A software engineer explores how AI agents writing code alongside minimal human oversight mirrors the reliability demands of modern systems. The author argues that if a system requires a sober, sharp developer to deploy safely, it's fundamentally fragile—and that comprehensive verification (tests, formal methods, security tracing) should enable confident deployment regardless of human state.