What Are Code Reviews Even For?

Brian Houck, an Applied Scientist at DX, wrote a thought-provoking piece titled What are code reviews even for?.

The article examines the hidden risks of delegating code review entirely to automated AI tools. Beyond the obvious risk of subtle bugs slipping through, offloading reviews erodes one of engineering teams’ most vital mechanisms for building shared knowledge, understanding architecture changes, and mentoring junior engineers on best practices.

Rather than arguing against AI assistance altogether, Houck advocates for using AI to handle routine mechanical checks while ensuring humans remain engaged in the conceptual and architectural review.

It raises an important question for modern engineering teams: what workflows let us leverage AI for review efficiency without losing connection to the codebase?