|
· Föderation EN Sa 24.05.2025 00:24:18 @Lucomo depending on the context, people responsible for reviewing code like this have to be allowed to decline the review if the programmers want to take responsibility for it. I say let it burn so people see what happens. It's so bad for code reviewers' mental health. |
|
Föderation EN Sa 24.05.2025 02:46:39 @990000 @Lucomo if you are not allowed to say no to PRs why are you spending time reviewing them? |
|
Föderation EN Sa 24.05.2025 03:38:39 @kevingranade @990000 @Lucomo The problem is that figuring out that e.g. auth was skipped takes way more than 10s. |
|
Föderation EN Sa 24.05.2025 04:39:17 @Anibyl @990000 @Lucomo yes the almost right ones are the worst, but it kind of sounds like they were doing subtle stuff like that and also absurd out of control nonsense at the same time and they're being told to review and merge it anyway. My point is the ai is ludicrously bad, but the real failure here is them apparently not being allowed to say no to egregiously bad PRs. |
|
Föderation EN Sa 24.05.2025 08:47:38 @kevingranade @Anibyl @990000 @Lucomo This also shows how broken the current way people treat PRs. I see this a lot in small companies and corporate alike. If you have a PR as a single unit of work, and commits are meaningful changes, then 5k LOC PRs shouldn’t even be possible—unless you’re switching something in the code formatter or renaming a core thing. But those PRs should be exceptional and so similar that other changes are rejected instantly. |
