I wonder where they place the line. Many IDE’s today use machine learning for autocomplete, it would be difficult to have a 100% AI free piece of software.
I once asked ChatGPT to find a bug in a piece of code that wasn’t working. It found it and I fixed it now that I knew what the problem was. Does that count as AI-assisted even if I didn’t use any AI generated code?
I don’t like the arbitrary exception for “well established” projects. It sounds like they want to keep their cake and eat it too. Ban AI while not losing any big popular projects. As most big popular projects will have some form of AI assisted content. If they are principled they should ban it for everyone. As it is, what they saying is that AI slop is welcome as long it’s in a popular project.
So caliber is getting booted?
Based
I wonder how that policy change affects library usage, because while I’ve still never felt a need to code anything using an LLM, I rely on e.g. Valibot for input validation.
Witches can be found everywhere once you start hunting for them.
It could be to be honest, who knows what absolutely mental policy they’ll come up with next
Good news for Snap though
Flathub now prohibits applications containing any AI-generated or AI-assisted code, documentation, or other content.
I would say this is a huge kick in the shins because AI is everywhere in coding now but I don’t think flathub is used that much anyway? I can’t think of anything I’ve been like, holy hell I better go to flathub to get this app, I can’t get it anywhere else
Just because they‘re used everywhere doesn’t mean that we just have to accept them. Also doesn’t mean that LLMs are a good thing.
I think LLMs can be used as an (additional!) cyber security analysis tool, that’s honestly the only area in which it seems to be actually useful (right now). And most projects don’t reach the size in which spotting security risks spanning across many different modules is a relevant skill to have. So it should be used sparingly, on things like the linux kernel. Then the cost of it might even be worth it (but I also don’t want to know about the amount of hallucinated bugs it finds).
And I want to add: even though LLMs can identify cybersecurity risks, it doesn’t mean they are good at cybersecurity. They’re probably just as bad as in any other area. Also questionable if the actual positives outweigh the labor required to flag all the false-positives.
It’s the main flatpak repo, so it does get a lot of use on Linux. Anyway, people who really want those apps will just have to add a third party repo most likely.
It’s what Bazaar uses on Bazzite.
I think the policy was made in good faith to stop vibe coded apps from being added, but there are actual developers using AI to speed up development. It can help fill in with weaknesses that slow developers down.
It can help fill in with weaknesses that slow developers down.
No no no no…
You don’t want AI replacing devs where they’re “weak”. That is literally the worst thing you can do with AI. All that does is mean the devs aren’t qualified to assess and debug those portions of code. The solution to devs being “weak” in an area is for them to gain experience doing that task.
Some programmers use AI to make boilerplate code they can easily check to save time without much issue. That’s about the only thing you can use AI for in software development with little risk without taking up excessive time checking/fixing what it shits out.
I’m weak in knowing APIs. I can use AI to generate the code that might include a library that I wasn’t aware of. I can then use my experience to validate syntax with linters as well as reading up on the library and finding if there are features that exist that might help solve the issue that I have.
It’s not perfect as AI can make up a library that doesn’t exist, but it lets me get started and fine tune the code for a solution that works the way that I want.
Instead of that, you could just search online and go through examples on how to use the API’s in question. Then you actually learn and gain experience in that task, which you wouldn’t get if you have AI get you 95% of the answer from the start.
While having an AI code something you’re unfamiliar with is more efficient for that specific task, over time it’s less effecient and detrimental to you as a programmer. Think of it like getting a teacher’s edition of a textbook for a class. You can answer questions more quickly by looking up the answer in the back of the book and copying it down. Sure, you absorb the answer a bit, but not nearly as much as if you worked through it yourself.
Now think about the differences between junior and senior dev jobs. What’s the primary difference? It’s experience. If you’re a junior dev that has AI do all the hard parts for you, then you’ll gain experience slower than another junior dev that does everything on their own. In the future, that other junior dev is more likely to get the lucrative senior dev promotion than you because they’ve built up more experience.