Struggles with large or legacy codebases

I have nothing against AI, I use it heavily myself, but what is annoying is people completely ignoring the years of work of existing projects that do the exact same thing - instead of just forking, but instead re-tokenizing the exact same work again.

fschuett · Oct 26, 2026 · comment in “PaperPilot – Zero-Cloud, Pure-Rust PDF Engine with WASM and MCP”

There have been over enthusiastic people (and trolls) bothering projects in issue trackers and mailing lists, but I've never seen a level of volume commensurate to the meme's fame. (Nowadays I see vibe-coded projects done in Rust, but those are also not imposing their codebase on other existing projects beyond sucking up the air out of the room and making it annoying to read news aggregators due to high volume and l…

estebank · Oct 14, 2026 · comment in “Problems with the Rust Language”

For example, they have css snippets in dev mode when you have a frame selected, but those aren't really that useful since it doesn't know anything about your existing code base or frameworks. Code connect was supposed to bridge this gap, allowing users to define how Figma components should generate relevant code snippets, but it only worked with React and they had some janky string based templating language for ever…

miguel-muniz · Oct 1, 2026 · comment in “Figma restricts MCP access to whitelisted clients, excluding Pi”

I canceled Claude for my personal account due to the Astra hype, but I'm back after a single month and the absolutely ridiculous context window is one of the many reasons. I barely compact at work in a very complex monorepo (neither with Fable 5.1 nor Opus 5.5), and yet in my personal greenfield project Astra keeps compacting all the time, to the point of it being unusable.

kaoD · Sep 30, 2026 · comment in “GPT 6.1 Sol: Near-Astra intelligence for a fifth of the price”

They only build prototypes and rough drafts and don’t understand that AI makes the first 80% easier and the last 20% harder, and makes maintainability a nightmare where a codebase that’s gone through vibe coding for 3 months is like the equivalent of a 10 year old legacy codebase.

an0malous · Sep 23, 2026 · comment in “I am done with this shit”

Indeed. To add to this, the fact that Claude so often tries to deviate from defined architecture can be really frustrating if you're working in a mature codebase.

tomrod · Sep 14, 2026 · comment in “GPT-5.6 Luna vs. GPT-6 Astra: Is a $1.20 Model Good Enough for Code Review?”

He did the right thing. The very idea of Claude copying an existing project to answer one of my prompts or a whole project in Claude Code is a nightmare scenario.

mark_l_watson · Aug 9, 2026 · comment in “Mea Culpa – Dark Hours”

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, ...

damienmeur · Aug 3, 2026 · story in “Show HN: Argot, a Rust AI guardrail based on your codebase AST patterns”

If you work in a large codebase it should always be your first attempt after a rebase, and that's how Ziggity works. It will ask to fallback to normal force-push if it didn't work, but you'll remain in control.

TheSorcerer · Jul 21, 2026 · comment in “Ziggity – A terminal UI for Git, written in Zig”

If you work in a large codebase it should always be your first attempt after a rebase, and that's how Ziggity works. It will ask to fallback to normal force-push if it didn't work, but you'll remain in control.

TheSorcerer · Jul 21, 2026 · comment in “Ziggity – A terminal UI for Git, written in Zig”

It's tuned for the kinds of tasks where "just try" doesn't get good results. A major complaint with AI code was that AIs struggle with complex codebases, don't respect existing conventions, reinvent functionality multiple times over, etc.

ACCount37 · Jul 12, 2026 · comment in “Claude Code sends 33k tokens before reading the prompt; OpenCode sends 7k”

Not that I disagree with anything here, but... I wish it were clearer in these kinds of posts how "I use AI code I don't understand" is so different from "I use libraries written by other people I don't understand", or "I work in a large codebase which was 99% written by other people, and I haven't seen all of it", or even "I use software written by other people I don't understand".

edanm · Jun 21, 2026 · comment in “When I reject AI code even if it works”

To get an AI to do a task well, you need to give it the right context and the right task description. I've had a lot of struggles working in a large codebase getting AI to follow the existing patterns and architecture.

espresso_dev · Jun 18, 2026 · story in “Show HN: writing tasks with focused context + instructions = sharper AI”

But it does not at all(!) compare to the frontier models. For a large project Claude's Opus 4.6+ just work, whereas local gemma tangles itself up, makes weird mistakes, and just can't handle it (for those cases it is faster if I do it myself).

linuxhansl · Jun 17, 2026 · comment in “Running local models is good now”

No experience in huge codebases because I only use them outside work, but they seem good enough about gathering info before they dive in that I'd expect them to grep around as they need. An annoying caveat: individual subscription plans, used heavily, are much cheaper than the API -- see https://she-llac.com/claude-limits -- which complicates any argument about cost.

twotwotwo · Jun 8, 2026 · comment in “DeepSeek V4 Pro beats GPT-5.5 Pro on precision”

Mmm....His position of "stopped using markdown altogether for almost everything" strikes me as overstated and selfpromotional. The 2-4x higher generation cost is non-trivial when you're operating under rate limits, and Git versioning friction is a real problem for your work in big and complex repos.

vdelpuerto · May 9, 2026 · comment in “Using Claude Code: The unreasonable effectiveness of HTML”

I've personally found it uses significantly less memory on large projects than VSCode. VSCode has historically been nigh-unusable for me on Linux, it gets incredibly sluggish.

sodacanner · Apr 29, 2026 · comment in “Zed 1.0”

Grep is effective for the most part, except for situations like when you have huge codebases and the thing you're looking for is used in too many places both as symbol and non-symbol. Another annoying thing about plain grep is, LLMs often end up pulling in bundled packages when using grep where 1 line is large enough to ruin the context window

GodelNumbering · Apr 27, 2026 · comment in “Show HN: OSS Agent I built topped the TerminalBench on Gemini-3-flash-preview”

I struggle to believe that a ton of seemingly intelligent software engineers are too dumb to figure out how to use Claude code to get reliable results, it seems much more likely to me that it can do well at isolated tasks or new projects but fails when pointed at large complex code bases because it just... is a token predictor lol.

ericmcer · Apr 7, 2026 · comment in “Claude Code login fails with OAuth timeout on Windows”

> I understand why this doesn't work well with legacy projects, but it's something that the language could strive towards.

embedding-shape · Mar 31, 2026 · comment in “Axios compromised on NPM – Malicious versions drop remote access trojan”

It really makes me question if "packages" should even exist in JavaScript, or if we could just be importing standard plain source code from a git repo. I understand why this doesn't work well with legacy projects, but it's something that the language could strive towards.

zdc1 · Mar 31, 2026 · comment in “Axios compromised on NPM – Malicious versions drop remote access trojan”

It’s a fast, local-first, open-source Git client for Linux, macOS, and Windows. We started building it after running into the same problem over and over: Git tools felt fine on small projects, but got painful on large repos and big diffs.

Havunen · Mar 18, 2026 · story in “Show HN: GitComet speedy Git GUI written in Rust end-to-end”

I built this after getting tired of writing CLAUDE.md files by hand every time I started a new project or onboarded to an existing codebase.

vexorkai · Mar 1, 2026 · comment in “Show HN: Rulegen – Auto-generate Claude.md and .cursorrules from your codebase”

Docs: https://vexp.dev/docs I'd love to hear feedback, especially from people working on large codebases (50k+ lines) where context management is a real bottleneck.

nicola_alessi · Feb 22, 2026 · story in “Show HN: Vexp – graph-RAG context engine, 65-70% fewer tokens for AI agents”

I’ve been playing around with a few AI automation stacks lately, and honestly, the shift from basic autocomplete to actual 'context-aware agents' is wild. The biggest headache isn't getting the AI to write a function, but making sure it doesn't break the overall architecture of a large project.

CodeBit26 · Feb 17, 2026 · comment in “I built a tool for software developers”

They're fast when they have what you need, but they don't cover every repo, they go stale, and there are hundreds of questions about any codebase that can't be pre-answered. GitHub MCP checks out files one at a time, which burns through context tokens fast and doesn't scale to large codebases.

instalabsai · Feb 13, 2026 · story in “Show HN: Instagit – MCP server that answers questions about any GitHub repo”

Why I built this While experimenting with MCP and AI agents, I found that integrating existing codebases was often the painful part.

justvugg · Feb 11, 2026 · story in “Show HN: PolyMCP – Expose Python functions as MCP tools”

Most tools tell you what the code does; I wanted to see how it's built. The Problem: Onboarding onto a large codebase is a nightmare.

treeliker · Feb 9, 2026 · story in “Show HN: A Satellite View for Python Code”

Looking at the reasoning traces, Opus 4.5 even writes down the correct information, yet somehow provides an incorrect final output anyways. What makes this so annoying is that in coding tasks, with extensive prompts that require far greater adherence to very specific requirements in a complex code base, Opus 4.5 does not show such a regression.

Topfi · Jan 29, 2026 · comment in “Claude Code daily benchmarks for degradation tracking”

Its also time-aware so you can ask it to recall from yesterday for example. What motivated me to build this tool is that I work with big codebases, I spend hours with claude digging into some functionality, and it was frustrating that I have to start from scratch next day/session.

AttentionBlock · Jan 15, 2026 · story in “Show HN: A-MEM – Memory for Claude Code that links and evolves on its own”

The Problem: A user buying a single item needs a completely different UX than a user planning a complex project. A standard "list of cards" doesn't work for both.

ahmedm24 · Jan 14, 2026 · story in “Show HN: LogiCart – Agentic shopping using Generative UI (A2UI pattern)”

Vs code became unusable on a not so big monorepo on a MacBook Pro M4, which is concerning.

coopykins · Jan 5, 2026 · comment in “I switched from VSCode to Zed”

With ChatGPT it's much easier for me to power through the annoying bullshit associated with any large project, and as a result I've been able to get a lot more projects past the finish line.

tombert · Jan 4, 2026 · comment in “Web development is fun again”

I dislike the official git completion script because it’s so slow when working with large repos. On macOS - because of vagaries of the way its file system works - with large repos it’s effectively unusable.

derriz · Dec 25, 2025 · comment in “Archiving Git branches as tags”

We’re three engineers who’ve built and maintained large codebases, and kept wanting to work on as many features in parallel as possible. Git worktrees [1] have been a useful solution for this task but they’re annoying to spin up and manage.

avipeltz · Dec 23, 2025 · story in “Show HN: Superset – Terminal to run 10 parallel coding agents”

It's weird, suspicious, and plain annoying. I like the the tool and my tests have shown it to be very powerful (if a bit rough and buggy), but this is ridiculous - I won't use it for any real world projects until this is fixed.

throwuxiytayq · Dec 18, 2025 · comment in “GPT-5.2-Codex”

I'm assuming this is the case where they are working in an existing codebase written by other humans. I've been in this situation a lot recently, and Copilot is a pretty big help to figure out particularly fiddly bits of syntax - but it's also really stupid suggests a lot of stuff that doesn't work at all.

QuercusMax · Dec 16, 2025 · comment in “Vibe coding creates fatigue?”

Hi HN, I’m Andrew, Founder of Zencoder. While building our IDE extensions and cloud agents, we ran into the same issue many of you likely face when using coding agents in complex repos: agents getting stuck in loops, apologizing, and wasting time.

andrewsthoughts · Dec 16, 2025 · story in “Show HN: Zenflow – orchestrate coding agents without "you're right" loops”

This stuff is a nightmare to manage, and with large code bases/products, you need a dedicated "devops" just to babysit the thing and avoid it becoming a liability for your devs.

kakwa_ · Dec 11, 2025 · comment in “GitHub Actions has a package manager, and it might be the worst”

Gemini seriously need a built-in "Plan" mode that will properly prevent it from doing any write operations. It's pretty good at code review and figuring out massive codebases, but its tendency to just rush in and "fix" things is annoying.

theshrike79 · Dec 1, 2025 · comment in “Gemini CLI tips and tricks for agentic coding”

It keeps the incomplete file in the cache where the sha hash fails until you wipe your entire cache. Which means people with shit internet connections and large projects basically waste hours several times a week doing updates that fail.

hinkley · Nov 24, 2025 · comment in “Shai-Hulud Returns: Over 300 NPM Packages Infected”

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.

boole1854 · Nov 19, 2025 · comment in “Building more with GPT-5.1-Codex-Max”

Just pointing an LLM at a large code base and expecting good things doesn't work.

marcus_holmes · Nov 19, 2025 · comment in “Gemini 3”

First impressions are that its still not really ready for prime time within existing complex code bases. It does not follow instructions.

mparis · Nov 18, 2025 · comment in “Gemini 3”