AI Agents Reframe Brooks’s Law
AI agents are changing how software projects divide work, but they have not made Brooks’s Law obsolete. Based on Frederick Brooks’s experience with large software projects, the law holds that adding people to a late project can make it later. Agents can absorb context quickly and avoid much of the training, meeting, and routine coordination required by human teams.
The Bottleneck Moves to Agent Orchestration
The difficult parts remain. Some tasks are inherently sequential, so multiple agents cannot compress a month of continuous work into a single day. Someone still has to define what “done” means, review code, detect hallucinations, and combine the agents’ outputs. Without clear orchestration, agents may repeatedly overwrite one another’s changes and create more rework.
Faster Answers Can Still Be Wrong
The key limitation is that AI agents may reduce some coordination costs without removing ambiguity, verification responsibility, or system complexity. The article’s proposed update is: “Adding agents to a project with unclear requirements gets you the wrong answer faster.” That remains a concise rule of thumb; real gains depend on whether work can be parallelized, whether requirements are clear, and whether people can review the results effectively.