# Addy Osmani - Agentic Code Review (Highlights)  ## Metadata **Review**:: [readwise.io](https://readwise.io/bookreview/61390452) **Source**:: #from/readwise #from/reader **Zettel**:: #zettel/fleeting **Status**:: #x **Authors**:: [[Addy Osmani]] **Full Title**:: Agentic Code Review **Category**:: #tweets #readwise/tweets **Category Icon**:: 🐦 **URL**:: [x.com](https://x.com/addyosmani/status/2066595308629594363/?rw_tt_thread=True) **Host**:: [[x.com]] **Highlighted**:: [[2026-06-17]] **Created**:: [[2026-06-20]] ## Highlights - I point Claude Code or Codex at a batch of incoming PRs and ask for a first pass: a high-level read of what looks safe to merge, what needs more work, and what is genuinely high-risk. I do not auto-merge on the result, and I do not lazy-merge whatever it approves. What it gives me is a way to allocate attention. ([View Highlight](https://read.readwise.io/read/01kv9e8y6w57fjf15cy5y4dppt)) ^1025989920 - This is how you stop being the first human to read the code. You push the intent-reconstruction work back onto whoever submitted it, where it is cheap, rather than absorbing it yourself, where it is expensive. ([View Highlight](https://read.readwise.io/read/01kv9dvjbrwfec234y6mbnh9m0)) ^1025989403 - **Read the test changes more carefully than the code.** This is the agent failure mode to watch. The agent changes behavior, then “fixes” the test by rewriting the assertion to match the new, broken behavior. ([View Highlight](https://read.readwise.io/read/01kv9dwrwxwnvwy50qndtx80b3)) ^1025989470