Skip to content
Main Site News Console

Quotes from Addy Osmani, Morgan Westlee Lunt, and Qing Wang

· Simon Willison Translated
技巧LLM

Sponsor: Teleport — AI agents (and the sub-agents they spawn) are granted the same access permissions as humans, yet they never sleep. Learn how to monitor and constrain their behavior using isolated ephemeral trusted runtimes.

October 16, 2026

We believe that “did a human review this code?” is a poor substitute for “what could go wrong with this change, and what kind of check—automated or human—is best suited to catch it?”

Asking this question also shifts when reviews happen. Historically, review was a gate at the end of the process: write code, open a pull request (PR), and wait for a peer to review it.

Today, review runs continuously, much like a spell checker while you type. Most changes have already been reviewed and fixed by Claude at least once before a human is ever asked to take a look. This catches the vast majority of routine errors and preserves the team’s attention for higher-level discussions and rigorous quality control.

— Addy Osmani, Morgan Westlee Lunt, and Qing Wang, How the Claude Code team does code review