What does it mean that the cost of agent sprawl?

Too many agents create coordination overhead that can exceed their collective value.

Example: You built a morning planning agent, then an evening review agent, then a weekly reflection agent, then a reading curation agent, then a health tracking agent, then a relationship maintenance agent, then a financial review agent. Each one seemed worthwhile when you created it. But now you spend forty minutes a day just running the agents — checking outputs, feeding inputs between them, resolving conflicts when the health agent says to sleep early and the reading agent has queued three articles for tonight. The agents are not failing individually. The system they form is failing collectively, because the overhead of coordinating seven agents exceeds the value any three of them produce.

Try this: List every active agent in your current cognitive infrastructure — every habit, routine, system, delegation, or automated process you maintain. For each one, estimate two numbers: (1) the value it produces per week in minutes saved, decisions improved, or outcomes achieved, and (2) the coordination cost per week — time spent maintaining it, feeding it inputs, checking its outputs, resolving conflicts with other agents. Now sum the coordination costs. If the total exceeds 30 percent of the total value produced, you have agent sprawl. Identify the two or three agents with the worst value-to-coordination-cost ratio and ask: would retiring these improve the system even though each one, in isolation, still works?