Comment by tossandthrow

If I go into a PR that requires a lot of feedback, I will usually stop after 5 - 10 pieces of feedback and let the author know that it is not ready for review, and needs to be further developed.

I am OK with the author using AI heavily. But if I continue to see slop, I will continue to review less and send it back.

In the end, if the engineer is fiddling around for too long, they don't get any work in, which is a performance issue.

I am always available to help out the colleague to write understand the system and write code.

For me, the key is to not accept to review AI slop just like I do not accept reviewing other types of slop.

If something is recognized as slop, it is not ready to be reviews.

This puts an upwards pressure on developers to deliver better code.