I personally think their criteria for no AI is a bit too much. While I am 100% against slop code, if someone does a PR on my code that adds a constructor or a few getters, I wouldn’t refuse it just because “it’s tainted with AI.”
Yes I would have preferred it if it wasn’t. Yes I will give them a talking to. But I’m not going to refuse their PR either as a human would do the exact same thing as the LLM produced. It’s just unproductive, wasted time and compute for no reason.
And with a fully against AI policy, that just makes people hide their usage of AI, making it even harder to find the slop code.
Idk, given that we can’t really conclusively prove that anything post-2022 is slop-free, don’t you think setting a no-AI policy is still good for setting social expectations? And making people spend effort to conceal AI work is not a bad thing, since one problem with slop-PRs is that they are low-effort.
And I use tons of old projects that haven’t made an anti-AI declaration (and have to use a bunch that are slop-permissive e.g. the kernel), but I still appreciate every one that does.
I’m not against a policy. I’m just choosing a policy like “99.9% of slop will get rejected” policy .
Although I never really experienced bad low effort PRs. Either it was small changes with no other ways to do X, which I merged with a “please don’t use AI in the future, it’s bad blah blah blah”… or the entire project forked, updated to an unstable version, and full on slop that I reject
I personally think their criteria for no AI is a bit too much. While I am 100% against slop code, if someone does a PR on my code that adds a constructor or a few getters, I wouldn’t refuse it just because “it’s tainted with AI.”
Yes I would have preferred it if it wasn’t. Yes I will give them a talking to. But I’m not going to refuse their PR either as a human would do the exact same thing as the LLM produced. It’s just unproductive, wasted time and compute for no reason.
And with a fully against AI policy, that just makes people hide their usage of AI, making it even harder to find the slop code.
I would most certainly give them a talking to for trying to add constructors, getters, and setters!
Idk, given that we can’t really conclusively prove that anything post-2022 is slop-free, don’t you think setting a no-AI policy is still good for setting social expectations? And making people spend effort to conceal AI work is not a bad thing, since one problem with slop-PRs is that they are low-effort.
And I use tons of old projects that haven’t made an anti-AI declaration (and have to use a bunch that are slop-permissive e.g. the kernel), but I still appreciate every one that does.
I’m not against a policy. I’m just choosing a policy like “99.9% of slop will get rejected” policy .
Although I never really experienced bad low effort PRs. Either it was small changes with no other ways to do X, which I merged with a “please don’t use AI in the future, it’s bad blah blah blah”… or the entire project forked, updated to an unstable version, and full on slop that I reject