Anthropic’s NEW Claude Code Review Agent (Full Open Source Workflow)

In this monologue podcast, Patrick discusses Anthropic's open-sourced agentic review system, which replaces manual code reviews with AI agents. He explains how Anthropic uses Claude Code to build Claude Code, achieving significant velocity by automating code reviews and security audits. Patrick breaks down the AI Code Review process, highlighting the roles of both AI and human reviewers, and introduces frameworks like inner and outer loops for efficient development. He emphasizes the importance of context, tools, and validators for AI agents and shares high-leverage mental frameworks, including engineering process encapsulation and customization. Patrick also demonstrates the use of slash commands, subagents, and GitHub Action Runners, providing examples and resources for implementing these workflows.