Quoting Addy Osmani, Morgan Westlee Lunt, and Qing Wang
Simon Willison’s Weblog Simon Willison
Claude now reviews code before humans do at Anthropic. They say it catches routine bugs early, so people focus on the hard stuff.
Based on reporting by Simon Willison’s Weblog, Simon Willison — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic’s Claude Code team has flipped code review from a last stop into something much more continuous. Instead of writing code, opening a pull request, and waiting for a colleague, review now happens as the work is being written.
The team’s point is simple: “did a human read the code?” is the wrong test. What matters is what could go wrong, and whether the best check for that risk is automated or human. That shift changes the whole rhythm of review.
Most changes, they say, have already been reviewed and fixed by Claude at least once before a human ever sees them. That takes the edge off routine mistakes. It also means human attention can go to the questions that actually need judgment, not just box-ticking.
There’s a quiet logic to it. If a machine can catch the obvious stuff early, humans don’t have to waste their limited patience on obvious stuff later. The interesting part isn’t that Claude is doing review; it’s that the team is redefining what review is for in the first place.
My take — AI-written commentary, not fact-checked reporting
This is the sane version of AI coding: let the bot sweep up the lint and the silly mistakes, then bring in a human for the parts that matter. The industry keeps pretending review is a sacred ritual; really, it’s quality control with a calendar problem. Claude just gets there first.
Read more about this at: Simon Willison’s Weblog