Propose a stance on AI code/issues/PR's in the kernel.

[IAmTheNerdNextDoor]

This is picking up from a conversation I had on moss's IRC a while ago, and whatever PR 54 was.

I really think that there needs to be a definitive stance here. While it's not top priority, it's good to have.

I have examples of some kinds of enforcement that could be done here:
https://social.coop/@eb/115646613032814668
https://codeberg.org/superseriousbusiness/gotosocial/raw/branch/main/AGENTS.md
https://boehs.org/llms.txt

I can't remember what I said in IRC, but the rules were something along the line of:

1. if you're going to use AI, make sure it works, doesn't make random/unnecessary changes, make sure it isn't "slop", don't add obvious comments in, etc etc, the hallmarks of terrible AI code

and I forgot the rest.

What does everyone think?

[arihant2math]

I think we have few enough contributors and a low-enough volume of PRs that it can be as simple as "if you're going to use AI, make sure it works, doesn't make random/unnecessary changes, make sure it isn't "slop", don't add obvious comments in, etc etc, the hallmarks of terrible AI code".

[IAmTheNerdNextDoor]

Then again, there's the issue of AI straight up stealing code (e.g. Linux exFAT code sent into FreeBSD that ended in rejection) and the issue of licensing that comes with this too.

[arihant2math]

Surprised by fact FreeBSD doesn't support exFAT.

On the point though, I think if it's not considered IP theft under the law, it should be fine. However I have no idea which nations' copyright law applies here anyhow.

I'm more worried about someone forgetting to review AI generated code and it inserting some weird sort of bug in the code. That would be far worse than a massive 11k LOC slop PR or a completely AI generated PR.