The Mythical Man-Month in the Age of AI
Points and comments are a snapshot, not live.
Brooks's Law still applies to AI agents, only the bottlenecks have changed.
The author argues that while AI agents remove traditional software overhead like onboarding and meetings, Brooks's Law has mutated rather than vanished. Sequential tasks remain sequential, and new bottlenecks emerge: managing agents, defining "done," reviewing code, and catching hallucinations. The proposed 2026 update: "Adding agents to a project with unclear requirements gets you the wrong answer faster."
What commenters are saying
Commenters broadly agree that coordination and understanding remain the cost, not typing. Many note that AI makes code production cheaper but review and design coherence still demand human attention. A split emerges: some find small, well-bounded tasks with tests work well, while open-ended changes create review work that cancels speedups. One camp argues for not reading AI-generated code at all, but others counter that accountability requires understanding. A strong thread argues that rapid, wrong prototypes are now the best way to discover requirements, saving expensive stakeholder meetings.