Agents Write My Code. Agents Review It. I Referee.

This article explores the author's experience with AI agents that write and review code, transforming their role from a bug-catching reviewer to a referee overseeing multi-agent reviews. The author argues that as agents handle code generation and review, engineers are elevated to focus on product judgment, but caution that trusting agents requires adversarial review swarms and deterministic checks. The piece critiques Linear's argument that reviewers stop catching bugs, asserting that bug-catching is merely delegated to agents, not eliminated. It highlights the value of diverse agent perspectives for catching edge cases in areas like security, CSS, and SQL, while acknowledging shared blind spots and the need for human oversight. The article is relevant to IT/TECHNOLOGY as it discusses AI-assisted software development, code review practices, and engineering workflow evolution.