On Friday, 13 March 2026 18:56:13 UTC you wrote: > chris <[email protected]> writes: > > Actually, I'm not sure that I agree, especially regarding the usefulness > > aspect. I don't think there is any. > > If the contributor did check and understand all the code they suggest, > then they at least did a review themselves. > > If that’s what they do, and they explain the benefit of the contribution > themselves (I don’t mind them using an LLM to fix typos or grammar, but > the content of the description must be their own), then I’d consider > reviewing it. > > >> In other words, focus on quality, not on means. > > The contributor is the first who has to focus on quality. If they do > that, and they actually know that what they propose has high quality > (they can’t if they didn’t read it, they can’t if they don’t understand > it, they can’t if it’s a wall of code too big for anyone to read, they > can’t if they didn’t test it all themselves), then they clear the > minimum bar.
I tried very hard in my previous answer to draw some limits to what I was saying, but apparently I failed. This is what I said: ``` Actually, I'm not sure that I agree, especially regarding the usefulness aspect. I don't think there is any. First, there's bad code masquerading as good code, which creates a heavy workload for the maintainer and doesn't produce any useful work. It's akin to a DOS attack on reviewer. Second, clicking the button that says, "Please fix that bug and send the patch to the maintainer. Don't bother showing me the code; I don't care," doesn't carry out any work, and therefore isn't useful. It's useful if someone else does measurable work for you. However, clicking carries an area under the work curve equal to zero. Therefore, there is no help there. ``` In my opinion, code generated by an LLM that hasn't been reviewed by the contributor isn't useful, even if it's correct, good, or optimal by mere luck. Otherwise, I'm a total proponent of LLMs. However, we cannot allow LLMs to pour unmanageable amounts of unreviewed code into the codebase. Reviewing all the poor-quality code LLMs can generate would overwhelm the maintainers. > > Best wishes, > Arne
