I would be a lot more sympathetic to the "code reviews are not for finding bugs" noise if:

- I hadn't, this week, spotted a multiple bugs in multiple code reviews
- I hadn't, this week, spotted a bug in LLM-generated code
- There weren't multiple studies that showed how effective code review is at catching bugs.
- If Copilot reviews hadn't caught bugs in my code

I agree, code review is not _just_ about catching bugs. But this recent attempt to claim, in the face of all the facts, that we don't need to do reviews to catch bugs in LLM-generated code smacks of desperate rationalisation. That's the voice of the hype talking: LLMs will make you 10x, 100x faster, but you're slowing it down by doing all of these _tedious_ reviews.

#LLM #ai #code #bugs #codereview