Agent-Manager: Reviewing code while an AI agent is still rewriting it (agent-manager.dev)

A new code review approach, termed "Agent-Manager," has been introduced that allows developers to review code while an AI agent is still in the process of rewriting it. Traditionally, code reviews only happen after the code has been finalized, but with the advent of coding agents, developers can now provide input even while changes are ongoing. This method hinges on treating comments as concurrent states rather than static annotations, employing a system that intelligently re-anchors comments based on the text surrounding the edited lines. This development is significant for the AI/ML community as it enables more dynamic and efficient interactions between human developers and AI coding agents, potentially improving productivity and collaboration. Key technical innovations include maintaining the integrity of comments by capturing precise excerpts and utilizing session-based keys to prevent outdated information from overwriting current work. The result is a synchronized workflow where revisions and feedback can progress seamlessly, demonstrating a practical way to coexist with intelligent coding tools without hindering the review process.