Rubén Rubio
lux 🦊ΘΔ"You should judge PRs based on code quality not the fact i used LLMs" okay but if a construction worker closes their eyes and starts flailing their arms while holding hammers and says "you should judge my skill based on the nails i hit" you rightfully move aside and stop them before they knock someone unconscious
nicole4fox@SharpLimefox since when did we stop caring about the outcome when its about getting stuff done? As long as the person can exain everything and its an actually good PR?
Todd Knarr@nicole4fox @SharpLimefox Time spent. Most maintainers don't have unlimited time, so they have to prioritize which PRs are worth reviewing and which can be dropped for lack of time with the least impact. Note that maintainers already do this: if you consistently submit low-quality PRs, you'll soon find the maintainers just ignoring you.
@tknarr @SharpLimefox oh yeah thats why i was talking about high quality prs that still use llms
@nicole4fox @SharpLimefox A dev who submits lots of low-quality PRs can still submit the occasional good one. The thing is, the maintainers don't have the time to spend on all those bad PRs on the off chance of finding a good one. So they ignore the dev and accept that the loss of a few good PRs from them is worth it to free up time for PRs from devs with a better track record.
Same thing with genAI. It tends to produce lots of low-quality PRs and the devs most likely to use it...
@nicole4fox @SharpLimefox ... are the ones least likely to review their own code and clean it up to a higher standard. It makes sense then for maintainers to apply the same logic and drop it in favor of PRs what're more likely to be usable.
@tknarr @SharpLimefox I feel like i was only disagreeing with not reviewing high quality LLM commits by categorally excluding those. There are good devs out there who use LLMs and adhere to standards. To me it doesnt make sense to exclude PRs for the sole reason of being written by an LLM. Partly because you can train a local LLM based on your own coding style.
@nicole4fox @SharpLimefox The problem is then you get people submitting low-quality LLM-assisted PRs going "But you accepted them from HIM! Why not from ME!?". Even if the reason is quality, it turns into a time-wasting flame war about LLMs. Not specific to LLMs either, if there's any two criteria there's always a group who'll try to abuse the one that benefits them.
Which would be manageable by just focusing on quality, save that LLMs make it easy to generate a couple dozen bad...
@tknarr @SharpLimefox id just give them a LLM generated response about code quality and why this PR is bad and the other is not tbh
@nicole4fox @SharpLimefox Still burns maintainer time dealing with their arguments. Hence the hard and fast rule with no room for abuse.
@nicole4fox @SharpLimefox ... PRs for every decent one. That results in the maintainer getting overloaded and drowned in crap. They have to deal with that without burning all their time on stuff they can't use.
@tknarr @SharpLimefox i disagree tbh but i dont think we will agree on anything in that regard so ill just leave it be
Joe Brockmeier@nicole4fox @SharpLimefox is it "about getting stuff done", solely, though?
It kind of reeks of entitlement to demand that somebody take the time to evaluate or accept something if they've set criteria against it.
To make a point with a silly example: if a project says "we won't accept any contributions created with Vim", then I'll accept that. End of story. It's not for me to dictate to the project what its rules are, even if I think it's a stupid rule.
Even if the project has literally no way of telling whether I used Vim, Emacs, nano, or something else -- I'll respect their rule and not submit things coded or written in Vim. Because, IMO, open source is not merely about "getting stuff done" -- it's about collaborating with people to build community. And you can't build community without respecting people's boundaries.
Yes, that means that some communities are choosing to exclude LLM-driven stuff and, by extension, contributors who only want to submit LLM-assisted contributions.
Surely, if LLMs are so awesome, those contributors can start their own projects that welcome LLM-assisted contributions and outpace the human-driven projects because they'll be "getting stuff done", right?
@jzb @SharpLimefox i dont mind andhering to rules set by the maintainers and its completely valid to exclude LLMs from ones project i just think it doesnt make sense when the PRs are actually good and high quality to exclude them as a maintainer.
Mark Rotteveel@nicole4fox @SharpLimefox I can give you several reasons: 1) copyright: 1a) LLM output is not copyrightable (in the US for one), so you also can't license it, 1b) the training data can leak into the produced code, so you might open yourself up to copyright and/or license infringement by accepting it into your project, 2) LLMs produce seemingly high quality output, while being either subtly or outright wrong, the seemingly high quality makes it mentally *harder* to detect this (more review time).
dasdom@SharpLimefox Why exactly should I? My code, my rules. No AI code in my repositories.
kira for scale@SharpLimefox can't help but feel like if everyone on earth decided to become a vuvuzela-ist and they all started applying to conservatories and orchestras, they would all start turning down vuvuzela players on principle even if one or two of them happen to be the well-tempered even-handed one who could play a vuvuzela in an orchestra
64 Islands Aroha Cooperative@SharpLimefox fully 80% of my code review praxis is based on 'do i trust this dev to make good decisions' -- if the answer is no then it's going to take 3x as long to do the review
arihi 
@SharpLimefox hammer flailing is welcome in this construction project, as long as you let us know you were flailing around with the hammers while building the house. flailing around with hammers is the future of building so it would be unwise to limit ourselves to old-fashioned hammer-hitters.
Goupilleau@SharpLimefox If they didn't bother writing it, I'm not going to bother reading it.
ari 
@SharpLimefox it's a lot more like the construction worker, who figures hammering nails is beneath them, recently bought this nifty nail-hammering robot that hasn't been properly field-tested or regulated in any way to be remotely safe, and says "you should judge my skill based on the nails i hit".
except they're not actually hammering nails at all anymore, instead sitting in a sun chair sipping lemonade while the nail-hammering robot has started bludgeoning a random pedestrian in the background
except they're not actually hammering nails at all anymore, instead sitting in a sun chair sipping lemonade while the nail-hammering robot has started bludgeoning a random pedestrian in the background