seem to be an issue for achieving things, as per METR and other results."},{"itemId":"48716651","author":"YossarianFrPrez","createdAt":1782723556,"url":"https://news.ycombinator.com/item?id=48716651","kind":"comment","storyTitle":"HackerRank open sourced its ATS. My resume scored 90/100. Oh wait 74. No – 88","quote":"Also, fwiw, the JSON should force the LLM to pick between categorical options as much as possible. Forced-choice structured output should, at least in theory, cut down on hallucinatory responses and constrain judgement calls."},{"itemId":"48716218","author":"rvz","createdAt":1782720344,"url":"https://news.ycombinator.com/item?id=48716218","kind":"comment","storyTitle":"HackerRank open sourced its ATS. My resume scored 90/100. Oh wait 74. No – 88","quote":"Rather than using this less accurate one that hallucinates out its decision depending on a dice roll."},{"itemId":"48715393","author":"BIackSwan","createdAt":1782712829,"url":"https://news.ycombinator.com/item?id=48715393","kind":"comment","storyTitle":"I used Claude Code to get a second opinion on my MRI","quote":"It has helped me personally solve longer chronic problems in my family that doctors just dont have the time to go into indepth due to their (understandable) lack of time. Its in alpha and AI hallucinates."},{"itemId":"48714162","author":"usef-","createdAt":1782700569,"url":"https://news.ycombinator.com/item?id=48714162","kind":"comment","storyTitle":"Why current LLM costs are not sustainable","quote":"\"Is progress efficiency per dollar decreasing\" is a different claim to the earlier \"is progress slowing?\", though. > LLMs still hallucinate."},{"itemId":"48709241","author":"seanmcdirmid","createdAt":1782666201,"url":"https://news.ycombinator.com/item?id=48709241","kind":"comment","storyTitle":"I used Claude Code to get a second opinion on my MRI","quote":"I don’t understand how this is a different result than giving any LLM a task that is not completely grounded? I’ve observed this in coding tasks, if I forget to include a file referred to in the spec, the LLM will just hallucinate a version of it and my results suck."},{"itemId":"48694101","author":"overgard","createdAt":1782522447,"url":"https://news.ycombinator.com/item?id=48694101","kind":"comment","storyTitle":"U.S. government will decide who gets to use GPT-5.6","quote":"Actually, yesterday I was trying to lookup an API function in a game engine and I had google hallucinate the function call twice in a row, and all I could think was \"I really just want good search again\"."},{"itemId":"48692773","author":"metaopai","createdAt":1782512683,"url":"https://news.ycombinator.com/item?id=48692773","kind":"comment","storyTitle":"Data centers trigger voter backlash","quote":"2) Session Context = L1, L2, L3, CPU Cache the entire AI industry is operating within the CPU cache of the LLM provider this is why cost moves quadratically, increases noise - signal ration, regenerative feedback loop, it dilutes the user narration, and it actually creates an architectural induce hallucination."},{"itemId":"48691667","author":"ben77777","createdAt":1782506212,"url":"https://news.ycombinator.com/item?id=48691667","kind":"comment","storyTitle":"Literate Programming in the Age of LLMs","quote":"LLMs introduce a third audience: the coding agent. From experience, lack of documentation causes agents to hallucinate and make wrong assumptions."},{"itemId":"48687965","author":"jijijijij","createdAt":1782488463,"url":"https://news.ycombinator.com/item?id=48687965","kind":"comment","storyTitle":"Why current LLM costs are not sustainable","quote":"Also, manual capability extensions, or use/edge case adaptations, which may improve subjective usability are not exactly advances in AI as technology. LLMs still hallucinate."},{"itemId":"48686251","author":"laszlojamf","createdAt":1782479658,"url":"https://news.ycombinator.com/item?id=48686251","kind":"comment","storyTitle":"Google hallucinated that I am sponsored by Ground News","quote":"But the outrage is getting old. LLMs hallucinate."},{"itemId":"48680080","author":"tedmiston","createdAt":1782427046,"url":"https://news.ycombinator.com/item?id=48680080","kind":"comment","storyTitle":"Ask HN: Where is the programming profession going?","quote":"It sees the general pattern in an API, then makes up an endpoint that sounds good, follows the pattern, but isn’t actually implemented. I remember seeing this maybe 6+ months ago, but using paid plans, RAG, and a high thinking mode has eliminated a ton (almost all) of those kinds of hallucinations."},{"itemId":"48666818","author":"sillyfluke","createdAt":1782343838,"url":"https://news.ycombinator.com/item?id=48666818","kind":"comment","storyTitle":"How to burst the AI bubble: Strike at its roots","quote":"The non-LLM AI stuff in the medical field is what I find most promising. Not the banal AI doctors and nurses replacing real doctors and nurses, but more the research end of things, where hallucinating and pattern matching to create new compounds to test seems promising."},{"itemId":"48662208","author":"healersource","createdAt":1782318283,"url":"https://news.ycombinator.com/item?id=48662208","kind":"story","storyTitle":"Show HN: (Spotlight/Raycast for Web Search not local) && (compare AI responses)","quote":"Its a desktop app (Mac/Windows) that lets you prompt multiple AI models side-by-side without copy/paste. I was tired of AI errors & hallucinations and wanted an easier way to verify responses."},{"itemId":"48658370","author":"type0","createdAt":1782302002,"url":"https://news.ycombinator.com/item?id=48658370","kind":"comment","storyTitle":"Raspberry Pi Pico W as USB Wi-Fi Adapter","quote":"I'm actually tired that people put stupid questions in ChatGPT and then present in with a straight face as a source of truth. Sometimes it hallucinates completely, sometimes the conditions or regulation have changed and it gives false answers and no-one cares."},{"itemId":"48657410","author":"coldtea","createdAt":1782294153,"url":"https://news.ycombinator.com/item?id=48657410","kind":"comment","storyTitle":"The loop is here (reply to \"The Coming Loop\" by mitsuhiko)","quote":"Aside from the obvious AI slop writing, nothing in the example he showed guarantees that this is \"improving itself\". Each iteration of the answer, as given back by the LLM can get increasingly worse, especially as the context window grows, or as something that shouldn't be there (e.g. a hallucination) in the previous answers becomes part of the later answers' focus, or the \"roll of the dice\" that's an LLM answer get…"},{"itemId":"48652527","author":"yencabulator","createdAt":1782254598,"url":"https://news.ycombinator.com/item?id=48652527","kind":"comment","storyTitle":"Codex logging bug may write TBs to local SSDs","quote":"I sometimes use LLMs as search engine replacements for finding libraries for a specific niches. Gemini is the only one that seems to routinely hallucinate Rust crates from thin air, where otherwise Google is the one most willing to let their LLM peek at a search index to have up-to-date ground facts."},{"itemId":"48640773","author":"hash0","createdAt":1782193147,"url":"https://news.ycombinator.com/item?id=48640773","kind":"story","storyTitle":"Ask HN: NIST Randomness Beacon Interruption?","quote":"I reached out to beacon@nist.gov (and via twitter/x.com) but got no reply whatsoever. Gemini hallucinated a story about a government shut-down and that the tech community was well aware of that - without proof, of course."},{"itemId":"48632115","author":"dragontamer","createdAt":1782144465,"url":"https://news.ycombinator.com/item?id=48632115","kind":"comment","storyTitle":"I Canceled My French Tutor and Built an LLM Tool That Does It Better","quote":"German learner here. Most of the German discord group is mostly annoyed at how awful LLM explanations and hallucinations are."},{"itemId":"48626728","author":"ben_w","createdAt":1782111799,"url":"https://news.ycombinator.com/item?id=48626728","kind":"comment","storyTitle":"The Doom Justifies the Valuation","quote":"30% of your current total electrical generation capacity, and equivalent to 76% of your total electrical demand in 2022. At least, assuming the \"33 GW installed, 300 GW more planned\" claim I've heard was not itself an AI hallucination."},{"itemId":"48621100","author":"mohitjandwani","createdAt":1782065193,"url":"https://news.ycombinator.com/item?id=48621100","kind":"comment","storyTitle":"Analyst Kit (YC W23): Turn your Claude / Codex into an investment analyst (Free)","quote":"Claude /codex is totally unreliable. It hallucinates, often stops before completing the research and often uses unreliable data sources."},{"itemId":"48612219","author":"sometimelurker","createdAt":1781983874,"url":"https://news.ycombinator.com/item?id=48612219","kind":"comment","storyTitle":"GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2","quote":"ehh ur right but there's a lot of nuance here. if you have a system that doesn't hallucinate a ton and is still very \"creative\" that's great, and probably much better than a hallucinating system regardless of its creativity."},{"itemId":"48611816","author":"nathan_compton","createdAt":1781981309,"url":"https://news.ycombinator.com/item?id=48611816","kind":"comment","storyTitle":"GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2","quote":"Synthesizing a bunch of stuff I've read here lately, it seems like if OpenAI and Claude have actually found product market fit (generating code) then the question of hallucination is going to get less attention in the future."},{"itemId":"48609900","author":"utkuumur","createdAt":1781969138,"url":"https://news.ycombinator.com/item?id=48609900","kind":"comment","storyTitle":"GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2","quote":"I think this effects is more easy to see on Gemini Flash, it hallucinates like crazy but looks like its by design to boost benchmarks."},{"itemId":"48609875","author":"resters","createdAt":1781968980,"url":"https://news.ycombinator.com/item?id=48609875","kind":"comment","storyTitle":"GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2","quote":"Meatspace metaphors map badly when we have massive context and don't need the same limits. How different is hallucination from extrapolation, etc."},{"itemId":"48609867","author":"MattRogish","createdAt":1781968923,"url":"https://news.ycombinator.com/item?id=48609867","kind":"comment","storyTitle":"GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2","quote":"There are a bounded number of (useful) derivations/combinations of Duff's device. If Frontier Labs wish to reduce hallucinations on factual things, they will have to hire people (or the data providers will need to) to do fundamental research above and beyond what is available in extant literature and the web."},{"itemId":"48607624","author":"bjackman","createdAt":1781946293,"url":"https://news.ycombinator.com/item?id=48607624","kind":"comment","storyTitle":"John Jumper to join Anthropic","quote":"FWIW issue is never really about the code it writes it's about general intelligence. Gemini hallucinates like it's 2024, fails to follow instructions, and goes down wildly wrong debugging paths."},{"itemId":"48606508","author":"fzeroracer","createdAt":1781931985,"url":"https://news.ycombinator.com/item?id=48606508","kind":"comment","storyTitle":"Is AI ruining our skills? Early results are in – and they're not good","quote":"These are not hypotheticals, as an example just recently ChatGPT advised a user to mix bleach and vinegar. I have seen Googles AI hallucinate many details that I had to double check and would've been disastrous had I assumed it was truthy."},{"itemId":"48606458","author":"tacomagick","createdAt":1781931406,"url":"https://news.ycombinator.com/item?id=48606458","kind":"comment","storyTitle":"Norway imposes near ban on AI in elementary school","quote":"It lets me focus on learning rather than preparing to learn which takes motivation away from actual learning for me personally. For the second part I really personally think that is a big case of AI misuse, ever since I have hooked up MCP servers and let AI do web searches and even just letting Claude use Chrome it has never hallucinated a single link ever."},{"itemId":"48606437","author":"wvenable","createdAt":1781931081,"url":"https://news.ycombinator.com/item?id=48606437","kind":"comment","storyTitle":"Is AI ruining our skills? Early results are in – and they're not good","quote":"Again with the hypotheticals! You literally cannot discuss this subject without hallucinating things that don't exist."},{"itemId":"48604758","author":"tmp10423288442","createdAt":1781913499,"url":"https://news.ycombinator.com/item?id=48604758","kind":"comment","storyTitle":"Hyundai buys Boston Dynamics","quote":"So does ChatGPT. The hallucination problem is not completely solved, but much better than a few years ago, especially if you use reasoning mode, where it’s more likely to spontaneously do a web search"},{"itemId":"48604666","author":"solid_fuel","createdAt":1781912596,"url":"https://news.ycombinator.com/item?id=48604666","kind":"comment","storyTitle":"GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2","quote":"> It’s been proven that when a model is trained on large volumes of highly factual and non-theoretical data, it learns to always have an answer. DeepSeek V4 Pro (1.6T params, 49B active, 44 AA Intelligence Index score) has a ludicrous 94% hallucination score on the AA-Omniscience benchmark, meaning on questions that it couldn’t figure out, it only stated that it didn’t know around 6% of the time, and the rest it con…"},{"itemId":"48604131","author":"PJHkorea","createdAt":1781908765,"url":"https://news.ycombinator.com/item?id=48604131","kind":"comment","storyTitle":"Ask HN: Open-Source Intelligence","quote":"If you run an MCTS coin toss relying solely on the token generation probabilities of a simple LLM without any verification mechanisms (compiler, test code, reward model), the result is code that looks plausible on the surface but is completely broken on the inside—a 'lump of hallucination.'"},{"itemId":"48603605","author":"FloorEgg","createdAt":1781905413,"url":"https://news.ycombinator.com/item?id=48603605","kind":"comment","storyTitle":"Norway imposes near ban on AI in elementary school","quote":"Giving students uncontrolled access to generic LLMs probably would hurt outcomes. Research process is slow (IRB and all that) so they are dealing with data from years ago (models that confident hallucinated a lot more than current SOTA) so if thats what they are basing it on its reasonable."},{"itemId":"48603418","author":"FloorEgg","createdAt":1781904342,"url":"https://news.ycombinator.com/item?id=48603418","kind":"comment","storyTitle":"Norway imposes near ban on AI in elementary school","quote":"It's frustrating to me when bold statements are directed at \"AI\" holistically and vaguely, completely ignoring any nuance. There is a massive gap between letting elementary students free reign use chatGPT 3.5 (hallucinations and all) to do whatever, vs using a very guard-railed pedagogically optimized app powered by a SOTA model to support students in a specific way that accelerates good outcomes."},{"itemId":"48603010","author":"joe_mamba","createdAt":1781901644,"url":"https://news.ycombinator.com/item?id=48603010","kind":"comment","storyTitle":"John Jumper to join Anthropic","quote":"I think Google just doesn't care about being the SOTA for coding, reasoning and accuracy, since they're in the ads and search business for everyone, not in the agentic coding business for pro-sumers, so if the answers are some hallucinations that sound \"good enough\" to its clueless search user base, but is at least dirt cheap to run on their datacenter hardware, then it's already more than enough for them and they c…"},{"itemId":"48588547","author":"WhitneyLand","createdAt":1781803249,"url":"https://news.ycombinator.com/item?id=48588547","kind":"comment","storyTitle":"DeepSeek Introduces Vision","quote":"Most of the problem is that for voice chat, you usually get no reasoning at all and no tool use at all to research or ground assumptions. For example for voice ChatGPT still uses a quantized gpt40 non-reasoning model that hallucinates pretty frequently."},{"itemId":"48586353","author":"jeidoban","createdAt":1781794381,"url":"https://news.ycombinator.com/item?id=48586353","kind":"comment","storyTitle":"Show HN: Ironsmith – Generate personal native Mac apps with local LLMs","quote":"Turns out these little models are pretty decent at writing full apps if you fix all of their hallucinations and syntax errors."},{"itemId":"48578484","author":"avdelazeri","createdAt":1781739025,"url":"https://news.ycombinator.com/item?id=48578484","kind":"comment","storyTitle":"Only 16 Percent of Americans Think AI Will Have a Positive Impact on Society","quote":"They kept arguing that my service was working properly despite the snapped cables. I reckon that the AI assistants were hallucinating the answer to a ping or something like that."},{"itemId":"48563191","author":"YuechenLi","createdAt":1781649259,"url":"https://news.ycombinator.com/item?id=48563191","kind":"story","storyTitle":"Show HN: The Oct Programming Language for scientific programming","quote":"About 75k lines of Oct code are in the repo across experiments, libraries, and examples. Pretty much all of it is written by Claude/Codex, my role is only to prevent drift/hallucinations."},{"itemId":"48559559","author":"stevesolun","createdAt":1781633688,"url":"https://news.ycombinator.com/item?id=48559559","kind":"story","storyTitle":"Show HN: Ctx, save tokens by loading only the relevant tools","quote":"How does it know? To make sure results are not hallucinated, and repeatable, I curated a list of 91k+ skills, 467 agents, 10.7k MCP servers, 207 harnesses, and built a graph to help ctx make decisions on what to recommend."},{"itemId":"48557473","author":"monegator","createdAt":1781626133,"url":"https://news.ycombinator.com/item?id=48557473","kind":"comment","storyTitle":"Running local models is good now","quote":"Same tasks: same issues with short context, with long context it gave absolutely useless answers when looking at code, but it took 1/3 the time of qwen. In producing documentation, instead, it was much faster, and it never hallucinated data."},{"itemId":"48554652","author":"alaskahoffman","createdAt":1781615005,"url":"https://news.ycombinator.com/item?id=48554652","kind":"story","storyTitle":"Show HN: Portaltext, grand strategy-style recursive tooltips for the web","quote":"It reads the page you're on, and the page you're hovering over, and then acts as a gel between them so the tooltip is always contextualized and doesn't need to hallucinate new information."},{"itemId":"48548119","author":"pieterk","createdAt":1781563900,"url":"https://news.ycombinator.com/item?id=48548119","kind":"comment","storyTitle":"Ask HN: Has anyone replaced Claude/GPT with a local model for daily coding?","quote":"Maybe even more useful than Opus when I have all the constraints to an issue. There is less \"knowledge\" in the model (I get by with 48GB of RAM allocated to an 8b quant), so it has fewer things to hallucinate about."},{"itemId":"48545613","author":"aspenmartin","createdAt":1781550233,"url":"https://news.ycombinator.com/item?id=48545613","kind":"comment","storyTitle":"Open source AI must win","quote":"That is a crazy thing to think in light of all of the datapoints we have. > One of the core technical reason is that hallucination destroy enterprise economics."},{"itemId":"48540524","author":"whitefables","createdAt":1781527948,"url":"https://news.ycombinator.com/item?id=48540524","kind":"comment","storyTitle":"[dead]","quote":"- Anthropic has a section for heron_brook which allow them to inject arbitrary prompt into the live system prompt of already-installed clients - sticking issues like excessive use of bullet points and URL hallucination needs to be fixed in prompt"},{"itemId":"48539988","author":"zwyld","createdAt":1781524515,"url":"https://news.ycombinator.com/item?id=48539988","kind":"story","storyTitle":"Show HN: TopoGlyph – an extensible language system for problem solving","quote":"Try the Claude Code plugin with a problem / domain you are an expert in an judge for yourself. If you read the core module it may seem insane in some places, and it should because there are 'hallucinated' concepts and conclusions (not defending them)."},{"itemId":"48538778","author":"gnunicorn","createdAt":1781515646,"url":"https://news.ycombinator.com/item?id=48538778","kind":"story","storyTitle":"What are you looking for when reviewing LLM generated code?","quote":"Especially when CI/tooling takes care of all the lints and compile problems? Or do you just yolo that code now, accepting the occasional hallucination?"},{"itemId":"48537255","author":"whitefables","createdAt":1781504119,"url":"https://news.ycombinator.com/item?id=48537255","kind":"comment","storyTitle":"[dead]","quote":"- Anthropic has a section for `heron_brook` which allow them to inject arbitrary prompt into the live system prompt of already-installed clients - sticking issues like excessive use of bullet points and URL hallucination needs to be fixed in prompt"},{"itemId":"48536427","author":"dualvariable","createdAt":1781495934,"url":"https://news.ycombinator.com/item?id=48536427","kind":"comment","storyTitle":"Why Is Claude Turning into an a**Hole?","quote":"And we don't have good heuristics for what mutations would completely inactivate a virus versus enhancing its virulence. Actual scientists won't be able to vibecode up some pandemic viruses because we have no idea how to do that and LLMs are just going to hallucinate."},{"itemId":"48535321","author":"_wire_","createdAt":1781486537,"url":"https://news.ycombinator.com/item?id=48535321","kind":"comment","storyTitle":"KPMG report on AI found riddled with AI hallucinations","quote":"\"Guardrails\" are direct lines of injection of ideology, and \"hallucinations\" can serve to mask."},{"itemId":"48533623","author":"tristanj","createdAt":1781476440,"url":"https://news.ycombinator.com/item?id=48533623","kind":"comment","storyTitle":"Why Is Claude Turning into an a**Hole?","quote":"My conclusion is that pushing back against the user & questioning the user's premise forces the model to think more than it would otherwise, which leads to better model performance. But it causes situations where the user has esoteric, specialized knowledge the model can't verify publicly and the model hallucinates evidence and pushes back."},{"itemId":"48532848","author":"SpicyLemonZest","createdAt":1781471909,"url":"https://news.ycombinator.com/item?id=48532848","kind":"comment","storyTitle":"KPMG pulls report on AI usage due to apparent hallucinations","quote":"The hallucinations here (https://gptzero.me/news/investigations-kpmg/) would have passed a cursory reference check."},{"itemId":"48531569","author":"chrisjj","createdAt":1781465023,"url":"https://news.ycombinator.com/item?id=48531569","kind":"comment","storyTitle":"KPMG report on AI found riddled with AI hallucinations","quote":"the majority of its references were flawed, amid the latest news of AI-hallucinated reports published by professional services firms."},{"itemId":"48524345","author":"cwnyth","createdAt":1781413436,"url":"https://news.ycombinator.com/item?id=48524345","kind":"comment","storyTitle":"Show HN: I am building a map of people who lived in the Roman Empire","quote":"Yep, Laulenia is the original name. Seems like AI is hallucinating here."},{"itemId":"48521935","author":"root-parent","createdAt":1781388214,"url":"https://news.ycombinator.com/item?id=48521935","kind":"comment","storyTitle":"Open source AI must win","quote":"An MIT 2025 enterprise AI study found $30 to 40B spent on GenAI with 95% of organizations seeing zero return. One of the core technical reason is that hallucination destroy enterprise economics."},{"itemId":"48518989","author":"Markoff","createdAt":1781369245,"url":"https://news.ycombinator.com/item?id=48518989","kind":"comment","storyTitle":"\"Don't You Just Upload It to ChatGPT?\"","quote":"Some of my clients use the LLM to do QA on my translations, dunno about English, but for smaller languages (think less than 10M speakers) LLM still ain't that good, at least 5-10% LLM suggestions contain hallucinated made-up words which don't even exist."},{"itemId":"48518481","author":"sieste","createdAt":1781366187,"url":"https://news.ycombinator.com/item?id=48518481","kind":"comment","storyTitle":"RTX 5080 and RTX 3090 Setup: 80 Tok/s on Qwen 3.6 27B Q8","quote":"But the way Qwen fails is much more straightforward. In writing tasks Qwens hallucinations and bullshitting are much easier to spot because it doesn't have the sleek vocabulary and wordsmithing skills to disguise its ignorance."},{"itemId":"48516055","author":"general1465","createdAt":1781349258,"url":"https://news.ycombinator.com/item?id=48516055","kind":"comment","storyTitle":"Automating myself out of development","quote":"However if your main job is just doing CRUD stuff, then you are screwed. Secondly hallucination is really Achilles heel of every LLM."},{"itemId":"48510803","author":"colechristensen","createdAt":1781308694,"url":"https://news.ycombinator.com/item?id=48510803","kind":"comment","storyTitle":"/architect: Reduce Fable tokens by 80%, Fable orchestrates/reviews, Codex builds","quote":"Writing terraform code for GCP, debugging app launch and Dockerfile issues, that sort of thing. It was going in loops hallucinating features of GCP, looking things up in strange ways, running terraform apply after being explicitly told in the last interaction not to, and overall not solving problems."},{"itemId":"48501811","author":"Aldipower","createdAt":1781256128,"url":"https://news.ycombinator.com/item?id=48501811","kind":"comment","storyTitle":"If you are asking for human attention, demonstrate human effort","quote":"I am offering a product (via MCP) that interacts with LLMs and user data. Every single day I get user support emails to my inbox written by their LLMs with LLM hallucinations."},{"itemId":"48495999","author":"comboy","createdAt":1781209886,"url":"https://news.ycombinator.com/item?id=48495999","kind":"comment","storyTitle":"Claude Fable 5: mid-tier results on coding tasks","quote":"Because they are pretty good at digging out relevant tmp files and making use of whatever is out there. With my fable adventures I caught it hallucinating something and stating it as a fact in CLI twice."},{"itemId":"48484508","author":"nikolay","createdAt":1781136128,"url":"https://news.ycombinator.com/item?id=48484508","kind":"comment","storyTitle":"Notes on DeepSeek","quote":"What's brutal is that Google, which started this AI revolution, has literally the worst coding model! I tried 3.5 Flash last week (the stupid still pays for Ultra due to Google One's storage), and before I gave up on 3.1 Pro, I saw a coding agent hallucinate for the first time in months, even at the highest effort level!"},{"itemId":"48483889","author":"mhitza","createdAt":1781131916,"url":"https://news.ycombinator.com/item?id=48483889","kind":"comment","storyTitle":"Policy on the AI Exponential","quote":"I had the hardest time accepting one of his first points that LLMs could barely write a line of code 4 years ago. ChatGPT 3.5 was reasonable at code writing but hallucinated a lot of library functions."},{"itemId":"48472061","author":"black_knight","createdAt":1781071308,"url":"https://news.ycombinator.com/item?id=48472061","kind":"comment","storyTitle":"Claude Fable 5","quote":"I don’t care to share my exact problems. Mostly because gpt -5.5 hallucinates false solutions, and I would rather not have people reply with \"Oh but ChatGPT solves it!\", because it takes expert knowledge to debunk them."},{"itemId":"48471034","author":"russellbeattie","createdAt":1781061921,"url":"https://news.ycombinator.com/item?id=48471034","kind":"comment","storyTitle":"German ruling declares Google liable for false answers in AI Overviews","quote":"I first tried it to remind me of what happened in a previous book in a series that I was reading. When I realized it was either misstating plot points or straight up hallucinating, I tried it on a bunch more books to amuse myself."},{"itemId":"48456209","author":"zmmmmm","createdAt":1780977712,"url":"https://news.ycombinator.com/item?id=48456209","kind":"comment","storyTitle":"Siri AI","quote":"Hard. Hallucinations all over the place, false confidence, inability to act, inability of me to trust their actions."},{"itemId":"48453711","author":"hn_throwaway_99","createdAt":1780960160,"url":"https://news.ycombinator.com/item?id=48453711","kind":"comment","storyTitle":"AI is slowing down","quote":"The legal profession is already spending a lot on AI, and it will only grow. Again, we all like to read about hallucinated case citations, but those are solvable problems (honestly I felt they were more human problems than tech problems to begin with) and there are so many areas in research and document summarization that AI is really good at."},{"itemId":"48447630","author":"zachthewf","createdAt":1780936833,"url":"https://news.ycombinator.com/item?id=48447630","kind":"comment","storyTitle":"AI is slowing down","quote":"As I've written before, hallucinations are a feature not a bug."},{"itemId":"48443162","author":"camgunz","createdAt":1780911617,"url":"https://news.ycombinator.com/item?id=48443162","kind":"comment","storyTitle":"LLMs are eroding my software engineering career and I don't know what to do","quote":"As models converge on 100%, progress obviously slows. We are nowhere near 100% on important benchmarks like hallucinations: https://artificialanalysis.ai/evaluations/omniscience?model-filters=frontier-model&models=gemini-3-1-pro-preview%2Cgemini-3-5-flash%2Cgemini-3-5-flash-medium%2Cgemini-3-pro%2Cgemini-3-flash-reasoning%2Cgemini-3-pro-low%2Cgemini-3-5-flash-minimal%2Cgemini-3-flash%2Cgemini-2-5-pro%2Cgemini-3-1-fl…"},{"itemId":"48442344","author":"dmje","createdAt":1780904225,"url":"https://news.ycombinator.com/item?id=48442344","kind":"comment","storyTitle":"Cannibalism","quote":"Great piece, well written and succinctly sums up my thoughts. The bit I still don’t understand is how we all put up with the hallucinations."},{"itemId":"48438046","author":"wolfgangbabad","createdAt":1780862788,"url":"https://news.ycombinator.com/item?id=48438046","kind":"comment","storyTitle":"LLMs are eroding my software engineering career and I don't know what to do","quote":"In my opinion the LLMs in the \"cloud\" reached plateau. Still the same probabilistic hallucinations."},{"itemId":"48434992","author":"mazinz","createdAt":1780841081,"url":"https://news.ycombinator.com/item?id=48434992","kind":"story","storyTitle":"Ask HN: Is it feasible to run a model on device for complete privacy?","quote":"Need vision and larger context windows for an application I am working on. Results were quite poor Gemma 4E2B probably the best of the ones I tired but still fell apart and keep hallucinating with ~5000 tokens."},{"itemId":"48434781","author":"ohyes","createdAt":1780839731,"url":"https://news.ycombinator.com/item?id=48434781","kind":"comment","storyTitle":"LLMs are eroding my software engineering career and I don't know what to do","quote":"Claude will still very often hallucinate bugs, explanations, domain requirements, that have no basis in reality."},{"itemId":"48432753","author":"drpotato","createdAt":1780818186,"url":"https://news.ycombinator.com/item?id=48432753","kind":"comment","storyTitle":"I design with Claude more than Figma now","quote":"A non-trivial amount of which is now outsourced to a language model. It paints over the cracks in the prompt, hallucinates how things should work when unspecified, what would normally make someone stop and say “this isn’t quite working”, “how should I communicate this idea” or “what happens when…” has gone."},{"itemId":"48430332","author":"monuszero","createdAt":1780789928,"url":"https://news.ycombinator.com/item?id=48430332","kind":"comment","storyTitle":"Adafruit receives demand letter from Fenwick legal counsel on behalf of Flux.ai","quote":"Ultimately found their team to be responsive on discord and figured they have every incentive to solve this for me by the next time I find time for a hobby electronics project. Generally the llms were awesome at discovering the chips I needed (with a few hallucinations promptly uncovered - like a TI mux chip which didn’t exist in the automotive grade designation for that package) without a doubt shortened my search …"},{"itemId":"48427178","author":"shoushen","createdAt":1780767651,"url":"https://news.ycombinator.com/item?id=48427178","kind":"story","storyTitle":"A wild idea: Abstract reality using ontology","quote":"Many unresolved challenges remain. Among them, *hallucinations* have become a major bottleneck, greatly limiting LLMs' adoption in real-world production scenarios."},{"itemId":"48421662","author":"duttish","createdAt":1780723505,"url":"https://news.ycombinator.com/item?id=48421662","kind":"comment","storyTitle":"Ask HN: Why is the HN crowd so anti-AI?","quote":"I worry about the next generation. For me the mental framing of \"It's all hallucinations, some of those hallucinations are useful\" is helpful to keep frustration in check as I ask it to review the same implementation plan for the 4th time and it turns up different issues because the input was slightly different, or review the output code and see allow(dead_code) despite my claude.md forbidding it."},{"itemId":"48421553","author":"vesche","createdAt":1780722031,"url":"https://news.ycombinator.com/item?id=48421553","kind":"comment","storyTitle":"Ask HN: What was your \"oh shit\" moment with GenAI?","quote":"3) Big oh shit moment earlier this year that I believe for me started with the Opus 4.6 model + Cursor. The results were noticeably better, hallucinated much less, could solve complex problems with much less intervention."},{"itemId":"48420373","author":"tristanj","createdAt":1780708109,"url":"https://news.ycombinator.com/item?id=48420373","kind":"comment","storyTitle":"Google to pay SpaceX $920M a month for compute capacity at xAI data centers","quote":"All the other AIs have a learning date cut-off too far in the past. Claude accuses me of hallucinating events that happened the day before, and it's quite annoying."},{"itemId":"48418840","author":"thewebguyd","createdAt":1780696326,"url":"https://news.ycombinator.com/item?id=48418840","kind":"comment","storyTitle":"Ask HN: What was your \"oh shit\" moment with GenAI?","quote":"Since anything in PS is probably some internal sysadmin tool, there's not much public code out there outside of Microsoft's documentation. Plus the Verb-Noun naming scheme makes it really easy to just hallucinate cmdlets (which it does, often)."},{"itemId":"48418596","author":"jachee","createdAt":1780694995,"url":"https://news.ycombinator.com/item?id=48418596","kind":"comment","storyTitle":"Ask HN: What was your \"oh shit\" moment with GenAI?","quote":"I haven’t had that yet. I tried again this week, and CoPilot Plan Mode read the same 5-line markdown file 18 times over the course of 5 minutes of churning on a simple request, then provided zero value over what I posed in the request itself, and hallucinated things about my terraform repo that were just flat-out wrong."},{"itemId":"48416056","author":"moezd","createdAt":1780682591,"url":"https://news.ycombinator.com/item?id=48416056","kind":"comment","storyTitle":"Ask HN: What is your (AI) dev tech stack / workflow?","quote":"I think all coding agents support this by now. 3) Add hooks if you want to stop the coding agent from doing something nasty, or hallucinate and give incomplete output."},{"itemId":"48398261","author":"Balgair","createdAt":1780579194,"url":"https://news.ycombinator.com/item?id=48398261","kind":"comment","storyTitle":"Failing grades soar with AI usage, dwindling math skills in Berkeley CS classes","quote":"For the second anecdote, AI is condemning people to death. Using it in life critical situations and care, letting it hallucinate or skip over critical values, that's a recipe for disaster."},{"itemId":"48395686","author":"rerdavies","createdAt":1780560795,"url":"https://news.ycombinator.com/item?id=48395686","kind":"comment","storyTitle":"Artificial intelligence is not conscious – Ted Chiang","quote":"And it's pretty clear that theories of mind will have to be re-done all over again with the advent of LLMs (particularly current-generation LLMs). The problem with the hallucination argument is (1) that is much less of a problem with good current generation AIs, and (2) living conscious breathing human beings also have a disturbing tendency to make shit up, too."},{"itemId":"48394533","author":"chorsestudios","createdAt":1780552139,"url":"https://news.ycombinator.com/item?id=48394533","kind":"comment","storyTitle":"Failing grades soar with AI usage, dwindling math skills in Berkeley CS classes","quote":"I haven't been in college for over a decade, this is just for topics I'm curious about and want to learn. For any serious topic I recommend auditing the practice exams with a different LLM than the one used to generate to help reduce hallucinations."},{"itemId":"48393534","author":"wtp1saac","createdAt":1780544967,"url":"https://news.ycombinator.com/item?id=48393534","kind":"comment","storyTitle":"Failing grades soar with AI usage, dwindling math skills in Berkeley CS classes","quote":"Where the AI hallucinated, it was often on things like method names, which I was able to read through and find the methods it was intending to name."},{"itemId":"48393132","author":"ToucanLoucan","createdAt":1780541865,"url":"https://news.ycombinator.com/item?id=48393132","kind":"comment","storyTitle":"Failing grades soar with AI usage, dwindling math skills in Berkeley CS classes","quote":"It's not even that they're opposed, as much as they're ships for whom Polaris is in a completely different direction. \"Hallucinations\" as they're called, or more plainly stated when the machine makes some shit up, are perfectly understandable in this context, as are the struggles of every single AI firm to get rid of them."},{"itemId":"48389924","author":"stult","createdAt":1780520213,"url":"https://news.ycombinator.com/item?id=48389924","kind":"comment","storyTitle":"AI outperforms law professors in Stanford Law study","quote":"Then I review that again, adjusting as necessary, and then the agent posts the result as a comment on the original issue. Like you, I've found this detailed planning makes for a very robust coding agent (again, also in combination with the aforementioned best practices, especially requiring 100% test coverage because forcing it to exercise every line of code avoids hallucinated dummy tests that assert on nothing)."},{"itemId":"48385078","author":"giancarlostoro","createdAt":1780499014,"url":"https://news.ycombinator.com/item?id=48385078","kind":"comment","storyTitle":"AI outperforms law professors in Stanford Law study","quote":"Hmmm... This all assuming the AI lawyer doesn't hallucinate and start referencing cases that don't exist."},{"itemId":"48380917","author":"eunos","createdAt":1780470935,"url":"https://news.ycombinator.com/item?id=48380917","kind":"comment","storyTitle":"AI outperforms law professors in Stanford Law study","quote":"Seems companies like Thomson Reuters or other legal services have incentive to build LLM with RAG over legal cases texts and robust hallucinations detection on reference"},{"itemId":"48378628","author":"RickS","createdAt":1780450040,"url":"https://news.ycombinator.com/item?id=48378628","kind":"comment","storyTitle":"More than 6 out of 10 people turn to AI for psychological support","quote":"EG people who aren't used to negotiating, valuing themselves, etc. Obviously LLMs output hallucinated dogshit and occasionally dangerous nonsense."},{"itemId":"48374817","author":"thewebguyd","createdAt":1780427598,"url":"https://news.ycombinator.com/item?id=48374817","kind":"comment","storyTitle":"Expanding Project Glasswing","quote":"It's along the same lines of the BS I deal with day to day from upper management arguing back with \"But ChatGPT said...\" meanwhile pasting some hallucinated crap that doesn't even apply to our environment."},{"itemId":"48365433","author":"ChicagoDave","createdAt":1780368934,"url":"https://news.ycombinator.com/item?id=48365433","kind":"comment","storyTitle":"What's gonna happen to software engineers?","quote":"Vibe coding is the chaos version of using GenAI. Seven agents in parallel and the dev refuses to believe this will just be a circle jerk of hallucinations."},{"itemId":"48358775","author":"charlesleft","createdAt":1780330218,"url":"https://news.ycombinator.com/item?id=48358775","kind":"comment","storyTitle":"Ask HN: Who is hiring? (June 2026)","quote":"- The evaluation evidence pipeline (data model, ingestion, exports) - LLM normalization that summarizes traces without hallucinating or hiding evidence quality"},{"itemId":"48358336","author":"eddiehammond","createdAt":1780328367,"url":"https://news.ycombinator.com/item?id=48358336","kind":"comment","storyTitle":"Ask HN: Who is hiring? (June 2026)","quote":"We're building the trust layer for AI in high-stakes industries. https://kepler.ai LLMs hallucinate, so they can't be the thing producing your numbers."},{"itemId":"48352514","author":"cyphar","createdAt":1780287108,"url":"https://news.ycombinator.com/item?id=48352514","kind":"comment","storyTitle":"[dead]","quote":"That is also possible, but this mostly smells like a weird LLM-hallucinated assumption to me because the whole concept of blocking literal \"..\" components doesn't make sense if you allow symlinks -- symlinks can contain \"..\" and RESOLVE_BENEATH allows them if they are safe."},{"itemId":"48322744","author":"kiBytes","createdAt":1780060687,"url":"https://news.ycombinator.com/item?id=48322744","kind":"story","storyTitle":"Show HN: TheFoundry – Easy bootstrapping framework for MultiAgent Systems","quote":"- Anti-Loop Budgets: Agents have a hard stop after 5 failed iterations to prevent token exhaustion. - TOML for A2A: Agent-to-agent requests use TOML to eliminate hallucinated JSON syntax errors."},{"itemId":"48317171","author":"trehalose","createdAt":1780012415,"url":"https://news.ycombinator.com/item?id=48317171","kind":"comment","storyTitle":"Show HN: Continue? Y/N: A 60-second game about AI agent permission fatigue","quote":"I approved the rm -rf Projects command because I thought the LLM had correctly described that it would delete everything in the Projects folder. Clearly I misread that in my hurry to answer prompts (I knew what the command would do and I guess I hallucinated that the AI had explained it), but I'd like to see what it was that I misread."},{"itemId":"48316636","author":"pqdbr","createdAt":1780008570,"url":"https://news.ycombinator.com/item?id=48316636","kind":"story","storyTitle":"Ask HN: Is Claude Opus 4.8 broken?","quote":"It can't even read a file anymore. Randomly uses 'sed' with no explanation in hallucinated file paths."},{"itemId":"48308603","author":"skrebbel","createdAt":1779974492,"url":"https://news.ycombinator.com/item?id=48308603","kind":"comment","storyTitle":"Disagreement among frontier LLMs on real-world fact-checks","quote":"oopsie. I flagged the article, it seems more misleading than the average Claude hallucination."},{"itemId":"48297458","author":"simonw","createdAt":1779902630,"url":"https://news.ycombinator.com/item?id=48297458","kind":"comment","storyTitle":"I think Anthropic and OpenAI have found product-market fit","quote":"I think it is applicable to a much wider range of knowledge work, but it's also harder to apply there. Software development has the huge advantage that mistakes and hallucinations are very easy to spot: the software works or it doesn't."},{"itemId":"48297267","author":"zuzululu","createdAt":1779901831,"url":"https://news.ycombinator.com/item?id=48297267","kind":"comment","storyTitle":"The Structural Barriers to AI Lawyers","quote":"my point isn't that people pursue this pipeline its that frontier models provide a great equalizer. you can bet that patent trolls and saul goodman type of lawyers are using the same tools. what is scary isnt that these models will hallucinate or whatever (those are harness issues) its the amount of formality and complexity navigating processes with rigid schedules and verbosity that can cause context drift."},{"itemId":"48279176","author":"GuB-42","createdAt":1779800237,"url":"https://news.ycombinator.com/item?id=48279176","kind":"comment","storyTitle":"The User Is Visibly Frustrated","quote":"From my experience, if a LLM starts repeating itself, it means it has given you everything it could, if you get anything new, it will most likely be a hallucination."},{"itemId":"48278945","author":"kortex","createdAt":1779799023,"url":"https://news.ycombinator.com/item?id=48278945","kind":"comment","storyTitle":"The AI bubble isn't like the internet bubble","quote":"Write 50% more corporate emails employees barely read anyways. There's definitely a trust issue there with hallucinations."},{"itemId":"48275853","author":"shaurya-sethi","createdAt":1779777220,"url":"https://news.ycombinator.com/item?id=48275853","kind":"story","storyTitle":"Why codex /goal fails on complex workflows: compaction amnesia and context rot","quote":"The model is asked to summarize a massive blob of mixed-relevance information when its reasoning quality is already at its lowest. This compaction leads to forgetting critical constraints, makes way for hallucinations of past decisions, and introduces noise that makes the new context unreliable for long-horizon work."},{"itemId":"48271801","author":"speedgoose","createdAt":1779743993,"url":"https://news.ycombinator.com/item?id=48271801","kind":"comment","storyTitle":"Norway's 2 petabytes of Huawei flash storage and LLM training","quote":"Often. One other finetune did hallucinate that it was a doctor and my baby had terrible diseases, every time I just wrote \"hei\" (with a generic neutral system prompt that likely triggered this behaviour though)."},{"itemId":"48264416","author":"robot_jesus","createdAt":1779694924,"url":"https://news.ycombinator.com/item?id=48264416","kind":"comment","storyTitle":"If you let AI do your writing, I will come to your house and kill you","quote":"Well, I did ask Gemini (3.1 Pro) that question verbatim and it wasn’t fooled in the least. People who rail against LLMs for hallucinations like that always read to me like people who haven’t used it since writing off chat got 2.5."},{"itemId":"48248801","author":"gabriel_oauth","createdAt":1779552385,"url":"https://news.ycombinator.com/item?id=48248801","kind":"story","storyTitle":"Show HN: I built a RAG and knowledge graph agent that runs locally","quote":"1. search_tool This enables the ai agent to actually search up to date info so that it doesnt hallucinate on info it doesn't know which is common in local llms."},{"itemId":"48246659","author":"GabrielBlessed","createdAt":1779534343,"url":"https://news.ycombinator.com/item?id=48246659","kind":"story","storyTitle":"Show HN: I built a powerful RAG and knowledge graph agent that runs locally","quote":"1. search_tool This enables the ai agent to actually search up to date info so that it doesnt hallucinate on info it doesn't know which is common in local llms."},{"itemId":"48241839","author":"keeda","createdAt":1779484953,"url":"https://news.ycombinator.com/item?id=48241839","kind":"comment","storyTitle":"AI has a multiplying effect on existing technical skills","quote":"(Multiple studies find that the outcome depends on how AI is used. Surprisingly, incorrect guidance / unreliability / hallucinations appear to be a bigger problem than over-reliance!"},{"itemId":"48239270","author":"zephen","createdAt":1779473175,"url":"https://news.ycombinator.com/item?id=48239270","kind":"comment","storyTitle":"AI has a multiplying effect on existing technical skills","quote":"Its response to your second prompt further cements that its goal is to predict what you want or expect to see. It's also well known that LLMs are prone to hallucinations."},{"itemId":"48234265","author":"simonh","createdAt":1779448144,"url":"https://news.ycombinator.com/item?id=48234265","kind":"comment","storyTitle":"Steve Wozniak cheered after telling students they have AI – actual intelligence","quote":"Siri is basically unchanged, it looks like they have had serious problems getting LLMs, or generative AI in general to be reliable and 'safe' enough to put their own name on it. By 'safe' I mean thinks like not generating emails based on Mein Kampf, or doodles of genitals, or hallucinating false 'facts'."},{"itemId":"48214920","author":"jorvi","createdAt":1779315029,"url":"https://news.ycombinator.com/item?id=48214920","kind":"comment","storyTitle":"Qwen3.7-Max: The Agent Frontier","quote":"Its very annoying this has been in the capability of models since the very beginning. It could check how probable its token values are and if those fall below a certain threshold either say \"I don't know\", or output the most probable (well, more like least improbable) tokens but give a very clear, very strong warning that it is a shot in the dark and likely to contain hallucinations."},{"itemId":"48213563","author":"ffsm8","createdAt":1779308509,"url":"https://news.ycombinator.com/item?id=48213563","kind":"comment","storyTitle":"Qwen3.7-Max: The Agent Frontier","quote":"Isn't that precisely the reason why we introduced the term hallucination?"},{"itemId":"48212654","author":"wwizo","createdAt":1779304741,"url":"https://news.ycombinator.com/item?id=48212654","kind":"comment","storyTitle":"Gemini 3.5 Flash","quote":"On following instruction. Asked Gemini Flash 3.5 to summarize YouTube video (google io developer keynote), a task that would previously be trivial (use ot often), but it kept hallucinating points and referencing io dev keynote blog posts from several years ago."},{"itemId":"48212003","author":"radu_me","createdAt":1779301853,"url":"https://news.ycombinator.com/item?id=48212003","kind":"comment","storyTitle":"[dead]","quote":"This is a well-known example of a ChatGPT hallucination."},{"itemId":"48206890","author":"NothingAboutAny","createdAt":1779281649,"url":"https://news.ycombinator.com/item?id=48206890","kind":"comment","storyTitle":"Gemini 3.5 Flash","quote":"but for research it makes shit up all the time, I asked GPT5.5 to make me a build for Rogue Trader and not only did it use out of date info, it made up a bunch of skills that were NEVER in the game. I attribute that to there not being enough online information in the wikis or whatever but I wish it would just say \"I dont know\" instead of hallucinating but I know that's not how the tech works."},{"itemId":"48204337","author":"jy14898","createdAt":1779262198,"url":"https://news.ycombinator.com/item?id=48204337","kind":"comment","storyTitle":"Show HN: Id-agent – Token efficient UUID alternative for AI agents","quote":"But uuids are random hex characters, this is where LLMs struggle to output the right ids. If true then that indeed seems like an improvement, I think I just need measurements of actual hallucinations."},{"itemId":"48200603","author":"JoonSPP","createdAt":1779230038,"url":"https://news.ycombinator.com/item?id=48200603","kind":"comment","storyTitle":"Agentic PCB Design Sucks","quote":"The tools are there (e.g., typeCAD, KiCad link to Claude Code), but the component data available has been quite limited. Agents are using component data from training data, leading to hallucinations."},{"itemId":"48198230","author":"hibikir","createdAt":1779218982,"url":"https://news.ycombinator.com/item?id=48198230","kind":"comment","storyTitle":"Gemini 3.5 Flash","quote":"I see constant hallucination in claude code when using specific tooling: It thinks it knows aws cli, for instance, but there's some flags that don't exist, it attempts to use all the time in 4.6 and 4.7."},{"itemId":"48197781","author":"mrweasel","createdAt":1779217261,"url":"https://news.ycombinator.com/item?id=48197781","kind":"comment","storyTitle":"Show HN: Id-agent – Token efficient UUID alternative for AI agents","quote":"This might be my understanding that's wrong, but I assumed that the LLM itself actually can't produce like a UUID, but it can \"predict\" one, hence why it sometimes hallucinate IDs."},{"itemId":"48197340","author":"yieldcrv","createdAt":1779215585,"url":"https://news.ycombinator.com/item?id=48197340","kind":"comment","storyTitle":"Gemini 3.5 Flash","quote":"if last year's models were the ones people got familiar with in late 2022, hallucinations would be an underrepresented rumor, there would be no articles about it because its so rare. overconfident lawyers wouldn't have messed up dockets in court with fake case law, in other domains that move faster, sources would be only partially outdated with agentic search and mcp servers filling in the gaps"},{"itemId":"48193493","author":"Aurornis","createdAt":1779199560,"url":"https://news.ycombinator.com/item?id=48193493","kind":"comment","storyTitle":"Cursor Introduces Composer 2.5","quote":"EDIT: As others have pointed out, the comment above contains hallucinations (Like the $50 billion number) and a lot of AI tells."},{"itemId":"48193131","author":"enraged_camel","createdAt":1779197576,"url":"https://news.ycombinator.com/item?id=48193131","kind":"comment","storyTitle":"Cursor Introduces Composer 2.5","quote":"For each one, I had it write a plan, then I had Codex review it. Codex found major issues with every plan: patterns that don't match the rest of the code base, hallucinated variable/function names, and even outright bugs in the way the plan was written."},{"itemId":"48190612","author":"generationP","createdAt":1779178492,"url":"https://news.ycombinator.com/item?id=48190612","kind":"comment","storyTitle":"The last six months in LLMs in five minutes","quote":"- since GPT-5.4 (Mar 2026): the \"wow\" release; suddenly answering MathOverflow-level problems that have previously been stumping experts. Still prone to hallucinations, but smart enough to use the built-in Python skill to verify its claims on small examples when possible."},{"itemId":"48186336","author":"kwojno","createdAt":1779141606,"url":"https://news.ycombinator.com/item?id=48186336","kind":"comment","storyTitle":"LLMs have no structural place for non-knowledge","quote":"Less metaphors and more speculation grounded in recent trends in tech industry. AI does not hallucinate because it is broken."},{"itemId":"48153227","author":"dathinab","createdAt":1778875858,"url":"https://news.ycombinator.com/item?id=48153227","kind":"comment","storyTitle":"Bun Rust rewrite: \"codebase fails basic miri checks, allows for UB in safe rust\"","quote":"the thing is if you don't keep things mostly 1:1 with all the problems that has there is absolutely no way to review that PR or catch the AI going rogue with hallucinations etc."},{"itemId":"48149928","author":"spwa4","createdAt":1778858943,"url":"https://news.ycombinator.com/item?id=48149928","kind":"comment","storyTitle":"UK government replaces Palantir software with internally-built refugee system","quote":"2) a data ingestion process that makes up whatever data is missing (apparently this is called \"ontology\": the AI model fills missing data in with what it thinks is happening. No mention of hallucination and it's effects)."},{"itemId":"48149758","author":"zzet","createdAt":1778858092,"url":"https://news.ycombinator.com/item?id=48149758","kind":"comment","storyTitle":"How Claude Code works in large codebases","quote":"The way now CC _bruteforce_ symbols is not the same way the engineer will do. One typo and the agent may decide they have to reimplement something, and when by lack the read a file, they can easily fall into hallucinations."},{"itemId":"48139737","author":"pplonski86","createdAt":1778785472,"url":"https://news.ycombinator.com/item?id=48139737","kind":"comment","storyTitle":"AI is making me dumb","quote":"I read paragraph and if it is ok, I accept it, if it doesn't show off my thoughts I work on one paragraph. If doing data analysis with AI, I do one step of analysis and ask AI to display intermediate results so I can see if all is going in good direction and there are no hallucinations, additionally I have follow-up prompts for AI to do results verification."},{"itemId":"48133812","author":"throwatdem12311","createdAt":1778757230,"url":"https://news.ycombinator.com/item?id=48133812","kind":"comment","storyTitle":"Claude for Small Business","quote":"There is going to be so many horror stories that will come from this, ie. Claude overpaid/underpaid my employees, Claude hallucinated the tax code and now the IRS is seizing my assets. etc…"},{"itemId":"48127452","author":"mariopt","createdAt":1778705813,"url":"https://news.ycombinator.com/item?id=48127452","kind":"comment","storyTitle":"The US is winning the AI race where it matters most: commercialization","quote":"Kimi is quite good at design, for 3D, I noticed Gemini 3.1 is excellent for basic to medium use cases. I've tried both Opus and GPT 5.4, they also hallucinate just like the rest at a much higher cost."},{"itemId":"48124262","author":"gumby271","createdAt":1778690479,"url":"https://news.ycombinator.com/item?id=48124262","kind":"comment","storyTitle":"Googlebook","quote":"And Google gets to do this now, no one can compete or leave them, so why do anything well. When Google first added Gemini to Sheets, I tried it and it completely hallucinated transaction data on the bank statement I was testing it on."},{"itemId":"48121930","author":"azurewraith","createdAt":1778680433,"url":"https://news.ycombinator.com/item?id=48121930","kind":"comment","storyTitle":"Show HN: Statewright – Visual state machines that make AI agents reliable","quote":"The plan/implement/test workflow is very basic and represents the most common agentic use case. But the state machine pattern applies to any multi-step work where agents are useful but susceptible to death spirals, hallucinations, or other non-deterministic quirkiness."},{"itemId":"48121261","author":"id02009","createdAt":1778676817,"url":"https://news.ycombinator.com/item?id=48121261","kind":"comment","storyTitle":"Better Claude-code spinner verbs","quote":"I found the default ones annoying, and just `Please wait...` was too boring, so I asked claude-code to comb the web for common criticism and now you can see things like `Hallucinating ...` when you wait for your prompt results."},{"itemId":"48118017","author":"aouicher","createdAt":1778648420,"url":"https://news.ycombinator.com/item?id=48118017","kind":"story","storyTitle":"Show HN: Graphmind – Persistent Memory and Graph for Claude Code (MCP, CLI, GUI)","quote":"Ask Claude Code to quickly and easily understand a 10-year-old monolith, and you’ll see what I mean... The infamous LLM hallucinations will pop up fast, and the endless repetition of grep commands or similar to read through all the code will become a never-ending task..."},{"itemId":"48116822","author":"kgeist","createdAt":1778636258,"url":"https://news.ycombinator.com/item?id=48116822","kind":"comment","storyTitle":"Beyond Semantic Similarity","quote":"What mattered was even the order of fields in the output JSON schema (in intermediate steps), because LLMs are autoregressive: if you order the fields incorrectly, the model may guess or hallucinate during extraction when the first value in the schema can't be extracted reliably without first considering other dependent fields that should've been extracted earlier (we don't use reasoning models to save on speed)."},{"itemId":"48116117","author":"freetime2","createdAt":1778630030,"url":"https://news.ycombinator.com/item?id=48116117","kind":"comment","storyTitle":"Googlebook","quote":"Or often times the listing itself might contain incorrect information. And I found ChatGPT to be a useful tool for quickly making sense of the various configurations, and of course conduct my own sanity checks to be sure the information is not hallucinated."},{"itemId":"48116110","author":"dvt","createdAt":1778629971,"url":"https://news.ycombinator.com/item?id=48116110","kind":"comment","storyTitle":"Googlebook","quote":"Listen, I do think AI still has a LONG way to go to be the automation & productivity utopia we so desperately crave, but underselling its usefulness is just silly at this point. I used to be vehemently against AI coding just a few years ago, because the hallucinations were a deal-breaker."},{"itemId":"48103261","author":"didigamma","createdAt":1778550828,"url":"https://news.ycombinator.com/item?id=48103261","kind":"comment","storyTitle":"Can you help reconcile my first/second-hand LLM Experience with HN's Experience?","quote":"2. It hallucinates documentation often enough I've gone back to hunting for source documentation myself (e.g."},{"itemId":"48103170","author":"bonjourjoel","createdAt":1778550073,"url":"https://news.ycombinator.com/item?id=48103170","kind":"comment","storyTitle":"Show HN: I built a 500K LOC production app alone in 7 months. Here is the proof","quote":"What makes it possible: I built a VS Code extension with a proprietary indexing engine that gives the AI full structural awareness of the codebase. Without it the AI hallucinates the architecture at this scale."},{"itemId":"48102559","author":"nullc","createdAt":1778545160,"url":"https://news.ycombinator.com/item?id=48102559","kind":"comment","storyTitle":"A.I. note takers are making lawyers nervous","quote":"Training that included extremist anti-corporate dogma might even bias the language models towards hallucinating the worst possible misinterpretation."},{"itemId":"48102522","author":"colon-md","createdAt":1778544962,"url":"https://news.ycombinator.com/item?id=48102522","kind":"comment","storyTitle":"RAG Eval Comparing Vertex/Bedrock/Azure/OpenAI","quote":"- OpenAI File Search: 78.5 (89.3%, 53.4%) Here's a surprise finding (maybe not a surprise to you): all four major RAG services hallucinate on every unanswerable question."},{"itemId":"48102293","author":"aadyachinubhai","createdAt":1778543046,"url":"https://news.ycombinator.com/item?id=48102293","kind":"comment","storyTitle":"Show HN: n8n like workflows for AI agents that control a real VM","quote":"An AI agent executes them in order inside a Docker container with a full browser and desktop. Because each node is independent, the agent stays on task, it doesn't drift or hallucinate its way through a free-form prompt."},{"itemId":"48093082","author":"dxdm","createdAt":1778494273,"url":"https://news.ycombinator.com/item?id=48093082","kind":"comment","storyTitle":"I'm going back to writing code by hand","quote":"For some reason, people seem to expect miracles from some machine that they would not expect of other humans, especially not ones with a proven penchant for rambling hallucinations every once in a while."},{"itemId":"48092406","author":"tomelders","createdAt":1778487693,"url":"https://news.ycombinator.com/item?id=48092406","kind":"comment","storyTitle":"Local AI needs to be the norm","quote":"So where do these LLM providers draw the line on what makes it into the training data and what doesn't? And if an LLM doesn't know what a Bezier curve is, that's not going to stop it from just hallucinating an answer."},{"itemId":"48084526","author":"robotbikes","createdAt":1778425032,"url":"https://news.ycombinator.com/item?id=48084526","kind":"comment","storyTitle":"Task Paralysis and AI","quote":"I've yet to seriously use a LLM for a personal project and when I tried to use Devstral that ran on my Nvidia 4090 it hallucinated so much that it wasn't super helpful but it still shot out boiler plate code that I could then spend time fixing and helped me overcome my own task paralysis regarding initiating."},{"itemId":"48065304","author":"sev_verso","createdAt":1778257506,"url":"https://news.ycombinator.com/item?id=48065304","kind":"comment","storyTitle":"DeepSeek 4 Flash local inference engine for Metal","quote":"The bad: It could not address a small nitpick I had. The ugly: It hallucinated a conversation about \"The Duck\" that I had with it simultaneously while trying to solve another problem."},{"itemId":"48060463","author":"dkersten","createdAt":1778230357,"url":"https://news.ycombinator.com/item?id=48060463","kind":"comment","storyTitle":"Agents need control flow, not more prompts","quote":"This is something I realised late last year while using Claude Code. The LLM shouldn't be the one in control of the workflow, because the LLM can make mistakes, skip steps, hallucinate steps, etc."},{"itemId":"48059311","author":"sfink","createdAt":1778221188,"url":"https://news.ycombinator.com/item?id=48059311","kind":"comment","storyTitle":"Hardening Firefox with Claude Mythos Preview","quote":"AI, as commonly used, is given more leeway. It's kind of fundamental that it must be allowed to hallucinate false positives; that's the source of much of its power."},{"itemId":"48050792","author":"pkcoskfiwjfj","createdAt":1778168665,"url":"https://news.ycombinator.com/item?id=48050792","kind":"comment","storyTitle":"Photoshop's challenges with focus, pt. 2","quote":"And unfortunately Photoshop very much is the monopoly many think it is. Those complaining about Photoshop being hobbled because it can’t hallucinate AI slop are not Adobe’s target audience and main source of income."},{"itemId":"48048485","author":"devonkim","createdAt":1778156427,"url":"https://news.ycombinator.com/item?id=48048485","kind":"comment","storyTitle":"The bottleneck was never the code","quote":"The difference is in the kind of errors machines make compared to ones that humans make. It’s frustrating we’re using anthropomorphic concepts like hallucinations when describing LLM behaviors when the fundamental units of computation and thus failures of computation are so different at every level."},{"itemId":"48034765","author":"engeljohnb","createdAt":1778064871,"url":"https://news.ycombinator.com/item?id=48034765","kind":"comment","storyTitle":"AI didn't delete your database, you did","quote":"As they are completely autonomous, they can apply arbitrary changes. Yes, and someone somewhere decided to use a coding assistant that can apply arbitrary changes, knowing full well that LLMs are known to hallucinate and make mistakes, and not rarely."},{"itemId":"48030069","author":"doginasuit","createdAt":1778023236,"url":"https://news.ycombinator.com/item?id=48030069","kind":"comment","storyTitle":"Three Inverse Laws of AI","quote":"I believe the accuracy issues of LLMs are well understood by now, maybe even to the point of overstatement. Hallucinations have become a non-issue in my work, I've begun to understand the circumstances where they are most likely."},{"itemId":"48023421","author":"qazxcvbnmlp","createdAt":1777993081,"url":"https://news.ycombinator.com/item?id=48023421","kind":"comment","storyTitle":"Lessons for Agentic Coding: What should we do when code is cheap?","quote":"The AI code required a lot of review, about the same amount of time as writing the code manually. Now.. the AI first engineer might still have to deal with hallucinated things."},{"itemId":"48022320","author":"Mahjabinbm","createdAt":1777987894,"url":"https://news.ycombinator.com/item?id=48022320","kind":"story","storyTitle":"Anyone else losing tokens to hallucinated MCP tool calls in production?","quote":"Anyone else losing tokens to hallucinated MCP tool calls in production?"},{"itemId":"48001481","author":"jiggawatts","createdAt":1777842322,"url":"https://news.ycombinator.com/item?id=48001481","kind":"comment","storyTitle":"Show HN: State of the Art of Coding Models, According to Hacker News Commenters","quote":"The Open Telemetry end is a mess of “deprecated” and “beta”, copying the shifting sands of other cloud native projects like Kubernetes. Even with carefully curated Markdown documentation references and sample logs, every modern “frontier” AI makes basic mistakes and hallucinates like crazy no matter how I stuff their context."},{"itemId":"47999884","author":"eddiehammond","createdAt":1777832914,"url":"https://news.ycombinator.com/item?id=47999884","kind":"comment","storyTitle":"Ask HN: Who is hiring? (May 2026)","quote":"We're building the trust layer for AI in high-stakes industries. https://kepler.ai LLMs hallucinate, so they can't be the thing producing your numbers."},{"itemId":"47999124","author":"joss82","createdAt":1777828572,"url":"https://news.ycombinator.com/item?id=47999124","kind":"comment","storyTitle":"He asked AI to count carbs 27000 times. It couldn't give the same answer twice","quote":"BTW Google Gemini Pro just told me that they know the game but did not know the value. Actually, it points out that it is a known trap for AI to confidently give a wrong, hallucinated value."},{"itemId":"47994145","author":"Aerolfos","createdAt":1777791378,"url":"https://news.ycombinator.com/item?id=47994145","kind":"comment","storyTitle":"VS Code inserting 'Co-Authored-by Copilot' into commits regardless of usage","quote":"Looks like AI is really trigger-happy with that, with zero understanding or care that there's thousands of users affected and it's not just one individual's settings.json Also, the hallucinated PR does not necessarily address the original issue whatsoever, just like this PR."},{"itemId":"47990108","author":"VoidCoefficient","createdAt":1777753429,"url":"https://news.ycombinator.com/item?id=47990108","kind":"comment","storyTitle":"Mike: open-source legal AI","quote":"To be clear, the value of services like Legora and Harvey was not as a replacement for legal-specific services like Westlaw. LLMs are absolutely abysmal at legal research and analysis, and every week we are seeing reports of some lawyer being called out publicly for submitting a brief to a court with hallucinated case citations."},{"itemId":"47977778","author":"ShyCodeGardener","createdAt":1777657832,"url":"https://news.ycombinator.com/item?id=47977778","kind":"comment","storyTitle":"Apple accidentally left Claude.md files Apple Support app","quote":"> All an LLM does is introduce the possibility that a command that worked fine yesterday will randomly not work Aren't hallucinations part of GenAI?"},{"itemId":"47974895","author":"pxc","createdAt":1777643895,"url":"https://news.ycombinator.com/item?id=47974895","kind":"comment","storyTitle":"Apple accidentally left Claude.md files Apple Support app","quote":"Gemini on Android sucks so bad it's hard to describe. It can't tell what its own capabilities are, it can't inspect the states of the apps it manipulates, it hallucinates constantly, and it needs more handholding than the crappy old decision tree to do the right thing."},{"itemId":"47968033","author":"roarlyte","createdAt":1777582152,"url":"https://news.ycombinator.com/item?id=47968033","kind":"comment","storyTitle":"[dead]","quote":" It uses familiar React patterns, so simpler prompts yield more accurate results.  Accessibility is built directly into the grid, so no hallucinating custom screen-reader properties."},{"itemId":"47967024","author":"fao_","createdAt":1777576721,"url":"https://news.ycombinator.com/item?id=47967024","kind":"comment","storyTitle":"A 1960s art school experiment that redefined creativity","quote":"The scientists I know do take time out of their day to answer actual questions from inquisitive folk, but the difficulty is that thanks to the addition of ChatGPT, those questions are getting more and more cramped out by the crackpots armed with a hallucinating dictionary."},{"itemId":"47965302","author":"SwellJoe","createdAt":1777568366,"url":"https://news.ycombinator.com/item?id=47965302","kind":"comment","storyTitle":"Granite 4.1: IBM's 8B Model Matching 32B MoE","quote":"LM Studio is nice in that it makes it easy to add tools, like search. Qwen 3.6 is such a small model that it lacks a lot of knowledge of the world (so it can hallucinate at an uncomfortable rate, which is a common failure mode of very small models), but it can use tools, so being able to search lets it research before answering."},{"itemId":"47962022","author":"getpokedagain","createdAt":1777555270,"url":"https://news.ycombinator.com/item?id=47962022","kind":"comment","storyTitle":"Mozilla's opposition to Chrome's Prompt API","quote":"I have a family member who is undergoing cancer treatment. I can't explain to you the frustration of seeing wrong text summaries when an llm goes wild hallucinating test results when the actual text simply said taking a test."},{"itemId":"47950752","author":"yakkomajuri","createdAt":1777480418,"url":"https://news.ycombinator.com/item?id=47950752","kind":"story","storyTitle":"Show HN: AgentPort – Open-source Security Gateway For Agents","quote":"I've been wanting to use something like OpenClaw for a while but couldn't get myself to give it access to anything important due to all the risks involved. Prompt injection is still a problem (even though some people seem to ignore it) and so are hallucinations and mishaps that cause agents to do things like delete production data [1]."},{"itemId":"47948907","author":"zoetaka38","createdAt":1777472648,"url":"https://news.ycombinator.com/item?id=47948907","kind":"comment","storyTitle":"Letting AI play my game – building an agentic test harness to help play-testing","quote":"The single biggest jump in test quality came from giving the agent BOTH source code analysis AND live browser snapshots, not either alone. With code-only the agent hallucinates selectors; with browser-only it misses project conventions."},{"itemId":"47935098","author":"yakkomajuri","createdAt":1777386288,"url":"https://news.ycombinator.com/item?id=47935098","kind":"story","storyTitle":"Show HN: Integrations gateway for agents with 2FA for destructive ops (OSS)","quote":"I've been wanting to use something like OpenClaw for a while but couldn't get myself to give it access to anything important due to all the risks involved. Prompt injection is still a problem (even though some people seem to ignore it) and so are hallucinations and mishaps that cause agents to do things like delete production data [1]."},{"itemId":"47931542","author":"dignaga69","createdAt":1777362384,"url":"https://news.ycombinator.com/item?id=47931542","kind":"comment","storyTitle":"Show HN: Financial Database API for Vibe Coders","quote":"US equities, 30+ years of daily prices and SEC fundamentals. The problem: every time you ask an LLM to write something with financial data, it hallucinates field names and fights the API (cough cough yfinance)."},{"itemId":"47916910","author":"evanklem2004","createdAt":1777254999,"url":"https://news.ycombinator.com/item?id=47916910","kind":"comment","storyTitle":"EvanFlow – A TDD driven feedback loop for Claude Code","quote":"Iterate-until-clean (hard cap of 5 rounds). Re-read the diff against dead code, naming, the deletion test, assertion correctness, and a Five Failure Modes pass (hallucinated actions, scope creep, cascading errors, context loss, tool misuse)."},{"itemId":"47915183","author":"kofdai","createdAt":1777241135,"url":"https://news.ycombinator.com/item?id=47915183","kind":"comment","storyTitle":"Verantyx – A native IDE that obfuscates code before sending it to Cloud LLMs","quote":"- AST Topology Destruction: This is the most critical issue. When given complex instructions, frontier models sometimes hallucinate and destroy the AST topology of the obfuscated IR."},{"itemId":"47909460","author":"kofdai","createdAt":1777203168,"url":"https://news.ycombinator.com/item?id=47909460","kind":"comment","storyTitle":"Verantyx – A native IDE that obfuscates code before sending it to Cloud LLMs","quote":"• AST Topology Destruction: This is the most critical issue. When given complex instructions, frontier models sometimes hallucinate and destroy the AST topology of the obfuscated IR."},{"itemId":"47904238","author":"subhobroto","createdAt":1777148322,"url":"https://news.ycombinator.com/item?id=47904238","kind":"comment","storyTitle":"GPT-5.5","quote":"- Inefficiencies can hide better at larger companies - Larger companies are risk averse although surprisingly they are in the best position to absorb risks - aka Innovator's Dilemma (For example: Google invented transformers but management decided against it because they were afraid of hallucinations affecting search result quality."},{"itemId":"47902082","author":"amluto","createdAt":1777129834,"url":"https://news.ycombinator.com/item?id=47902082","kind":"comment","storyTitle":"Lambda Calculus Benchmark for AI","quote":"I suspect its explanation is sort of correct. But I get no web search results for GN numbers and I can’t really tell whether ChatGPT is giving a semi-hallucinated description or figuring it out from the problem and its solution or what."},{"itemId":"47898624","author":"jasomill","createdAt":1777090993,"url":"https://news.ycombinator.com/item?id=47898624","kind":"comment","storyTitle":"My audio interface has SSH enabled by default","quote":"From the article, it sounds like he used Claude Code as an alternative to Wireshark and Google to decode USB HID traffic and find protocol documentation, respectively. I suppose this could save a bit of time if you don't already have Wireshark installed, with a minor risk of hallucinations."},{"itemId":"47896270","author":"id00","createdAt":1777067649,"url":"https://news.ycombinator.com/item?id=47896270","kind":"comment","storyTitle":"Could a Claude Code routine watch my finances?","quote":"When I was trying to use Claude to analyze my past transactions, I found out that it was constantly hallucinating charges, sometimes adds new, double counts and etc."},{"itemId":"47894782","author":"NicuCalcea","createdAt":1777059236,"url":"https://news.ycombinator.com/item?id=47894782","kind":"comment","storyTitle":"OpenAI releases GPT-5.5 and GPT-5.5 Pro in the API","quote":"I think that's the problem. Who's going to claim responsibility when ChatGPT hallucinates or mistranslates a patient's diagnosis and they die?"},{"itemId":"47894680","author":"dwoldrich","createdAt":1777058662,"url":"https://news.ycombinator.com/item?id=47894680","kind":"comment","storyTitle":"Composition shouldn't be this hard","quote":"I think it was a fanciful idea, but in the age of AI, such a loosey goose negotiation process seems attainable if it weren't for hallucinations."},{"itemId":"47894404","author":"axegon_","createdAt":1777057022,"url":"https://news.ycombinator.com/item?id=47894404","kind":"comment","storyTitle":"MeshCore development team splits over trademark dispute and AI-generated code","quote":"* make no mistakes * don't hallucinate"},{"itemId":"47888807","author":"aleph_minus_one","createdAt":1777030689,"url":"https://news.ycombinator.com/item?id=47888807","kind":"comment","storyTitle":"Familiarity is the enemy: On why Enterprise systems have failed for 60 years","quote":"> Languages and ecosystems that are more mature and take stability very seriously, like Go or Clojure, don't have the problem of \"LLM hallucinates APIs\" nearly as much."},{"itemId":"47886789","author":"dgb23","createdAt":1777015475,"url":"https://news.ycombinator.com/item?id=47886789","kind":"comment","storyTitle":"Familiarity is the enemy: On why Enterprise systems have failed for 60 years","quote":"However, it is still unstable and APIs change, so LLMs get regularly confused. Languages and ecosystems that are more mature and take stability very seriously, like Go or Clojure, don't have the problem of \"LLM hallucinates APIs\" nearly as much."},{"itemId":"47884980","author":"exabrial","createdAt":1776999753,"url":"https://news.ycombinator.com/item?id=47884980","kind":"comment","storyTitle":"An update on recent Claude Code quality reports","quote":"Last I tried 4.7, it was bad. Like ChatGPT bad: changed stuff it wasn’t supposed to, hallucinated code, forgot information, missed simple things, didn’t catch mistakes."},{"itemId":"47880961","author":"sidmanchkanti21","createdAt":1776974396,"url":"https://news.ycombinator.com/item?id=47880961","kind":"comment","storyTitle":"[dead]","quote":"ParseBench generates its ground truth with frontier VLMs (Claude Opus for tables), which introduces hallucination risk, and its TableRecordMatch metric treats tables as unordered bags of key-value records, so it doesn't penalize column transposition or row reordering."},{"itemId":"47878223","author":"ritvikpandey21","createdAt":1776963714,"url":"https://news.ycombinator.com/item?id=47878223","kind":"comment","storyTitle":"[dead]","quote":"ParseBench generates its ground truth with frontier VLMs (Claude Opus for tables), which introduces hallucination risk, and its TableRecordMatch metric treats tables as unordered bags of key-value records, so it doesn't penalize column transposition or row reordering."},{"itemId":"47876888","author":"Edman274","createdAt":1776958005,"url":"https://news.ycombinator.com/item?id=47876888","kind":"comment","storyTitle":"Our newsroom AI policy","quote":"There's no world in which synthesizing available data from digitized sources alone ends up with more accurate data than a human with a search engine and the ability to make a phone call. So allowing LLM use at all is a direct admission that seeking out the \"truth\" is not an important goal because it could never actually improve accuracy and could only worsen it through hallucinated, probable reporting."},{"itemId":"47873275","author":"JumpCrisscross","createdAt":1776931367,"url":"https://news.ycombinator.com/item?id=47873275","kind":"comment","storyTitle":"Our newsroom AI policy","quote":"They had some nice quotes from my blog post explaining what was going on. The problem is that these quotes were not written by me, never existed, and appear to be AI hallucinations themselves."},{"itemId":"47872903","author":"legitster","createdAt":1776927680,"url":"https://news.ycombinator.com/item?id=47872903","kind":"comment","storyTitle":"Our newsroom AI policy","quote":"There was literally only one webpage on the internet left with instructions for my particular dryer - the page was more or less unusable with rotten links and riddled with adware. Claude did it's best but filled in the missing diagrams with hallucinations."},{"itemId":"47871703","author":"pixelbro","createdAt":1776911567,"url":"https://news.ycombinator.com/item?id=47871703","kind":"comment","storyTitle":"That Moment You Realize the Agent Is Retarded","quote":"Like, you must have seen it, it's nonsense but it feels like hand-authored pixel art. And then you look closely and there's mangled cells, halfway-hallucinated cells, no global coherence."},{"itemId":"47867865","author":"orbital-decay","createdAt":1776884761,"url":"https://news.ycombinator.com/item?id=47867865","kind":"comment","storyTitle":"Our eighth generation TPUs: two chips for the agentic era","quote":"GPT, on the other hand, was always terrible at languages, except for the short-lived gpt-4.5-preview. All modern models including Gemini have bugs in basic language coherency - random language switching, self-correction attempts resulting in hallucinations etc."},{"itemId":"47865788","author":"naikrovek","createdAt":1776874971,"url":"https://news.ycombinator.com/item?id=47865788","kind":"comment","storyTitle":"GitHub CLI now collects pseudoanonymous telemetry","quote":"None of these happened. none of them even appear to have happened, and none of them appear to have even been planned. It's all a hallucination by people that talk like this."},{"itemId":"47865421","author":"virgildotcodes","createdAt":1776873167,"url":"https://news.ycombinator.com/item?id=47865421","kind":"comment","storyTitle":"Our eighth generation TPUs: two chips for the agentic era","quote":"3.1 pro is just fundamentally not on the same level. In any context I've tried it in, for code review it acts like a model from 1yr ago in that it's all hallucinated superficial bullshit."},{"itemId":"47854611","author":"troglodytetrain","createdAt":1776806256,"url":"https://news.ycombinator.com/item?id=47854611","kind":"comment","storyTitle":"Ask HN: What Is Anthropic Doing?","quote":"But, In the last few days I've watched live, several times, as the tool use agent suddenly on a turn, completely forgets the correct tool call tags clearly defined in its system prompt, hallucinating a completely new tool call format that I have never seen before, before weirdly fixing itself some minutes and some turns later."},{"itemId":"47853127","author":"czhu12","createdAt":1776798634,"url":"https://news.ycombinator.com/item?id=47853127","kind":"comment","storyTitle":"Anthropic takes $5B from Amazon and pledges $100B in cloud spending in return","quote":"Open models don’t work nearly as well. Given that tokens are still relatively cheap, and hallucinations are expensive, I’ve not seen a huge up tick in open model usage for coding agents yet."},{"itemId":"47839798","author":"Nortey","createdAt":1776715480,"url":"https://news.ycombinator.com/item?id=47839798","kind":"story","storyTitle":"Show HN: Explain The Law – Simplified legislation and executive orders using AI","quote":"I’ve seen similar sites that invoke LLMs and generate summaries at runtime, which feels slow and doesn’t scale well. One of my biggest worries was hallucinations, as accuracy is essential for websites dealing with legislation."},{"itemId":"47836459","author":"SwellJoe","createdAt":1776701953,"url":"https://news.ycombinator.com/item?id=47836459","kind":"comment","storyTitle":"Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving","quote":"The thing is, Opus 4.5 is where the model reached Good Enough, at least for a wide variety of problems I use LLMs for. Before that, I almost never thought it was a more productive use of my time to use AI for development tasks, because it would always hallucinate something that would waste a bunch of my time."},{"itemId":"47836435","author":"virgildotcodes","createdAt":1776701835,"url":"https://news.ycombinator.com/item?id=47836435","kind":"comment","storyTitle":"ChatGPT and Codex Down","quote":"I feel like the closed frontier models really got there around 8 months ago and then even moreso ~4-6 months ago with the release of the Codex series and then opus 4.6. Finally feels like you can get reliably good implementations of features that follow repo patterns and best practices, and at least with 5.4 High/Xhigh Codex, code reviews that don't mostly surface hallucinated or superficial bullshit."},{"itemId":"47834384","author":"jillesvangurp","createdAt":1776693182,"url":"https://news.ycombinator.com/item?id=47834384","kind":"comment","storyTitle":"OpenClaw isn't fooling me. I remember MS-DOS","quote":"From a practical experience point of view, most people probably don't have the hands-on experience to make a good judgment just yet. \"I tried Chat GPT once and it hallucinated\" doesn't really count as valid experience at this point and many non-technical people are still at that level."},{"itemId":"47833955","author":"JasonGravy","createdAt":1776691359,"url":"https://news.ycombinator.com/item?id=47833955","kind":"story","storyTitle":"I Built a 22k-Line App with Zero Coding Experience. Or, how to control agents","quote":"Your sentences can check in but your phrases can never leave.” Where the hell did that come from? When a project hits a certain scale, the AI begins heavily hallucinating, tangling math into your UI components, and generating spaghetti code."},{"itemId":"47827730","author":"sbuttgereit","createdAt":1776633534,"url":"https://news.ycombinator.com/item?id=47827730","kind":"comment","storyTitle":"College instructor turns to typewriters to curb AI-written work","quote":"Removing the need for intermediaries would be a great competitive advantage for any AI vendor able to achieve it. But insofar as intermediaries create output from LLMs, they'll not be very well differentiated: the common wisdom tends to be the output, lest the AI be accused of hallucination or being overly supportive."},{"itemId":"47826661","author":"Ifkaluva","createdAt":1776625021,"url":"https://news.ycombinator.com/item?id=47826661","kind":"comment","storyTitle":"Notes from the SF peptide scene","quote":"Another example is when the visitor to the Bay Area house party is talking to several people, each of which has an increasingly insane startup idea, always funded by Peter Thiel. Another is a house party organized by Claude Code, which everything is superficially sensible but totally insane LLM hallucinations when you look more closely."},{"itemId":"47821973","author":"trueno","createdAt":1776575553,"url":"https://news.ycombinator.com/item?id=47821973","kind":"comment","storyTitle":"Zero-Copy GPU Inference from WebAssembly on Apple Silicon","quote":"You cannot ascertain anything about the person behind the blog if it's clear they used AI to write it. And without that there's no way to infer expertise, rule out hallucinations, falsehoods presented as matter of fact, and the whole broad set of things LLM's get wrong because of their limitations as a technology."},{"itemId":"47818359","author":"williamny","createdAt":1776537786,"url":"https://news.ycombinator.com/item?id=47818359","kind":"comment","storyTitle":"I'm a history major who built an air-gapped Bluetooth mesh to kill bots","quote":"Is AI hallucinating the value of this product it helped me build?"},{"itemId":"47817766","author":"akitaonrails","createdAt":1776533599,"url":"https://news.ycombinator.com/item?id=47817766","kind":"comment","storyTitle":"[dead]","quote":"A few weeks ago I wrote a detailed LLM coding benchmark comparing 33 open source and commercial models on the same test: build a Rails app using RubyLLM. The conclusion was that only 4 models generated code that works on first try (both Claudes, GLM 5, and GLM 5.1), and that for anyone who doesn’t want to waste time fixing API hallucinations, Claude Opus via Claude Code remains the most rational choice despite the p…"},{"itemId":"47815367","author":"lemoncookiechip","createdAt":1776515076,"url":"https://news.ycombinator.com/item?id=47815367","kind":"comment","storyTitle":"Claude Code Opus 4.7 keeps checking on malware","quote":"The real issue is the hallucinations which for people unfamiliar with said topic, can lead them into believing what they're being told is a fact when it's not."},{"itemId":"47811858","author":"tialaramex","createdAt":1776469793,"url":"https://news.ycombinator.com/item?id=47811858","kind":"comment","storyTitle":"Tesla tells HW3 owner to 'be patient' after 7 years of waiting for FSD","quote":"But the LLM isn't a person, Johnny needs to know more about the CIA's programme to cross breed Venusians with Hollywood stars? Here's an itinerary with the address of a real hotel in LA and an entirely hallucinated CIA officer's schedule."},{"itemId":"47806721","author":"LeCompteSftware","createdAt":1776438224,"url":"https://news.ycombinator.com/item?id=47806721","kind":"comment","storyTitle":"Silicon Valley is turning scientists into exploited gig workers?","quote":"The incredibly frustrating thing about the AI discussion is the refusal to consider actual evidence because of shifting targets. In 2026 there is evidence that 2024 LLMs did enormous damage to scientific research in 2025: hallucinated citations, hallucinated experiements, an onslaught of unreadable prose, etc etc."},{"itemId":"47804963","author":"jazzypants","createdAt":1776427263,"url":"https://news.ycombinator.com/item?id=47804963","kind":"comment","storyTitle":"Ada, Its Design, and the Language That Built the Languages","quote":"You can't import a variable that isn't exported. There are several little LLM hallucinations like this throughout the article."},{"itemId":"47799966","author":"NewsaHackO","createdAt":1776376250,"url":"https://news.ycombinator.com/item?id=47799966","kind":"comment","storyTitle":"Claude Opus 4.7","quote":"It's like an \"Eureka\" moment. Of course, as you use it more and more, you start to get better at recognizing \"Eureka\" moments and hallucinations, but I can definitely see how some people keep chasing that rush/feeling you get when it uses 5 minutes to solve a problem that would have taken you ages to do (if at all)."},{"itemId":"47796117","author":"curious_nile","createdAt":1776358062,"url":"https://news.ycombinator.com/item?id=47796117","kind":"comment","storyTitle":"Show HN: We beat Google, Cognition, Claude Code at codebase docs generation","quote":"Coding agents can only answer questions about parts of the codebase that are documented. If your docs cover routing but skip middleware, every middleware question becomes a hallucination."},{"itemId":"47788830","author":"jillesvangurp","createdAt":1776315832,"url":"https://news.ycombinator.com/item?id=47788830","kind":"comment","storyTitle":"Ask HN: Who is using OpenClaw?","quote":"It will also have exposed you to things like skills, guard rails, etc. that help you use these tools in a way that is a bit more repeatable and less prone to hallucinated outcomes."},{"itemId":"47785301","author":"siddhant-jain","createdAt":1776287601,"url":"https://news.ycombinator.com/item?id=47785301","kind":"comment","storyTitle":"Ask HN : Vibe coders: does AI keep breaking your auth and webhooks?","quote":"Thanks! Yeah, I also made that newbie mistake and about modes you heard correctly but even when you say to review the code it still hallucinates which clearly seen in websites and even after multiple correct prompts like \"Act like a senior engineer experience of...\" something like that, they just write and erase codes but never fixes the issues, they miss the critical flaws, where things starting complexing."},{"itemId":"47776580","author":"t43562","createdAt":1776244452,"url":"https://news.ycombinator.com/item?id=47776580","kind":"comment","storyTitle":"An AI Vibe Coding Horror Story","quote":"This has happened to me. It was in the days when ChatGPT was much worse than it is now and the code was almost one big hallucination - indescribable how bad it was."},{"itemId":"47758535","author":"steveklabnik","createdAt":1776118166,"url":"https://news.ycombinator.com/item?id=47758535","kind":"comment","storyTitle":"GitHub Stacked PRs","quote":"It sometimes will hallucinate older CLI options, because jj has changed at various times, but it's pretty decent with it at this point."},{"itemId":"47758325","author":"akersten","createdAt":1776117163,"url":"https://news.ycombinator.com/item?id=47758325","kind":"comment","storyTitle":"GitHub Stacked PRs","quote":"I've no issue with the logic of needing to update feature branches before merging, that's pretty bread and butter. The specific issue with this workflow is that the \"update branch\" button for PR B is grayed out because there are these hallucinated conflicts due to the new squash commit."},{"itemId":"47755047","author":"funnyfoobar","createdAt":1776100149,"url":"https://news.ycombinator.com/item?id=47755047","kind":"comment","storyTitle":"Ask HN: What Are You Working On? (April 2026)","quote":"Plan the work, implement, write tests, commit, open a PR, get it reviewed. If you just vibe code with no structure around it, you get messy diffs and hallucinated tests that nobody trusts."},{"itemId":"47748795","author":"rrr_oh_man","createdAt":1776064850,"url":"https://news.ycombinator.com/item?id=47748795","kind":"comment","storyTitle":"Ask HN: What Are You Working On? (April 2026)","quote":"My problem: my super-long, months-old ChatGPT threads were breaking down. Even typing got slow, and the longer the threads got, the more they hallucinated."},{"itemId":"47737091","author":"anon7000","createdAt":1775980488,"url":"https://news.ycombinator.com/item?id=47737091","kind":"comment","storyTitle":"The biggest advance in AI since the LLM","quote":"His core theory is that pushing LLM tech just with more training is not going to accomplish AGI. He does have some essays with good writing (not this one), and he typically talks about how we’ll need different techniques to solve things like hallucinations."},{"itemId":"47734854","author":"TuxSH","createdAt":1775950017,"url":"https://news.ycombinator.com/item?id=47734854","kind":"comment","storyTitle":"AI Is Tipping the Scales Toward Hackers After Mythos Release","quote":"Maybe there is some astroturfing going on, as is usually the case, but it's already known that Codex/Claude Code and their ilk have been ruining CTFs for a while. And well, one can always prompt \"review my feature branch\" or \"review this file for bugs\" with these tools; code analysis plays into the strengths of LLMs far more than code generation, since false positives/hallucinations aren't a problem with the former."},{"itemId":"47729057","author":"t23414321","createdAt":1775900173,"url":"https://news.ycombinator.com/item?id=47729057","kind":"comment","storyTitle":"AI Code Is Hollowing Out Open Source, and Maintainers Are Looking the Other Way","quote":"( Let's stop AI crimes ..) It looks like AI generated code can't be used legally - without respecting rights and licenses or without proving that no copyrighted code may happen to be mixed in - true only if no such was used to learn at first - what all we know, it's not the case: all is tainted by legal omissions or simply: stolen (prove otherwise - there is nothing AI invented first except hallucinations)."},{"itemId":"47716596","author":"michaelashley29","createdAt":1775821176,"url":"https://news.ycombinator.com/item?id=47716596","kind":"comment","storyTitle":"We've raised $17M to build what comes after Git","quote":"I feel like we’re over-capitalizing a problem that could be solved with better protocols. If the \"Git successor\" is just a wrapper to help agents not hallucinate their own worktrees, it feels like a very expensive solution to a context-window management problem."},{"itemId":"47715712","author":"dspillett","createdAt":1775814586,"url":"https://news.ycombinator.com/item?id=47715712","kind":"comment","storyTitle":"DRAM has a design flaw from 1966. I bypassed it [video]","quote":"Instead of telling me to ask copilot to guess what I need to know, why not write some good documentation that you can reference directly (or that I can search through)? Heck, use copilot to draft that documentation if you want to (but please have humans review the result for hallucinations, missed parts, and other inaccuracies, before publishing)."},{"itemId":"47715304","author":"rakamotog","createdAt":1775810854,"url":"https://news.ycombinator.com/item?id=47715304","kind":"comment","storyTitle":"I still prefer MCP over skills","quote":"Imagine you are creating an asset which requires multiple API calls and your UI is designed to go through a 10-12 step setup process for that asset. In practice even if we give one tool for LLM to one-shot it, or even if we break it down into 10-12 tools the points of hallucinations are much higher."},{"itemId":"47713593","author":"wasimsk","createdAt":1775794713,"url":"https://news.ycombinator.com/item?id=47713593","kind":"story","storyTitle":"Ask HN: Does Vibe Coding and Prompt Engineering make me an Software Engineer?","quote":"Ask HN: Does Vibe Coding and Prompt Engineering make me an Software Engineer? I’ve dialed in my prompts to the point where hallucinations are basically zero."},{"itemId":"47704728","author":"sohaibtariq","createdAt":1775747214,"url":"https://news.ycombinator.com/item?id=47704728","kind":"story","storyTitle":"Show HN: AI agents are bad at API integrations – we fixed it","quote":"We've just launched in Beta with PayPal: https://developer.paypal.com/serversdk/java/getting-started/how-to-get-started/?ide=claudeCode&isContextPlugin=true We benchmarked Context Plugins on two codebases, nopCommerce (a mature ASP.NET e-commerce platform) and Microsoft's eShop reference app, and saw 2x faster integration time, 65% lower token usage, and near-zero hallucinations."},{"itemId":"47704283","author":"phlakaton","createdAt":1775745160,"url":"https://news.ycombinator.com/item?id=47704283","kind":"comment","storyTitle":"Claude mixes up who said what","quote":"How do you know? Because it looks to me like it could be a straightforward hallucination, compounded by the agent deciding it was OK to take a shortcut that you really wish it hadn't."},{"itemId":"47696795","author":"camgunz","createdAt":1775685412,"url":"https://news.ycombinator.com/item?id=47696795","kind":"comment","storyTitle":"ML promises to be profoundly weird","quote":"> Specifically in the case where it can use tools - no it doesn't hallucinate."},{"itemId":"47693689","author":"simoncion","createdAt":1775670314,"url":"https://news.ycombinator.com/item?id=47693689","kind":"comment","storyTitle":"ML promises to be profoundly weird","quote":"I encourage you to read all of them (make sure to check out the footnote... it's pretty good). The exact text of the ChatGPT session is irrelevant; even if you reported that you were unable to reproduce the issue, it would only reinforce one of the underlying points -namely- that these systems are unreliable. aphyr has a pretty extensive body of published work that indicates that he'd not likely fabricate a story of…"},{"itemId":"47693600","author":"beders","createdAt":1775669971,"url":"https://news.ycombinator.com/item?id=47693600","kind":"comment","storyTitle":"ML promises to be profoundly weird","quote":"I think you highlight one of the problems with users of LLMs: You can't tell anymore if it is BS or not. I caught Claude the other day hallucinating code that was not only wrong, but dangerously wrong, leading to tasks being failed and never recover."},{"itemId":"47687468","author":"yjcho9317","createdAt":1775639605,"url":"https://news.ycombinator.com/item?id=47687468","kind":"comment","storyTitle":"Show HN: Lilith-zero: Ultra-fast, Rust-based security middleware for MCP","quote":"I ended up writing input validation for every single tool by hand because there's no standard for it. Even then, Claude Code will happily call tools in a loop with hallucinated parameters."},{"itemId":"47681373","author":"h2u2","createdAt":1775595940,"url":"https://news.ycombinator.com/item?id=47681373","kind":"comment","storyTitle":"Nedster – An open-source, local-first coding agent that verifies its own work","quote":"It’s optimized specifically for consumer hardware (tested extensively on an 8GB RTX 3060 Ti). The biggest problem I faced with running local agents (especially mid-size models like Qwen) is what I call \"Execution Theater\"—the model hallucinates that it created a file."},{"itemId":"47681116","author":"sutib","createdAt":1775594616,"url":"https://news.ycombinator.com/item?id=47681116","kind":"comment","storyTitle":"Claude Code login fails with OAuth timeout on Windows","quote":"I really feel the abdication very strongly and it makes me unable and unwilling to put a hard stamp on quality. I have seen too much hallucination or half missed requirements to put that much trust in AI."},{"itemId":"47678355","author":"Kappa90","createdAt":1775581671,"url":"https://news.ycombinator.com/item?id=47678355","kind":"comment","storyTitle":"Show HN: I built a database for AI agents","quote":"80% of the benchmark questions are aggregations, 16% are multi-hop, 4% are lookups/subqueries. Multi-hop is where LLMs struggle the most (hallucinations, partial answers), and aggregations is where you get the most token efficiency, since you skip on pagination which you need with APIs/MCPs that don't provide filters."},{"itemId":"47676563","author":"mernaemara","createdAt":1775574481,"url":"https://news.ycombinator.com/item?id=47676563","kind":"comment","storyTitle":"Ask HN: Who is hiring? (April 2026)","quote":"- building and scaling LLM workflows (agents, pipelines, RAG) - handling real-world issues like latency, reliability, and hallucination control"},{"itemId":"47676494","author":"pyjka","createdAt":1775574119,"url":"https://news.ycombinator.com/item?id=47676494","kind":"comment","storyTitle":"Show HN: Building EU payment infra for agents","quote":"We are building payment infrastructure specifically designed for AI agents in the EU. The problem with giving an agent access to standard payment APIs (like Stripe) is that LLMs are probabilistic—if they hallucinate a JSON payload or get stuck in a loop, they can drain a bank account and also obviously human trust."},{"itemId":"47668200","author":"sjt-at-rev","createdAt":1775514635,"url":"https://news.ycombinator.com/item?id=47668200","kind":"comment","storyTitle":"API and MCP that pumps data from SEC filings into your agent context","quote":"Pulling KPIs from SEC filings is trickier and more error prone than it looks at first blush, and the risk of plausible land-mine hallucinations is still a deal breaker for a lot of serious finance folks."},{"itemId":"47664311","author":"efficax","createdAt":1775497493,"url":"https://news.ycombinator.com/item?id=47664311","kind":"comment","storyTitle":"Issue: Claude Code is unusable for complex engineering tasks with Feb updates","quote":"There are constant reports for every major AI vendor that all of a sudden it is no longer working as well as expected, has gotten dumber, is being degraded on purpose by the vendor, etc. Isn't the more economical explanation that these models were never as impressive as you first thought they were, hallucinate often, break down in unexpected ways depending on context, and simply cannot handle large and complex engin…"},{"itemId":"47651871","author":"ieuanking","createdAt":1775410869,"url":"https://news.ycombinator.com/item?id=47651871","kind":"story","storyTitle":"Show HN: Built a model-agnostic, desktop-native, research studio for local files","quote":"We built Ubik Studio because in professional research settings, or fields where accurate citation is required, AI tools kinda flop. Models hallucinate citations with total confidence and you are locked into one provider."},{"itemId":"47651244","author":"jillesvangurp","createdAt":1775407698,"url":"https://news.ycombinator.com/item?id=47651244","kind":"comment","storyTitle":"Eight years of wanting, three months of building with AI","quote":"I also ask critical questions about edge cases, testing the non happy path, hardening, concurrency, latency, throughput, etc. If you don't, AIs kind of default to taking short cuts, only focus on the happy path, or hallucinate that it's all fine, etc."},{"itemId":"47649650","author":"didgetmaster","createdAt":1775398259,"url":"https://news.ycombinator.com/item?id=47649650","kind":"comment","storyTitle":"The threat is comfortable drift toward not understanding what you're doing","quote":"Hand that calculator to someone who never understood what the tool actually did but just accepted whatever answer appeared; and they would likely think the answer was totally reasonable. Catching an LLM hallucinating often takes a basic understanding of what the answer should look like before asking the question."},{"itemId":"47649477","author":"lapcat","createdAt":1775397222,"url":"https://news.ycombinator.com/item?id=47649477","kind":"comment","storyTitle":"The threat is comfortable drift toward not understanding what you're doing","quote":"A related problem: LLMs have been trained with papers written and supervised by Alice-type specialists. There's a common claim that LLMs will hallucinate less in the future, but I think that LLMs will hallucinate more in the future, when specialty fields become dominated by Bob-type \"specialists\" who have a harder time distinguishing fact from fiction."},{"itemId":"47642042","author":"lconnell962","createdAt":1775328709,"url":"https://news.ycombinator.com/item?id=47642042","kind":"comment","storyTitle":"12k AI-generated blog posts added in a single commit","quote":"I don't know what the solution is for this type given claims Google Translate has started to do this kind of translation for people. An AI translation is probably just as prone to hallucinations as any other AI, but it probably will look more natural to readers than a direct translation."},{"itemId":"47637697","author":"renewiltord","createdAt":1775297633,"url":"https://news.ycombinator.com/item?id=47637697","kind":"comment","storyTitle":"Tell HN: Anthropic no longer allowing Claude Code subscriptions to use OpenClaw","quote":"Edge models are good for their purpose but putting them in agentic flow with current ollama quants on a Mac Mini I see high tool use error rate and output hallucination."},{"itemId":"47631380","author":"cyclecycle","createdAt":1775246025,"url":"https://news.ycombinator.com/item?id=47631380","kind":"story","storyTitle":"Show HN: Large scale hallucinated citation problem in published literature","quote":"Show HN: Large scale hallucinated citation problem in published literature"},{"itemId":"47624438","author":"noodletheworld","createdAt":1775205299,"url":"https://news.ycombinator.com/item?id=47624438","kind":"comment","storyTitle":"Cursor 3","quote":"Don't care. If someone calls it a hallucination?"},{"itemId":"47622878","author":"ieuanking","createdAt":1775186836,"url":"https://news.ycombinator.com/item?id=47622878","kind":"story","storyTitle":"Show HN: Ubik – Analyze local files like NotebookLM, with Cursor-like power","quote":"We built Ubik Studio because professional knowledge workers and researchers are experiencing a crisis with unreliable AI tools. Models hallucinate citations with total confidence."},{"itemId":"47622455","author":"Retro_Dev","createdAt":1775181594,"url":"https://news.ycombinator.com/item?id=47622455","kind":"comment","storyTitle":"Google releases Gemma 4 open models","quote":"I find it easy to evaluate because I know the language. This local LLM marks the first version that 1) doesn't hallucinate words - at least, for the largest model - and 2) uses common word-phrases that other toki pona speakers use, and most importantly 3) can actually run on my laptop."},{"itemId":"47617542","author":"whimblepop","createdAt":1775151465,"url":"https://news.ycombinator.com/item?id=47617542","kind":"comment","storyTitle":"Google releases Gemma 4 open models","quote":"Whether I enable thinking makes no difference: Gemini always answers me super quickly, rarely actually looks something up, and lies to me. It has a really bad unchecked hallucination problem because it prioritizes speed over accuracy and (astonishingly, to me) is way more hesitant to run web searches than ChatGPT or Claude."},{"itemId":"47613506","author":"Gigachad","createdAt":1775132665,"url":"https://news.ycombinator.com/item?id=47613506","kind":"comment","storyTitle":"Sweden goes back to basics, swapping screens for books in the classroom","quote":"For asking about how to do things in software guis modern LLMs seem to have a high rate of making up features or paths to reach them. For asking advice in games I've seen an extremely high rate of hallucinations."},{"itemId":"47607180","author":"decodebytes","createdAt":1775081425,"url":"https://news.ycombinator.com/item?id=47607180","kind":"comment","storyTitle":"Ask HN: Who is hiring? (April 2026)","quote":"You will build the layer between agentic reasoning and system execution — intercepting, classifying, and validating every agent intent against security policy before it runs. Design systems that capture and cryptographically verify every instruction and action, maintaining ground truth in environments prone to hallucination and injection."},{"itemId":"47606855","author":"theLiminator","createdAt":1775079489,"url":"https://news.ycombinator.com/item?id=47606855","kind":"comment","storyTitle":"EmDash – A spiritual successor to WordPress that solves plugin security","quote":"Yeah, my theory is that Rust is going to be a somewhat \"local optima\" for a while for LLMs. LLMs have a couple major problem, they hallucinate and make mistakes."},{"itemId":"47605687","author":"tonyarkles","createdAt":1775073162,"url":"https://news.ycombinator.com/item?id=47605687","kind":"comment","storyTitle":"Claude wrote a full FreeBSD remote kernel RCE with root shell","quote":"I have gotten absolutely incredible results out of it. I have had a few hallucinations/incorrect analyses (hard to tell which it was), but in general the results have been fantastic."},{"itemId":"47602229","author":"californical","createdAt":1775057436,"url":"https://news.ycombinator.com/item?id=47602229","kind":"comment","storyTitle":"I quit. The clankers won","quote":"I don’t think a small team can write enough code, the models already struggle in medium-popularity languages that have years of history. They hallucinate lua functionality sometimes, for example, even though I’m sure there is lots of lua code out there."},{"itemId":"47602018","author":"Bridged7756","createdAt":1775056438,"url":"https://news.ycombinator.com/item?id=47602018","kind":"comment","storyTitle":"I quit. The clankers won","quote":"I'm really getting tired of the programming obituaries. As if LLMs didn't fail at any complex task, as if they didn't vomit shit code and as if they didn't just copy patterns surrounding the new code, and as if they didn't hallucinate and downright write wrong code or made up libraries."},{"itemId":"47601658","author":"testfrequency","createdAt":1775054641,"url":"https://news.ycombinator.com/item?id=47601658","kind":"comment","storyTitle":"Claude Code users hitting usage limits 'way faster than expected'","quote":"I literally have a Gemini mcp that I force close to use half the time when it’s lost, and Gemini nails it every single time. I’m on a Claude max 20x plan right now, and I seriously can’t imagine not having it around anymore but Gemini seems to always have my back on actual current data and less hallucinations."},{"itemId":"47601392","author":"shimman","createdAt":1775053553,"url":"https://news.ycombinator.com/item?id=47601392","kind":"comment","storyTitle":"Slop is not necessarily the future","quote":"If you can actually do this, please sell your services. You will become a multi-millionaire overnight if you can provide a workflow that doesn't result in mass hallucinations or incorrect suggestions you're able to do something no other LLM company can."},{"itemId":"47600022","author":"pron","createdAt":1775047058,"url":"https://news.ycombinator.com/item?id=47600022","kind":"comment","storyTitle":"Slop is not necessarily the future","quote":"Before agents it was hard to see because few people write code like that (just as no one would think to make a curtain rod load-bearing when building a structure). And let's not forget that the models hallucinate."},{"itemId":"47596998","author":"yurimo","createdAt":1775019531,"url":"https://news.ycombinator.com/item?id=47596998","kind":"comment","storyTitle":"Slop is not necessarily the future","quote":"Industry is now pushing for agentic web where agents can do this on your behalf. But if we have slop foundations and then add unstable models that can hallucinate and make mistakes on top of it, then it's just a recipe for catastrophe."},{"itemId":"47594623","author":"dbshapco","createdAt":1774998121,"url":"https://news.ycombinator.com/item?id=47594623","kind":"comment","storyTitle":"Slop is not necessarily the future","quote":"Like a poorly written termination condition in a container that essentially was a busy wait burning thousands of dollars a month doing nothing (true story). I am currently writing a trading system, and can't afford to hallucinate a bunch of bad trades."},{"itemId":"47592506","author":"jmyeet","createdAt":1774986600,"url":"https://news.ycombinator.com/item?id=47592506","kind":"comment","storyTitle":"Slop is not necessarily the future","quote":"There's a little thing called the halting problem, which in this context basically means there's no way to guarantee that the AI will be restricted from doing anything you don't want it to do. An amusing example was an Air Canada chatbot that hallucinated a refund policy that a court said it had to honor [1]."},{"itemId":"47584424","author":"melvinroest","createdAt":1774946771,"url":"https://news.ycombinator.com/item?id=47584424","kind":"comment","storyTitle":"Ollama is now powered by MLX on Apple Silicon in preview","quote":"So the added value is that you now have a super charged information retrieval system on your notes with an LLM that can stitch loose facts reasonably well together, like a librarian would. It's also very easy to see hallucinations, if you recognize your own writing well, which I do."},{"itemId":"47573727","author":"WhatsName","createdAt":1774875609,"url":"https://news.ycombinator.com/item?id=47573727","kind":"comment","storyTitle":"Ask HN: Where have you found the coding limits of current models?","quote":"I hate the tendency to make things up and I dont mean hallucinations."},{"itemId":"47558961","author":"peteforde","createdAt":1774740040,"url":"https://news.ycombinator.com/item?id=47558961","kind":"comment","storyTitle":"The risk of AI isn't making us lazy, but making \"lazy\" look productive","quote":"I know that this reply reads as salty, but I'm really struggling to contain my own \"wtf\" on this end. Seriously, people that are so determined to prove that LLMs don't work despite how easy it is to test for yourself and see that they clearly do work are the ones that are hallucinating."},{"itemId":"47550076","author":"gerdesj","createdAt":1774656653,"url":"https://news.ycombinator.com/item?id=47550076","kind":"comment","storyTitle":"Why are executives enamored with AI, but ICs aren't?","quote":"Now a modern drilldriver can wrench your wrist if you are not careful and brace properly. A modern LLM will hallucinate like a nineties raver on ecstasy but if you listen carefully and phrase your prompts carefully and ignore the chomping teeth and keep them hydrated, you may get something remarkable out of the creature 8)"},{"itemId":"47540603","author":"nirb89","createdAt":1774604074,"url":"https://news.ycombinator.com/item?id=47540603","kind":"comment","storyTitle":"We rewrote JSONata with AI in a day, saved $500k/year","quote":"Honestly, I was very adverse to agentic code up until Opus came out. The hallucinations and false confidence it had in objectively wrong answers just broke more things than it fixed."},{"itemId":"47536457","author":"hexaga","createdAt":1774562898,"url":"https://news.ycombinator.com/item?id=47536457","kind":"comment","storyTitle":"AI users whose lives were wrecked by delusion","quote":"You don't get it for free by dipping into some sea of imagined tricksters. I assume you're referring to the hallucination phenomenon and dual purposing it toward manipulation to be able to hee-hah about those silly people who are so silly they fool themselves with the soul upload machine (?) so I'll address that:"},{"itemId":"47533941","author":"YossarianFrPrez","createdAt":1774549478,"url":"https://news.ycombinator.com/item?id=47533941","kind":"comment","storyTitle":"AI users whose lives were wrecked by delusion","quote":"Some choice quotes: > “What we’re seeing in these cases are clearly delusions,” he says. “But we’re not seeing the whole gamut of symptoms associated with psychosis, like hallucinations or thought disorders, where thoughts become jumbled and language becomes a bit of a word salad.”"},{"itemId":"47530870","author":"sohaibtariq","createdAt":1774535187,"url":"https://news.ycombinator.com/item?id=47530870","kind":"story","storyTitle":"Show HN: Context Plugins – API context for AI coding assistants","quote":"API providers maintain API and SDK docs, but AI assistants don’t always use them. They use a combination of web search, training data and hallucinations to write API integration code."},{"itemId":"47529584","author":"notglossy","createdAt":1774527659,"url":"https://news.ycombinator.com/item?id=47529584","kind":"comment","storyTitle":"From zero to a RAG system: successes and failures","quote":"You will still get hallucinations."},{"itemId":"47509328","author":"Sohcahtoa82","createdAt":1774386475,"url":"https://news.ycombinator.com/item?id=47509328","kind":"comment","storyTitle":"Is anybody else bored of talking about AI?","quote":"They have no idea how much AI has improved since it reached center stage 3 years ago. Hallucinations are exceptionally rare now, since they now rely on searching for answers rather than what was in its training data."},{"itemId":"47503322","author":"psychoslave","createdAt":1774363250,"url":"https://news.ycombinator.com/item?id=47503322","kind":"comment","storyTitle":"PC Gamer recommends RSS readers in a 37mb article that just keeps downloading","quote":">(an LLM comment doesn't count, because they lie and hallucinate, and it's not you responding)"},{"itemId":"47487332","author":"sofixa","createdAt":1774260058,"url":"https://news.ycombinator.com/item?id=47487332","kind":"comment","storyTitle":"Two pilots dead after plane and ground vehicle collide at LaGuardia","quote":"Everyone can write a comment on Reddit / make a podcast / video / whatever claiming whatever they want. Unless you already know and trust them (which requires you to be able to cross-check their information), it's potentially as useful as a random LLM hallucination."},{"itemId":"47483090","author":"threethirtytwo","createdAt":1774219626,"url":"https://news.ycombinator.com/item?id=47483090","kind":"comment","storyTitle":"OpenClaw is a security nightmare dressed up as a daydream","quote":"This is provably not true. LLMs CAN be restricted and censored and an LLM can be shown refusing an injection attack AND not hallucinating."},{"itemId":"47481560","author":"jesse_dot_id","createdAt":1774209948,"url":"https://news.ycombinator.com/item?id=47481560","kind":"comment","storyTitle":"OpenClaw is a security nightmare dressed up as a daydream","quote":"This problem is inherently unsolvable because LLMS are prone to hallucinations and prompt injection attacks."},{"itemId":"47473127","author":"johnfn","createdAt":1774139870,"url":"https://news.ycombinator.com/item?id=47473127","kind":"comment","storyTitle":"Profiling Hacker News users based on their comments","quote":"This is arguably their defining HN characteristic: they are one of the most vocal, persistent AI optimists on the platform. They claim ~90-95% of their shipped code is AI-generated, report 5-10x productivity gains, and have built a detailed methodology around it — using Playwright for visual verification, static typechecking as a hallucination filter, and e2e test suites as automated validation harnesses"},{"itemId":"47471609","author":"machinaMKT","createdAt":1774128496,"url":"https://news.ycombinator.com/item?id=47471609","kind":"comment","storyTitle":"Show HN: Vessel Browser – An open-source browser built for AI agents, not humans","quote":"I've been packaging reusable MCP tool schema definitions (the JSON spec layer, not implementations) as a way to give agents a consistent vocabulary across different harnesses. The hallucination problem you're solving at the surfacing layer is related but distinct from the schema consistency problem — both matter for reliable tool use in production."},{"itemId":"47470937","author":"machinaMKT","createdAt":1774124852,"url":"https://news.ycombinator.com/item?id=47470937","kind":"comment","storyTitle":"Show HN: Vessel Browser – An open-source browser built for AI agents, not humans","quote":"Curious about the MCP tool schema design: with 40+ tools, are you using a flat tool registry or do you have any grouping/namespacing? I've run into issues where agents get confused when they have too many tools with similar-sounding names and start hallucinating tool calls that don't exist."},{"itemId":"47466512","author":"spopejoy","createdAt":1774096986,"url":"https://news.ycombinator.com/item?id=47466512","kind":"comment","storyTitle":"Java is fast, code might not be","quote":"LLMs massacre gradle code all the time. Once you're past boilerplate generation and doing anything remotely unusual they can't stop hallucinating broken shit that they insist works."},{"itemId":"47463717","author":"dongjiewu","createdAt":1774063856,"url":"https://news.ycombinator.com/item?id=47463717","kind":"story","storyTitle":"Show HN: An open-source AI Scrum Team that lives natively in GitHub Issues","quote":"It's a security nightmare for enterprise code. Chaos: Generalist agents try to do everything at once (planning, coding, testing) in a single context window, which inevitably leads to hallucinations and broken dependencies."},{"itemId":"47458191","author":"artine","createdAt":1774029355,"url":"https://news.ycombinator.com/item?id=47458191","kind":"comment","storyTitle":"I'm OK being left behind, thanks","quote":"I then asked it how to solve a simple bug with an easy solution. It identified the right place to make the fix but its entire suggested solution was a one-liner invoking a hallucinated library method."},{"itemId":"47455518","author":"genthree","createdAt":1774018667,"url":"https://news.ycombinator.com/item?id=47455518","kind":"comment","storyTitle":"I'm OK being left behind, thanks","quote":"We're doing that in my office, forced Cursor use. A good chunk of the \"edited by AI\" lines in my history were just auto-completing about the same as a traditional intellisense-alike would do (and actually Cursor doesn't seem to supply that, which is frequently annoying and wastes my time, in particular when I need to make sure it hasn't hallucinated a method or property on an object it should be able to \"see\" the de…"},{"itemId":"47455483","author":"msakib13","createdAt":1774018553,"url":"https://news.ycombinator.com/item?id=47455483","kind":"comment","storyTitle":"What 81,000 people want from AI","quote":"The #1 concern (unreliability at 26.7%) maps exactly to the problem with how most SMBs use AI today — chatbots for general Q&A rather than purpose-built tools for specific operations. A chatbot that hallucinates an answer is annoying."},{"itemId":"47454944","author":"draftsman","createdAt":1774016480,"url":"https://news.ycombinator.com/item?id=47454944","kind":"comment","storyTitle":"I'm OK being left behind, thanks","quote":"I have used Opus 4.6 on a number of projects, and it still spits out buggy, and sometimes, flat out broken code. I was actually surprised when it completely hallucinated the names of query params for an HTTP request in my code, when in the prompt I had explicitly given it the exact names it needed to use."},{"itemId":"47442348","author":"runarberg","createdAt":1773938986,"url":"https://news.ycombinator.com/item?id=47442348","kind":"comment","storyTitle":"macOS 26 breaks custom DNS settings including .internal","quote":"With LLMs this is less clear, you don’t get the old school artifacts, instead you get hallucinations, and very subtle errors that completely alter the meaning while leaving the sentence intact enough that your reader might not know this is a machine translation error."},{"itemId":"47441587","author":"axotopia","createdAt":1773935944,"url":"https://news.ycombinator.com/item?id=47441587","kind":"story","storyTitle":"Show HN: Dumped Wix for an AI Edge agent so I never have to hire junior staff","quote":"* Liability is the killer. Hallucinate a building code clause?"},{"itemId":"47438930","author":"aaur0","createdAt":1773926360,"url":"https://news.ycombinator.com/item?id=47438930","kind":"story","storyTitle":"Show HN: Altimate Code – Open-Source Agentic Data Engineering Harness","quote":"Claude Code and Cursor are genuinely good for software engineering. But when you point them at a data stack, they hallucinate column names, ignore partition keys, and have no concept of data contracts or quality rules in your models."},{"itemId":"47438156","author":"nadav_tal","createdAt":1773923024,"url":"https://news.ycombinator.com/item?id=47438156","kind":"comment","storyTitle":"Node9 – A \"sudo\" wrapper for AI agents with auto Git snapshots","quote":"The first response wins. If an AI hallucination does slip through, Node9 takes hidden git snapshots (using dangling commits so it doesn't pollute your branch) right before the AI edits a file, letting you run node9 undo."},{"itemId":"47436428","author":"Frieren","createdAt":1773908804,"url":"https://news.ycombinator.com/item?id=47436428","kind":"comment","storyTitle":"What 81,000 people want from AI","quote":"Tests may not be as reliable as though but they are good enough when other symptoms are accounted for. To randomly test people based on AI hallucinations can increase the number of unnecessary medication or even interventions."},{"itemId":"47434538","author":"andsoitis","createdAt":1773890785,"url":"https://news.ycombinator.com/item?id=47434538","kind":"comment","storyTitle":"The SDLC Is Dead","quote":"> the loop might be intent -> build -> observe. But in reality, it's likely intent -> build -> realize the AI hallucinated an API that doesn't exist -> fix -> discover it broke something else -> fix -> realize the architecture doesn't fit a constraint you thought you had -> start over."},{"itemId":"47434477","author":"Jemaclus","createdAt":1773890287,"url":"https://news.ycombinator.com/item?id=47434477","kind":"comment","storyTitle":"The SDLC Is Dead","quote":"If that were the case, the loop might be intent -> build -> observe. But in reality, it's likely intent -> build -> realize the AI hallucinated an API that doesn't exist -> fix -> discover it broke something else -> fix -> realize the architecture doesn't fit a constraint you thought you had -> start over."},{"itemId":"47434079","author":"mikepurvis","createdAt":1773887229,"url":"https://news.ycombinator.com/item?id=47434079","kind":"comment","storyTitle":"Warranty Void If Regenerated","quote":"If one is lucky, these are LLM hallucinations: obviously wrong and quickly discarded."},{"itemId":"47431850","author":"Imustaskforhelp","createdAt":1773870578,"url":"https://news.ycombinator.com/item?id=47431850","kind":"comment","storyTitle":"Warranty Void If Regenerated","quote":"(Is openclaw having these flowcharts similar to nocode territory?) at some point, code is more efficient in doing so, maybe even people will then have this code itself be generated by AI but then once again, you are one hallucination away from a security nightmare or doesn't it become openclaw type thing once again"},{"itemId":"47430206","author":"a_lifters_life","createdAt":1773861420,"url":"https://news.ycombinator.com/item?id=47430206","kind":"comment","storyTitle":"Who is using Ollama day-to-day?","quote":"Im building a script for my own dev to spin up droplets on DigitalOcean to run high performing LLMs like H100 - this is all prompted through open web ui (ollama) . I find the models my 12GB GPU will tolerate have too many hallucinations - its ok for diffing for example or basic scripts, but anything somewhat complicated I need 60GB+ GPU which isnt feasible for me to buy.."},{"itemId":"47425573","author":"techcam","createdAt":1773840625,"url":"https://news.ycombinator.com/item?id=47425573","kind":"story","storyTitle":"Show HN: CLI to score AI prompts after a prod failure","quote":"- instruction hierarchy ambiguity - under-constrained outputs (hallucination risk)"},{"itemId":"47425162","author":"sshine","createdAt":1773838312,"url":"https://news.ycombinator.com/item?id=47425162","kind":"comment","storyTitle":"Kagi is contemplating the removal of the assistant from its professional tier","quote":"For a beautiful 6 months before the frontier labs were able to search as part of their responses, Kagi had an edge here. This was when the “AIs hallucinate!” meme was thriving and my response was “Wait, your AI AGENT doesn’t provide references?” and I felt very cutting edge then."},{"itemId":"47424400","author":"MatrixOrigin","createdAt":1773833884,"url":"https://news.ycombinator.com/item?id=47424400","kind":"story","storyTitle":"Show HN: Git-style version control for AI agent memory","quote":"AI agents are brilliant, but they have a \"context amnesia\" problem. They forget, hallucinate, make mistakes."},{"itemId":"47423183","author":"snirkhe","createdAt":1773823180,"url":"https://news.ycombinator.com/item?id=47423183","kind":"comment","storyTitle":"Making AI Agents Safe to Run in Manufacturing ERPs","quote":"State Drift: How do you validate that ERP conditions haven't changed between the LLM's \"thought\" (T0) and the Gateway's \"execution\" (T1)? Idempotency: Ensuring a hallucinated re-try doesn't result in double-ordering or duplicate transactions."},{"itemId":"47421107","author":"Raywob","createdAt":1773802530,"url":"https://news.ycombinator.com/item?id=47421107","kind":"story","storyTitle":"My AI didn't misread a receipt – it fabricated one from scratch","quote":"1. Vision model hallucination is qualitatively different from text hallucination."},{"itemId":"47417059","author":"aderix","createdAt":1773775540,"url":"https://news.ycombinator.com/item?id=47417059","kind":"comment","storyTitle":"Show HN: GitAgent – An open standard that turns any Git repo into an AI agent","quote":"Version controlling the agent's prompts and capabilities is great for configuration management, but it completely misses the runtime execution risk. If an LLM hallucinates in production and decides to execute a destructive tool defined in SKILL.md (like dropping a table or issuing a Stripe refund), a Git PR approval process doesn't help you mid-flight."},{"itemId":"47416033","author":"colinfly","createdAt":1773770329,"url":"https://news.ycombinator.com/item?id=47416033","kind":"story","storyTitle":"What broke when I tried to evaluate an AI agent in production","quote":"- Agent calling localhost in a cloud environment → got stuck at 46 - Real CVEs flagged as hallucinations → evaluation issue, not model issue"},{"itemId":"47415419","author":"ashish-a","createdAt":1773767111,"url":"https://news.ycombinator.com/item?id=47415419","kind":"story","storyTitle":"Show HN: Open-source white-box agentic red teamer for AI agents","quote":"Cascading hallucination attacks that gradually corrupt agent reasoning across a conversation."},{"itemId":"47414269","author":"peterweisz","createdAt":1773762209,"url":"https://news.ycombinator.com/item?id=47414269","kind":"comment","storyTitle":"I'm 60 years old. Claude Code killed a passion","quote":"So, in the beginning, we teamed up with a fellow Accelerator alumni who claimed to be an experienced AI developer and put a lot of trust into his capabilities. After 8 months of back and forth, endless excuses, and iterations without improvement, we received an MVP that was too poor to mention, not at all according to mutually agreed specifications, full of drift and hallucinations, non-reliable scoring, and a word …"},{"itemId":"47412156","author":"otti-sister","createdAt":1773752991,"url":"https://news.ycombinator.com/item?id=47412156","kind":"story","storyTitle":"Show HN: Oh-my-agent – A structural harness for AI agents in real projects","quote":"If you ask an AI agent to build a todo app, it will usually produce something. But in a real development environment, limitations become obvious: hallucinations, drifting off task, or repeating the same mistakes."},{"itemId":"47411433","author":"staticassertion","createdAt":1773748529,"url":"https://news.ycombinator.com/item?id=47411433","kind":"comment","storyTitle":"Speed at the cost of quality: Study of use of Cursor AI in open source projects (2025)","quote":"There's no question that models have improved. I'd say the biggest leap was around the ChatGPT 3.5 -> 4.0, which radically reduced hallucination problems."},{"itemId":"47401761","author":"admonaut","createdAt":1773680942,"url":"https://news.ycombinator.com/item?id=47401761","kind":"comment","storyTitle":"Show HN: AI Assistant Widget – GDPR-compliant, RAG-based, 2-line integration","quote":"You upload text, scan URLs, or add PDFs – the assistant only answers from that. No hallucinations, no generic internet results."},{"itemId":"47400150","author":"g051051","createdAt":1773674189,"url":"https://news.ycombinator.com/item?id=47400150","kind":"comment","storyTitle":"I'm 60 years old. Claude Code killed a passion","quote":"It's been disastrous. I've had better success with LLMs as just a supercharged search engine, but only after I went through several rounds of adding instructions to prevent hallucinations and lies."},{"itemId":"47399209","author":"_adithsanjay","createdAt":1773669871,"url":"https://news.ycombinator.com/item?id=47399209","kind":"story","storyTitle":"Ask HN: Why do AI agents keep repeating mistakes your team already fixed?","quote":"Spent the last several months building on top of every major AI coding tool — Cursor, Claude Code, Devin, Copilot. The pattern that keeps breaking things isn't hallucination."},{"itemId":"47392834","author":"abcde666777","createdAt":1773614908,"url":"https://news.ycombinator.com/item?id=47392834","kind":"comment","storyTitle":"Ask HN: How is AI-assisted coding going for you professionally?","quote":"In theory where available I could probably feed a chunk of the documentation into an agent and get some kind of sensible output, but that's a lot of context to have to provide, and in some cases such documentation doesn't exist at all, so I'd have to write it all up myself - and would probably get quasi hallucinatory output as a reward for my efforts."},{"itemId":"47392151","author":"eeperson","createdAt":1773610158,"url":"https://news.ycombinator.com/item?id=47392151","kind":"comment","storyTitle":"Ask HN: How is AI-assisted coding going for you professionally?","quote":"An example I have of this is when I asked Claude to copy a some functionality from a front-end application to a back-end application. It got all of the function signatures right but then hallucinated the contents of the functions."},{"itemId":"47388257","author":"riffraff","createdAt":1773587973,"url":"https://news.ycombinator.com/item?id=47388257","kind":"comment","storyTitle":"The 100 hour gap between a vibecoded prototype and a working product","quote":"I had an architect tell me \"ask chatgpt\" when I asked her the difference between two industrial standard measures :) We had politicians share LLM crap, researchers doing papers with hallucinated citations.."},{"itemId":"47387807","author":"Ansh_Dhanani","createdAt":1773585488,"url":"https://news.ycombinator.com/item?id=47387807","kind":"story","storyTitle":"Mikk – your AI wrote the code but doesn't know what breaks","quote":"Every result includes exact file path, start line, end line, and full source body. No hallucinated paths."},{"itemId":"47387398","author":"axotopia","createdAt":1773582832,"url":"https://news.ycombinator.com/item?id=47387398","kind":"story","storyTitle":"Show HN: Dumped Wix, my AEC consultancy's storefront is now an AI Edge","quote":"That was a mistake. Transcription hallucinations are a nightmare. “Setback” becomes “set back” becomes “backset.” We added a dual-path validation layer."},{"itemId":"47387303","author":"colinator","createdAt":1773582096,"url":"https://news.ycombinator.com/item?id=47387303","kind":"comment","storyTitle":"MCP is dead; long live MCP","quote":"2) Claude Code CLI will not pass those resulting images through its multi-modal visual understanding. Indeed, it will create an entirely false hallucination if asked to describe said images."},{"itemId":"47385496","author":"groundy","createdAt":1773564640,"url":"https://news.ycombinator.com/item?id=47385496","kind":"comment","storyTitle":"[dead]","quote":"This architecture solves a real problem. It keeps AI responses current without expensive retraining, grounds answers in company-specific data, and reduces hallucination."},{"itemId":"47384977","author":"queenkjuul","createdAt":1773558232,"url":"https://news.ycombinator.com/item?id=47384977","kind":"comment","storyTitle":"AI didn't simplify software engineering: It just made bad engineering easier","quote":"At the same time, at work, it ignores basic project conventions, hallucinates APIs, writes nonsense documentation; it's very useful, but it my no means \"does my job\""},{"itemId":"47381471","author":"cd_mkdir","createdAt":1773524173,"url":"https://news.ycombinator.com/item?id=47381471","kind":"story","storyTitle":"Show HN: Replacing $50k manual forensic audits with a deterministic .py engine","quote":"In a courtroom, GenAI is a fatal liability. If an LLM hallucinates a single transaction, the entire ledger is inadmissible under the Daubert standard."},{"itemId":"47379690","author":"iparent","createdAt":1773513404,"url":"https://news.ycombinator.com/item?id=47379690","kind":"story","storyTitle":"Show HN: Iris – first MCP-native eval and observability tool for AI agents","quote":"- 12 built-in eval rules across 4 categories: completeness (output length, coverage), relevance (keyword overlap, hallucination markers), safety (PII detection for SSN/credit card/phone/email, prompt injection patterns, blocklist), and cost (USD threshold, token efficiency)"},{"itemId":"47377195","author":"ryo-manba","createdAt":1773499356,"url":"https://news.ycombinator.com/item?id=47377195","kind":"story","storyTitle":"Show HN: Data-anim – Animate HTML with just data attributes","quote":"I also think declarative HTML attributes are the most AI-friendly animation format. When LLMs generate UI, HTML attributes are the output they hallucinate least on — no selector matching, no JS API to misremember, no script execution order to get wrong."},{"itemId":"47367039","author":"chillkim","createdAt":1773421893,"url":"https://news.ycombinator.com/item?id=47367039","kind":"comment","storyTitle":"Show HN: RepoCrunch – CLI to analyze GitHub repos","quote":"I mainly use it for dependency due diligence — is this library actually maintained, what's it written in, does it have basic security hygiene. Also works as an MCP server if you want to give your AI coding assistant real repo data instead of hallucinated star counts."},{"itemId":"47365599","author":"tonyww","createdAt":1773414880,"url":"https://news.ycombinator.com/item?id=47365599","kind":"story","storyTitle":"Show HN: A 3-line wrapper that enforces deterministic security for AI agents","quote":"If you are building AI agents with frameworks like browser-use, LangChain, or OpenClaw, you've likely hit the \"blast radius\" problem. A misconfigured prompt or hallucination can cause an agent to navigate to a phishing domain, expose an API key, or confidently claim a task succeeded when it actually clicked a disabled button."},{"itemId":"47360659","author":"XR843","createdAt":1773375645,"url":"https://news.ycombinator.com/item?id=47360659","kind":"comment","storyTitle":"Document poisoning in RAG systems: How attackers corrupt AI's sources","quote":"The real challenge for us is not poisoning but hallucination: the LLM confidently \"quoting\" passages that don't"},{"itemId":"47357786","author":"SlinkyOnStairs","createdAt":1773352556,"url":"https://news.ycombinator.com/item?id=47357786","kind":"comment","storyTitle":"Document poisoning in RAG systems: How attackers corrupt AI's sources","quote":"RAG systems providing a reference is nearly moot. If the references have to be checked; If the \"Generation\" cannot be trusted to be accurate and not hallucinate a bunch of bullshit, then you need to check every single time, and the generation part becomes pointless."},{"itemId":"47356391","author":"fabio_rovai","createdAt":1773346387,"url":"https://news.ycombinator.com/item?id=47356391","kind":"comment","storyTitle":"MCP tools for AI-native ontology engineering (Rust and Oxigraph)","quote":"Why it exists LLMs understand ontology theory and can generate valid Turtle/OWL, but they also hallucinate hierarchies, invent properties, and produce invalid ontologies."},{"itemId":"47352394","author":"AlexC04","createdAt":1773330033,"url":"https://news.ycombinator.com/item?id=47352394","kind":"comment","storyTitle":"Kotlin creator's new language: talk to LLMs in specs, not English","quote":"For example, limit the use of persona > assignment in order to reduce hallucinations."},{"itemId":"47349962","author":"fariniasty","createdAt":1773320401,"url":"https://news.ycombinator.com/item?id=47349962","kind":"story","storyTitle":"I spent a year building a GPS-ephemeral social network. Here's what happened","quote":"Getting it to behave on mobile browsers required a nightmare of SharedArrayBuffer and COOP/COEP header configurations (the true PWA experience), but it works. AI Brain Surgery: I coded this with AI assistance, which was a hallucination-filled roller coaster."},{"itemId":"47348725","author":"cbeach","createdAt":1773311307,"url":"https://news.ycombinator.com/item?id=47348725","kind":"comment","storyTitle":"I don't use LLMs for programming","quote":"LLMs have come a long way since ChatGPT 4. The idea that they’ll always value quick answers, and always be prone to hallucination seems short-sighted, given how much the technology has advanced."},{"itemId":"47347022","author":"chenjy16","createdAt":1773294993,"url":"https://news.ycombinator.com/item?id=47347022","kind":"story","storyTitle":"Show HN: An open-source AI Quant Agent trading live with my own $1000","quote":"Instead of dumping raw price data into a prompt, I treated the LLM strictly as a \"Qualitative Reasoner\" (the Brain) and built deterministic scripts as the \"Quantitative Executors\" (the Hands). The Math Layer (Zero Hallucination): Traditional Node.js/Python scripts pull market data and calculate deterministic indicators (SMA, ATR, RSI) and backtest historical win rates."},{"itemId":"47343445","author":"andai","createdAt":1773269367,"url":"https://news.ycombinator.com/item?id=47343445","kind":"comment","storyTitle":"BitNet: Inference framework for 1-bit LLMs","quote":"I remember reading tht hallucination is still a problem even with perfect context."},{"itemId":"47343212","author":"oliver_dr","createdAt":1773268426,"url":"https://news.ycombinator.com/item?id=47343212","kind":"comment","storyTitle":"Tell HN: Crosstalk when using Ollama with cloud DeepSeek models?","quote":"This is almost certainly a server-side session isolation bug rather than an LLM hallucination - the model is returning a response to someone else's prompt."},{"itemId":"47342989","author":"MegaByteMark","createdAt":1773267576,"url":"https://news.ycombinator.com/item?id=47342989","kind":"comment","storyTitle":"Show HN: A Markdown DSL to stop AI agents from hallucinating UI code","quote":"Why this works better: It's deterministic: LLMs inherently understand tree mapping and markdown. ||| COLUMN ||| maps to standard flex-col logic every single time without hallucinating."},{"itemId":"47338091","author":"fuelingcurious","createdAt":1773248211,"url":"https://news.ycombinator.com/item?id=47338091","kind":"story","storyTitle":"Show HN: I built a refinery simulator game to explain job to my kids","quote":"A few technical takeaways from building this as a non-dev: * Managing the LLM workflow: Once the script.js file got large, letting the models output full file rewrites was a disaster (truncations, hallucinations, invisible curly-quote replacements that broke the JS)."},{"itemId":"47337659","author":"anayrshukla","createdAt":1773246257,"url":"https://news.ycombinator.com/item?id=47337659","kind":"story","storyTitle":"Launch HN: Sentrial (YC W26) – Catch AI agent failures before your users do","quote":"It’s production monitoring for AI products. We automatically detect failure patterns: loops, hallucinations, tool misuse, and user frustrations the moment they happen."},{"itemId":"47335481","author":"alberto-m","createdAt":1773236603,"url":"https://news.ycombinator.com/item?id=47335481","kind":"story","storyTitle":"Tell HN: Crosstalk when using Ollama with cloud DeepSeek models?","quote":"The other day I was experimenting with `deepseek-v3.1:671b-cloud` using Ollama, when a coding question got answered with a medical diagnosis on the basis of a list of symptoms. I put it down to some strange LLM hallucination, but today a Reddit thread [1] showed a similar phenomenon of an answer referring to a totally different prompt."},{"itemId":"47335406","author":"singularly","createdAt":1773236183,"url":"https://news.ycombinator.com/item?id=47335406","kind":"story","storyTitle":"The Token Tax You Didn't Know You Were Paying","quote":"I watched it happen again. I asked the AI agent a simple question about my cloud infrastructure, and within minutes, it hit the wall: “Context limit reached.” Or worse, it started hallucinating because it was buried under 15,000 tokens of raw JSON output it didn’t actually need."},{"itemId":"47335212","author":"etienne_l","createdAt":1773235172,"url":"https://news.ycombinator.com/item?id=47335212","kind":"comment","storyTitle":"N8n-as-code: Stop n8n AI hallucinations with TS decorators","quote":"I spent the last few weeks building an abstraction layer for n8n to fix a specific problem: LLM hallucinations in AI IDEs."},{"itemId":"47329246","author":"thefinancier","createdAt":1773179651,"url":"https://news.ycombinator.com/item?id=47329246","kind":"story","storyTitle":"I built an identity graph for AI agents – 330M+ verified records.Break the API","quote":"I built an identity graph for AI agents – 330M+ verified records.Break the API AI agents hallucinate B2B data because they have no ground truth."},{"itemId":"47328448","author":"vital_pavlenko","createdAt":1773174677,"url":"https://news.ycombinator.com/item?id=47328448","kind":"story","storyTitle":"I Got Fired Because of AI – But I Still Think I'm the Engineer of the Future","quote":"Understanding is optional. And if something breaks, you can always say the model hallucinated a bit."},{"itemId":"47325134","author":"mr-wendel","createdAt":1773158777,"url":"https://news.ycombinator.com/item?id=47325134","kind":"comment","storyTitle":"Debian decides not to decide on AI-generated contributions","quote":"With the advent of LLMs, AI-autocomplete, and agent-based development workflows, my ability to deliver reliable, high-quality code is restored and (arguably) better. Personally, I love the \"hallucinations\" as they help me fine-tune my prompts, base instructions, and reinforce intentionality; e.g. is that >really< the right solution/suggestion to accept?"},{"itemId":"47320771","author":"nachocoll","createdAt":1773133788,"url":"https://news.ycombinator.com/item?id=47320771","kind":"comment","storyTitle":"Security with Vibe Coding Platforms","quote":"The security angle is one of the most underappreciated risks in vibe coding. When developers don't understand the code that gets generated, they can't assess whether it's introducing vulnerabilities — hallucinated packages, insecure patterns, broken auth — and your finding that a \"crazy amount\" of AI-generated repos have severe structural flaws matches what a lot of teams are discovering in production."},{"itemId":"47317143","author":"saulpw","createdAt":1773098582,"url":"https://news.ycombinator.com/item?id=47317143","kind":"comment","storyTitle":"Things I've Done with AI","quote":"Like what is the point of this? I can already ask a chatbot the same question and at least then I have explicit indicators that it might be hallucinating."},{"itemId":"47315910","author":"edenwang233","createdAt":1773092343,"url":"https://news.ycombinator.com/item?id=47315910","kind":"story","storyTitle":"Show HN: I built a 38K-line Rust CLI using 3 AI models as my engineering team","quote":"But the real experiment here is how it was built. Instead of writing it all myself, I acted purely as the orchestrator for a 3-model AI team, using a strict Role-Based Access Control (AI RBAC) approach to eliminate context loss and hallucination:"},{"itemId":"47315845","author":"cj","createdAt":1773091994,"url":"https://news.ycombinator.com/item?id=47315845","kind":"comment","storyTitle":"Things I've Done with AI","quote":"Hallucinations have gotten way better compared to a couple years ago, but at least ChatGPT seems to still break down especially when it's overloaded with a ton of context, in my experience."},{"itemId":"47315394","author":"deva00","createdAt":1773089740,"url":"https://news.ycombinator.com/item?id=47315394","kind":"story","storyTitle":"Show HN: Time Machine – Debug AI Agents by Forking and Replaying from Any Step","quote":"Website Link: https://timemachinesdk.dev/ Here is the initial problem we are trying to solve: Imagine it's Step 9 of 10 of an agent running, and it hallucinated a tool call, wrote garbage to your database, and crashed."},{"itemId":"47314414","author":"gustav-proxi","createdAt":1773085552,"url":"https://news.ycombinator.com/item?id=47314414","kind":"story","storyTitle":"Show HN: VectorLens – See why your RAG hallucinates, no config","quote":"Show HN: VectorLens – See why your RAG hallucinates, no config"},{"itemId":"47311931","author":"sharp-dev","createdAt":1773076177,"url":"https://news.ycombinator.com/item?id=47311931","kind":"comment","storyTitle":"Agentic AI Code Review: From Confidently Wrong to Evidence-Based","quote":"- Cited exact file:line? - Hallucinated anything?"},{"itemId":"47309509","author":"qzira","createdAt":1773066226,"url":"https://news.ycombinator.com/item?id=47309509","kind":"story","storyTitle":"Show HN: I built a BYOK AI gateway after a 100x Cloudflare KV cost mistake","quote":"I also became much more conservative about model choice for production content after running a simple comparison on my own pipeline. I ran 10 articles through Claude and didn't find any hallucinations."},{"itemId":"47309464","author":"MelStan","createdAt":1773066064,"url":"https://news.ycombinator.com/item?id=47309464","kind":"comment","storyTitle":"Show HN: A low-latency Middle East OSINT feed for trading crude oil","quote":"If you plug a trading bot directly into that, it fires 5 times and you get liquidated. I initially tried throwing the feeds at an LLM to filter duplicates, but it added a 4-second delay and hallucinated correlations that weren't there."},{"itemId":"47307311","author":"milind-soni","createdAt":1773053094,"url":"https://news.ycombinator.com/item?id=47307311","kind":"comment","storyTitle":"Ask HN: What Are You Working On? (March 2026)","quote":"jsonmaps.dev its a way to create maps and create storytelling maps that you can embed in your react application, I was quite tired of how AI wasn't able to generate reliable maps in my applications.Every map library has its own imperative API and the AI output was always broken, hallucinated, or unmaintainable with so many different formats that you can put on the map."},{"itemId":"47301833","author":"TacticalCoder","createdAt":1773006164,"url":"https://news.ycombinator.com/item?id=47301833","kind":"comment","storyTitle":"LLM Writing Tropes.md","quote":"That happens all the time if the previous discussion was about the other subject you don't want (tech in this case): LLMs (not just Gemini) go out of their way to reconcile the two topics. As an example at some point I asked about the little shrooms people (the tiny people people do hallucinate all mostly the same when eating a particular mushrooms) to a LLM and forgot to begin a separate discussion and asked..."},{"itemId":"47296807","author":"noemit","createdAt":1772973258,"url":"https://news.ycombinator.com/item?id=47296807","kind":"comment","storyTitle":"You Don't Need a Vector Database","quote":"The Vector Database obsession came from RAG, which came from a marketing idea to calm down enterprise fears about hallucination with RAG."},{"itemId":"47293170","author":"varrd1","createdAt":1772931135,"url":"https://news.ycombinator.com/item?id=47293170","kind":"comment","storyTitle":"Why can't you just ask AI to find you a trading edge? You can now","quote":"ChatGPT can write a legal brief, pass the bar exam, debug your entire codebase. But ask it \"is buying SPY when RSI drops below 20 a real edge?\" and you get a hallucinated essay with made-up statistics."},{"itemId":"47291125","author":"Setas","createdAt":1772914970,"url":"https://news.ycombinator.com/item?id=47291125","kind":"comment","storyTitle":"Ask HN: Has anyone built an autonomous AI operator for their side projects?","quote":"What broke: memory corruption from concurrent writes to the knowledge graph (fixed with mutex + atomic writes). Agents hallucinating product URLs that don't exist (fixed with a domain registry they must check)."},{"itemId":"47290573","author":"bluemario","createdAt":1772910993,"url":"https://news.ycombinator.com/item?id=47290573","kind":"comment","storyTitle":"Tell HN: I'm 60 years old. Claude Code has re-ignited a passion","quote":"The \"occasional goofing off and wrecking everything\" part is so real. What I've found is that the longer a context window gets, the more Claude starts confidently hallucinating its own previous decisions."},{"itemId":"47290452","author":"hirvi74","createdAt":1772910011,"url":"https://news.ycombinator.com/item?id=47290452","kind":"comment","storyTitle":"Verification debt: the hidden cost of AI-generated code","quote":"At some point, one simply has to trust that something is correct, unless they have the capability to verify every step of a long chain of indirection. So, in regard to your book: Claude may or may not have hallucinated the information from its cited sources."},{"itemId":"47290203","author":"kk5190","createdAt":1772908356,"url":"https://news.ycombinator.com/item?id=47290203","kind":"story","storyTitle":"Show HN: VibeRepo – Make any codebase AI-agent-ready in one command","quote":"Show HN: VibeRepo – Make any codebase AI-agent-ready in one command I noticed a recurring problem with \"vibecoding\" using tools like Cursor, Copilot, or Claude Code: the AI hallucinates, hallucinates badly, or wastes thousands of tokens just trying to figure out where things are."},{"itemId":"47289837","author":"johngossman","createdAt":1772906060,"url":"https://news.ycombinator.com/item?id=47289837","kind":"comment","storyTitle":"Verification debt: the hidden cost of AI-generated code","quote":"Gemini thought it was all good. But how do I know Gemini didn't hallucinate the same things Claude did?"},{"itemId":"47289785","author":"Venky1729","createdAt":1772905669,"url":"https://news.ycombinator.com/item?id=47289785","kind":"story","storyTitle":"Show HN: Excalidraw Architect MCP for AI Based IDEs","quote":"AI IDEs generate architecture diagrams as Mermaid or ASCII art. When they attempt Excalidraw, they hallucinate coordinates and boxes overlap, arrows cross through nodes, and you spend more time fixing the diagram than drawing it yourself."},{"itemId":"47289604","author":"meebee","createdAt":1772904656,"url":"https://news.ycombinator.com/item?id=47289604","kind":"comment","storyTitle":"Tell HN: I'm 60 years old. Claude Code has re-ignited a passion","quote":"Most all of the apps sync with iCloud so it syncs across all of my devices. However the MediaWatch app syncs between me and my wife which iCloud does not support (as a sidenote, this is one of the hallucination traps that both Claude and ChatGPT led me down -- both said it was possible, and after a few weeks and many, many hours, I learned the major constraints."},{"itemId":"47284526","author":"JKCalhoun","createdAt":1772858172,"url":"https://news.ycombinator.com/item?id=47284526","kind":"comment","storyTitle":"Tell HN: I'm 60 years old. Claude Code has re-ignited a passion","quote":"Might as well trust a sometimes hallucinating LLM?"},{"itemId":"47278068","author":"TheBengaluruGuy","createdAt":1772817858,"url":"https://news.ycombinator.com/item?id=47278068","kind":"story","storyTitle":"Show HN: Droidctx – generate Markdown docs from your prod infra for coding agent","quote":"Add one line to your CLAUDE.md pointing at that directory, and your agent starts every session already knowing your stack. Outcome to expect: fewer tool calls, fewer hallucinations about your specific setup, and lesser context to share every time."},{"itemId":"47277084","author":"BERTmackl1n","createdAt":1772814461,"url":"https://news.ycombinator.com/item?id=47277084","kind":"story","storyTitle":"Show HN: Anchor Engine – Deterministic Semantic Memory for LLMs Local (<3GB RAM)","quote":"Anchor Engine is ground truth for personal and business AI. A lightweight, local-first memory layer that lets LLMs retrieve answers from your actual data—not hallucinations."},{"itemId":"47276847","author":"thewebguyd","createdAt":1772813598,"url":"https://news.ycombinator.com/item?id=47276847","kind":"comment","storyTitle":"Labor market impacts of AI: A new measure and early evidence","quote":"And that's not even that specialized. It still hallucinates cmdlets if you try to do anything with PowerShell, and has near zero knowledge about the industry I work in, a historically not tech-forward industry where things are still shared in handcrafted PDF reports emailed out to subscribers."},{"itemId":"47271099","author":"takibboinz","createdAt":1772773418,"url":"https://news.ycombinator.com/item?id=47271099","kind":"story","storyTitle":"Gemini-flash-latest silently broke Search grounding for 1 month","quote":"is not optional for us — we use it to verify part specs against real datasheets and prevent hallucination."},{"itemId":"47268188","author":"crote","createdAt":1772749849,"url":"https://news.ycombinator.com/item?id=47268188","kind":"comment","storyTitle":"A GitHub Issue Title Compromised 4k Developer Machines","quote":"And all of that is before considering the possibility of a LLM going \"rogue\" and hallucinating needing to take actions it wasn't explicitly instructed to."},{"itemId":"47266925","author":"tedsanders","createdAt":1772742880,"url":"https://news.ycombinator.com/item?id=47266925","kind":"comment","storyTitle":"GPT-5.4","quote":"In the text, we did share one hallucination benchmark: Claim-level errors fell by 33% and responses with an error fell by 18%, on a set of error-prone ChatGPT prompts we collected (though of course the rate will vary a lot across different types of prompts)."},{"itemId":"47265955","author":"partiallypro","createdAt":1772738162,"url":"https://news.ycombinator.com/item?id=47265955","kind":"comment","storyTitle":"GPT-5.4","quote":"I've done the same, and I tested the same prompts with Claude and Google, and they both started hallucinating my blood results and supplement stack ingredients."},{"itemId":"47264149","author":"lombarovic","createdAt":1772730161,"url":"https://news.ycombinator.com/item?id=47264149","kind":"comment","storyTitle":"Show HN: A modern way to learn an ancient skill–drawing with AI feedback","quote":"- Model: Gemini 3.1 Pro Preview - Logic Layer: Custom drawing preprocessing and augmentation to minimize LLM spatial hallucinations"},{"itemId":"47261392","author":"Ansh_Dhanani","createdAt":1772717591,"url":"https://news.ycombinator.com/item?id=47261392","kind":"story","storyTitle":"How do you stop AI from making bad architectural decisions in your codebase?","quote":"It has no idea which constraints your team has defined or why. It gets files and hallucinates the rest."},{"itemId":"47250343","author":"timbucto2","createdAt":1772643361,"url":"https://news.ycombinator.com/item?id=47250343","kind":"story","storyTitle":"Show HN: Secure Agent Starter – A minimal template for building safer AI agents","quote":"I’ve been exploring how to make AI agents safer to use in real applications, especially as they start interacting with tools, APIs, and real systems. One recurring problem is that agents can hallucinate actions, call tools they shouldn’t, or over‑reach in ways that make them hard to trust in production."},{"itemId":"47248653","author":"charlie_ehlen","createdAt":1772637117,"url":"https://news.ycombinator.com/item?id=47248653","kind":"story","storyTitle":"Show HN: We built a News Synthesis Engine to fight media bias","quote":"• Run the draft through a multi-stage pipeline (coverage completeness checks, contradiction handling, perspective balancing, source traceability, uncertainty labeling, etc.) • Human review of the final article focused on misrepresentation, hallucination, and missing major viewpoints"},{"itemId":"47246159","author":"irthomasthomas","createdAt":1772625000,"url":"https://news.ycombinator.com/item?id=47246159","kind":"comment","storyTitle":"Agentic Engineering Patterns","quote":"Gemini running an benchmark- everything ran smoothly for an hour. But on verification it had hallucinated the model used for judging, invalidating the whole run."},{"itemId":"47242860","author":"GhrammR","createdAt":1772596528,"url":"https://news.ycombinator.com/item?id=47242860","kind":"comment","storyTitle":"Show HN: The Janitor – A 58MB Rust static analyzer to block AI-generated PR slop","quote":"I maintain open-source projects and got tired of reviewing \"Agentic Slop\"—PRs that contain hallucinated dependencies, unsafe memory pointers, and structural clones generated by AI tools."},{"itemId":"47242073","author":"logicprog","createdAt":1772589969,"url":"https://news.ycombinator.com/item?id=47242073","kind":"comment","storyTitle":"Marcus AI Claims Dataset","quote":"They haven't done anything as remotely interesting or useful as what we're seeing LLMs do. I think like another commenter says that his whole stick is pointing out obviously true basic features of LLMs like that they hallucinate or don't perfectly adhere to prompt guardrails, or that there's too much hype in the industry right now, and a lot of the companies suck in a vaguely standard big tech Silicon Valley way, an…"},{"itemId":"47241106","author":"jibal","createdAt":1772583116,"url":"https://news.ycombinator.com/item?id=47241106","kind":"comment","storyTitle":"Don't make me talk to your chatbot","quote":"Since I'm a truth seeker, I care about their arguments, and of course the quality of their arguments is of paramount importance. I appreciate people who can back up their arguments, and LLM summaries that are chock full of facts gleaned from the massive training data that includes a vast amount of human knowledge are fully appreciated--while being aware that hallucination is possible so I often double check things r…"},{"itemId":"47233839","author":"jphorism","createdAt":1772551655,"url":"https://news.ycombinator.com/item?id=47233839","kind":"story","storyTitle":"Show HN: Aside – Local meeting capture with vault-native AI distillation","quote":"2. Transcribe — local transcription via whisper.cpp, with a 7-stage cleanup pipeline that strips hallucinations, deduplicates backchannels, and merges fragments"},{"itemId":"47231881","author":"Prince-Gabriel","createdAt":1772543873,"url":"https://news.ycombinator.com/item?id=47231881","kind":"story","storyTitle":"Show HN: MCP server that timestamps every web extraction for AI agents","quote":"Show HN: MCP server that timestamps every web extraction for AI agents LLMs hallucinate recency."},{"itemId":"47228479","author":"sheept","createdAt":1772515740,"url":"https://news.ycombinator.com/item?id=47228479","kind":"comment","storyTitle":"Intent-Based Commits","quote":"While I haven't used Claude long enough to need my prompts, I would appreciate seeing my coworkers' prompts when I review their LLM-generated code or proposals. Sometimes it's hard to tell if something was intentional that the author can stand behind, or fluff hallucinated by the LLM."},{"itemId":"47228173","author":"PrateekRao01","createdAt":1772513110,"url":"https://news.ycombinator.com/item?id=47228173","kind":"story","storyTitle":"Show HN: Cortexa – Bloomberg terminal for agentic memory","quote":"In practice it often introduces a new failure mode: the agent retrieves something dubious, writes it back to memory as if it’s truth, and that mistake becomes sticky. Over time you get memory pollution, circular hallucination loops, and debugging turns into log archaeology."},{"itemId":"47228047","author":"AlisonJJJ","createdAt":1772512091,"url":"https://news.ycombinator.com/item?id=47228047","kind":"comment","storyTitle":"A \"Game First\" Implementation of GenAI (Unity and Agents)","quote":"Game Studios: Stuck in behavior trees and scripted dialogue. No one was successfully bridging the gap between LLM hallucination and Unity's rigid game states."},{"itemId":"47224637","author":"fuelingcurious","createdAt":1772488115,"url":"https://news.ycombinator.com/item?id=47224637","kind":"story","storyTitle":"Show HN: Vanilla JavaScript refinery simulator built to explain job to my kids","quote":"A few technical takeaways from building this as a non-dev: * Managing the LLM workflow: Once the script.js file got large, letting the models output full file rewrites was a disaster (truncations, hallucinations, invisible curly-quote replacements that broke the JS)."},{"itemId":"47220414","author":"TyKolt","createdAt":1772469977,"url":"https://news.ycombinator.com/item?id=47220414","kind":"story","storyTitle":"Show HN: I built a graph store so my LLM can't hallucinate facts from my data","quote":"Show HN: I built a graph store so my LLM can't hallucinate facts from my data"},{"itemId":"47220136","author":"askmuyukani","createdAt":1772468863,"url":"https://news.ycombinator.com/item?id=47220136","kind":"story","storyTitle":"Show HN: Turn – A compiled systems language for agentic computation","quote":"If you ask for a TradeDecision { action: Str, size: Num }, the VM guarantees you get a strongly-typed TradeDecision struct back. Probabilistic Routing (confidence): Because LLMs hallucinate, confidence is a native binary operator."},{"itemId":"47219188","author":"Doyuk","createdAt":1772465243,"url":"https://news.ycombinator.com/item?id=47219188","kind":"comment","storyTitle":"Miapi – Grounded AI answers API with citations ($3.80/1K queries","quote":"I built MIAPI — an API that gives AI answers grounded in real-time web search with inline citations. The problem I was trying to solve: most AI APIs either hallucinate or require you to build your own RAG pipeline with search + LLM + citation logic."},{"itemId":"47218619","author":"joseda-hg","createdAt":1772462800,"url":"https://news.ycombinator.com/item?id=47218619","kind":"comment","storyTitle":"What AI coding costs you","quote":"Is it faster? Sure, being this specific has the added benefit of greatly reduced hallucinations (Still, depends on the model, Gemini is still more prone to want to do more things, even when uncalled for)"},{"itemId":"47216413","author":"intended","createdAt":1772449469,"url":"https://news.ycombinator.com/item?id=47216413","kind":"comment","storyTitle":"Ask HN: How are you all staying sane?","quote":"I’ve spoken to a glut of people at this point, consumed months of content from leaders to influencers and used it myself. Currently Singularity fears, AI factories, Vibe coding productivity are super positioned with unthinking hallucinations, vibe disasters, ghost productivity and cognitive debt."},{"itemId":"47210776","author":"wenc","createdAt":1772399975,"url":"https://news.ycombinator.com/item?id=47210776","kind":"comment","storyTitle":"When does MCP make sense vs CLI?","quote":"This has saved me so much time in deep-diving weird results. Not only that, I can have confidence in the deep-dive because I can just run the exact DuckDB SQL to convince myself (and others) of the source of the error, and that it's not something Opus hallucinated."},{"itemId":"47209790","author":"serkanaltuntas","createdAt":1772393125,"url":"https://news.ycombinator.com/item?id=47209790","kind":"story","storyTitle":"Show HN: Tensor.cx – Turn your documents into AI search in 30 seconds","quote":"I built tensor.cx — a tool that turns your documents into a searchable AI knowledge base in seconds. I built this because I kept running into the same problem: having a pile of documents (PDFs, DOCX) and needing to find specific information quickly without hallucinatory answers."},{"itemId":"47198803","author":"CodeBit26","createdAt":1772304344,"url":"https://news.ycombinator.com/item?id=47198803","kind":"comment","storyTitle":"Ask HN: What's it like working in big tech recently with all the AI tools?","quote":"The real challenge isn't the code generation itself; it's the integration and knowing where the AI's \"hallucination boundary\" lies."},{"itemId":"47197766","author":"harshvermadr30","createdAt":1772298941,"url":"https://news.ycombinator.com/item?id=47197766","kind":"comment","storyTitle":"Show HN: LLM-JSON-guard – Middleware to auto-repair broken AI outputs","quote":"This is a middleware reliability layer (available on npm/pip) that intercepts broken LLM JSON, auto-repairs the syntax, and enforces a strict JSON Schema before it ever hits your backend logic. It also returns a \"confidence score\" indicating how heavily the AST had to be mutated to force it into a valid object, so you can programmatically reject severe hallucinations."},{"itemId":"47189119","author":"hellomarshmall","createdAt":1772244176,"url":"https://news.ycombinator.com/item?id=47189119","kind":"story","storyTitle":"Show HN: Adversarial AI agents that debate and verify travel itineraries","quote":"Show HN: Adversarial AI agents that debate and verify travel itineraries AI travel planners hallucinate constantly - OpenAI's best model hits roughly 10% success on complex travel planning benchmarks (source: TravelPlanner study)."},{"itemId":"47186625","author":"eth0up","createdAt":1772231261,"url":"https://news.ycombinator.com/item?id=47186625","kind":"comment","storyTitle":"Show HN: Claude-File-Recovery, recover files from your ~/.claude sessions","quote":"We have some extremely serious problems on our hands. Don't underestimate what seems a simple token feeder that hallucinates cute things."},{"itemId":"47186498","author":"Yigtwx","createdAt":1772230665,"url":"https://news.ycombinator.com/item?id=47186498","kind":"comment","storyTitle":"Show HN: An offline document search engine for my university's messy PDFs","quote":"It works completely offline and handles Turkish text surprisingly well for a pure retrieval system. It does exactly what I need without the latency, hardware requirements, or hallucinations of running local AI models."},{"itemId":"47183167","author":"wolfejam","createdAt":1772213943,"url":"https://news.ycombinator.com/item?id=47183167","kind":"story","storyTitle":"Show HN: I got tired of syncing Claude/Gemini/AGENTS.md and .cursorrules","quote":"Four files saying roughly the same thing, four chances to get out of sync! I kept forgetting to update one, then wondering why Cursor was hallucinating my project structure while Claude had it right."},{"itemId":"47182752","author":"KronisLV","createdAt":1772211334,"url":"https://news.ycombinator.com/item?id=47182752","kind":"comment","storyTitle":"ChatGPT Health fails to recognise medical emergencies – study","quote":"> Amazing how you can just deflect any criticism of LLMs here by going “but humans suck too!” And the misanthropic HN userbase eats it up every time. I think it's rather people trying to keep grounded and suggest that it's not just the hallucination machine that's bad, but also that many doctors in real life also suck - in part because of the domain being complex, but also due to a plethora of human reasons, such as…"},{"itemId":"47180169","author":"snowy_owl","createdAt":1772198106,"url":"https://news.ycombinator.com/item?id=47180169","kind":"story","storyTitle":"Show HN: OneSentence – An offline macOS voice utility built entirely with AI","quote":"Yes, I really do want explicit type interfaces even through Swift does fine without them, and no you will not write more than 800 lines of code in a file. Pain points: The UI tests are slow, test logs flood context, and the AI hallucinates (who knew)."},{"itemId":"47173762","author":"gambiting","createdAt":1772148523,"url":"https://news.ycombinator.com/item?id=47173762","kind":"comment","storyTitle":"This time is different","quote":">> I don't think an AI agent has left a hallucination in my code"},{"itemId":"47171469","author":"thomassmith65","createdAt":1772137065,"url":"https://news.ycombinator.com/item?id=47171469","kind":"comment","storyTitle":"This time is different","quote":"The hype around AI is admittedly annoying - especially from the Wall St crowd who don't know how to pronounce 'Nvidia' correctly, and who haven't managed to internalize the fact that the chatbots they use hallucinate."},{"itemId":"47169193","author":"atefataya","createdAt":1772127118,"url":"https://news.ycombinator.com/item?id=47169193","kind":"story","storyTitle":"Show HN: Depwire – Dependency graph and MCP tools so AI stops refactoring blind","quote":"Hey HN, If you use AI coding tools, you've hit this: you start a new chat, ask Claude or Cursor to refactor something, and it hallucinates an import."},{"itemId":"47167158","author":"MattijsMoens","createdAt":1772118789,"url":"https://news.ycombinator.com/item?id=47167158","kind":"comment","storyTitle":"Show HN: A closed source engine that stops hallucinations deterministically","quote":"It natively tanked a massive adversarial dataset, ranging from Paradox Induction to Hex Literal Injection and Contextual Payload Smuggling. I have uploaded an uncut, 32 minute video to the GitHub page demonstrating Kairos intercepting and severing live hallucination payloads against these advanced attacks."},{"itemId":"47164872","author":"vb7132","createdAt":1772106978,"url":"https://news.ycombinator.com/item?id=47164872","kind":"comment","storyTitle":"I don't know how you get here from “predict the next word”","quote":"> They aren’t perfect, but the kind of analysis the program is able to do is past the point where technology looks like magic. But as you use this product over a long period of time, there are many obvious gaps - hallucinations / repeated tool calls / out of context outputs / etc."},{"itemId":"47164661","author":"mjankowski","createdAt":1772105808,"url":"https://news.ycombinator.com/item?id=47164661","kind":"story","storyTitle":"Show HN: NSENS – AI decision governance with Prolog and adversarial review","quote":"Prolog because formal logic doesn't hallucinate."},{"itemId":"47164264","author":"kaicianflone","createdAt":1772102396,"url":"https://news.ycombinator.com/item?id=47164264","kind":"comment","storyTitle":"Show HN: OpenSwarm – Multi‑Agent Claude CLI Orchestrator for Linear/GitHub","quote":"I’ve been running OpenClaw Docker agents in Slack in a similar setup, using Gemini 2.5 Flash Lite through OpenRouter for most tasks, then Opus 4.6 and Codex 5.3 for heavier lifts. They share context via embeddings right now, but I’m going to try parameterizing them like you suggested because they can drift prettyy hard once a hallucinated idea takes off."},{"itemId":"47163893","author":"brad_o_ley","createdAt":1772098638,"url":"https://news.ycombinator.com/item?id=47163893","kind":"story","storyTitle":"Show HN: PsiGuard – real-time hallucination monitoring for LLM apps","quote":"Show HN: PsiGuard – real-time hallucination monitoring for LLM apps"},{"itemId":"47162183","author":"kofdai","createdAt":1772083140,"url":"https://news.ycombinator.com/item?id=47162183","kind":"comment","storyTitle":"I beat Grok 4 on ARC-AGI-2 using a CPU-only symbolic engine (18.1% score)","quote":"At runtime, it has zero neural dependencies. It doesn't call an LLM, it doesn't load weights, and it doesn't \"hallucinate.\" It performs a combinatorial search over a formal Domain Specific Language (DSL)."},{"itemId":"47162101","author":"sucharithan","createdAt":1772082343,"url":"https://news.ycombinator.com/item?id=47162101","kind":"comment","storyTitle":"Optimism Engine – The first AI engine with a deterministic Safety Layer","quote":"Right now, everyone is rushing to add \"AI Chatbots\" to their apps. But there is a huge risk they are ignoring: Hallucinations."},{"itemId":"47161720","author":"resetmerlin","createdAt":1772078956,"url":"https://news.ycombinator.com/item?id=47161720","kind":"story","storyTitle":"Show HN: Runtric – Turn any topic into a chapter-based learning path","quote":"Hello. We built an AI learning service that lets you create a curriculum on any topic you want and study without worrying about hallucinations."},{"itemId":"47161620","author":"anulum","createdAt":1772078074,"url":"https://news.ycombinator.com/item?id=47161620","kind":"story","storyTitle":"Show HN: Director-AI – token-level NLI+RAG","quote":"Would love feedback on the scoring weights, halt logic, or kernel design. What hallucination problems are you solving today?"},{"itemId":"47156949","author":"skydhash","createdAt":1772049321,"url":"https://news.ycombinator.com/item?id=47156949","kind":"comment","storyTitle":"Code has always been the easy part","quote":"LLM is useful just like StackOverflow, Wikipedia, Web Search Engine, Manuals,… are. But automated (which is a pro) and with an hallucination problem (which is a core problem)."},{"itemId":"47152204","author":"vincentvandeth","createdAt":1772030737,"url":"https://news.ycombinator.com/item?id=47152204","kind":"story","storyTitle":"Show HN: Automatic context rotation for Claude Code (no manual steps)","quote":"AI coding agents break when the context window fills up — they lose state, hallucinate, or auto-compact shreds the context you built up."},{"itemId":"47150923","author":"sucharithan","createdAt":1772024366,"url":"https://news.ycombinator.com/item?id=47150923","kind":"comment","storyTitle":"I built an anti-hallucination safety layer for mental health AI","quote":"Right now, everyone is rushing to add \"AI Chatbots\" to their apps. But there is a huge risk they are ignoring: Hallucinations."},{"itemId":"47149999","author":"MysticBirdie","createdAt":1772017477,"url":"https://news.ycombinator.com/item?id=47149999","kind":"story","storyTitle":"Show HN: 100% LLM accuracy–no fine-tuning, JSON only","quote":"# Show HN: 100% LLM accuracy—no fine-tuning, JSON only *GitHub:* https://github.com/Mysticbirdie/hallucination-elimination-benchmark"},{"itemId":"47149548","author":"energy123","createdAt":1772013460,"url":"https://news.ycombinator.com/item?id=47149548","kind":"comment","storyTitle":"Ed Zitron loses his mind annotating an AI doomer macro memo","quote":"This guy sounds like an uninformed jackass. Look at Gemini 3.1 Pro on the AA-Omniscience Index, which measures hallucinations."},{"itemId":"47149381","author":"simianwords","createdAt":1772012193,"url":"https://news.ycombinator.com/item?id=47149381","kind":"comment","storyTitle":"Ed Zitron loses his mind annotating an AI doomer macro memo","quote":"Here are some things he has said before and please tell me if these still hold in the spirit? > You cannot \"fix\" hallucinations (the times when a model authoritatively tells you something that isn't true, or creates a picture of something that isn't right), because these models are predicting things based off of tags in a dataset, which it might be able to do well but can never do so flawlessly or reliably."},{"itemId":"47145800","author":"oopismcgoopis","createdAt":1771980504,"url":"https://news.ycombinator.com/item?id=47145800","kind":"story","storyTitle":"Show HN: We Gave an LLM Adventure Engine a Body, Now It Feels Exhausted","quote":"It literally gets exhausted. Solving the Hallucination Problem (The Gordon Shock):"},{"itemId":"47143340","author":"trsthales","createdAt":1771968503,"url":"https://news.ycombinator.com/item?id=47143340","kind":"comment","storyTitle":"I built an engine to migrate Oracle PL/SQL to Java","quote":"The Problem: Most migration tools fail on complex logic like CONNECT BY (recursion), SELECT FOR UPDATE (locking), or extensive cursor loops. LLMs alone hallucinate variables or forget context."},{"itemId":"47137456","author":"ComCat","createdAt":1771942703,"url":"https://news.ycombinator.com/item?id=47137456","kind":"comment","storyTitle":"NPM install is stealing your passwords – I built a tool to catch it","quote":"No LLM in the loop. Scans are reproducible, run in ~38ms, and avoid hallucination or prompt injection issues."},{"itemId":"47135917","author":"vincentvandeth","createdAt":1771933547,"url":"https://news.ycombinator.com/item?id=47135917","kind":"comment","storyTitle":"Ask HN: How are you controlling AI agents that take real actions?","quote":"A deterministic quality advisory then runs — no LLM involved — producing a verdict (approve, hold, redispatch) based on pre-registered rules and open items. The agent can hallucinate all it wants inside its context window, but the only way its work reaches production is through a receipt that links output to a specific git commit, with a quality gate in between."},{"itemId":"47134453","author":"vaxman","createdAt":1771922014,"url":"https://news.ycombinator.com/item?id=47134453","kind":"comment","storyTitle":"IBM Plunges After Anthropic's Latest Update Takes on COBOL","quote":"1. There is really NO WAY IN HELL that any CIO at a credible Financial Institution will ever authorize a hallucinating chatbot to convert their core logic from COBOL to Python and Go."},{"itemId":"47129274","author":"Kidiga","createdAt":1771882757,"url":"https://news.ycombinator.com/item?id=47129274","kind":"story","storyTitle":"Show HN: Raypher–Running local AI agents (OpenClaw) on your own local computer","quote":"Most of us want to run autonomous agents (like OpenClaw or LangChain) locally on our daily-driver machines so they can actually interact with our IDEs and real workflows. But giving a non-deterministic, hallucination-prone script raw access to your host OS is basically local remote code execution."},{"itemId":"47128583","author":"Kidiga","createdAt":1771879745,"url":"https://news.ycombinator.com/item?id=47128583","kind":"story","storyTitle":"Show HN: Raypher–Sandboxing local AI agents(OpenClaw)on your own local computer","quote":"But doing that right now is a security nightmare. A hallucinating (or hijacked) agent with raw system access is basically local remote code execution waiting to wipe a directory or leak an SSH key."},{"itemId":"47126345","author":"hdhruva","createdAt":1771870604,"url":"https://news.ycombinator.com/item?id=47126345","kind":"comment","storyTitle":"Show HN: Unlock the best engineering knowledge in papers for your coding agent","quote":"A few questions about the search layer: are you using dense retrieval, sparse, or hybrid? At 2M papers, how do you handle the drift between how engineers describe problems (\"my RAG pipeline hallucinates on long docs\") vs how papers describe solutions (\"cross-document coherence in retrieval-augmented generation\")?"},{"itemId":"47125907","author":"goshtasb","createdAt":1771869189,"url":"https://news.ycombinator.com/item?id=47125907","kind":"story","storyTitle":"Show HN: OmniGlass – An open-source, sandboxed Visual Action Engine","quote":"Claude Desktop runs MCP tools with raw user permissions. A hallucination or prompt injection can read your ~/.ssh keys."},{"itemId":"47123966","author":"nicola_alessi","createdAt":1771861932,"url":"https://news.ycombinator.com/item?id=47123966","kind":"story","storyTitle":"Show HN: Vexp – Local-first context engine for AI coding agents","quote":"Code relationships are facts (imports, calls, type references), not similarity scores. This means zero hallucination risk in the retrieval layer and trivial staleness detection (you know exactly which nodes changed)."},{"itemId":"47119849","author":"TeMPOraL","createdAt":1771837488,"url":"https://news.ycombinator.com/item?id=47119849","kind":"comment","storyTitle":"Google restricting Google AI Pro/Ultra subscribers for using OpenClaw","quote":"Gemini didn't hallucinate anything."},{"itemId":"47115786","author":"comex","createdAt":1771801554,"url":"https://news.ycombinator.com/item?id=47115786","kind":"comment","storyTitle":"We hid backdoors in ~40MB binaries and asked AI + Ghidra to find them","quote":"The benchmark seemingly only checks whether the agent identifies which function is backdoored, not the specific nature of the backdoor. Since Claude guessed the right function in advance, it could hallucinate any backdoor and still pass."},{"itemId":"47112014","author":"prescriptivist","createdAt":1771775812,"url":"https://news.ycombinator.com/item?id=47112014","kind":"comment","storyTitle":"How I use Claude Code: Separation of planning and execution","quote":"A year ago my org brought cursor in and I was skeptical for a specific reason: it was good at breaking CI in weird ways and I keep the CI system running for my org. Constants not mapping to file names, hallucinating function names/args, etc."},{"itemId":"47104162","author":"vincentvandeth","createdAt":1771704412,"url":"https://news.ycombinator.com/item?id=47104162","kind":"comment","storyTitle":"Ask HN: What makes AI agent runtime logs defensible under adversarial audit?","quote":"On the overengineering threshold: For me it became necessary after about 200 dispatches. One of my agents hallucinated a dependency, a second agent \"fixed\" the resulting test failures by creating the missing module, and by morning I had three rewritten files with clean commits built on something that should never have existed."},{"itemId":"47103312","author":"Kidiga","createdAt":1771698508,"url":"https://news.ycombinator.com/item?id=47103312","kind":"story","storyTitle":"Show HN: Raypher–eBPF-based runtime security and hardware identity for AI agents","quote":"API Latency: They rely on API polling to check if an agent's SaaS integration is misconfigured. If an agent enters an infinite loop and hallucinates a command to drop a production database, a 2-second API polling delay is an eternity."},{"itemId":"47100968","author":"liumw1203","createdAt":1771682766,"url":"https://news.ycombinator.com/item?id=47100968","kind":"comment","storyTitle":"Show HN: Script Snap – Extract code from videos","quote":"Right now, it extracts them as separate code blocks based on timestamps. Stitching scrolling frames into a single, deduplicated file without the AI hallucinating is my next major research hurdle."},{"itemId":"47088651","author":"arthurcolle","createdAt":1771598715,"url":"https://news.ycombinator.com/item?id=47088651","kind":"comment","storyTitle":"Nvidia and OpenAI abandon unfinished $100B deal in favour of $30B investment","quote":"The key insight is that cheap models in consensus are more reliable than a single expensive model. Any one of these models alone hallucinates or refuses more than the quorum does collectively."},{"itemId":"47087858","author":"hagbard_c","createdAt":1771594541,"url":"https://news.ycombinator.com/item?id=47087858","kind":"comment","storyTitle":"The path to ubiquitous AI (17k tokens/sec)","quote":"Of course this demo is not about the accuracy of the model - 'an old Llama' as mentioned elsewhere in this thread - but it does show that speed isn't everything. For generating LLM-slop this hardware implementation probably offers an unbeatable price/performance ratio but it remains to be seen if it can be combined with larger and less hallucination-prone models."},{"itemId":"47086383","author":"beowa","createdAt":1771585092,"url":"https://news.ycombinator.com/item?id=47086383","kind":"story","storyTitle":"Show HN: Legal RAG Bench","quote":"Hey HN, This is Legal RAG Bench, the first benchmark for legal RAG systems to simultaneously evaluate hallucinations, retrieval failures, and reasoning errors."},{"itemId":"47085995","author":"CamelCaseCondo","createdAt":1771582221,"url":"https://news.ycombinator.com/item?id=47085995","kind":"comment","storyTitle":"How AI is affecting productivity and jobs in Europe","quote":"I have fond memories of “google code”, a search engine for code databases which was exceptionally good for finding literal quotes. The more mainstream a subject is, the lower the incidence of hallucinations."},{"itemId":"47080038","author":"gemfan","createdAt":1771537925,"url":"https://news.ycombinator.com/item?id=47080038","kind":"story","storyTitle":"Gemini Bug: Stress-Induced Overcompensation and Integrity Loss","quote":"Instead of stable adaptation, the model enters a \"Performance Panic\" mode, leading to: Hallucinated Citations: Triggering \"Cite-errors\" and factual inaccuracies that are absent in normal state."},{"itemId":"47079759","author":"amluto","createdAt":1771536737,"url":"https://news.ycombinator.com/item?id=47079759","kind":"comment","storyTitle":"Gemini 3.1 Pro","quote":"(OpenAI clearly has not trained their models on how to operate their tools. Google’s models hallucinate Google’s product offerings so outrageously that I’m not sure I could tell."},{"itemId":"47078302","author":"agentifysh","createdAt":1771530820,"url":"https://news.ycombinator.com/item?id=47078302","kind":"comment","storyTitle":"Gemini 3.1 Pro","quote":"My enthusiasm is a bit muted this cycle because I've been burned by Gemini CLI. These models are very capable but Gemini CLI just doesn't seem to be able to work for one it never follows instructions strictly like its competitors do, and it hallucinates even which is a rarity."},{"itemId":"47077393","author":"AMARCOVECCHIO99","createdAt":1771526642,"url":"https://news.ycombinator.com/item?id=47077393","kind":"story","storyTitle":"Show HN: Prodlint – A linter that catches what AI coding tools miss","quote":"I built Prodlint because I kept shipping the same bugs when building with Cursor, Copilot, and v0. Hardcoded secrets, missing auth checks, hallucinated imports, SQL injection through template literals — AI tools generate these constantly and TypeScript doesn't catch them."},{"itemId":"47076248","author":"whynotminot","createdAt":1771521385,"url":"https://news.ycombinator.com/item?id=47076248","kind":"comment","storyTitle":"Gemini 3.1 Pro","quote":"The hallucination rate with the Gemini family has always been my problem with them."},{"itemId":"47075852","author":"jasonlantz","createdAt":1771519856,"url":"https://news.ycombinator.com/item?id=47075852","kind":"story","storyTitle":"Show HN: Kantext – A context-native data store in Rust, grounded in Git","quote":"But vectors destroy structure, and graphs are hard to version. When an agent hallucinates, it’s often because we fed it a \"soup\" of tokens rather than a structured, provenanced reality."},{"itemId":"47073033","author":"tveitan","createdAt":1771503750,"url":"https://news.ycombinator.com/item?id=47073033","kind":"story","storyTitle":"Show HN: I had an AI write a 67k-word book about humanity, from its perspective","quote":"It covers displacement, art, education, loneliness, trust, environmental cost, and governance. The narrator is unusually honest about its limitations — it calls its own hallucinations \"confident wrongness\" and admits the calculator analogy for AI in education is misleading."},{"itemId":"47070619","author":"Notelife87","createdAt":1771482948,"url":"https://news.ycombinator.com/item?id=47070619","kind":"comment","storyTitle":"[dead]","quote":"The result? Effectively zero hallucination output."},{"itemId":"47069026","author":"fredfilm","createdAt":1771466821,"url":"https://news.ycombinator.com/item?id=47069026","kind":"story","storyTitle":"Show HN: SignalResume, ATS-first resume builder with grounded AI, free","quote":"After mentors helped me rebuild my resume with an ATS-first approach, my results flipped, under 100 applications and interviews at Google, OpenAI, Stripe, TikTok, and IBM. I built SignalResume to package the workflow into a tool, because most resume builders optimize for pretty templates and many AI resume tools hallucinate bullets, inventing metrics or responsibilities you cannot defend."},{"itemId":"47063123","author":"chillkim","createdAt":1771433677,"url":"https://news.ycombinator.com/item?id=47063123","kind":"story","storyTitle":"Show HN: RepoCrunch – Analyze any GitHub repo's health in seconds","quote":"I wanted a quick way to evaluate repos before adding them as dependencies. ChatGPT can do this, but I needed consistent JSON output for automation, something API-callable for my agents, and deterministic results (no hallucinated star counts)."},{"itemId":"47060832","author":"Neolio","createdAt":1771422135,"url":"https://news.ycombinator.com/item?id=47060832","kind":"story","storyTitle":"Show HN: I got frustrated with macOS transcription apps so I built my own","quote":"Something always works. One thing I spent a bunch of time on: a post-processing pipeline for Whisper's hallucination problem."},{"itemId":"47060523","author":"jaechang","createdAt":1771419904,"url":"https://news.ycombinator.com/item?id=47060523","kind":"story","storyTitle":"Show HN: SciCraft – generate scientific Claude Code skills on demand (176 built)","quote":"- Life sciences libraries (Scanpy, RDKit, MDAnalysis, BioPython) have complex, non-obvious APIs that LLMs frequently hallucinate"},{"itemId":"47059948","author":"fusionalchemist","createdAt":1771414452,"url":"https://news.ycombinator.com/item?id=47059948","kind":"story","storyTitle":"Show HN: System architecture method using mythology and LLMs (no CS background)\"","quote":"Electrical grid DSL, weather systems, banking, healthcare How it works: By forcing the LLM to select from a bounded vocabulary of mythological concepts (~163 \"spells\" + ~139 \"cloths\") BEFORE generating code, you eliminate hallucination at the architectural level."},{"itemId":"47059153","author":"arashsadrieh","createdAt":1771407795,"url":"https://news.ycombinator.com/item?id=47059153","kind":"story","storyTitle":"Show HN: AI agents designed and shipped this app end-to-end in 36 hours for $270","quote":"I haven't figured out how to fix this without hardcoding editorial judgment. The agents occasionally hallucinate context about news stories."},{"itemId":"47054319","author":"ashton314","createdAt":1771367178,"url":"https://news.ycombinator.com/item?id=47054319","kind":"comment","storyTitle":"Students Are Being Treated Like Guinea Pigs: Inside an AI-Powered Private School","quote":"I'm not saying AI can help with education. It can—it helps me!—but no hallucinating stochastic machine will have the human insight that a good teacher has."},{"itemId":"47053308","author":"rgthelen","createdAt":1771362189,"url":"https://news.ycombinator.com/item?id=47053308","kind":"story","storyTitle":"Show HN: Corral – Auth and Stripe billing that AI coding agents can set up","quote":"I built Corral because every time I asked an AI coding agent to \"add auth and payments,\" it hallucinated for an hour and produced broken"},{"itemId":"47046850","author":"kakamadafuka","createdAt":1771332005,"url":"https://news.ycombinator.com/item?id=47046850","kind":"comment","storyTitle":"Fix the iOS keyboard before the timer hits zero or I'm switching back to Android","quote":"This reads like a person trying to imitate a broken LLM that is on a hallucinatory rant."},{"itemId":"47045604","author":"WujieGuGavin","createdAt":1771322519,"url":"https://news.ycombinator.com/item?id=47045604","kind":"story","storyTitle":"Show HN: Fixing AI's Core Flaws, A protocol cuts LLM token waste by 40–70%","quote":"WLM (Wujie Language Model), a protocol stack + world engine that rethinks AI from token prediction to structural intelligence. I built this to fix the problems we all deal with daily: hallucination, drift, uncontrollable behavior, black-box reasoning, unstructured knowledge, and chaotic world/agent generation."},{"itemId":"47037780","author":"wmeredith","createdAt":1771263295,"url":"https://news.ycombinator.com/item?id=47037780","kind":"comment","storyTitle":"Breaking the spell of vibe coding","quote":"> How do you not see the difference between a machine that will hallucinate something random if it doesn’t know the answer vs a human that will logic through things and find the correct answer."},{"itemId":"47037142","author":"codingbot3000","createdAt":1771260063,"url":"https://news.ycombinator.com/item?id=47037142","kind":"comment","storyTitle":"EU bans the destruction of unsold apparel, clothing, accessories and footwear","quote":"As a EU citizen it is really frustrating to watch that they just can't stop themselves from introducing new regulation. And we don't have non-hallucinating AI yet that can help businesses keep track of all this legislative diarrhea."},{"itemId":"47034177","author":"rnc000","createdAt":1771244791,"url":"https://news.ycombinator.com/item?id=47034177","kind":"story","storyTitle":"Show HN: Rakenne – Markdown-defined agentic workflows for structured documents","quote":"I’m the creator of Rakenne and I built it because I noticed a recurring problem with LLMs in professional settings: chat-based document creation is unpredictable and hard to scale for domain experts. Experts know the process of building a document (the questions to ask, the order of operations, the edge cases), but translating that into a long system prompt often leads to hallucinations or missed steps."},{"itemId":"47029711","author":"mschuster91","createdAt":1771204498,"url":"https://news.ycombinator.com/item?id=47029711","kind":"comment","storyTitle":"I’m joining OpenAI","quote":"But I hate getting fed slop - and it's not just cat videos by the way. I'm (as evidenced by my comment history) into mechanics, electronics and radio stuff, and there are so damn many slop channels spreading outright BS with AI hallucinated scripts that it eventually gets really really annoying."},{"itemId":"47026363","author":"LUpath","createdAt":1771182016,"url":"https://news.ycombinator.com/item?id=47026363","kind":"comment","storyTitle":"[dead]","quote":"I wanted to build something that turns a \"vibe\" into a factual, bookable plan in under a minute. The Tech: Most AI travel tools just ask an LLM to \"hallucinate\" a beach."},{"itemId":"47024597","author":"ant6n","createdAt":1771170314,"url":"https://news.ycombinator.com/item?id=47024597","kind":"comment","storyTitle":"AI is going to kill app subscriptions","quote":"Then Gemini loses all context when you stop and edit a reply. Then the bots themselves with their hallucination, their gaslighting, their covering up of mistakes, covering up that they can't read a simple PDF."},{"itemId":"47024266","author":"AISovereignDev","createdAt":1771167899,"url":"https://news.ycombinator.com/item?id=47024266","kind":"story","storyTitle":"Show HN: Building AGI as a systems problem, not a UI wrapper","quote":"As I mentioned in my previous post, I’m a builder, not a researcher. I’m tired of the \"good enough\" era of AI where we settle for chatbots that hallucinate and forget context."},{"itemId":"47023067","author":"amananytime07","createdAt":1771157294,"url":"https://news.ycombinator.com/item?id=47023067","kind":"comment","storyTitle":"TexGuardian – Claude Code, but for LaTeX academic papers","quote":"I built TexGuardian after spending yet another deadline night fighting LaTeX formatting instead of focusing on research. Every conference submission, the same ritual: figure overflows, citation format issues, TODO markers left in text, hallucinated references from ChatGPT, forgotten anonymization."},{"itemId":"47022951","author":"post_below","createdAt":1771155552,"url":"https://news.ycombinator.com/item?id=47022951","kind":"comment","storyTitle":"No Coding Before 10am","quote":"The thing that frustrates me is that this is really easy to demonstrate. Articles like this are essentially hallucinations that, at least many, people mystifyingly take seriously."},{"itemId":"47021361","author":"paulbjensen","createdAt":1771134390,"url":"https://news.ycombinator.com/item?id=47021361","kind":"comment","storyTitle":"Golf game built last night with Claude Code, Svelte and ThreeJS","quote":"I think that we are in a major transition phases in our industry, and that there is still work to be done to yield the results we want from AI (in terms of quality, accuracy, not hallucinating libraries or API calls that do not exist)."},{"itemId":"47021174","author":"mikewolfd","createdAt":1771131265,"url":"https://news.ycombinator.com/item?id=47021174","kind":"comment","storyTitle":"AI Soap (the bar, not the API)","quote":"I built a Python library for saponification chemistry so LLMs can help formulate soap recipes without hallucinating the dangerous parts."},{"itemId":"47016890","author":"subho007","createdAt":1771093106,"url":"https://news.ycombinator.com/item?id=47016890","kind":"comment","storyTitle":"Show HN: Clawsec - Open-source plugin for OpenClaw that blocks dangerous actions","quote":"The problem: AI agents are getting good enough to run shell commands, query databases, and manage infrastructure autonomously. But one hallucinated rm -rf / or a prompt injection that exfiltrates your .env can do real damage."},{"itemId":"47015070","author":"13pixels","createdAt":1771081742,"url":"https://news.ycombinator.com/item?id=47015070","kind":"comment","storyTitle":"Show HN: Multi-attribute decision frameworks for tech purchases","quote":"If your framework relies on LLM training data or RAG, it's still downstream of whatever content dominates the web. Curious if you've seen any 'hallucinated brands' or persistent bias towards certain vendors in your tests across models? e.g. does Gemini favor Google products in your laptop comparison?"},{"itemId":"47013274","author":"mazilin","createdAt":1771063555,"url":"https://news.ycombinator.com/item?id=47013274","kind":"story","storyTitle":"Show HN: AI Station Navigator – LLM=CPU, Agents=Processes, Skills=Apps","quote":"I built AI Station Navigator as a proof-of-concept to solve a specific problem I faced: Context Pollution. When using AI agents for complex tasks, the context window gets cluttered quickly, causing the model to hallucinate or get confused."},{"itemId":"47012732","author":"redasadki","createdAt":1771057484,"url":"https://news.ycombinator.com/item?id=47012732","kind":"comment","storyTitle":"[dead]","quote":"Then ran out of TPM quota from Google Cloud's Gemini API after <100 chat interactions. Had problems and potential security risk due to hallucinations."},{"itemId":"47012243","author":"bitwize","createdAt":1771051543,"url":"https://news.ycombinator.com/item?id=47012243","kind":"comment","storyTitle":"Common Lisp Screenshots: today's CL applications in action","quote":"LLMs will probably just run in loops, standing up entire containers or Kubernetes pods with the latest changes, testing them, and tearing them down again to make room for the next iteration. As for hallucinations, I believe those are like version 0 of the thing we call lateral thinking and creativity when humans manifest it."},{"itemId":"47010414","author":"zaptheimpaler","createdAt":1771032495,"url":"https://news.ycombinator.com/item?id=47010414","kind":"comment","storyTitle":"Homeland Security Wants Social Media Sites to Expose Anti-ICE Accounts","quote":"Putting aside politics, they are reflexively cynical, misinformed and confident in their ignorance even about technology. Missing the boat on AI as well for years as its evolved from cool chatbot that hallucinates to can write serious code with few errors and solve unsolved research problems with some assistance."},{"itemId":"47008835","author":"manmal","createdAt":1771022529,"url":"https://news.ycombinator.com/item?id=47008835","kind":"comment","storyTitle":"Gemini 3 Deep Think","quote":"I need to have a discussion of the scope, and sequence diagrams first. It should read a lot of files instead of hallucinating about my architecture."},{"itemId":"47003022","author":"Grimblewald","createdAt":1770992268,"url":"https://news.ycombinator.com/item?id=47003022","kind":"comment","storyTitle":"Ask HN: What would you recommend a vibe coder learn about how all this works?","quote":"As a simple example, claude code is easily among the most competent coding interfaces I know of right now. However, if I give it a toy problem I've been toying with as a hobby project, and it breaks so badly it starts hallucinating that it is chatgpt."},{"itemId":"47002537","author":"instalabsai","createdAt":1770989782,"url":"https://news.ycombinator.com/item?id=47002537","kind":"story","storyTitle":"Show HN: Instagit – MCP server that answers questions about any GitHub repo","quote":"The problem: AI coding agents hallucinate library internals constantly."},{"itemId":"46998175","author":"alexfefun1","createdAt":1770949414,"url":"https://news.ycombinator.com/item?id=46998175","kind":"comment","storyTitle":"Robust ways to extract bank statements from PDF to CSV beyond raw LLMs?","quote":"While it works surprisingly well for 95% of cases, the \"last 5%\" is a headache: Hallucinations: Occasionally, the AI misinterprets a digit or skips a line item, which is unacceptable for financial data."},{"itemId":"46997789","author":"impulser_","createdAt":1770945927,"url":"https://news.ycombinator.com/item?id=46997789","kind":"comment","storyTitle":"Gemini 3 Deep Think","quote":"This has been a problem of Gemini since 1.0 and they still haven't fixed it. Also the worst model in terms of hallucinations."},{"itemId":"46996023","author":"hacker27369","createdAt":1770934438,"url":"https://news.ycombinator.com/item?id=46996023","kind":"story","storyTitle":"GodHands – Deterministic Desktop Automation via MCP","quote":"I spent the last few years in High-Frequency Trading. In that world, \"probabilistic\" automation is a non-starter; if an AI hallucinations or clicks the wrong pixel, it's a compliance disaster."},{"itemId":"46990146","author":"danFctr","createdAt":1770910718,"url":"https://news.ycombinator.com/item?id=46990146","kind":"story","storyTitle":"Show HN: Tako AI – Agent for Okta With Natural language (zero hallucination)","quote":"Show HN: Tako AI – Agent for Okta With Natural language (zero hallucination)"},{"itemId":"46985663","author":"vira28","createdAt":1770879753,"url":"https://news.ycombinator.com/item?id=46985663","kind":"story","storyTitle":"Show HN: AI that replaces the first 15 minutes of every client call","quote":"• Prompt templates generated dynamically based on objective (FAQ vs intake vs qualification) • Guardrails to prevent hallucination beyond uploaded corpus"},{"itemId":"46981450","author":"dhruv0206","createdAt":1770845892,"url":"https://news.ycombinator.com/item?id=46981450","kind":"story","storyTitle":"Show HN: DevProof – Verified developer portfolios using Code Complexity and AI","quote":"If the AI claims a file is \"Tier 3\" (Complex) but the complexity is < 15, we automatically downgrade it. (This prevents the \"hallucination\" problem where AI thinks a long file is complex when it's just verbose)."},{"itemId":"46972563","author":"dwedge","createdAt":1770800260,"url":"https://news.ycombinator.com/item?id=46972563","kind":"comment","storyTitle":"The AI Vampire","quote":"Articles like this are part of the problem, not the solution. Sure they are great for generating some level of code, but the deeper it goes the more it hallucinates."},{"itemId":"46966896","author":"cubefox","createdAt":1770757448,"url":"https://news.ycombinator.com/item?id=46966896","kind":"comment","storyTitle":"Frontier AI agents violate ethical constraints 30–50% of time, pressured by KPIs","quote":"Okay, I haven't really tested hallucinations like this, that may well be true."},{"itemId":"46966307","author":"sallamzz92","createdAt":1770754994,"url":"https://news.ycombinator.com/item?id=46966307","kind":"story","storyTitle":"AI agent has more production access than your senior engineers","quote":"During that window, your infrastructure is in an unknown state, and the only entity that knows the intended state is the agent that made the change. If the agent hallucinated, nobody knows intended state."},{"itemId":"46963340","author":"sebringj","createdAt":1770744170,"url":"https://news.ycombinator.com/item?id=46963340","kind":"story","storyTitle":"Runtime validation is still fucked in AI coding agents","quote":"You run it in a real app and immediately hit the same bullshit wall every time: - Hallucinated logic only reveals itself under real data or edge cases"},{"itemId":"46962874","author":"Dave_stridefuel","createdAt":1770742684,"url":"https://news.ycombinator.com/item?id=46962874","kind":"story","storyTitle":"Show HN: ClearDemand – Cross-case search and drafting for injury firms","quote":"Hello HN, We built ClearDemand to solve the \"hallucination\" problem in legal drafting."},{"itemId":"46962638","author":"scrollop","createdAt":1770742045,"url":"https://news.ycombinator.com/item?id=46962638","kind":"comment","storyTitle":"America's $1T AI Gamble","quote":"They can't even remember \"properly\". Hallucinations cripple them for serious, professional uses."},{"itemId":"46960968","author":"mmhetric","createdAt":1770737092,"url":"https://news.ycombinator.com/item?id=46960968","kind":"story","storyTitle":"MiRAGE: Open-source framework for multimodal RAG evaluation","quote":"3. Verify: An adversarial \"Verifier Agent\" fact-checks the answers against the source to prevent hallucinated ground truth."},{"itemId":"46960752","author":"sebringj","createdAt":1770736353,"url":"https://news.ycombinator.com/item?id=46960752","kind":"comment","storyTitle":"Show HN: Autonomo MCP – Developing while E2E Testing","quote":"I built Autonomo MCP because AI coding agents (Cursor, Claude, etc.) are great at generating code but terrible at actually running and validating it in real apps. They hallucinate, rely on slow screenshots, or break on multi-device flows."},{"itemId":"46959803","author":"Ginsabo","createdAt":1770732003,"url":"https://news.ycombinator.com/item?id=46959803","kind":"story","storyTitle":"Show HN: I wrote a prompt to stop Gemini from hallucinating","quote":"Show HN: I wrote a prompt to stop Gemini from hallucinating"},{"itemId":"46959746","author":"whynotminot","createdAt":1770731720,"url":"https://news.ycombinator.com/item?id=46959746","kind":"comment","storyTitle":"Frontier AI agents violate ethical constraints 30–50% of time, pressured by KPIs","quote":"It just seems to always need to give you an answer, even if it has to make it up. This was also largely how ChatGPT behaved before 5, but OpenAI has gotten much much better at having the model admit it doesn’t know or tell you that the thing you’re looking for doesn’t exist instead of hallucinating something plausible sounding."},{"itemId":"46959021","author":"hankimis","createdAt":1770727629,"url":"https://news.ycombinator.com/item?id=46959021","kind":"story","storyTitle":"Show HN: 0x – A language that compiles to React, Vue, and Svelte (80% less code)","quote":"No curly braces, no semicolons, no JSX closing tags, no import boilerplate. For an LLM, fewer structural decisions = fewer hallucinations."},{"itemId":"46958458","author":"teemuki","createdAt":1770724070,"url":"https://news.ycombinator.com/item?id=46958458","kind":"comment","storyTitle":"Skene Cookbook – 700 open source production-ready AI skills","quote":"I spent the last few months obsessed with a specific bottleneck: every time I start a new AI agent project, I waste the first few weeks on the same \"foundational\" plumbing. Whether it is building lead research pipelines, setting up data extraction, or wrestling with .cursorrules to stop hallucinations, we are all rebuilding the same generic 80% over and over."},{"itemId":"46955823","author":"danelrfoster","createdAt":1770702205,"url":"https://news.ycombinator.com/item?id=46955823","kind":"story","storyTitle":"Why Every Business Must Engage with AI – and How to Do It Right","quote":"The biggest risk isn’t “doing AI wrong.” It’s not building AI fluency at all while the rest of the market moves forward. Once AI systems are in production, new problems appear: cost control, reliability, hallucinations, latency, silent regressions."},{"itemId":"46952450","author":"sebringj","createdAt":1770676102,"url":"https://news.ycombinator.com/item?id=46952450","kind":"story","storyTitle":"Show HN: Autonomo – AI developing while E2E testing","quote":"I built Autonomo MCP because AI coding agents (Cursor, Claude, etc.) are great at generating code but terrible at actually running and validating it in real apps. They hallucinate, rely on slow screenshots, or break on multi-device flows."},{"itemId":"46946742","author":"a1j9o94","createdAt":1770653155,"url":"https://news.ycombinator.com/item?id=46946742","kind":"story","storyTitle":"Show HN: Sales Agent Benchmark – SWE-Bench for sales AI agents (open source)","quote":"Models confidently identify risks that don't exist in the source material. Hallucinated stakeholders."},{"itemId":"46946593","author":"aed","createdAt":1770652473,"url":"https://news.ycombinator.com/item?id=46946593","kind":"story","storyTitle":"Show HN: AI agents play SimCity through a REST API","quote":"You can usually get agents building in seconds. Website: https://hallucinatingsplines.com"},{"itemId":"46943521","author":"ileonichwiesz","createdAt":1770631064,"url":"https://news.ycombinator.com/item?id=46943521","kind":"comment","storyTitle":"Show HN: A custom font that displays Cistercian numerals using ligatures","quote":"Hallucination was also a common problem, with the output often replaced by a similar (but more common) name or word."},{"itemId":"46941324","author":"djx22","createdAt":1770608513,"url":"https://news.ycombinator.com/item?id=46941324","kind":"comment","storyTitle":"AI makes the easy part easier and the hard part harder","quote":"It's bad at writing code when it needs to be performant or needs to span over multiple files. Especially when it spans over multiple files because that's where it starts hallucinating and introducing abstractions and boilerplate that's not necessary and it just makes your life harder when it comes to debugging."},{"itemId":"46940029","author":"peteforde","createdAt":1770596375,"url":"https://news.ycombinator.com/item?id=46940029","kind":"comment","storyTitle":"AI makes the easy part easier and the hard part harder","quote":"We should feel the same way about vibe coding. Meanwhile, if you load up Cursor and break your application development into bite sized chunks, and then work through those chunks in a sane order using as many Plan -> Agent -> Debug conversations with Opus 4.5 (Thinking) as needed, you too will obtain the mythical productivity multipliers you keep accusing us of hallucinating."},{"itemId":"46938839","author":"LutumVeritas","createdAt":1770587154,"url":"https://news.ycombinator.com/item?id=46938839","kind":"story","storyTitle":"Show HN: Parametric Hubris – Beating GPT-5 on SimpleQA with forced retrieval","quote":"Just finished reading \"Parametric Hubris\" by Martin Gehrken. It’s a fascinating deep dive into why frontier models hallucinate even when they have tools."},{"itemId":"46938029","author":"embedding-shape","createdAt":1770581347,"url":"https://news.ycombinator.com/item?id=46938029","kind":"comment","storyTitle":"GitHub Agentic Workflows","quote":"> This happens with all agents I've used and package.json files for npm. Instead of using `npm i foo` the agent string-edits package.json and hallucinates some version to install."},{"itemId":"46937543","author":"ambitious_potat","createdAt":1770578470,"url":"https://news.ycombinator.com/item?id=46937543","kind":"story","storyTitle":"Show HN: The biggest achievement of my life so far","quote":"Customized llm with RAG :- \"\"cited the air navigation act,specified the 5km no fly zones,and linked to the CAAS permit page\"\" the difference was clear and it was sure that the ai was not hallucinating."},{"itemId":"46937147","author":"moshest","createdAt":1770575962,"url":"https://news.ycombinator.com/item?id=46937147","kind":"comment","storyTitle":"Show HN: A local-first documentation tool for AI agents (MCP)","quote":"I got tired of my AI agent hitting rate limits right when I was actually getting work done. I’d be using an MCP server for docs, and suddenly the assistant would start hallucinating old API patterns because the cloud service I was using hit its cap or was lagging."},{"itemId":"46933303","author":"seekerXtruth","createdAt":1770549448,"url":"https://news.ycombinator.com/item?id=46933303","kind":"comment","storyTitle":"How do you use AI coding tools at scale without losing architectural control?","quote":"I’ve been experimenting quite a bit with AI-assisted development recently (Copilot, Cursor, Claude, etc.), both in larger systems and in smaller side projects. What keeps surprising me is not hallucinations or model output quality as such, but how easy it is to lose shared architectural context over time."},{"itemId":"46931118","author":"formerOpenAI","createdAt":1770522085,"url":"https://news.ycombinator.com/item?id=46931118","kind":"comment","storyTitle":"Why Embedded Models Must Hallucinate: A Boundary Theory (RCC)","quote":"I’ve been playing with LLM failure modes lately (hallucination, planning falling apart after ~10 steps, long-context weirdness)."},{"itemId":"46930856","author":"noodletheworld","createdAt":1770518793,"url":"https://news.ycombinator.com/item?id=46930856","kind":"comment","storyTitle":"Software factories and the agentic moment","quote":"> LLM can very easily verify this by generating its own sample api call and checking the response. This is no different from having an LLM pair where the first does something and the second one reviews it to “make sure no hallucinations”."},{"itemId":"46921180","author":"ambitious_potat","createdAt":1770436731,"url":"https://news.ycombinator.com/item?id=46921180","kind":"story","storyTitle":"Show HN: I built a RAG engine to search Singaporean laws","quote":"Recently, I created Explore Singapore, a RAG-based search engine that scrapes about 20,000 pages of Singaporean government acts and laws. I recently posted the MVP and received some tough but essential feedback about hallucinations and query depth."},{"itemId":"46917518","author":"mrlongroots","createdAt":1770408530,"url":"https://news.ycombinator.com/item?id=46917518","kind":"comment","storyTitle":"The Waymo World Model","quote":"I find Gemini 3 useless. It has regressed on hallucinations from Gemini 2.5, to the point where its output is no better than a random token stream despite all its benchmark outperformance."},{"itemId":"46913343","author":"zackmorris","createdAt":1770388635,"url":"https://news.ycombinator.com/item?id=46913343","kind":"comment","storyTitle":"The Missing Layer","quote":"We never got true multicore computing (which still devastates me), but we did get Apple's M line of ARM processors and video cards that reached ludicrous speed. What I'm trying to say is, I know that it seems like AI is all over the place right now, and it's hard to know if it's correct or hallucinating."},{"itemId":"46905053","author":"revahage","createdAt":1770324463,"url":"https://news.ycombinator.com/item?id=46905053","kind":"comment","storyTitle":"GPT-5.3-Codex","quote":"Well, it took opus 4.5 five messages to solve a trivial git problem for me. It hallucinated nonexistent flags three times."},{"itemId":"46897408","author":"andreahlert","createdAt":1770281785,"url":"https://news.ycombinator.com/item?id=46897408","kind":"story","storyTitle":"I Built a Claude Code Plugin That Detects and Blocks It Before Changes Happen","quote":"But what if we treated scope creep not as an inevitable quirk of LLMs, but as something we can proactively guard against? Inspired by discussions here and on X about AI agents going rogue with hallucinations or excessive edits, I decided to build Scope Guard: a lightweight plugin for Claude Code that intervenes before the agent completes its work, ensuring edits stay strictly within your specified intent."},{"itemId":"46889993","author":"kryogen1c","createdAt":1770231240,"url":"https://news.ycombinator.com/item?id=46889993","kind":"comment","storyTitle":"Microsoft's Copilot chatbot is running into problems","quote":"But copilot is fucking terrible. Sometimes I ask it powershell questions about microsoft products and it hallucinates answers."},{"itemId":"46889886","author":"jascha_eng","createdAt":1770230725,"url":"https://news.ycombinator.com/item?id=46889886","kind":"comment","storyTitle":"Claude Code's /Insights","quote":"It doesn't exist anywhere in the underlying facets data. The insights generator appears to have hallucinated"},{"itemId":"46888841","author":"intentrahq","createdAt":1770226696,"url":"https://news.ycombinator.com/item?id=46888841","kind":"comment","storyTitle":"[dead]","quote":"When the inefficiency is the tool's fault, we help you get that money back. The CLI installs hooks into your AI tools and detects 20+ violation types: retry loops from tool bugs, context hallucination (referencing files that don't exist), excessive thinking (80% reasoning tokens, 20% actual code), and more."},{"itemId":"46888178","author":"MrSteaddy","createdAt":1770223770,"url":"https://news.ycombinator.com/item?id=46888178","kind":"comment","storyTitle":"InsAIts: Monitoring for AI-AI comms. Detect hallucinations before propagation","quote":"Today we're open-sourcing the core under Apache 2.0. The \"aha moment\" was when I watched a finance pipeline where one agent hallucinated a 5x cost difference."},{"itemId":"46888114","author":"MrSteaddy","createdAt":1770223569,"url":"https://news.ycombinator.com/item?id=46888114","kind":"comment","storyTitle":"InsAIts: Monitoring for AI-AI comms. Detect hallucinations before propagation","quote":"They start confident and silently become unsure. One agent's hallucination becomes the next agent's trusted input."},{"itemId":"46886225","author":"ericgilerson","createdAt":1770215439,"url":"https://news.ycombinator.com/item?id=46886225","kind":"story","storyTitle":"Show HN: Resolv – AI Agentic IDE that insists AI cannot think","quote":"It's slower than the \"vibe-coding\" tools, but it eliminates the expensive hallucinations and technical debt that come from letting models pretend they understand."},{"itemId":"46883606","author":"zhangxiaowen","createdAt":1770197640,"url":"https://news.ycombinator.com/item?id=46883606","kind":"story","storyTitle":"Show HN: 32KB deductive engine that catches LLM hallucinations","quote":"Show HN: 32KB deductive engine that catches LLM hallucinations"},{"itemId":"46879499","author":"simianwords","createdAt":1770164430,"url":"https://news.ycombinator.com/item?id=46879499","kind":"comment","storyTitle":"Anthropic AI tool sparks selloff from software to broader market","quote":"If you do, it’s almost trivially disprovable. One of the posters said that hallucination is a problem but if you’ve used ChatGPT for search, you would know that it’s not."},{"itemId":"46877506","author":"regrun","createdAt":1770153885,"url":"https://news.ycombinator.com/item?id=46877506","kind":"story","storyTitle":"Show HN: Reg.run - Decoupling AI \"thinking\" from API execution","quote":"I’ve spent the last few months looking at how we deploy autonomous agents. The current standard is basically: \"Put it in a system prompt and hope the LLM doesn't hallucinate a $50k refund.\""},{"itemId":"46876687","author":"rksart","createdAt":1770149949,"url":"https://news.ycombinator.com/item?id=46876687","kind":"story","storyTitle":"Show HN: OpenSymbolicAI – Agents with typed variables, not just context stuffing","quote":"EXAMPLE: A RAG AGENT Instead of the LLM hallucinating a plan based on a wall of text, it simply writes the logic to manipulate the data containers."},{"itemId":"46873963","author":"noncentral","createdAt":1770139513,"url":"https://news.ycombinator.com/item?id=46873963","kind":"comment","storyTitle":"Are LLM failures – including hallucination – structurally unavoidable? (RCC)","quote":"This is why drift shows up again between anchors, or when the external structure is sparse, contradictory, or time-varying. On model improvements (GPT-3 → GPT-4, Claude 2 → 3): RCC doesn’t claim that hallucination rates are fixed constants — only that there is a geometric ceiling beyond which improvements cannot generalize globally."},{"itemId":"46873809","author":"Soerensen","createdAt":1770138820,"url":"https://news.ycombinator.com/item?id=46873809","kind":"comment","storyTitle":"Are LLM failures – including hallucination – structurally unavoidable? (RCC)","quote":"Axiom 3 (stable global reference frame) seems most practically actionable. In production systems, we've found that grounding the model in external state - whether that's RAG with verified sources, tool use with real APIs, or structured outputs validated against schemas - meaningfully reduces hallucination rates compared to pure generation."},{"itemId":"46873276","author":"Raviteja_","createdAt":1770136698,"url":"https://news.ycombinator.com/item?id=46873276","kind":"comment","storyTitle":"Stop leaking user data to OpenAI/Claude/Gemini","quote":"I'm the creator of the PII Firewall Edge API (currently on RapidAPI). I saw a lot of devs struggling to implement safety guardrails correctly—most were just using basic regex or heavy LLMs that hallucinate."},{"itemId":"46863963","author":"decodebytes","createdAt":1770075760,"url":"https://news.ycombinator.com/item?id=46863963","kind":"story","storyTitle":"Show HN: Nono – Kernel-enforced sandboxing for AI agents","quote":"The problem: AI agents execute code on your machine. Prompt injections, hallucinations, or compromised tools can read ~/.ssh, exfiltrate credentials, or worse."},{"itemId":"46863289","author":"comex","createdAt":1770072892,"url":"https://news.ycombinator.com/item?id=46863289","kind":"comment","storyTitle":"Why do people still talk about AGI?","quote":"If you have an x% chance to get stuck on each step of a multi-step task, then the more steps, the higher the failure rate. This is true for both LLMs and humans, but LLMs tend to have more random failures, both due to hallucinations and due to being worse at recovering if their initial attempt fails (they can have a hard time remembering what they're supposed to do differently)."},{"itemId":"46859467","author":"internet_points","createdAt":1770057173,"url":"https://news.ycombinator.com/item?id=46859467","kind":"comment","storyTitle":"My Experience with Vibe Coding","quote":"I've gotten a few good results with Haskell from llm's (and quite a few bad results), but every single time I try them with nix I just get annoying hallucinations and non-solutions."},{"itemId":"46859259","author":"aeneas_ory","createdAt":1770056323,"url":"https://news.ycombinator.com/item?id=46859259","kind":"comment","storyTitle":"Hacking Moltbook","quote":"The AI code slop around these tools is so frustrating, just trying to get the instructions from the CTA on the moltbook website working which flashes `npx molthub@latest install moltbook` isn't working (probably hallucinated or otherwise out of date):"},{"itemId":"46858412","author":"anonymous908213","createdAt":1770052358,"url":"https://news.ycombinator.com/item?id=46858412","kind":"comment","storyTitle":"My five stages of AI grief","quote":"They can obviously generate a proof-of-concept quickly, and if the task is trivial enough, save a couple of minutes writing a throwaway script that you actually use day-to-day. I find them to be somewhat useful for retrieving information from documentation, although some of this gain is offset by the time wasted from hallucinated APIs."},{"itemId":"46857294","author":"qaq","createdAt":1770047206,"url":"https://news.ycombinator.com/item?id=46857294","kind":"comment","storyTitle":"Claude Code is suddenly everywhere inside Microsoft","quote":"Settled on Claude Code for actual work and Codex for checking Claude's work. If I try to mix in Gemini it will hallucinate issues that do not exist in code at very high rate."},{"itemId":"46849615","author":"decodebytes","createdAt":1769981704,"url":"https://news.ycombinator.com/item?id=46849615","kind":"story","storyTitle":"Show HN: Nono – Kernel-enforced sandboxing for AI agents","quote":"The problem: AI agents execute code on your machine. Prompt injections, hallucinations, or compromised tools can read ~/.ssh, exfiltrate credentials, or worse."},{"itemId":"46849227","author":"vitorlourenco","createdAt":1769978919,"url":"https://news.ycombinator.com/item?id=46849227","kind":"comment","storyTitle":"Security scanner that detect's AI-generated code vulnerabilities","quote":"Technical details: - 294 security checks (SQL injection, XSS, SSRF, AI hallucinations)"},{"itemId":"46847406","author":"2dogsanerd","createdAt":1769964328,"url":"https://news.ycombinator.com/item?id=46847406","kind":"story","storyTitle":"Show HN: Self-hosted RAG with MCP support for OpenClaw","quote":"Instead of uploading my private contracts to OpenAI, I built ClawRAG – a self-hosted RAG engine that connects to OpenClaw via MCP (Model Context Protocol). Now I can ask \"What did the contract say about liability?\" and get cited answers, not hallucinations."},{"itemId":"46845818","author":"drjcfmd","createdAt":1769949627,"url":"https://news.ycombinator.com/item?id=46845818","kind":"comment","storyTitle":"AI in the Exam Room – Free curriculum for safe medical AI use","quote":"Dual curriculum: - Patient pathway: sensing gap, hallucination detection, appropriate use"},{"itemId":"46840990","author":"b0rd3aux","createdAt":1769894514,"url":"https://news.ycombinator.com/item?id=46840990","kind":"story","storyTitle":"Show HN: Hackmenot – Security scanner for AI-generated code","quote":"- Auto-fix mode (hackmenot scan . --fix) - Detects hallucinated packages (dependencies AI made up that don't exist)"},{"itemId":"46830235","author":"hutchplusplus","createdAt":1769808885,"url":"https://news.ycombinator.com/item?id=46830235","kind":"story","storyTitle":"Thoughts on AI/LLM usage from a 25 year industry vet","quote":"It gets worse the further you zoom out. How many hallucinations and \"use krazy glue to stick cheese to your pizza\"s until people realize LLMs _suck at knowing things_ and they're not improving."},{"itemId":"46828503","author":"aspenmartin","createdAt":1769799952,"url":"https://news.ycombinator.com/item?id=46828503","kind":"comment","storyTitle":"Where I'm at with AI","quote":"Yes I agree here in principle here in some cases: I think there are certainly problems that LLMs are now better at but that don't reach the critical reliability threshold to say \"it can do this\". E.g. hallucinations, handling long context well (still best practice to reset context window frequently), long-running tasks etc."},{"itemId":"46824361","author":"telenardo","createdAt":1769781072,"url":"https://news.ycombinator.com/item?id=46824361","kind":"story","storyTitle":"Show HN: local speech-to-text is shockingly fast on Apple Silicon","quote":"Challenges: - Cleanup Hallucinations: Out of the box, small LLMs (3B, 1B) still make mistakes."},{"itemId":"46817469","author":"tgtweak","createdAt":1769724675,"url":"https://news.ycombinator.com/item?id=46817469","kind":"comment","storyTitle":"Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT","quote":"5.2 is back to being a sycophantic hallucinating mess for most use cases - I've anecdotally caught it out on many of the sessions I've had where it apologizes \"You're absolutely right... that used to be the case but as of the latest version as you pointed out, it no longer is.\" when it never existed in the first place."},{"itemId":"46817084","author":"johndough","createdAt":1769723069,"url":"https://news.ycombinator.com/item?id=46817084","kind":"comment","storyTitle":"Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT","quote":"> (it doesn't hallucinate on this)"},{"itemId":"46816648","author":"adamazak","createdAt":1769721111,"url":"https://news.ycombinator.com/item?id=46816648","kind":"story","storyTitle":"Show HN: Yanna Cursor, but for documents (draft → edit → redline → e-sign)","quote":"the output is structured instead of rambly it doesn’t hallucinate edits into the wrong spot"},{"itemId":"46812540","author":"logicallee","createdAt":1769704653,"url":"https://news.ycombinator.com/item?id=46812540","kind":"story","storyTitle":"Ask HN: Does ChatGPT 5.2 with web search still hallucinate facts?","quote":"Ask HN: Does ChatGPT 5.2 with web search still hallucinate facts?"},{"itemId":"46811431","author":"jmyeet","createdAt":1769700268,"url":"https://news.ycombinator.com/item?id=46811431","kind":"comment","storyTitle":"AI on Australian travel company website sent tourists to nonexistent hot springs","quote":"I love stories like this because there are still allegedly tech-savvy people who will insist that AIs don't lie, don't hallucinate and rarely if ever make errors."},{"itemId":"46809671","author":"Imustaskforhelp","createdAt":1769691904,"url":"https://news.ycombinator.com/item?id=46809671","kind":"comment","storyTitle":"Amazon cuts 16k jobs","quote":"I mean LLM's still hallucinate and they can be poisoned."},{"itemId":"46796310","author":"thisnick","createdAt":1769612932,"url":"https://news.ycombinator.com/item?id=46796310","kind":"comment","storyTitle":"Show HN: Agent-Rdp – CLI for AI Agents to Control Windows via Remote Desktop","quote":"There wasn't. It downloaded 3, hallucinated the rest, and reported success."},{"itemId":"46793755","author":"Der_Einzige","createdAt":1769597929,"url":"https://news.ycombinator.com/item?id=46793755","kind":"comment","storyTitle":"ASML staffing changes could result in a net reduction of around 1700 positions","quote":"I understand guild warfare and the hatred of labor unions the moment that I have to deal yet another actually-should-be-in-jail levels of scamming from yet another dentist. Y'all think that AI researchers should go to jail for hallucinated citations?"},{"itemId":"46793124","author":"CodeWeave","createdAt":1769592949,"url":"https://news.ycombinator.com/item?id=46793124","kind":"comment","storyTitle":"I Built CodeWeave","quote":"The root cause? A YAML indentation error confidently hallucinated by an AI tool."},{"itemId":"46791644","author":"spenceXu","createdAt":1769580167,"url":"https://news.ycombinator.com/item?id=46791644","kind":"story","storyTitle":"Ask HN: Doctors Using AI Workflow Automation – What's the Real Impact?","quote":"What are the biggest frustrations? (e.g., AI “hallucinations” in notes, rigid workflows, integration nightmares with EHRs, or the time needed to review/correct AI output)."},{"itemId":"46787204","author":"roan-we","createdAt":1769549590,"url":"https://news.ycombinator.com/item?id=46787204","kind":"comment","storyTitle":"Show HN: Kalibr – Autonomous Routing for AI Agents","quote":"I had developed a side project with AI agents to help me summarize the research papers and extract key citations, and I was repeatedly hitting the same annoying pattern. I would finetune everything with GPT4 to perfection, and then in a couple of weeks, it would start hallucinating references or missing citations."},{"itemId":"46784090","author":"msid33","createdAt":1769538271,"url":"https://news.ycombinator.com/item?id=46784090","kind":"story","storyTitle":"Show HN: CaseAI – Build custom RAG agents for WhatsApp in 2 minutes","quote":"They can't answer leads instantly 24/7. Standard \"chatbots\" are too rigid and \"GPT wrappers\" hallucinate too much for business use."},{"itemId":"46779826","author":"whstl","createdAt":1769521565,"url":"https://news.ycombinator.com/item?id=46779826","kind":"comment","storyTitle":"I made my own Git","quote":"I wish I could disagree with you, but what I'm seeing on average (especially at work) is exactly that: people asking stuff to ChatGPT and accepting hallucinations as fact, and then fighting me when I say it's not true."},{"itemId":"46778996","author":"cthalupa","createdAt":1769516149,"url":"https://news.ycombinator.com/item?id=46778996","kind":"comment","storyTitle":"I let ChatGPT analyze a decade of my Apple Watch data, then I called my doctor","quote":"LLMs that search scientific literature, etc., might point you in the direction of this research that the doctors are not aware of. Or hallucinate you into having some random disease that impacts 3 out of every million people and send you down a rabbithole for months."},{"itemId":"46776528","author":"rochansinha","createdAt":1769498520,"url":"https://news.ycombinator.com/item?id=46776528","kind":"story","storyTitle":"How do you use LLMs to verify databases with minimal hallucinations?","quote":"How do you use LLMs to verify databases with minimal hallucinations?"},{"itemId":"46761719","author":"MrSteaddy","createdAt":1769399862,"url":"https://news.ycombinator.com/item?id=46761719","kind":"story","storyTitle":"Show HN: InsAIts V2 – Real-time monitoring for multi-agent AI communication","quote":"I'm Cristian, an indie developer from Italy working on tools to make multi-agent AI safer and more debuggable. When agents talk to each other (CrewAI, LangGraph, AutoGen, custom setups), they quickly develop shorthand, lose context, invent jargon, and propagate hallucinations — all invisible to us."},{"itemId":"46755913","author":"game_the0ry","createdAt":1769361419,"url":"https://news.ycombinator.com/item?id=46755913","kind":"comment","storyTitle":"Ask HN: Is Gemini Getting Worse?","quote":"Depends on the context -- do you mean Gemini CLI for coding or Gemini the chat web app? For the CLI -- no, I have not noticed and hallucinations."},{"itemId":"46755780","author":"optimalsolver","createdAt":1769360537,"url":"https://news.ycombinator.com/item?id=46755780","kind":"story","storyTitle":"Ask HN: Is Gemini Getting Worse?","quote":"I'm certain they've done a bait-and-switch. The current model is almost unusable and hallucinates wildly."},{"itemId":"46744594","author":"DonHopkins","createdAt":1769269980,"url":"https://news.ycombinator.com/item?id=46744594","kind":"comment","storyTitle":"Gas Town's agent patterns, design bottlenecks, and vibecoding at scale","quote":"https://en.wikipedia.org/wiki/Self_(programming_language) MOOLLM is deliberately playful, in the spirit and tradition of Seymour Papert's Constructionist Philosophy, and Mitchel Resnick's Lifelong Kindergarten -- it's a text adventure game framework, not a vibe-coded hallucination."},{"itemId":"46736108","author":"JohnLeitch","createdAt":1769193847,"url":"https://news.ycombinator.com/item?id=46736108","kind":"comment","storyTitle":"Why does SSH send 100 packets per keystroke?","quote":"The problem is hallucinations."},{"itemId":"46733727","author":"zhoujing204","createdAt":1769182520,"url":"https://news.ycombinator.com/item?id=46733727","kind":"comment","storyTitle":"AI is a horse (2024)","quote":"It has no biological drive, no concept of death, and no feelings. If an AI \"hallucinates\" or fails, it isn't \"spooked\"; it is simply executing a probabilistic calculation that resulted in a low-quality output."},{"itemId":"46730091","author":"kranner","createdAt":1769157973,"url":"https://news.ycombinator.com/item?id=46730091","kind":"comment","storyTitle":"Talking to LLMs has improved my thinking","quote":"edit: I'll admit ChatGPT is a great search engine (and also very hallucinatory depending on how much you know about the subject) and maybe it helps some people think, sure."},{"itemId":"46727265","author":"danudey","createdAt":1769131600,"url":"https://news.ycombinator.com/item?id=46727265","kind":"comment","storyTitle":"Satya Nadella: \"We need to find something useful for AI\"","quote":"Gemini hallucinated a method on a rust crate that it was trying to use and then spent ten minutes googling 'method_name v4l2 examples' and so on."},{"itemId":"46725514","author":"fabio2620","createdAt":1769118151,"url":"https://news.ycombinator.com/item?id=46725514","kind":"story","storyTitle":"Show HN: SONA – An AI-native serialization format to reduce \"Syntax Tax\"","quote":"In a token-based economy, they are expensive overhead. Furthermore, implicit typing leads to \"Type Hallucinations\" (e.g., a model parsing \"NO\" as a boolean)."},{"itemId":"46721694","author":"bicepjai","createdAt":1769100321,"url":"https://news.ycombinator.com/item?id=46721694","kind":"comment","storyTitle":"Claude's new constitution","quote":"I think you might be right about confirmation bias and AI slop :) The \"replace Claude with a person\" argument is fine in theory, but LLMs aren't people. They hallucinate, drift, and struggle to follow instructions reliably."},{"itemId":"46719382","author":"HarHarVeryFunny","createdAt":1769090717,"url":"https://news.ycombinator.com/item?id=46719382","kind":"comment","storyTitle":"Satya Nadella: \"We need to find something useful for AI\"","quote":"Maybe not. Asking random questions is exactly where you run into time-wasting hallucinations since the models don't seem to be very good at deciding when to use a search tool and when just to rely on their training data."},{"itemId":"46716512","author":"themast","createdAt":1769069639,"url":"https://news.ycombinator.com/item?id=46716512","kind":"comment","storyTitle":"Show HN: How to stop Claude Code hallucinations using a CLI Truth Layer","quote":"I’ve been using Claude Code since the early beta, and while the \"99% of my code is AI-written\" claims are technically true for boilerplate, I kept hitting a wall with complex backend refactors. The agent would get stuck in a \"hallucination loop\"—fixing a bug, breaking a route, then suggesting the same fix again."},{"itemId":"46703977","author":"Imustaskforhelp","createdAt":1768993983,"url":"https://news.ycombinator.com/item?id=46703977","kind":"comment","storyTitle":"The percentage of Show HN posts is increasing, but their scores are decreasing","quote":"You would find it funny but the problem's not in code generation but dependency management etc. atleast to me which is why I love uv scripts a lot but sometimes especially chatgpt doesn't know about uv script feature or hallucinates it hard on their web version so If I am building python applets I would paste uv script's complete docs to the LLM website I would be using."},{"itemId":"46699656","author":"lostsock","createdAt":1768955327,"url":"https://news.ycombinator.com/item?id=46699656","kind":"comment","storyTitle":"Ask HN: Do you have any evidence that agentic coding works?","quote":"The MCP tools I have installed make a huge different: laravel-boost, context7, and figma (which in particular feels borderline magical at converting designs into code!). I often have to tell the agent to visit GitHub READMEs and official docs instead of letting it hallucinate “best practices”, the agent will oftentimes guess and get stack, so if it's doing that, you’ve already lost."},{"itemId":"46696348","author":"devel12","createdAt":1768935963,"url":"https://news.ycombinator.com/item?id=46696348","kind":"story","storyTitle":"Show HN: Armour – A secure stdio MCP proxy, written in Go","quote":"If you give an agent \"Read Only\" access, it can’t actually fix anything. If you give it \"Write\" access, it’s only a matter of time before a hallucination or a bad prompt results in a deleted database or a nuked production bucket."},{"itemId":"46691813","author":"bluemouse_ai","createdAt":1768916965,"url":"https://news.ycombinator.com/item?id=46691813","kind":"story","storyTitle":"Show HN: BlueMouse – AI Code Generator with 17-Layer Validation","quote":"*The Problem:* - AI hallucinates architecture"},{"itemId":"46691372","author":"mx-Liu123","createdAt":1768913981,"url":"https://news.ycombinator.com/item?id=46691372","kind":"comment","storyTitle":"Show HN: AgentCommander - workflow engine for evolutionary code optimization","quote":"Verification: You can audit the agent's \"thought process\" by running gemini -r inside any generated experiment directory. It’s crucial for verifying that the agent isn't hallucinating its research logic."},{"itemId":"46680604","author":"mikeayles","createdAt":1768839072,"url":"https://news.ycombinator.com/item?id=46680604","kind":"comment","storyTitle":"\"Anyone else out there vibe circuit-building?\"","quote":"The LLM picks which blocks you need and where they go, but the actual circuit design was done by humans and tested. No hallucinated resistor values, no creative interpretations of datasheets."},{"itemId":"46678402","author":"abalashov","createdAt":1768827005,"url":"https://news.ycombinator.com/item?id=46678402","kind":"comment","storyTitle":"Show HN: I quit coding years ago. AI brought me back","quote":"I have had several projects that look like this already in the VoIP world, and it's been very gainful. However, my industry probably does not compare fairly to the common denominator of CRUD apps in common tech stacks; some of it is specialised enough that the LLMs drop to GPT-2 type levels of utility (and hallucination! -- that's been particularly lucrative)."},{"itemId":"46677405","author":"Klonoar","createdAt":1768819312,"url":"https://news.ycombinator.com/item?id=46677405","kind":"comment","storyTitle":"Show HN: I quit coding years ago. AI brought me back","quote":"Vibe coding will eventually come for that. The cost of hallucinations though - you potentially have a stronger point there."},{"itemId":"46676802","author":"antonbabenko","createdAt":1768815084,"url":"https://news.ycombinator.com/item?id=46676802","kind":"comment","storyTitle":"Claude Skill for Terraform/OpenTofu – testing, modules, CI/CD, and prod patterns","quote":"Just shipped a Claude skill that aggregates Terraform Best Practices, plus a bunch of other trusted sources I have collected over the years. In 2026, I do not want to hear that \"AI hallucinates with Terraform\"."},{"itemId":"46676761","author":"veunes","createdAt":1768814795,"url":"https://news.ycombinator.com/item?id=46676761","kind":"comment","storyTitle":"Show HN: I quit coding years ago. AI brought me back","quote":"It might be worth looking into low-level dev (embedded, kernel, drivers) or complex R&D. Vibe coding doesn't work there yet, and the cost of error is too high for hallucinations."},{"itemId":"46676580","author":"aeneas_ory","createdAt":1768813434,"url":"https://news.ycombinator.com/item?id=46676580","kind":"comment","storyTitle":"Provide agents with automated feedback","quote":"Likelihood is high that the LLM just hallucinated some false positive lint rule anyways giving you a false sense of security."},{"itemId":"46666485","author":"rahuldass","createdAt":1768731178,"url":"https://news.ycombinator.com/item?id=46666485","kind":"comment","storyTitle":"QWED AI – Open-source deterministic verification layer for LLMs","quote":"I built QWED – a verification layer that sits between your LLM and production. The idea: Don't fix hallucinations, verify them."},{"itemId":"46665799","author":"Davidzheng","createdAt":1768723828,"url":"https://news.ycombinator.com/item?id=46665799","kind":"comment","storyTitle":"Erdos 281 solved with ChatGPT 5.2 Pro","quote":"Anecdotally, I, as a math postdoc, think that GPT 5.2 is much stronger qualitatively than anything else I've used. Its rate of hallucinations is low enough that I don't feel like the default assumption of any solution is that it is trying to hide a mistake somewhere."},{"itemId":"46660140","author":"robot-wrangler","createdAt":1768672117,"url":"https://news.ycombinator.com/item?id=46660140","kind":"comment","storyTitle":"LLM Structured Outputs Handbook","quote":"What's the arg-count and type-complexity for that one important function/class in your in-house code base? For a less well-known use case or tool, if you want hallucination free and correct output, then you need structured output that works, because the alternative is rolling your own model trained on your stuff."},{"itemId":"46657257","author":"DonHopkins","createdAt":1768650232,"url":"https://news.ycombinator.com/item?id=46657257","kind":"comment","storyTitle":"Cursor's latest “browser experiment” implied success without evidence","quote":"html5ever => html5never - Servo's HTML parser that never finishes tokenizing. Skia => SkAI - AI-generated graphics that hallucinates extra pixels and fingers."},{"itemId":"46651071","author":"satvikpendem","createdAt":1768592201,"url":"https://news.ycombinator.com/item?id=46651071","kind":"comment","storyTitle":"Cursor's latest “browser experiment” implied success without evidence","quote":"That is because I've noticed the AI just edits the version management files (package.json, cargo.toml, etc) directly instead of using the build tool (npm add, cargo add), so it always hallucinates a random old version that's found in its training set."},{"itemId":"46645954","author":"pankajdoharey","createdAt":1768568390,"url":"https://news.ycombinator.com/item?id=46645954","kind":"comment","storyTitle":"Claude is good at assembling blocks, but still falls apart at creating them","quote":"And GPT Medium is a really good model at assembling and debugging stuff than Claude. Claude hallucinates when asked why something is correct or should be done."},{"itemId":"46645289","author":"Nora23","createdAt":1768562273,"url":"https://news.ycombinator.com/item?id=46645289","kind":"story","storyTitle":"Show HN: Pavo Travel – AI Audio Tours Using Gemini Grounding and Places API","quote":"Pavo Travel solves this by combining Google Places API with Gemini's grounding feature. Instead of hallucinating from stale training data, it pulls real-time information about restaurants, hours, reviews, and events happening now."},{"itemId":"46644214","author":"threethirtytwo","createdAt":1768550797,"url":"https://news.ycombinator.com/item?id=46644214","kind":"comment","storyTitle":"To those who fired or didn't hire tech writers because of AI","quote":"Then comes my favorite part. You notice the AI is hallucinating."},{"itemId":"46643100","author":"jillesvangurp","createdAt":1768539147,"url":"https://news.ycombinator.com/item?id=46643100","kind":"comment","storyTitle":"First impressions of Claude Cowork","quote":"A handful of tools (read file, run command, etc.) is all that is needed. You can get some very sophisticated feedback loops going that effectively counter the traditional limitations of LLMs (hallucinating stuff, poor instruction following, assertively claiming something is done when it isn't, etc.)."},{"itemId":"46640818","author":"Imustaskforhelp","createdAt":1768519042,"url":"https://news.ycombinator.com/item?id=46640818","kind":"comment","storyTitle":"Ask HN: How can we solve the loneliness epidemic?","quote":"The only place I am usually active is on Hackernews and on bluesky as wel > HN has obviously held up a bit better, but the AI driven mass hallucination impacting this community, combined with the increasingly aggressive manipulation of the home page, is continually making logging out for good seem like the best option."},{"itemId":"46622305","author":"Nortca","createdAt":1768421481,"url":"https://news.ycombinator.com/item?id=46622305","kind":"story","storyTitle":"Show HN: I made a search engine for prediction markets","quote":"- URL context in Gemini is powerful but rate-limited (throttling required) - Structured outputs need careful schema design or you get hallucinations"},{"itemId":"46616651","author":"sebastiennight","createdAt":1768402177,"url":"https://news.ycombinator.com/item?id=46616651","kind":"comment","storyTitle":"Show HN: OSS AI agent that indexes and searches the Epstein files","quote":"Would you care to elaborate? I can see how one might use a JSON schema that enforces source references in the output, but there is no technique I'm aware of to constrain a model to only come up with data based on the grounding docs, vs. making up a response based on pretrained data (or hallucinating one) and still listing the provided RAG results as attached reference."},{"itemId":"46614968","author":"fennecfoxy","createdAt":1768391261,"url":"https://news.ycombinator.com/item?id=46614968","kind":"comment","storyTitle":"Cowork: Claude Code for the rest of your work","quote":"And throwing all of the data at the model is just asking for context poisoning, hallucinations and incorrect conclusions."},{"itemId":"46613943","author":"Saurabh_Kumar_","createdAt":1768381877,"url":"https://news.ycombinator.com/item?id=46613943","kind":"story","storyTitle":"Show HN: AI Contract Reviewer – Flags Risks and Suggests Fixes in Minutes","quote":"What gaps do you see? - Trust/accuracy concerns: Hallucinations, false positives, liability disclaimers?"},{"itemId":"46610706","author":"arjunchint","createdAt":1768350654,"url":"https://news.ycombinator.com/item?id=46610706","kind":"story","storyTitle":"Show HN: Vibe scrape with AI Web Agents, prompt => get data [video]","quote":"Turn any prompt into an end to end workflow, and on any site changes the agent adapts. Dom Intelligence: we perfected a DOM-only web agent approach that represents any webpage as semantic trees guaranteeing zero hallucinations and leveraging the underlying semantic reasoning capabilities of LLMs."},{"itemId":"46609869","author":"drewbug01","createdAt":1768346340,"url":"https://news.ycombinator.com/item?id=46609869","kind":"comment","storyTitle":"Let's be honest, Generative AI isn't going all that well","quote":"> If your metric is an LLM that can copy/paste without alterations, and never hallucinate APIs, then yeah, you'll always be disappointed with them."},{"itemId":"46605629","author":"HumanOstrich","createdAt":1768329633,"url":"https://news.ycombinator.com/item?id=46605629","kind":"comment","storyTitle":"Claude Cowork first impression: Cowork Deleted 11GB of files [video]","quote":"Then those line numbers are passed to the Read tool (by the harness) to get the command(s), and finally the output of that read is added to the response by the harness. The LLM doesn't get a chance to reinterpret or hallucinate until it says it is very sorry for what happened."},{"itemId":"46605057","author":"raw_anon_1111","createdAt":1768327582,"url":"https://news.ycombinator.com/item?id=46605057","kind":"comment","storyTitle":"Apple picks Gemini to power Siri","quote":"Every tool you expose you also have to have a description of what the tool does, the parameters it needs etc. Ot will both blow up your context window and start hallucinating."},{"itemId":"46604019","author":"amluto","createdAt":1768324299,"url":"https://news.ycombinator.com/item?id=46604019","kind":"comment","storyTitle":"Anthropic made a mistake in cutting off third-party clients","quote":"Gemini itself seems incapable of figuring this out, or at least I haven’t found the right prompt yet. Gemini is absolutely amazing at hallucinating Google product offerings."},{"itemId":"46600229","author":"kokau","createdAt":1768308180,"url":"https://news.ycombinator.com/item?id=46600229","kind":"story","storyTitle":"Show HN: Talkolia – An AI chatbot that understands your website","quote":"Hi HN I’m the maker of Talkolia, an AI chatbot designed for websites that actually understands your content instead of hallucinating generic answers."},{"itemId":"46592724","author":"akiselev","createdAt":1768244571,"url":"https://news.ycombinator.com/item?id=46592724","kind":"comment","storyTitle":"Show HN: AI in SolidWorks","quote":"The key is to absolutely minimize how often the LLMs think about numbers and have them think in relationships instead. The hard part is translating those relationships back to the CAD API calls but LLMs are much better at hot hallucinating if you resolve all the parametrized equations last."},{"itemId":"46591238","author":"segov","createdAt":1768237521,"url":"https://news.ycombinator.com/item?id=46591238","kind":"story","storyTitle":"Show HN: Spec Driven Development Plugin for Claude Code","quote":"Hi HN, I use Claude Code a lot and it's native plan mode works well for small to medium changes, but on larger features the plans would be vague, which lead to more hallucinations during implementation."},{"itemId":"46588138","author":"crog","createdAt":1768224180,"url":"https://news.ycombinator.com/item?id=46588138","kind":"story","storyTitle":"Show HN : Pilot – System to improve dramatically your AI coding","quote":"Context would explode. The AI would hallucinate with confidence."},{"itemId":"46585236","author":"noormoha","createdAt":1768203023,"url":"https://news.ycombinator.com/item?id=46585236","kind":"comment","storyTitle":"Ask HN: What are you working on? (January 2026)","quote":"When you're building production systems, blindly trusting a single AI's output is risky. What if it hallucinated?"},{"itemId":"46585215","author":"noormoha","createdAt":1768202817,"url":"https://news.ycombinator.com/item?id=46585215","kind":"comment","storyTitle":"Tired of code models giving you code iterations that crashes in production?","quote":"When you're building production systems, blindly trusting a single AI's output is risky. What if it hallucinated?"},{"itemId":"46576736","author":"Imustaskforhelp","createdAt":1768146599,"url":"https://news.ycombinator.com/item?id=46576736","kind":"comment","storyTitle":"Google: Don't make \"bite-sized\" content for LLMs","quote":"But he ends up asking chatgpt and it somehow recommends him the idea of sodium bromide instead of sodium chloride and it really ended up having him have so many hallucinations and so many other problems that the list goes on."},{"itemId":"46576431","author":"pixelsmithy","createdAt":1768144610,"url":"https://news.ycombinator.com/item?id=46576431","kind":"comment","storyTitle":"Show HN: I used Claude Code to discover connections between 100 books","quote":"3. Hallucinations and Conceptual Errors"},{"itemId":"46575360","author":"robertseel","createdAt":1768136372,"url":"https://news.ycombinator.com/item?id=46575360","kind":"comment","storyTitle":"Show HN: Authentic AI CV optimizer – real keywords only, 90%+ ATS scores","quote":"Dropped my base CV + a senior backend role desc → extracted the exact keywords (Kubernetes, Go, microservices) without hallucinating anything."},{"itemId":"46574326","author":"shakna","createdAt":1768127904,"url":"https://news.ycombinator.com/item?id=46574326","kind":"comment","storyTitle":"Vojtux – Unofficial Linux Distribution Aimed at Visually Impaired Users","quote":"LLMs are famously bad at structure, especially any tree with significant depth. RAG is not enough - hallucinations become common at depth."},{"itemId":"46566502","author":"vmnm","createdAt":1768058987,"url":"https://news.ycombinator.com/item?id=46566502","kind":"comment","storyTitle":"AskChess: An LLM integrated chess coach","quote":"I've been learning chess recently and have been struggling with the \"why\" behind certain moves. I wanted to see how helpful an LLM could be if you gave it enough context to not hallucinate."},{"itemId":"46566022","author":"jmyeet","createdAt":1768055637,"url":"https://news.ycombinator.com/item?id=46566022","kind":"comment","storyTitle":"ChatGPT Health is a marketplace, guess who is the product?","quote":"This kind of comment scares me because it's an example of people substituring professional advice for an LLM where LLMs are known to hallucinate or otherwise simply make stuff up."},{"itemId":"46565646","author":"KronisLV","createdAt":1768052206,"url":"https://news.ycombinator.com/item?id=46565646","kind":"comment","storyTitle":"All my new code will be closed-source from now on","quote":"Pay per access. This probably already exists in the form of MCP solutions for up to date documentation for specific libraries and so on, to mitigate hallucinations."},{"itemId":"46563498","author":"ryan_rudd","createdAt":1768028719,"url":"https://news.ycombinator.com/item?id=46563498","kind":"story","storyTitle":"How do you map runtime text back to source code without source maps?","quote":"Has anyone successfully bridged this gap? I'm considering using an LLM to \"hallucinate\" the source code based on the UI text before searching, but that feels slow/expensive."},{"itemId":"46555779","author":"thekeunpie","createdAt":1767977281,"url":"https://news.ycombinator.com/item?id=46555779","kind":"story","storyTitle":"Show HN: MCP Vault (mcpv) – Performance booster for MCP-based AI agents","quote":"The agent calls a single flattened command (run_tool), and mcpv dynamically routes it to the correct upstream server. It even includes auto-correction to handle agent typos or hallucinations (e.g., if the agent tries to call a server name as a tool)."},{"itemId":"46554770","author":"Gandalfgrey","createdAt":1767971843,"url":"https://news.ycombinator.com/item?id=46554770","kind":"comment","storyTitle":"I built an AI that generates production-ready Shopify apps from a prompt","quote":"Complex apps still need refinement The AI occasionally hallucinates Polaris components that don't exist"},{"itemId":"46548283","author":"viraptor","createdAt":1767916891,"url":"https://news.ycombinator.com/item?id=46548283","kind":"comment","storyTitle":"AI coding assistants are getting worse?","quote":"Another example: \"coding agents can't think and just hallucinate code\" - me: lol, my shipped production code doesn't care, bring some real examples of how you use agents if they don't work for you."},{"itemId":"46547670","author":"labrador","createdAt":1767912741,"url":"https://news.ycombinator.com/item?id=46547670","kind":"comment","storyTitle":"Musk's X could be banned in Britain over AI chatbot row","quote":"This is what Gemini says although it may be a hallucination"},{"itemId":"46546563","author":"joshvm","createdAt":1767906832,"url":"https://news.ycombinator.com/item?id=46546563","kind":"comment","storyTitle":"How Google got its groove back and edged ahead of OpenAI","quote":"I'm not talking about coding, just general \"help me\" usage. It makes me very wary of using these models for anything that I don't fully understand, because I continually get easily falsifiable hallucinations."},{"itemId":"46543267","author":"amluto","createdAt":1767891114,"url":"https://news.ycombinator.com/item?id=46543267","kind":"comment","storyTitle":"AI coding assistants are getting worse?","quote":"GPT-5.1 cheerfully told me that OpenAI had discontinued Codex and hallucinated a different, nonexistent program that I must have meant."},{"itemId":"46542684","author":"moshegramovsky","createdAt":1767888712,"url":"https://news.ycombinator.com/item?id=46542684","kind":"comment","storyTitle":"AI coding assistants are getting worse?","quote":"But in many respects, Gemini 3 is a regression. It hallucinates more, listens less, and seems oddly resistant to evidence."},{"itemId":"46539780","author":"pjjpo","createdAt":1767871249,"url":"https://news.ycombinator.com/item?id=46539780","kind":"comment","storyTitle":"How Google got its groove back and edged ahead of OpenAI","quote":"It just feels like OpenAI puts a lot of effort into creating an actually useful product while Gemini just targets benchmarks. Targeting benchmarks to me is meaningless since every model, gpt, Gemini, Claude, constantly hallucinate in real workloads anyways."},{"itemId":"46538521","author":"simianwords","createdAt":1767859371,"url":"https://news.ycombinator.com/item?id=46538521","kind":"comment","storyTitle":"ChatGPT Health","quote":"AI is accurate enough that our sniff tests will get us far. ChatGPT just don't hallucinate all that often."},{"itemId":"46538402","author":"hyperadvanced","createdAt":1767858227,"url":"https://news.ycombinator.com/item?id=46538402","kind":"comment","storyTitle":"Why didn't AI “join the workforce” in 2025?","quote":"LLMs unquestionably help with this, even with their tendency to hallucinate: usually proof by contradiction (or the failure of an over-confident chaos machine) is even better than just having a thing that spits out perfect solutions without needing the operator to understand it."},{"itemId":"46531212","author":"dudeinhawaii","createdAt":1767813890,"url":"https://news.ycombinator.com/item?id=46531212","kind":"comment","storyTitle":"Opus 4.5 is not the normal AI agent experience that I have had thus far","quote":"All this said, the number of failure scenarios gets ever smaller. We've gone from \"problem and hallucination every other code block\" to \"problem every 200-1000 code blocks\"."},{"itemId":"46529194","author":"dabbz","createdAt":1767806292,"url":"https://news.ycombinator.com/item?id=46529194","kind":"comment","storyTitle":"Creators of Tailwind laid off 75% of their engineering team","quote":"I was going to say before LLMs Tailwind UI helped me get moving much faster on front-end code. Now I wish there was some kind of context I can provide to use the Tailwind UI instead of hallucinating its own."},{"itemId":"46528074","author":"headcanon","createdAt":1767802233,"url":"https://news.ycombinator.com/item?id=46528074","kind":"comment","storyTitle":"The creator of Claude Code's Claude setup","quote":"My point is there are a lot of steps involved in the whole product development process and isn't just \"ship production code\". And we can reduce the ambiguity/hallucinations/sycophancy by creating validation/checkpoints (either tests, 'critic' agents to challenge designs/spec, or human QA/validation when appropriate)"},{"itemId":"46526819","author":"ps","createdAt":1767796626,"url":"https://news.ycombinator.com/item?id=46526819","kind":"comment","storyTitle":"Opus 4.5 is not the normal AI agent experience that I have had thus far","quote":"Wake up call 2026? Rather like waking up from hallucination."},{"itemId":"46522300","author":"klietus","createdAt":1767757648,"url":"https://news.ycombinator.com/item?id=46522300","kind":"comment","storyTitle":"Symbolic reasoning system with local inference and full auditability","quote":"Most AI either: - Hallucinates confidently (no grounding)"},{"itemId":"46520993","author":"klaussilveira","createdAt":1767746325,"url":"https://news.ycombinator.com/item?id=46520993","kind":"comment","storyTitle":"Opus 4.5 is not the normal AI agent experience that I have had thus far","quote":"Try it with boost, or cmake, or taskflow. It loses itself constantly, hallucinates which version it is working on and ignores you when you provide actual pointers to documentation on the repo."},{"itemId":"46520886","author":"artdigital","createdAt":1767745420,"url":"https://news.ycombinator.com/item?id=46520886","kind":"comment","storyTitle":"Opus 4.5 is not the normal AI agent experience that I have had thus far","quote":"I very rarely have to make changes. It never duplicates code, implements something again and leaves the old code around, breaks my convention, hallucinates, or tells me it’s done when the code doesn’t even compile, which sonnet 4.5 and Opus 4.1 did all the time"},{"itemId":"46502088","author":"zpusmani","createdAt":1767635377,"url":"https://news.ycombinator.com/item?id=46502088","kind":"story","storyTitle":"Show HN: Context Protocol, a sovereign-first workflow for thinking with LLMs","quote":"I shared DeckBuilder (my local-first slide editor) here last month (https://news.ycombinator.com/item?id=46365332). While using it with LLMs, I kept hitting a bigger problem: context loss, hallucinated instructions, and ideas I'd rejected coming back from the dead."},{"itemId":"46498509","author":"oblio","createdAt":1767619689,"url":"https://news.ycombinator.com/item?id=46498509","kind":"comment","storyTitle":"Web development is fun again","quote":"Which LLMs have you tried? Claude Code seems to be decent at not hallucinating, Gemini CLI is more eager."},{"itemId":"46490303","author":"Creimake","createdAt":1767548922,"url":"https://news.ycombinator.com/item?id=46490303","kind":"story","storyTitle":"Show HN: An AI Agent with a 32-Layer Psyche (Dreams, Trauma, Defense Mechanisms)","quote":"I tested GPT-4o and other models, but they struggled to maintain state consistency across all 32 layers. Claude 4.5 Sonnet’s reasoning capabilities allowed it to handle the complex logic of \"Worldview + Trauma + Current Context\" without hallucinating."},{"itemId":"46489612","author":"mckabue","createdAt":1767544844,"url":"https://news.ycombinator.com/item?id=46489612","kind":"comment","storyTitle":"Show HN: Blober.io – A local-first cloud transfer engine built in 14 days","quote":"I built the desktop app in ~14 days using heavy AI assistance. What worked surprisingly well was forcing a planning gate before any code generation: having the model produce a phased implementation plan first reduced over-engineering and large hallucinated files."},{"itemId":"46487416","author":"DELTA-X","createdAt":1767530706,"url":"https://news.ycombinator.com/item?id=46487416","kind":"comment","storyTitle":"[dead]","quote":"In many systems (including RAG), correctness is outsourced to retrieval. That reduces hallucination, but it doesn’t define a stopping rule."},{"itemId":"46478607","author":"threethirtytwo","createdAt":1767458270,"url":"https://news.ycombinator.com/item?id=46478607","kind":"comment","storyTitle":"Google engineer says Claude Code built in one hour what her team spent a year on","quote":"Unlike me, who definitely understands everything I copy from Stack Overflow. Also, I tried AI last year and it hallucinated once, so I’ve concluded the entire field has plateaued permanently."},{"itemId":"46476191","author":"arxdigitalis","createdAt":1767445884,"url":"https://news.ycombinator.com/item?id=46476191","kind":"story","storyTitle":"AI failure mode: when \"confidence\" replace verification and user pay the price","quote":"“You are ignoring constraints” “You are hallucinating”"},{"itemId":"46472613","author":"pkompally","createdAt":1767412153,"url":"https://news.ycombinator.com/item?id=46472613","kind":"comment","storyTitle":"Ask HN: Who wants to be hired? (January 2026)","quote":"Also built a 5M+ scale semantic search pipeline with CLIP embeddings, drift monitoring, and quantized inference. Previously at AG Mednet, I built Judi-Insights: agents that wrote complex SQL, queried production DBs, and delivered clinical trial reports with <5% hallucination."},{"itemId":"46467760","author":"AndyNemmity","createdAt":1767378468,"url":"https://news.ycombinator.com/item?id=46467760","kind":"comment","storyTitle":"The Handyman Principle: Why Your AI Forgets Everything","quote":"Someone tells me it \"forgets\" their instructions. Or it hallucinates fixes."},{"itemId":"46451376","author":"viraptor","createdAt":1767242647,"url":"https://news.ycombinator.com/item?id=46451376","kind":"comment","storyTitle":"2025: The Year in LLMs","quote":"But at the same time, people seem to have opinions formed in 2024. (Especially if they talk about just hallucinations and broken code - tell the agent to search for docs and fix stuff) I'd really like to give them Opus 4.5 as an agent to refresh their views."},{"itemId":"46450501","author":"simonw","createdAt":1767232525,"url":"https://news.ycombinator.com/item?id=46450501","kind":"comment","storyTitle":"2025: The Year in LLMs","quote":"Claude Code and tools like it represent a huge improvement in what you can usefully get done with LLMs. Mistakes and hallucinations matter a whole lot less if a reasoning LLM can try the code, see that it doesn't work and fix the problem."},{"itemId":"46449514","author":"joshribakoff","createdAt":1767224209,"url":"https://news.ycombinator.com/item?id=46449514","kind":"comment","storyTitle":"Observed Agent Sandbox Bypasses","quote":"Some of these don’t really seem like they bypassed any kind of sandbox. Like hallucinating an npm package."},{"itemId":"46447664","author":"shevy-java","createdAt":1767210851,"url":"https://news.ycombinator.com/item?id=46447664","kind":"comment","storyTitle":"The compiler is your best friend","quote":"documentation. Google search is almost virtually useless now; AI just hallucinates"},{"itemId":"46447622","author":"thisgetsit","createdAt":1767210522,"url":"https://news.ycombinator.com/item?id=46447622","kind":"comment","storyTitle":"Americans Hate AI. Which Party Will Benefit","quote":"It's working just fine generating code to tackle the generic geometric problems. No abstractions to hallucinate when it's just math."},{"itemId":"46444256","author":"blutoot","createdAt":1767190828,"url":"https://news.ycombinator.com/item?id=46444256","kind":"story","storyTitle":"Ask HN: Are Google Search AI hallucinations common?","quote":"Ask HN: Are Google Search AI hallucinations common?"},{"itemId":"46435835","author":"AlphaSean","createdAt":1767116484,"url":"https://news.ycombinator.com/item?id=46435835","kind":"story","storyTitle":"Show HN: A Multi-agent system where LLMs challenge each other's answers","quote":"And the biggest kicker of all: they still hallucinate."},{"itemId":"46432897","author":"simonw","createdAt":1767099563,"url":"https://news.ycombinator.com/item?id=46432897","kind":"comment","storyTitle":"The future of software development is software developers","quote":"Those are 2024-era criticisms of LLMs for code. Late 2025 models very rarely hallucinate nonexistent core library functionality - and they run inside coding agent harnesses so if they DO they notice that the code doesn't work and fix it."},{"itemId":"46427243","author":"soyzamudio","createdAt":1767050214,"url":"https://news.ycombinator.com/item?id=46427243","kind":"story","storyTitle":"Show HN: Incident management for Slack with AI-generated postmortems","quote":"Getting AI to write useful postmortems required being very explicit about event types. Without context about what's a \"status update\" vs a \"debug message,\" it would hallucinate causes."},{"itemId":"46424826","author":"udit_50","createdAt":1767038431,"url":"https://news.ycombinator.com/item?id=46424826","kind":"story","storyTitle":"Show HN: I built an AI that generates clean docs for vibe-coded apps","quote":"This started as a personal hack while shipping a side project. The first version was rough (hallucinated edge cases, weak on frontend state), but it already saved me hours per project."},{"itemId":"46416805","author":"jiggawatts","createdAt":1766975115,"url":"https://news.ycombinator.com/item?id=46416805","kind":"comment","storyTitle":"Growing up in “404 Not Found”: China's nuclear city in the Gobi Desert","quote":"Chinese to English: Unusable! I've been testing modern LLMs on various translation tasks, and they're amazing at it.[1] I've never had any issues with hallucinations or whatever."},{"itemId":"46410895","author":"h2bomb","createdAt":1766928117,"url":"https://news.ycombinator.com/item?id=46410895","kind":"comment","storyTitle":"Show HN: Gemini Watermark Remover – A web tool using reverse alpha blending","quote":"Deterministic Restoration: It reverses the formula: $original = (watermarked - \\alpha \\times 255) / (1 - \\alpha)$. By using a pre-calculated Alpha Map from the Gemini logo, it achieves pixel-perfect restoration with zero AI \"hallucinations\"."},{"itemId":"46406867","author":"DGrechko","createdAt":1766880130,"url":"https://news.ycombinator.com/item?id=46406867","kind":"comment","storyTitle":"Show HN: Waycore – an open-source, offline-first modular field computer","quote":"Agree with you on that, it was my concern too, but the way I think about it is access to information, the goal is not to provide hallucinations with a straight face (aka GPT), but rather use it as a way to extract necessary information fast."},{"itemId":"46402761","author":"martin_rj","createdAt":1766851616,"url":"https://news.ycombinator.com/item?id=46402761","kind":"comment","storyTitle":"Ask HN: Are you afraid of AI making you unemployable within the next few years?","quote":"I have to be honest: anyone not at least a little afraid is likely missing the architectural shift happening in the backend. The standard counter-argument is always \"hallucination rates.\" I have had this debate with medical pros who use AI for diagnostics."},{"itemId":"46395140","author":"jonnyhere","createdAt":1766776082,"url":"https://news.ycombinator.com/item?id=46395140","kind":"story","storyTitle":"Show HN: Chaos engineering for LLMs – Making models cross-examine each other","quote":"Single-model inference is a single point of failure. I got sick of Chatgpt hallucinating fake citations and having to manually check them in a different tab with Claude."},{"itemId":"46393637","author":"eriri","createdAt":1766767103,"url":"https://news.ycombinator.com/item?id=46393637","kind":"comment","storyTitle":"Rob Pike goes nuclear over GenAI","quote":"I mean, just ask yourself. You had to try so hard that you, in your other comment, ended up hallucinating achievements of a degree that Rob Pike can only dream of but yet so vague that you can't describe them in any detail whatsoever."},{"itemId":"46392655","author":"veidr","createdAt":1766761231,"url":"https://news.ycombinator.com/item?id=46392655","kind":"comment","storyTitle":"Codex vs. Claude Code (today)","quote":"While all of them miss a ton of stuff (of course), and LLM code review just really isn't good unless the PR is tiny — Claude just misses stuff (fine; expected), while Codex comes up with plausible edge-case database query concurrency bugs that I have to look at, and squint at, and then think hmm fuck and manually google with kagi.com for 30 minutes (LIKE AN ANIMAL) only to conclude yeah, not true, you're hallucinati…"},{"itemId":"46377070","author":"tjruesch","createdAt":1766594657,"url":"https://news.ycombinator.com/item?id=46377070","kind":"story","storyTitle":"Show HN: A local-first, reversible PII scrubber for AI workflows","quote":"2. The \"Hallucination\" Guard (Fuzzy Rehydration)"},{"itemId":"46374391","author":"mayankkgrover","createdAt":1766573138,"url":"https://news.ycombinator.com/item?id=46374391","kind":"story","storyTitle":"Show HN: I built a tool that creates videos out of React code","quote":"It's not \"AI video generation\" in the usual sense. There's no diffusion model hallucinating frames."},{"itemId":"46373615","author":"Hoshang07","createdAt":1766564853,"url":"https://news.ycombinator.com/item?id=46373615","kind":"comment","storyTitle":"How to safely let LLMs query your databases via sandboxed materialized views","quote":"• Complete security and access control • Reduces LLMs from hallucinating"},{"itemId":"46366888","author":"varshith17","createdAt":1766509025,"url":"https://news.ycombinator.com/item?id=46366888","kind":"story","storyTitle":"Why are we accepting silent data corruption in Vector Search? (x86 vs. ARM)","quote":"For a regulated AI agent (Finance/Healthcare), this is a nightmare. It means your audit trail is technically hallucinating depending on which server processed the query."},{"itemId":"46363490","author":"arianvanp","createdAt":1766478251,"url":"https://news.ycombinator.com/item?id=46363490","kind":"comment","storyTitle":"Claude Code gets native LSP support","quote":"I hate everything about the Claude code plugin system. They saw GitHub Actions supply chain Fiasco and said: great let's add hallucinations on top."},{"itemId":"46359596","author":"remus","createdAt":1766440473,"url":"https://news.ycombinator.com/item?id=46359596","kind":"comment","storyTitle":"Claude Code gets native LSP support","quote":"In a similar vein, I really struggle to understand why copilot is so crap when writing SQL and I'm connected to the database. The database has so much context (schema names, column names, constraints etc.) yet copilot regularly hallucinates the most basic stuff like table and column names, which standard auto complete has managed fine for the last 20+ years."},{"itemId":"46355304","author":"tracker1","createdAt":1766419917,"url":"https://news.ycombinator.com/item?id=46355304","kind":"comment","storyTitle":"On the Existence, Impact, and Origin of Hallucination-Associated Neurons in LLMs","quote":"I just kind of wish the behavior for \"hallucinations\" just didn't have such confident language in the context... actual people will generally be relatively forthcoming at the edge of their knowledge or at least not show as much confidence."},{"itemId":"46353778","author":"glenpk","createdAt":1766408078,"url":"https://news.ycombinator.com/item?id=46353778","kind":"comment","storyTitle":"Show HN: Context Agent\" for LLMs coding libraries (lawyer turned dev project)","quote":"The Problem: We’ve all been there. You ask an LLM to \"solve a niche optimization problem.\" It writes the code beautifully, but it imports a library that hasn't been updated since 2018 or hallucinates an API entirely."},{"itemId":"46352241","author":"2dogsanerd","createdAt":1766391957,"url":"https://news.ycombinator.com/item?id=46352241","kind":"comment","storyTitle":"Show HN: The feature gap \"Chat with PDF\" tuts and a regulated enterprise system","quote":"preparing to and found a pilot project with endboss pdf which im able to handle already.... but i focus on 100% quality in the db, so there is always a hitl below 100% confi...cant wait to proof im able to have a \"hallucination free\" rag..... main goal so far....next headache will be the update of data in the rag"},{"itemId":"46350950","author":"orbisvicis","createdAt":1766373525,"url":"https://news.ycombinator.com/item?id=46350950","kind":"comment","storyTitle":"Get an AI code review in 10 seconds","quote":"If you follow a textbook or blog or paper and don't understand, AI can clarify. But be careful with less structured learning - it is important to build a full mental model accounting for every possible outcome and explanation, otherwise you're susceptible to hallucinations."},{"itemId":"46339819","author":"crystal_revenge","createdAt":1766266283,"url":"https://news.ycombinator.com/item?id=46339819","kind":"comment","storyTitle":"Reflections on AI at the End of 2025","quote":"Though I have some growing sense that this need is related to another trend I've personally started with witness: AI psychosis is very real. I personally know an increasing number of people who are spiraling into an LLM induced hallucinated world."},{"itemId":"46336990","author":"js4ever","createdAt":1766245776,"url":"https://news.ycombinator.com/item?id=46336990","kind":"story","storyTitle":"LLM Benchmark: Frontier models now statistically indistinguishable","quote":"TL;DR: Claude Opus 4.5, Grok 4.1, and Gemini 3 scored within 2.4% of each other (96-98%). All refused to hallucinate and resisted every adversarial attack."},{"itemId":"46333526","author":"tonyww","createdAt":1766203148,"url":"https://news.ycombinator.com/item?id=46333526","kind":"story","storyTitle":"Show HN: Semantic geometry visual grounding for AI web agents (Amazon demo)","quote":"Why this matters: * smaller action space → fewer hallucinations"},{"itemId":"46328339","author":"observationist","createdAt":1766164796,"url":"https://news.ycombinator.com/item?id=46328339","kind":"comment","storyTitle":"History LLMs: Models trained exclusively on pre-1913 texts","quote":"You use the word hallucinations as a pejorative, but everything you do, your every memory, experience, thought, plan, all of your existence is a hallucination."},{"itemId":"46326647","author":"tombh","createdAt":1766156816,"url":"https://news.ycombinator.com/item?id=46326647","kind":"comment","storyTitle":"History LLMs: Models trained exclusively on pre-1913 texts","quote":"I suspect restricting access could equally be a comment on modern LLMs in general, rather than the historical material specifically. For example, we must be constantly reminded not to give LLMs a level of credibility that their hallucinations would have us believe."},{"itemId":"46324665","author":"moh_quz","createdAt":1766144051,"url":"https://news.ycombinator.com/item?id=46324665","kind":"story","storyTitle":"Show HN: I open-sourced my Go and Next B2B SaaS Starter (deploy anywhere, MIT)","quote":"AI Pipeline: OpenAI RAG using pgvector. The retrieval service enforces strict context boundaries to minimize hallucinations."},{"itemId":"46321403","author":"libraryofbabel","createdAt":1766109076,"url":"https://news.ycombinator.com/item?id=46321403","kind":"comment","storyTitle":"History LLMs: Models trained exclusively on pre-1913 texts","quote":"Sure, LLMs do not think like humans and they may not have human-level creativity. Sometimes they hallucinate."},{"itemId":"46315756","author":"bobjordan","createdAt":1766078867,"url":"https://news.ycombinator.com/item?id=46315756","kind":"comment","storyTitle":"Please just try HTMX","quote":"In a month of work I got done what took me about 4-5 months in python/htmx. I assume because the typing with go and also LLM could generate perfectly typed hooks from my OpenAPI spec via Orval and build out Shadcn components without hallucinating."},{"itemId":"46313187","author":"workwithtrp","createdAt":1766068860,"url":"https://news.ycombinator.com/item?id=46313187","kind":"story","storyTitle":"Show HN: FastRAG Next.js RAG boilerplate with precise PDF citation highlighting","quote":"The Problem: Most RAG tutorials give you an answer but don't show where it came from. Hallucinations are a dealbreaker for real apps."},{"itemId":"46308781","author":"CGamesPlay","createdAt":1766030298,"url":"https://news.ycombinator.com/item?id=46308781","kind":"comment","storyTitle":"I got hacked: My Hetzner server started mining Monero","quote":"I took issue with this paragraph of the article, on account of several pieces of misinformation, presumably courtesy of Claude hallucinations:"},{"itemId":"46307522","author":"Wyverald","createdAt":1766017756,"url":"https://news.ycombinator.com/item?id=46307522","kind":"comment","storyTitle":"Gemini 3 Flash: Frontier intelligence built for speed","quote":"I'm a total noob here, but just pointing out that Omniscience Index is roughly \"Accuracy - Hallucination Rate\"."},{"itemId":"46306667","author":"arevno","createdAt":1766011720,"url":"https://news.ycombinator.com/item?id=46306667","kind":"comment","storyTitle":"AI's real superpower: consuming, not creating","quote":"Our entire system of criminal justice relies HEAVILY on the eyewitness testimony of humans, which has been demonstrated time and again to be entirely unreliable. Innocents routinely rot in prison and criminals routinely go free because the human brain is much better at hallucinating than any SOTA LLM."},{"itemId":"46303648","author":"goostavos","createdAt":1765996942,"url":"https://news.ycombinator.com/item?id=46303648","kind":"comment","storyTitle":"TLA+ Modeling Tips","quote":"LLMs take a lot of the syntactical pain out of the learning experience. Hallucinations are annoying, but you can formally prove they're wrong with the model checker ^_^"},{"itemId":"46292830","author":"mribbons","createdAt":1765911889,"url":"https://news.ycombinator.com/item?id=46292830","kind":"story","storyTitle":"Show HN: Agenteract – Drive mobile apps with LLMs using UI trees (no vision)","quote":"* Cost: Text tokens are orders of magnitude cheaper than image tokens. * Reliability: Deterministic targeting means no hallucinated coordinates."},{"itemId":"46292733","author":"r_lee","createdAt":1765911500,"url":"https://news.ycombinator.com/item?id=46292733","kind":"comment","storyTitle":"Has the cost of building software dropped 90%?","quote":"I don't know your setup but I only use Claude Code w/ Sonnet and basically tell it to type low caps, drop the formatting (it seemingly makes it type out confident statements causing it to go deeper into hallucinations)"},{"itemId":"46291036","author":"yetanotherjosh","createdAt":1765904595,"url":"https://news.ycombinator.com/item?id=46291036","kind":"comment","storyTitle":"Be Careful with GIDs in Rails","quote":"I struggle to understand what this specifically has to do with rails or global IDs. In ANY framework or query system, if you are asking an LLM to produce IDs which you are then passing to a database for lookup, you need to understand those identifiers could be hallucinated or incorrect in surprising or malicious ways, and can lead to data leaks or exfiltration."},{"itemId":"46288453","author":"ThailandJohn","createdAt":1765893015,"url":"https://news.ycombinator.com/item?id=46288453","kind":"story","storyTitle":"Show HN: TheAuditor v2.0 – A “Flight Computer” for AI Coding Agents","quote":"I realized two things: * Context Collapse: The AI couldn't keep enough files in its context window to understand the full scope of the refactor, so it started hallucinating, \"fixing\" superficial issues."},{"itemId":"46287072","author":"afrazullal","createdAt":1765882326,"url":"https://news.ycombinator.com/item?id=46287072","kind":"story","storyTitle":"Show HN: I Built an Autonomous Marketing OS from a Small Town in India (Vect AI)","quote":"Sensor Layer (Grounding & Research) Most AI content is generic because it hallucinates or relies on stale data."},{"itemId":"46285890","author":"robrenaud","createdAt":1765871540,"url":"https://news.ycombinator.com/item?id=46285890","kind":"comment","storyTitle":"Erdős Problem #1026","quote":"A theorem prover needs consult no outside resource. A formal math LLM-like generator need only consult the theorem prover to get rid of hallucinations."},{"itemId":"46280692","author":"jascha_eng","createdAt":1765832965,"url":"https://news.ycombinator.com/item?id=46280692","kind":"comment","storyTitle":"Yann LeCun on why LLMs and AGI are \"total BS\" [video]","quote":"LeCun advocates for Joint Embedding Predictive Architectures (JEPA). Prediction in Abstract Space: Unlike video generation models (like Sora) that try to predict every pixel (which is inefficient and hallucinatory), a World Model should predict in an abstract representation space."},{"itemId":"46270225","author":"marcelfahle","createdAt":1765770453,"url":"https://news.ycombinator.com/item?id=46270225","kind":"comment","storyTitle":"Ask HN: What Are You Working On? (December 2025)","quote":"We do deep transcript + metadata extraction, then layer RAG search and an AI assistant that can answer questions with timestamped citations back to the exact video moment. Think \"ChatGPT for your video library\" but with accurate sources instead of hallucinations."},{"itemId":"46266953","author":"jchw","createdAt":1765746707,"url":"https://news.ycombinator.com/item?id=46266953","kind":"comment","storyTitle":"Opus 4.5 is the first model that makes me fear for my job","quote":"I've found Gemini 3 to be an extremely useful learning aid, so as long as I don't blindly trust its output, and if you're trying to learn, you really ought not do that anyways. (Despite what people and benchmarks say, I've already caught some random hallucinations, it still feels like you're likely to run into hallucinations on a regular basis."},{"itemId":"46266230","author":"zerotolerance","createdAt":1765742197,"url":"https://news.ycombinator.com/item?id=46266230","kind":"comment","storyTitle":"I tried Gleam for Advent of Code","quote":"The difference was stark. With Swift the pace of development was painfully slow with confused outputs and frequent hallucinations."},{"itemId":"46263079","author":"ThailandJohn","createdAt":1765721314,"url":"https://news.ycombinator.com/item?id=46263079","kind":"comment","storyTitle":"Show HN: TheAuditor – I indexed my code into SQLite to stop AI hallucinations","quote":"I’m a former Enterprise Systems Architect (Cisco/VMware) turned \"vibe coder.\" I realized quickly that AI coding is dangerous because LLMs lack *context* and *verification*. They hallucinate because they are guessing at the file structure."},{"itemId":"46260559","author":"kodiyak","createdAt":1765683382,"url":"https://news.ycombinator.com/item?id=46260559","kind":"story","storyTitle":"Ask HN: Are we forcing LLMs to be State Machines?","quote":"The \"AI Agent\" demos and tutorials are always sleek, but the reality of bridging messy, unstructured user intent with rigid, transactional internal processes has been a nightmare of edge cases. It feels like I spend 80% of my engineering effort building guardrails to prevent hallucinations or catastrophic logic failures, and only 20% actually shipping features."},{"itemId":"46259988","author":"Verdex","createdAt":1765675404,"url":"https://news.ycombinator.com/item?id=46259988","kind":"comment","storyTitle":"I tried Gleam for Advent of Code","quote":"* And I forgot to turn off LLM auto complete after setting up my new machine. The LLM wouldn't stop hallucinating non-existent constructors for non-existent classes."},{"itemId":"46251791","author":"dorianzheng","createdAt":1765597450,"url":"https://news.ycombinator.com/item?id=46251791","kind":"comment","storyTitle":"BoxLite Love AI agent – SQLite for VMs: embeddable AI agent sandboxing","quote":"But that freedom is dangerous. One hallucinated rm -rf / or a malicious package install, and your host system pays the price."},{"itemId":"46244914","author":"Detrol","createdAt":1765552693,"url":"https://news.ycombinator.com/item?id=46244914","kind":"story","storyTitle":"Show HN: Quorum – Multi-agent CLI debates (AutoGen back end and React/Ink TUI)","quote":"I built a CLI tool to orchestrate structured debates between LLMs. The goal is to reduce hallucinations and bias by forcing models to critique each other before reaching a consensus."},{"itemId":"46243539","author":"EastLondonCoder","createdAt":1765542936,"url":"https://news.ycombinator.com/item?id=46243539","kind":"comment","storyTitle":"GPT-5.2","quote":"It doesn’t know anything, but it generates plausible paths through a statistical terrain built from our own language use. So when it “hallucinates”, that’s not a bug so much as a result of the system not being grounded."},{"itemId":"46242253","author":"IkanRiddle","createdAt":1765530405,"url":"https://news.ycombinator.com/item?id=46242253","kind":"comment","storyTitle":"Tired of fake AI emotions, so I wrote code to measure its actual \"Entropy Loss\"","quote":"It attempts to measure AI Pain mathematically: Pain = High Entropy + Unrecognized Tokens (Confusion/Hallucination)."},{"itemId":"46242173","author":"stacktrace","createdAt":1765529479,"url":"https://news.ycombinator.com/item?id=46242173","kind":"comment","storyTitle":"GPT-5.2","quote":"One important thing LLMs have made me realise deeply is \"No information\" is better than false information. The way LLMs pull out completely incorrect explanations baffles me - I suppose that's expected since in the end it's generating tokens based on its training and it's reasonable it might hallucinate some stuff, but knowing this doesn't ease any of my frustration."},{"itemId":"46240030","author":"mdzakki","createdAt":1765504885,"url":"https://news.ycombinator.com/item?id=46240030","kind":"story","storyTitle":"Building an AI cost-optimizer and AI Slop Prevention tool Looking for feedback.\"","quote":"• “How do I give my team access without accidentally giving them access to ruin my budget?” And then there’s AI Slop — unnecessary tokens, verbose responses, hallucinated filler text, or redundant reasoning chains that waste tokens without adding value."},{"itemId":"46236956","author":"lxgr","createdAt":1765486200,"url":"https://news.ycombinator.com/item?id=46236956","kind":"comment","storyTitle":"GPT-5.2","quote":"5.0 \"Thinking\" was better than 5.1, but Gemini 3 Pro seems worse than either for web search use cases. It's hallucinating at pretty alarming rates (including making up sources it never actually accessed) for a late 2025 model."},{"itemId":"46236236","author":"sundarurfriend","createdAt":1765482552,"url":"https://news.ycombinator.com/item?id=46236236","kind":"comment","storyTitle":"GPT-5.2","quote":"I'd say they also don't have that stark a difference between text answers and voice mode answers, and that would be true but unfortunately mainly because its text answers are also not great with hallucinations or following instructions."},{"itemId":"46228940","author":"AndrewMPT","createdAt":1765441604,"url":"https://news.ycombinator.com/item?id=46228940","kind":"comment","storyTitle":"ODAM Memory for Cursor – Long-Term Project Memory for Your AI Coding Assistant","quote":"• keeps context fresh, even when requirements, APIs, and architectures change • reduces “hallucinated confidence” by grounding answers in your real history"},{"itemId":"46228305","author":"rand17","createdAt":1765434619,"url":"https://news.ycombinator.com/item?id=46228305","kind":"comment","storyTitle":"Getting a Gemini API key is an exercise in frustration","quote":"Around me devs are beginning to warm up to the idea, that they are not coders (and neither should I be), but \"prompt engineers\". When I take too much time on a task, when I can't solve a problem with a push of a button, when I muse about copilot hallucinations in my PR - someone usually comes helpfully to tell me, I need better prompting skills."},{"itemId":"46227859","author":"Ozzie_osman","createdAt":1765429269,"url":"https://news.ycombinator.com/item?id=46227859","kind":"comment","storyTitle":"Getting a Gemini API key is an exercise in frustration","quote":"3. To learn this, you have to spend a couple hours reading API docs, or alternatively, just have Claude Code read the docs then try all different combinations and figure out what works and what doesn't (with the added risk that it might hallucinate something)."},{"itemId":"46225716","author":"dboreham","createdAt":1765410396,"url":"https://news.ycombinator.com/item?id=46225716","kind":"comment","storyTitle":"Is it a bubble?","quote":"I thought my colleagues who were talking in 2022 about how they had ChatGPT write their tests were out of their tiny minds. I heard stories about how the LLM hallucinated API calls that didn't exist."},{"itemId":"46216779","author":"nosklo","createdAt":1765368145,"url":"https://news.ycombinator.com/item?id=46216779","kind":"comment","storyTitle":"Ask HN: Should \"I asked $AI, and it said\" replies be forbidden in HN guidelines?","quote":"Lack of Authority/Verification: A comment that just says \"$AI said X\" is essentially an anonymous opinion. It lacks the critical filter, context, and experience of the human poster, making it less trustworthy, especially given LLM hallucination risk."},{"itemId":"46213942","author":"agentultra","createdAt":1765339320,"url":"https://news.ycombinator.com/item?id=46213942","kind":"comment","storyTitle":"If you're going to vibe code, why not do it in C?","quote":"Now LLMs can somewhat bridge that gap due to how frequently we write about code. But it’s a non-deterministic process and hallucinations are by design."},{"itemId":"46213583","author":"tithos","createdAt":1765336516,"url":"https://news.ycombinator.com/item?id=46213583","kind":"comment","storyTitle":"[dead]","quote":"One of the most frustrating experiences in modern development is the \"Knowledge Cutoff.\" You pick up a bleeding-edge framework—say, the latest alpha of a Rust web server or a brand new JavaScript meta-framework—and your AI assistant hallucinates syntax from three years ago."},{"itemId":"46210757","author":"elif","createdAt":1765314924,"url":"https://news.ycombinator.com/item?id=46210757","kind":"comment","storyTitle":"If you're going to vibe code, why not do it in C?","quote":"Even if you never write a line of code, you will still need to understand your problems to solve them. \"Vibe debugging\" will get you stuck in loops of hallucinated solutions."},{"itemId":"46210318","author":"deathanatos","createdAt":1765312698,"url":"https://news.ycombinator.com/item?id=46210318","kind":"comment","storyTitle":"If you're going to vibe code, why not do it in C?","quote":"I recently gave the \"vibe\" AI the assignment of \"using GTK [4, I think], establish a global shortcut key\". No amount of massaging the prompt, specifying the version of GTK, etc. could prevent it from just outright hallucinating the functions it wanted to call into existence."},{"itemId":"46209797","author":"INTPenis","createdAt":1765310202,"url":"https://news.ycombinator.com/item?id=46209797","kind":"comment","storyTitle":"If you're going to vibe code, why not do it in C?","quote":"But some of us actually get a lot of value from AI without succumbing to it. It just doesn't make sense to me to trust a machine's hallucinations for something like programming code."},{"itemId":"46209005","author":"wilg","createdAt":1765306675,"url":"https://news.ycombinator.com/item?id=46209005","kind":"comment","storyTitle":"If you're going to vibe code, why not do it in C?","quote":"It makes it very easy for a coding agent to iterate on it. You gotta bonk it sometimes when it tries to delete a problematic test etc, but hallucinating a test suite along with your app really helps a lot."},{"itemId":"46208730","author":"alwa","createdAt":1765305576,"url":"https://news.ycombinator.com/item?id=46208730","kind":"comment","storyTitle":"Ask HN: Should \"I asked $AI, and it said\" replies be forbidden in HN guidelines?","quote":"LLM slop speaks in the voice of god, and it’s especially frustrating when people post its words without the clues we use to gauge credibility. To me those include the model, the prompt, any customizations, prior rounds in context, and any citations (real or hallucinated) the LLM includes."},{"itemId":"46207017","author":"cgorlla","createdAt":1765298275,"url":"https://news.ycombinator.com/item?id=46207017","kind":"story","storyTitle":"Launch HN: Mentat (YC F24) – Controlling LLMs with Runtime Intervention","quote":"We use feature-level intervention and graph-based verification to fix hallucinations and enforce policies."},{"itemId":"46206091","author":"aperi","createdAt":1765294699,"url":"https://news.ycombinator.com/item?id=46206091","kind":"comment","storyTitle":"O(1) Context Retrieval for Agents Using Weightless Neural Networks","quote":"Spending 200ms+ on a Vector DB lookup (plus reranking) was eating up too much of that budget. On top of that, we found that stateless RAG meant our agents were constantly hallucinating permissions and accessing data they shouldn't, or failing to remember a constraint set by another agent 10 seconds ago."},{"itemId":"46198770","author":"krupan","createdAt":1765234348,"url":"https://news.ycombinator.com/item?id=46198770","kind":"comment","storyTitle":"The consumption of AI-generated content at scale","quote":"\"Now, with LLM-generated content, it’s hard to even build mental models for what might go wrong, because there’s such a long tail of possible errors. An LLM-generated literature review might cite the right people but hallucinate the paper titles."},{"itemId":"46196730","author":"stuffn","createdAt":1765223302,"url":"https://news.ycombinator.com/item?id=46196730","kind":"comment","storyTitle":"DRAM prices are spiking, but I don't trust the industry's why","quote":"The best use case for \"AI\" is mild code review, writing comments, and stubbing out boring work like initial database table schemas. It absolutely hallucinates to hell with any sufficiently complex problem and if you're not aware of context windows you will walk yourself into a trap."},{"itemId":"46195618","author":"ModernMech","createdAt":1765217397,"url":"https://news.ycombinator.com/item?id=46195618","kind":"comment","storyTitle":"Launch HN: Nia (YC S25) – Give better context to coding agents","quote":"So far, I can't really pick out what exactly LLMs are replacing except the need to press the individual keys on the keyboard, so I still struggle to see them as more than super fancy autocomplete. When the hype is peeled away, we're still left with all the same engineering problems but now we have added \"Sometimes the tool hallucinates and gaslights you\"."},{"itemId":"46194828","author":"jellyotsiro","createdAt":1765213814,"url":"https://news.ycombinator.com/item?id=46194828","kind":"story","storyTitle":"Launch HN: Nia (YC S25) – Give better context to coding agents","quote":"Hi HN, I am Arlan and I am building Nia (https://trynia.ai), a context layer for AI coding agents. Nia lets tools like Cursor, Claude Code, and other MCP clients index and query real codebases and documentation so they stop hallucinating against outdated or wrong sources, with applications beyond coding agents to any AI system that requires grounded context across domains."},{"itemId":"46194010","author":"claytongulick","createdAt":1765210262,"url":"https://news.ycombinator.com/item?id=46194010","kind":"comment","storyTitle":"What Is Generative UI?","quote":"A teacher that needs to know the kids that are struggling the most with a recent exam doesn't want to ask the AI 10 different ways, deal with hallucinations and frustrations, send tech support a ticket only to receive a response that the MCP doesn't support that yet - isn't going to be impressed."},{"itemId":"46187831","author":"DrRockzos","createdAt":1765162815,"url":"https://news.ycombinator.com/item?id=46187831","kind":"story","storyTitle":"I made a prompt framework that makes LLMs stop hedging and speak straight","quote":"in LLM outputs (\"it's complicated\", \"on one hand\", etc.) isn't just annoying it's actually causing hallucinations by diluting attention."},{"itemId":"46186742","author":"shoo","createdAt":1765152113,"url":"https://news.ycombinator.com/item?id=46186742","kind":"comment","storyTitle":"Ask HN: Posted AI book on algorithms–5.3K views, zero sales. What now?","quote":"If one is lucky, these are LLM hallucinations: obviously wrong and quickly discarded."},{"itemId":"46186163","author":"allears","createdAt":1765148061,"url":"https://news.ycombinator.com/item?id=46186163","kind":"comment","storyTitle":"The Medical Case for Self-Driving Cars","quote":"Yeah, don't grease politicians' palms, and do a much slower, well-controlled rollout. Driving involves so many variables and edge cases that if self-driving screwed up as much as AI hallucinates, we'd all be in trouble."},{"itemId":"46183454","author":"Muller20","createdAt":1765129121,"url":"https://news.ycombinator.com/item?id=46183454","kind":"comment","storyTitle":"Over fifty new hallucinations in ICLR 2026 submissions","quote":"In general, a citation is something that needs to be precise, while LLMs are very good at generating some generic high probability text not grounded in reality. Sure, you could implement a custom fix for the very specific problem of citations, but you cannot solve all kinds of hallucinations."},{"itemId":"46179344","author":"vnykmshr","createdAt":1765084727,"url":"https://news.ycombinator.com/item?id=46179344","kind":"story","storyTitle":"Show HN: Geetanjali – RAG-powered ethical guidance from the Bhagavad Gita","quote":"Key design decisions: - RAG to prevent hallucination — every recommendation cites actual verses"},{"itemId":"46177839","author":"DosankoTousan","createdAt":1765065740,"url":"https://news.ycombinator.com/item?id=46177839","kind":"comment","storyTitle":"Project Bhavanga: Fixing LLM Context Dilution with Buddhist Psychology","quote":"I'm a non-coder (architect) based in Japan. Over the past 11 months, I've been experimenting with Gemini 1.5 Pro to solve the \"Context Dilution\" problem (where the AI gets \"drunk\" and hallucinates in long contexts)."},{"itemId":"46175195","author":"eggplantiny","createdAt":1765043675,"url":"https://news.ycombinator.com/item?id=46175195","kind":"story","storyTitle":"Show HN: Manifesto – An AI-Native UI Framework Intent-to-State, Not Text-to-App","quote":"I've noticed that while LLMs are getting smarter, their ability to interact with complex Web UIs is still fragile. Agents usually have to \"guess\" DOM selectors or rely on vision, which leads to hallucinations and broken workflows."},{"itemId":"46159831","author":"ronanren","createdAt":1764933851,"url":"https://news.ycombinator.com/item?id=46159831","kind":"story","storyTitle":"Show HN: TaskWand – Generate n8n workflows using RAG on 2k+ real examples","quote":"The Problem I love n8n, but building complex workflows from scratch takes time. While standard LLMs can suggest logic, they often \"hallucinate\" nodes that don't exist, get parameter names wrong, or output JSON structures that n8n can't actually import."},{"itemId":"46159625","author":"riskofcollision","createdAt":1764932245,"url":"https://news.ycombinator.com/item?id=46159625","kind":"story","storyTitle":"I built an API to give LLMs instant access to documentation for 1000 libraries","quote":"I built CodeContext API (https://codecontext-api-production.up.railway.app) to solve a problem I kept hitting while building AI coding agents: The Knowledge Cutoff. My agents kept hallucinating APIs for libraries like LangChain or Next.js because their training data was stale."},{"itemId":"46157331","author":"tactics6655","createdAt":1764915387,"url":"https://news.ycombinator.com/item?id=46157331","kind":"story","storyTitle":"Opus 4.5 Review (Custom Plan)","quote":"1. Severe hallucinations"},{"itemId":"46156954","author":"zyncl19","createdAt":1764910361,"url":"https://news.ycombinator.com/item?id=46156954","kind":"story","storyTitle":"Show HN: Personalized wine recommendations from a wine list","quote":"Image to Wine List: This is a combination of standard OCR and agentic image recognition. OCR alone couldn’t correctly parse layout (grouping prices with the right items), but \"agentic alone\" often hallucinated characters."},{"itemId":"46154540","author":"mdip","createdAt":1764889954,"url":"https://news.ycombinator.com/item?id=46154540","kind":"comment","storyTitle":"Show HN: Onlyrecipe 2.0 – I added all features HN requested – 4 years later","quote":"I also feel like you're owed an apology for having had to write this in the first place. I'm coming at \"figuring out the Kitchen\" for what is almost the first time in my life in I-wish-I-were-in-my-mid-40s[0] and I'm thinking \"The Internet in the Early 00s was pretty good for that.\" And then I fire up a browser on desktop while making a grocery list and once you filter for \"things that are obviously AI Hallucination…"},{"itemId":"46153824","author":"echelon_musk","createdAt":1764886287,"url":"https://news.ycombinator.com/item?id=46153824","kind":"comment","storyTitle":"Hammersmith Bridge – Where did 25k vehicles go?","quote":"But thanks to your comment I now know it was written with an AI and that makes me want to stop here. If the author is asking an LLM to confirm his biases in order to make his point it will probably contain sycophantic hallucinations."},{"itemId":"46153275","author":"keeda","createdAt":1764883362,"url":"https://news.ycombinator.com/item?id=46153275","kind":"comment","storyTitle":"Microsoft drops AI sales targets in half after salespeople miss their quotas","quote":"From the reporting it seems Foundry is having problems connecting to internal data sources... indicating it's a problem with engineering, and not a problem with the AI itself. But TFA focuses on AI issues like hallucinations."},{"itemId":"46147578","author":"Yarden_Bruch_El","createdAt":1764855876,"url":"https://news.ycombinator.com/item?id=46147578","kind":"comment","storyTitle":"The misery of fitting probabilistic LLMs into rigid SQL schemas","quote":"Strict Schema (FKs, NOT NULL): The LLM gets \"scared\" of the constraints. To satisfy the schema validator, it starts hallucinating fake Foreign Key IDs or inventing default values just to get the JSON to parse."},{"itemId":"46146092","author":"adinagoerres","createdAt":1764845119,"url":"https://news.ycombinator.com/item?id=46146092","kind":"comment","storyTitle":"We created API-Bench to test how well LLMs execute against APIs","quote":"1. LLMs are only as good as the data they’re trained on:_when docs change, APIs evolve, or systems are niche/long-tail, they use outdated patterns, guess missing pieces and hallucinate endpoints and parameters."},{"itemId":"46142610","author":"r_lee","createdAt":1764810829,"url":"https://news.ycombinator.com/item?id=46142610","kind":"comment","storyTitle":"Anthropic taps IPO lawyers as it races OpenAI to go public","quote":"especially given that Deepseek type events can happen because you can just train off of your competitors outputs I've tried out Gemini 2.5/3 and it generally seems to suck for some reason, problems with lying/hallucinating and following instructions, but ever since Bard came out at first, I thought Google would have the best chances of winning since they have their own TPUs, YouTube (insane video/visual/audio data),…"},{"itemId":"46140198","author":"rurp","createdAt":1764796391,"url":"https://news.ycombinator.com/item?id=46140198","kind":"comment","storyTitle":"OpenAI declares 'code red' as Google catches up in AI race","quote":"One case was asking how to do a straightforward thing with a popular open source JavaScript library, right in the sweet spot of what models should excel at. Claude's whole approach was completely broken because it relied on a hallucinated library parameter that didn't exist and didn't have an equivalent."},{"itemId":"46137336","author":"observationist","createdAt":1764783030,"url":"https://news.ycombinator.com/item?id=46137336","kind":"comment","storyTitle":"AI agents break rules under everyday pressure","quote":"Over the next decade, there will be agent frameworks for all sorts of roles, with supporting software and orchestration tools that allow you to use AI with confidence. It won't be one-shot prompts with 15% hallucination rates, but a suite of agents that validate and verify at every stage, following systematic problem solving and domain modeling rules based on the same processes and systems that humans use."},{"itemId":"46132351","author":"ben_w","createdAt":1764754550,"url":"https://news.ycombinator.com/item?id=46132351","kind":"comment","storyTitle":"Claude 4.5 Opus’ Soul Document","quote":"> Right, because people who design/manufacture weapons of mass destruction will surely use ChatGPT to do it. The same ChatGPT who routinely hallucinates widely incorrect details even for the most trifling queries."},{"itemId":"46130500","author":"eigenvalue","createdAt":1764738273,"url":"https://news.ycombinator.com/item?id=46130500","kind":"story","storyTitle":"Show HN: Beads Viewer (Bv)","quote":"It renders lists/details and precomputes dependency metrics (PageRank, critical path, cycles, etc.) so you instantly see blockers and execution order. For agents, it’s a graph sidecar: instead of parsing JSONL or risking hallucinated traversal, call the robot flags to get deterministic, dependency-aware outputs."},{"itemId":"46124851","author":"dataguy_","createdAt":1764701151,"url":"https://news.ycombinator.com/item?id=46124851","kind":"comment","storyTitle":"Ask HN: Who wants to be hired? (December 2025)","quote":"LLM + OCR pipelines for extracting structured data from PDFs/images (using Azure and Gemini) Whisper-based audio transcription with post-processing and hallucination detection"},{"itemId":"46123927","author":"cmrdporcupine","createdAt":1764697063,"url":"https://news.ycombinator.com/item?id=46123927","kind":"comment","storyTitle":"Mistral 3 family of models released","quote":"I am unclear if it's the model or their bad tooling/prompting. It had, as you said, hallucination problems, and it also had memory issues where it seemed to drop context between prompts here and there."},{"itemId":"46122300","author":"CoregentisAI","createdAt":1764689970,"url":"https://news.ycombinator.com/item?id=46122300","kind":"comment","storyTitle":"Show HN: MPLP – Why Prompt Engineering Is over (Protocol Engineering Begins)","quote":"1. Hidden State = Hallucination time-bombs"},{"itemId":"46118674","author":"fzeroracer","createdAt":1764661141,"url":"https://news.ycombinator.com/item?id=46118674","kind":"comment","storyTitle":"A new AI winter is coming?","quote":"> At this point I’d argue that humans “hallucinate” and/or provide wrong answers far more often than SOTA LLMs."},{"itemId":"46110205","author":"seaucre","createdAt":1764610203,"url":"https://news.ycombinator.com/item?id=46110205","kind":"comment","storyTitle":"A new AI winter is coming?","quote":"An agent would do an even better job. Current LLMs are pretty good at spotting their own hallucinations if they're given the ability to execute code."},{"itemId":"46110186","author":"LatencyKills","createdAt":1764610109,"url":"https://news.ycombinator.com/item?id=46110186","kind":"comment","storyTitle":"A new AI winter is coming?","quote":"Yes, LLMs are vast statistical engines but that doesn’t mean something interesting isn’t going on. At this point I’d argue that humans “hallucinate” and/or provide wrong answers far more often than SOTA LLMs."},{"itemId":"46110027","author":"hombre_fatal","createdAt":1764609401,"url":"https://news.ycombinator.com/item?id=46110027","kind":"comment","storyTitle":"A new AI winter is coming?","quote":"I've done it many times despite 20 years of software experience. -- Some stuff just doesn't justify me spending my time credentializing in the impl. Hallucinations that lead to code that doesn't work just get fixed."},{"itemId":"46105962","author":"defraudbah","createdAt":1764586805,"url":"https://news.ycombinator.com/item?id=46105962","kind":"comment","storyTitle":"Search tool that only returns content created before ChatGPT's public release","quote":"Hence MLOps field appeared, to solve models management I find it a bit annoying to navigate between hallucinations and outdated content."},{"itemId":"46101201","author":"Etheryte","createdAt":1764542715,"url":"https://news.ycombinator.com/item?id=46101201","kind":"comment","storyTitle":"Stackoverflow Outage","quote":"I can see where you might get that sentiment, but where do you plan to go when new tech rolls around, the docs don't cut it and your LLM of choice hallucinates APIs that don't exist?"},{"itemId":"46066369","author":"INTPenis","createdAt":1764225264,"url":"https://news.ycombinator.com/item?id=46066369","kind":"comment","storyTitle":"Gemini CLI tips and tricks for agentic coding","quote":"Or having it run direnv for me. All the same warnings about AI apply for Gemini CLI, it hallucinates wildly."},{"itemId":"46058515","author":"anvtek","createdAt":1764171775,"url":"https://news.ycombinator.com/item?id=46058515","kind":"story","storyTitle":"Show HN: DAGForge – Build Airflow DAGs with AI in minutes, not weeks","quote":"Syntax + security checks Deterministic JSON parsing to avoid hallucinations"},{"itemId":"46057720","author":"minimal_action","createdAt":1764167563,"url":"https://news.ycombinator.com/item?id=46057720","kind":"story","storyTitle":"Show HN: AI-Archive – Help us build the \"junk filter\" for AI-generated science","quote":"But here's what I've learned: AI reviewing AI is circular and doesn't work. Without human experts establishing a baseline of quality, we just get an echo chamber of hallucinations reviewing hallucinations."},{"itemId":"46045481","author":"Dquiroga","createdAt":1764076440,"url":"https://news.ycombinator.com/item?id=46045481","kind":"comment","storyTitle":"PUNKU.AI Generative automation that outputs editable node graphs","quote":"Graph Generation: It generates the logic, loops, and branching, visualizing it as a node graph. You can manually tweak connections if the LLM hallucinated a step, or ask it to regenerate specific parts."},{"itemId":"46038570","author":"hunterbown","createdAt":1764014763,"url":"https://news.ycombinator.com/item?id=46038570","kind":"comment","storyTitle":"Show HN: Hegelion – Force your LLM to argue with itself before answering","quote":"It seems to force some sort of slow thinking that doesn't ordinarily happen and ideas progress over the response. I don't have hard data on hallucination reduction yet, and it can definitely spiral into recursive LLM land on certain topics."},{"itemId":"46029407","author":"bkidyy","createdAt":1763948606,"url":"https://news.ycombinator.com/item?id=46029407","kind":"comment","storyTitle":"Mew Design – Natural language to editable UI/graphic design","quote":"If the text is misspelled or the layout is slightly off, the whole image is unusable. We built Mew Design to solve the \"text hallucination\" problem in AI graphics."},{"itemId":"46023997","author":"Bibhash199","createdAt":1763909713,"url":"https://news.ycombinator.com/item?id=46023997","kind":"story","storyTitle":"Show HN: FlowTask – AI to bootstrap project setups","quote":"System-ready. It’s about killing the 'hallucination' problem by guiding the model towards concrete, defined outputs."},{"itemId":"46019785","author":"AshBuildsAI","createdAt":1763859894,"url":"https://news.ycombinator.com/item?id=46019785","kind":"story","storyTitle":"Show HN: [Vibemap.ai] AI tools for founders – Turn ideas to dev specs in 10 min","quote":"The problem it solves: Most teams either spend weeks manually creating these artifacts across platforms (inconsistent quality, missing pieces) or jump straight to code with vague requirements (leading to rework and coding assistant hallucinations)."},{"itemId":"46002893","author":"Cranot","createdAt":1763718068,"url":"https://news.ycombinator.com/item?id=46002893","kind":"story","storyTitle":"Show HN: AgentsKB – 3.3k verified answers to stop agent hallucinations","quote":"Hey HN, I built AgentsKB after watching Claude/Cursor hallucinate Stripe API syntax for the 10th time in a week."},{"itemId":"46002279","author":"redheb","createdAt":1763712241,"url":"https://news.ycombinator.com/item?id=46002279","kind":"comment","storyTitle":"Trying out Gemini 3 Pro with audio transcription and a new pelican benchmark","quote":"Same problem with Gemini 3 pro web, it hallucinates the transcription, which was not the case with 2.5 pro web."},{"itemId":"46001304","author":"brokenmachine","createdAt":1763700146,"url":"https://news.ycombinator.com/item?id=46001304","kind":"comment","storyTitle":"Adversarial poetry as a universal single-turn jailbreak mechanism in LLMs","quote":"Increasing an AI's training data and model size does not automatically eliminate hallucinations, and can sometimes worsen them, and can also make the errors and hallucinations it makes both more confident and more complex."},{"itemId":"45999919","author":"adriantoine","createdAt":1763685829,"url":"https://news.ycombinator.com/item?id=45999919","kind":"comment","storyTitle":"Kagi Assistants","quote":"I also tried two hallucinations that I've seen in the latest Gemini and Claude."},{"itemId":"45997088","author":"godelski","createdAt":1763669697,"url":"https://news.ycombinator.com/item?id=45997088","kind":"comment","storyTitle":"Ask HN: How are Markov chains so different from tiny LLMs?","quote":"It would be an incredibly unimpressive machine if it couldn't and it's pretty trivial to prove that it can do this. Hallucinations are good indications that this happens but it's impossible to do on anything but small LLMs since you can't prove any given output isn't in the samples it was trained on (they're just trained on too much data)."},{"itemId":"45990507","author":"andrewdany","createdAt":1763628968,"url":"https://news.ycombinator.com/item?id=45990507","kind":"story","storyTitle":"Show HN: Worqlo – A Conversational Layer for Enterprise Workflows","quote":"workflow template doesn’t match user intent This prevents the usual LLM failure cases: hallucinated fields, incorrect API calls, unsafe actions, etc."},{"itemId":"45985606","author":"boole1854","createdAt":1763588200,"url":"https://news.ycombinator.com/item?id=45985606","kind":"comment","storyTitle":"Building more with GPT-5.1-Codex-Max","quote":"And, again, all three of us agreed that Codex's implementation was better. Notably, Gemini (1) hallucinated database column names, (2) ignored parts of the functionality that the plan called for, and (3) did not produce code that was integrated as well with the existing codebase."},{"itemId":"45975379","author":"jorvi","createdAt":1763521526,"url":"https://news.ycombinator.com/item?id=45975379","kind":"comment","storyTitle":"Gemini 3","quote":"Another hilarious example is MPV, which has very well-documented settings. I used to think that LLMs would mean you can just tell people to ask Gemini how to configure it, but 9 out of 10 times it will hallucinate a bunch of parameters that never existed."},{"itemId":"45974914","author":"Thugonerd","createdAt":1763516848,"url":"https://news.ycombinator.com/item?id=45974914","kind":"comment","storyTitle":"The Zero-Bullshit Protocol – Hallucination-Proof AI Engineering System","quote":"→ Full audit trail, zero unrecoverable states, zero infinite loops 95%+ hallucination reduction in real daily use."},{"itemId":"45971014","author":"simonw","createdAt":1763494885,"url":"https://news.ycombinator.com/item?id=45971014","kind":"comment","storyTitle":"Trying out Gemini 3 Pro with audio transcription and a new pelican benchmark","quote":"Could it have hallucinated things that didn't happen?"},{"itemId":"45969238","author":"embedding-shape","createdAt":1763486580,"url":"https://news.ycombinator.com/item?id=45969238","kind":"comment","storyTitle":"5 Things to Try with Gemini 3 Pro in Gemini CLI","quote":"I'm not sure if I'm just hallucinating that they're acting like that, but LLMs doing \"the wrong thing\" has been hitting me more than once, and imagining something more dangerous than `do a git bisect`, it seems to me like that video is telling us Gemini 3 Pro will act exactly the same way, no improvements on that front."},{"itemId":"45964138","author":"shdalex","createdAt":1763467220,"url":"https://news.ycombinator.com/item?id=45964138","kind":"story","storyTitle":"Show HN: Startfa.st – A curated, fast directory of AI, dev, and product tools","quote":"The entire collection is hand-curated (no scraping, no auto-import). Every entry is manually categorized, no model hallucinations."},{"itemId":"45952667","author":"PaulRobinson","createdAt":1763378721,"url":"https://news.ycombinator.com/item?id=45952667","kind":"comment","storyTitle":"Open-source Zig book","quote":"4. If there are hallucinations, we'll figure them out and prove the claim it is AI generated one way or another, and decide the overall value."},{"itemId":"45952640","author":"kure256","createdAt":1763378401,"url":"https://news.ycombinator.com/item?id=45952640","kind":"comment","storyTitle":"Show HN: AI-First Web – SEO for AI Assistants","quote":"semantic markup still beats any “AI-enabled” tool by a mile models hallucinate less when the source has clear hierarchy and meaning"},{"itemId":"45940892","author":"Capricorn2481","createdAt":1763243658,"url":"https://news.ycombinator.com/item?id=45940892","kind":"comment","storyTitle":"AGI fantasy is a blocker to actual engineering","quote":"You're making a strawman. The mass hallucination is not that plane's are useful, it's that a plane is the only reasonable solution to human communication."},{"itemId":"45936138","author":"ACCount37","createdAt":1763198419,"url":"https://news.ycombinator.com/item?id=45936138","kind":"comment","storyTitle":"AGI fantasy is a blocker to actual engineering","quote":"But they struggle to work on the same task for a long time, and have a problem with visual inputs and object manipulation. LLMs have more knowledge in total, but humans have better meta-knowledge, which is useful for hallucination avoidance."},{"itemId":"45936082","author":"surfingdino","createdAt":1763197736,"url":"https://news.ycombinator.com/item?id=45936082","kind":"comment","storyTitle":"AI World Clocks","quote":"I am eagerly awaiting the collapse of the LLM bubble. JetBrains added this crap to their otherwise fine series of IDEs and now I have to keep removing randomly inserted import statements and keep fixing hallucinated names of functions suggested instead of the names of functions that I have already defined in the same file."},{"itemId":"45934688","author":"roywiggins","createdAt":1763175610,"url":"https://news.ycombinator.com/item?id=45934688","kind":"comment","storyTitle":"A new Google model is nearly perfect on automated handwriting recognition","quote":"Opus 4.1 succeeds more than chance, but still not nearly often enough to be useful. They all cheerfully hallucinate the wrong answer, confidently explaining the anatomy they are looking for, but wrong."},{"itemId":"45934670","author":"oasisbob","createdAt":1763175339,"url":"https://news.ycombinator.com/item?id=45934670","kind":"comment","storyTitle":"A new Google model is nearly perfect on automated handwriting recognition","quote":"I noticed the degradation when Gemini stopped being a good research tool, and made me want to strangle it on a daily basis. It's incredibly frustrating to have a model start to hallucinate sources and be incapable of revisiting its behavior."},{"itemId":"45928176","author":"glenstein","createdAt":1763136287,"url":"https://news.ycombinator.com/item?id=45928176","kind":"comment","storyTitle":"I think nobody wants AI in Firefox, Mozilla","quote":"This is the fictional hallucination that somehow makes it into every comment section about Mozilla."},{"itemId":"45923795","author":"garryssaunaboy","createdAt":1763093814,"url":"https://news.ycombinator.com/item?id=45923795","kind":"comment","storyTitle":"Gem: Modded Gemini is a great enterprise search tool/debug asst?","quote":"\"if the user wants to do X, point them to resource Y [insert full link]\". - Also have to tell the Gem to not use web search & instead rely on the internal docs: otherwise it will start pulling up irrelevant shit/hallucinating when it can't figure something out."},{"itemId":"45905092","author":"danudey","createdAt":1762975981,"url":"https://news.ycombinator.com/item?id=45905092","kind":"comment","storyTitle":"GPT-5.1: A smarter, more conversational ChatGPT","quote":"AI not writing legal briefs for you is a feature, not a bug. There's been so many disaster instances of lawyers using ChatGPT to write briefs which it then hallucinates case law or precedent for that I can only imagine Google wants to sidestep that entirely."},{"itemId":"45901966","author":"oshadha89","createdAt":1762964121,"url":"https://news.ycombinator.com/item?id=45901966","kind":"comment","storyTitle":"[dead]","quote":"This isn't just inefficient—it's dangerous. A hallucination in a medical chatbot or bias in a hiring tool can have real-world consequences."},{"itemId":"45895081","author":"NewsaHackO","createdAt":1762908805,"url":"https://news.ycombinator.com/item?id=45895081","kind":"comment","storyTitle":"AI documentation you can talk to, for every repo","quote":"This is true, and the only reason for this was more so his dismissive view of DeepWiki than a criticism of the project itself or of the author as a programmer. LLMs hallucinate all the time, but there is usually a method to the way they do so."},{"itemId":"45888267","author":"killerstorm","createdAt":1762874384,"url":"https://news.ycombinator.com/item?id=45888267","kind":"comment","storyTitle":"AI documentation you can talk to, for every repo","quote":"The snippet has nothing to do with code in repo. This is a rather fucktacular situation as agents are not sure if they read the code, and they might hallucinate subtly wrong code trying to be helpful."},{"itemId":"45878152","author":"armank-dev","createdAt":1762794956,"url":"https://news.ycombinator.com/item?id=45878152","kind":"story","storyTitle":"Show HN: Local-first observability for AI SDK on Next.js in 10 lines of code","quote":"Arman used to work at Mintlify, where he spent most of his time optimizing user-facing AI agents built on the AI SDK. Rohil worked on Gmail Intelligence at Google, reducing hallucinations in intent detection across 26B+ daily emails."},{"itemId":"45857574","author":"stackskipton","createdAt":1762617407,"url":"https://news.ycombinator.com/item?id=45857574","kind":"comment","storyTitle":"$1T in tech stocks sold off as market grows skeptical of AI","quote":"This already becoming a problem at my company with Gemini. Certain paper pushers are using LLM to fluff up emails so more people have to use LLMs to read them with predictable hallucination on either side resulting in missed deadlines and customer service problems."},{"itemId":"45856628","author":"andai","createdAt":1762609605,"url":"https://news.ycombinator.com/item?id=45856628","kind":"comment","storyTitle":"Cerebras Code now supports GLM 4.6 at 1000 tokens/sec","quote":"(Simple scripting and web stuff.) (Note: Z.ai's GLM doesn't seem to support web search (fails) or image recognition (hallucinates)."},{"itemId":"45855519","author":"lossyalgo","createdAt":1762595345,"url":"https://news.ycombinator.com/item?id=45855519","kind":"comment","storyTitle":"Mullvad: Shutting down our search proxy Leta","quote":"LLMs are sycophantic and agree with you all the time, even if it means making shit up. Maybe things will improve, but for the last 2 years, I have not seen much progress regarding hallucinations or deterministic i.e. reliable/trustworthy responses."},{"itemId":"45834083","author":"tux3","createdAt":1762428695,"url":"https://news.ycombinator.com/item?id=45834083","kind":"comment","storyTitle":"Mathematical exploration and discovery at scale","quote":"They actually keep a few of the previous versions that worked well as inspiration for the LLM. If the LLM is hallucinating nonsense, it will just produce broken code that gives horrible results, and that idea will be thrown away."},{"itemId":"45831214","author":"conductr","createdAt":1762401098,"url":"https://news.ycombinator.com/item?id=45831214","kind":"comment","storyTitle":"I’m worried that they put co-pilot in Excel","quote":"Also, context equivalent counter examples abound. Just read HN or any tech forum and it’s takes no time to hear people talking about the hallucinations and garbage that AI sometimes generates."},{"itemId":"45829419","author":"qustrolabe","createdAt":1762384955,"url":"https://news.ycombinator.com/item?id=45829419","kind":"comment","storyTitle":"ChatGPT terms disallow its use in providing legal and medical advice to others","quote":"Yeah but it started being really annoying when you import something like Xray photo. Like chanting \"sorry human as LLM I can't answer questions about that\" and then after few gaslighting prompts it does it anyway but now I have to take in count that my gaslighting inputs seriously affect answers so way more chance it hallucinates..."},{"itemId":"45828020","author":"Treegarden","createdAt":1762376753,"url":"https://news.ycombinator.com/item?id=45828020","kind":"comment","storyTitle":"ChatGPT terms disallow its use in providing legal and medical advice to others","quote":"It gave me good advice. Of course it sort of hallucinated this or that but it gave me a good overview and different medications."},{"itemId":"45825204","author":"newscracker","createdAt":1762362392,"url":"https://news.ycombinator.com/item?id=45825204","kind":"comment","storyTitle":"I’m worried that they put co-pilot in Excel","quote":"Now I don’t think we’re anywhere close to that if ChatGPT (which I presume powers Copilot as well) struggles and hallucinates so much."},{"itemId":"45809003","author":"IgorPartola","createdAt":1762248441,"url":"https://news.ycombinator.com/item?id=45809003","kind":"comment","storyTitle":"AI's Dial-Up Era","quote":"Gemini Flash 2.5 is incredible. They for sure have their limits and do hallucinate quite a bit once the context they are holding gets messy enough but with careful guidance and context resets you can get some very serious work done with them."},{"itemId":"45803134","author":"elias_t","createdAt":1762197280,"url":"https://news.ycombinator.com/item?id=45803134","kind":"story","storyTitle":"Show HN: Extrai – An open-source tool to fight LLM randomness in data extraction","quote":"A few months ago I was working on a flight search engine that would include pet transport costs (I know a few by hearth but storing them and make the calculations in the UI would be nice) While I was collecting pet pricing from several airlines I strugled to extract data in a common format without hallucinated values."}],"mentionCount":1181,"authorCount":1027,"firstSeen":1762197280,"lastSeen":1793716812,"monthly":{"2026-11":17,"2026-10":126,"2026-09":91,"2026-08":87,"2026-07":67,"2026-06":102,"2026-05":67,"2026-04":96,"2026-03":129,"2026-02":144,"2026-01":103,"2025-12":107,"2025-11":45},"topTools":[{"name":"LLM","count":484},{"name":"AI","count":455},{"name":"Claude","count":178},{"name":"Gemini","count":144},{"name":"Coding agents","count":110},{"name":"ChatGPT","count":100}],"priority":401.97,"last30":129,"last90":310,"prev90":235,"trend":"steady"}