One of things i find myself doing a lot these days is getting chatgpt to write a function and then using another chat instance to check the code. I’ll do this over and over until all the bugs are caught. Since each chat is like a new code reviewer it often finds all of the bugs after many iterations.

I’ll even give the code review from one of the AIs to the one who wrote the code, it will then give a different implementation based on the ai review. I then give that implementation to the reviewers implementation and ask if its better or not. I can very quickly get to an optimized and less buggy function this way.

This feels very *human*. This is exactly how code review tends to play out in meatspace.

Demo: