Asking for a codereview, how many times?

I'm curious what your experiences are with this. I often find myself working on a feature then after completion ask for a codereview on the uncommitted changes. This is something that I do; but with capable models of today it can keep identifying findings. As a QA myself I look at this, and then often do some planning around that findings scope to best solve the issues. However with the powerful models that we have it can keep iterating and iterating and finding stuff 🤭 I'm wondering if you have experienced some of the same and is there maybe a number of iterations you max do; lets say 3? Or in your AI harness / skills do 1 or 2 rounds on this and identify the feature "finished"?

3 comments

Asking for a codereview, how many times?