Eugene  2d ago

LOL, the first vibe-coded commit landed in the FreeBSD. The fun part — in this commit was changed literally one line in one file. And this required the use of LLM, LMAO?!   

#FreeBSD #LLM

Show threadsome quirky girl2d ago
@evgandr i know for the fact this is not the "first" LLM commit in FreeBSD. and anyway, the change seems correct. so who cares?
Show threadEugene  1d ago

@lw I care. If there will be a lot of commits from "passer-by committers", who don't want to understand how things work and don't care about code they are changing — because they just want to obtain a stripe "Look ma, I committed to the FreeBSD" and nothing more — then we are doomed  

With LLM-conmits it is pretty easy to overload reviewers  

Also, it makes questions about developer's intent — they used LLM because they don't have fun from programming and/or debugging (so, why contributing to FOSS?)? Or they were pushed to release new features faster (like in corporation)? Or smth likewise?

Show threadPiotr Smyrak
@evgandr @lw Wasn't there some sort of moratorium on AI based contributions to the project?
Mar 26 at 12:06pmWeb
Show threadEugene  10h ago
@piero @lw As I remember, the policy for such contributions is still in development and for now the definitely "slopcode" (i.e. committer didn't understand the code and just copy-paste the LLM output everywhere) is prohibited
Show threadsome quirky girl7h ago

@evgandr @piero there is no policy from core on LLM-generated commits yet. we hope they will produce such a policy, but i would put money on that policy *not* prohibiting LLM-generated code.

i don't have any insider knowledge there, that's just based on my understand of how FreeBSD developers view LLMs from talking to them.

code which is bad or which the submitter doesn't understand should not pass review regardless of whether it's LLM-generated or not.