Ask HN: How Do You Review Your Code in the Age of AI?

With AI copilots becoming a major part of our dev workflows, I've been wondering how are developers and teams evolving their code review practices?

Traditionally, code reviews have been about readability, maintainability, performance, and correctness. But now with AI-generated code (from GitHub Copilot, ChatGPT, etc.), I'm noticing new patterns and challenges:

The code "works," but the logic is sometimes unfamiliar or subtly incorrect.

It's harder to gauge intent when the code is AI-assisted.

Reviewers often assume the AI got it right—dangerous!

AI can write very clever code—but should it?

So, I'm curious:

How do you personally approach reviewing code in this AI-assisted era?

Do you or your team have specific rules or red flags for AI-generated code?

Are there tools or processes you're using to catch silent failures, hallucinations, or overly complex solutions?

Do you still prioritize peer review or rely more on automated/static analysis tools?

Looking to learn from how others are adapting—especially for production-grade systems.