Aside from bug/issue spam where maintainers are stuck dealing with bogus problems raised by LLMs and posted by lazy humans, I don’t think there ought to be an issue. Maintainer burnout is no joke though; a bad day topped off with a bogus bug report might be the straw breaking a camel’s back.

The other issue is licensed code with stricter copy left protections can be stolen for training data and unwittingly regurgitated into a code base elsewhere. Unfortunately this would require lawyers to actually bring these cases to court, which is not happening at the moment. Seems more manageable as an issue though since there is a path to resolving it. I think I need to make an EFF donation to help with that process.

Open Source will always be on life-support under capitalism.

No genuine good open source project is dying because it is getting replaced by LLM slop code. People that need stable and secure software and know what they are doing would never use LLM made software.

What LLMs do destroy is peoples patience in dealing with bug reports and forum communication. Those are getting overrun with LLM generated garbage replies.

It creates a problems for copy-left licensed projects. Someone may be unintentionally breaking the license by using LLM bots, or even LLM autocomplete.

It’s being badly damaged by thousands of lazy AI-written merge requests to the extent that many have stopped accepting them entirely.

It’s also effectively DDOS’ing many open source sites (and the rest of the open internet), straining what limited funding they have.

So I would say a resounding yes.

As far as my applications for open-sourcing goes, AI has actually done a good number on assisting it.

I’m a DIY sort of person, and use a lot of software for things like ESP32 boards to complete niche tasks. The problem is that very many applications just didn’t have some preexisting code made for it, so it took a much larger load for me to try programming it by hand. In recent years, I’ve had a much easier time finding software for things, and sure enough, many of these projects have some mention or disclaimer about AI.

I know AI brings its own problems with it, namely that of code produced with lesser-optimized techniques, but the alternative I had to deal with was simply no premade code at all.

That being said, many of these projects did die out after AI was implemented, but not because the community was less interested, or the developers were less caring. These projects died because they reached their end goal, they did exactly what you needed it to do, no more or less. Far as I’m aware, that sounds like a successful outcome.

It’s one thing for a company to train a model with your code and then create a better copy of what you made and sell it for profit (which I think is an unrealistic thing to happen if their codebase is depending on AI slop code), and it’s another thing that an AI is providing access to public information (the code) that you previously monetized to help people understand it better. I really don’t see how that monetization model would have worked regardless of AI existing, at some point there are going to be enough people out there that understand the code that can build documentation of their own for free. I’m not a lawyer but I don’t see how this violates a GPL license either.

The only thing FOSS projects have to be wary of about AI is slop pull requests, but code review still had to be done before LLMs existed anyway.

Also my two cents about the threads regarding Tailwind is that, what FOSS devs wanting to live doing what they do should really hate is not AI making it harder for them to monetize their projects in odd ways, but capitalism requiring them to monetize anything they do for them to be able to live while doing it. FOSS devs should be able to hand out their creations to society without worrying about putting food on the table, their work is no less valuable than that of any engineer working for the big corporations.

It depends on the project. Not all open source project uses tailwind’s revenue model, which relied on that people are reading their docs.

LLMs won’t destroy open source, but it will change the landscape. Probably not for the better though.

No. Libraries will always be useful. Not everyone has the credits to generate the tokens necessary to rewrite every lib in existence.

If I can replace an open source project with shitty AI scripts, what is that project actually even doing?

Not defending the ethics of AI, but shit like “left-pad” and other low hanging packages shouldn’t exist.