Addy Osmani, Morgan Westlee Lunt and Qing Wang say the Claude Code team uses Claude to review and fix most code changes before asking a person to look. They describe review as a continuous check during coding, rather than a gate that happens only after a pull request is opened.
The team says this catches routine mistakes and leaves human attention for broader questions and closer scrutiny. It argues that the key question is what could go wrong with a change, and which automated or human check is suited to catch it.
Why it matters
For developers on the Claude Code team, this approach may leave more review time for higher-level concerns; the source does not say how well the process works outside that team.
