Addy Osmani says agent-written code needs automated constraints, not human review

Addy Osmani argues that code review does not scale to code written by AI agents, so quality checks should move into the harness and environment around the agent.

In an O'Reilly Radar piece, he describes quality gates that decide whether an agent's change can ship, including unit, property and acceptance tests, mutation testing and complexity metrics. He says human attention should be reserved for nuanced problems, with people brought in only when those automated constraints fail.