Ask HN: How much better are AI code reviews vs. human code reviews?

Has anyone here using AI-powered code reviews (e.g. coderabbit, greptile , codeant, korbit etc.) to human reviews in big projects?

I'm curious about:

- How useful feedback is from AI reviewers - Whether it really catches bugs or just nitpicks - How it fits into your workflow (PR comments, suggestions, etc.) - If you've stopped asking teammates for reviews because of it

Is it good enough to trust? Or still just a supplement to human eyes?

Would love to hear your experience.