How to deal with fully LLM-generated merge requests

[imurx]

I don’t know why but it’s harder for me to read this last wall of text than the whole email chain. I have this problem rather than it being troublesome to read a lot of text it’s because the text is redacted in such a way that it makes its harder to understand.

I’m also a non-native English speaker, I’m from Argentina and speak Spanish. I know a lot of people that were already using machine translation because they don’t know much English and it’s not easy to learn a new language. The problem with machine translation sometimes (or LLM always) is that it makes text feel very formal? corporative? I don’t know how to call it but it’s harder to get the message to the other side with it. And this becomes very tiring with time, having to read these kind of “meat proxy” texts because you start noticing that you have to put more effort into reading their text than the one who didn’t even write it, just asked an AI to do it for them.

You also mention that people were not reading policies before so why would they read them now, well LLMs can mention that you are going against a policy because they would have read it if we go the AGENTS.md way.

Aside from all that, I support having a policy for LLMs so maintainer burden can be reduced to some degree. And we can see with time what needs changing so it can get improved further.

- Uriel