In 1975, Fred Brooks published The Mythical Man-Month and gave us Brooks’s Law: adding people to a late software project makes it later. The reasoning was simple. Every person you add to a coupled effort brings communication overhead that grows faster than their contribution. Two people have one communication path. Five have ten. Ten have forty-five. The math is quadratic and unforgiving. But for decades, the tradeoff made sense — coordination cost went up, but so did throughput, and throughput was the thing you couldn’t fake. You needed humans to write the code, review it, test it, ship it. More people meant more output, even after the overhead tax.
That math just broke. Agentic engineering — the practice of orchestrating AI coding agents like Claude Code, Codex, OpenCode, and others to do the bulk of implementation work — has collapsed the throughput gap between small and large teams. One builder directing agents can produce many multiples of what a single engineer used to. Each additional engineer might individually produce multiples of what they used to — but the coordination drag between them scales with speed the way wind resistance scales with velocity. The faster each person can go, the more punishing the overhead of staying in sync becomes. Brooks’s quadratic coordination cost is the same as it ever was, but at agentic speeds, it hits like a wall.
The result is an inversion, and I think it’s worth naming:
The Agentic Inversion Principle
In agentic engineering, individual throughput is no longer the bottleneck — coordination is. In coupled domains, smaller aligned teams will outperform larger ones working the same problem space.
This isn’t necessarily an argument against building large engineering organizations. It’s an argument for how you structure them. You scale by parallelizing small, autonomous teams across clean domain boundaries — not by adding headcount within them. Two engineers own a service. Two or three more own another. Clear interfaces, minimal overlap, each pair moving at the speed agentic tooling enables without dragging each other down. The instinct to staff up when something is important? That’s the old math. In the agentic era, importance means you put your best pair on it, deeply understand the what and the why and get out of their way.