AI code review can be useful, but I always review the code it produces, coupled with intentional prompting with detailed tasks.

> But I wouldn't run my production apps—that actually make money or could cause harm if they break—on unreviewed AI code.

I hope no one is actually letting unreviewed code through. AI can, and _will_ make mistakes.

Nowadays > 90% of my code tasks are handled by AI. I still review and guide it to produce what I intended to do myself.