Vibe coding kills open source • Hacker News Board

Re: Vibe coding kills open source

#301

Post by vrighter » Thu, Jan 29, 2026, 3:38 PM UTC

px43 wrote: post #90

tomaytotomato wrote: post #6

I have been trying to use Claude code to help improve my opensource Java NLP location library. However trying to get it to do anything other than optimise code or fix small issues it struggles. It struggles with high level abstract issues. For example I currently have an issue with ambiguity collisions e.g. Input: "California" Output: "California, Missouri" California is a state but also city in Missouri - https://gi…

> it struggles It does not struggle, you struggle. It is a tool you are using, and it is doing exactly what you're telling it to do. Tools take time to learn, and that's fine. Blaming the tools is counterproductive. If the code is well documented, at a high level and with inline comments, and if your instructions are clear, it'll figure it out. If it makes a mistake, it's up to you to figure out where the communicati…

A tool is something I can tightly control. A thing that may or may not work today, and if it does, might stop working tomorrow when the model gets updated without any notification to anyone, the output of which I have to very carefully scrutinize anyway, is not a tool. It's a toy.

Re: Vibe coding kills open source

#302

Post by vrighter » Thu, Jan 29, 2026, 3:45 PM UTC

antirez wrote: post #130

m000 wrote: post #32

Earlier quoted context omitted.

The real question is how much of the new wave of vibe-coded software will be able to graduate from pet project to community-maintained project. It feels that vibe coding may exacerbate fragmentation (10 different vibe-coded packages for the same thing) and abandonment (made it in a weekend and left it to rot) for open source software.

I believe the process of accumulation of knowledge / fixes / interesting ideas will be still valid, so there will be a tons of small projects doing things that you can replicate and throw away, but the foundational libraries / tools will be still collaborative. But I don't agree with the idea of fragmentation, AI is very good at merging stuff from different branches, even when they diverged significantly.

What accumulation of knowledge? LLMs are marketed as "democratizing" (I hate this term) software development specifically by claiming to remove the need for knowledge.

Software development hasn't been democratized. The term doesn't even make sense. Anyone could learn whatever using the same resources as anyone else. I would (and have, in the past) gladly teach anyone who asks me to. And most of my friends would too. The fact that non-devs couldn't write code was not because devs were gatekeeping, but because they didn't want to learn. And they still don't.

Re: Vibe coding kills open source

#303

Post by munksbeer » Thu, Jan 29, 2026, 8:57 PM UTC

koakuma-chan wrote: post #200

simonw wrote: post #158

Earlier quoted context omitted.

Have you looked at antirez's code? https://github.com/antirez/flux2.c

Generating code from scratch and modifying existing code are two different things, obviously the latter being where AI doesn't do great. Carefully managing and compressing context can somewhat help, but that is far from being a perfect solution.

Every time someone asks for good examples of AI written code, and they are given one, they either move the goalposts or simply don't reply.

I went from using copilot autocomplete and mostly just disabling it because it was not great, to now using claude most of my day. I gets stuff wrong. It does generate inefficient code. It definitely has a way to go, but otherwise it just saves so much time.

Re: Vibe coding kills open source

#304

Post by koakuma-chan » Thu, Jan 29, 2026, 10:59 PM UTC

munksbeer wrote: post #303

koakuma-chan wrote: post #200

Earlier quoted context omitted.

Generating code from scratch and modifying existing code are two different things, obviously the latter being where AI doesn't do great. Carefully managing and compressing context can somewhat help, but that is far from being a perfect solution.

Every time someone asks for good examples of AI written code, and they are given one, they either move the goalposts or simply don't reply. I went from using copilot autocomplete and mostly just disabling it because it was not great, to now using claude most of my day. I gets stuff wrong. It does generate inefficient code. It definitely has a way to go, but otherwise it just saves so much time.

That's assuming the above-referenced code is good. I know next to nothing about flux, so I cannot tell. What I can tell though is that despite people claiming that AI can write code in Rust, I have to rewrite virtually all AI-generated code myself, because it's fucking garbage. It cannot even consistently follow basic guidelines. I concede that AI can save time, but my original point stands.