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

A lot of people have been trying out AI code review tools, but what I see most often is feedback that sounds smart but isn’t actually helpful. Either it flags way too much, or it misses the important stuff. In the end, it just adds more noise to the PR.

Teams that rely purely on LLMs like ChatGPT or Gemini run into this all the time — generic comments, no real understanding of the project, and sometimes subtle mistakes. It’s hard to trust, let alone use it as a real part of your review workflow.

That’s exactly why we built Kodus. It’s an AI-powered code review platform that runs directly on your PRs, combining generative intelligence with deterministic rules based on how your team already works. The result is less noise, more consistent feedback, and more time to focus on what actually matters.

If you want to give it a try and share feedback, we’d love to hear it. We’re improving the platform based on real-world usage https://kodus.io