Fifty review rounds overnight

Source: How the Claude Code team does code review, by Addy Osmani, Morgan Westlee Lunt, and Qing Wang (Oct 15, 2026). Discussion: Hacker News.

When a coding agent writes a pull request and a review agent reviews it, the two can keep each other busy. A person waves off a nit in seconds. An agent author fixes it, which means another commit and another round of review.

The case

On the Claude Code team, one CLI pull request went through more than fifty rounds of review and fixes overnight before its author stopped it. When the team ran git blame on what the reviewer kept flagging, about 59% of it was code that had been there since the reviewer's first pass. The reviewer was rarely finding new bugs; it was finding a few more old ones each round.

What the team changed

• A round cap: after three rounds, the reviewer asks the author to step back and reconsider the approach.
• Diff-only re-review after the first few passes.
• Flag the class of bug, so the author fixes every instance, not one line.
• No repeats of findings already posted.

The numbers after (internal pull requests)

measure	before	after
pass after one review and fix round	65%	78%
need five or more rounds	8.9%	1.5%
mean reviews to pass	2.08	1.32

Why it matters for an engineering lead

The team tracks rounds-to-pass alongside recall, because noise costs more when the author is an agent: every extra finding becomes a commit and a new review. The post also describes risk tiers for human approval and turning past incidents into an eval set for the reviewer; the source has the full process.