Every time we switch between agents our brain’s executive functions incur a switching cost. The prefrontal cortex spends its energy reorienting, rebuilding enough context to monitor an agent’s progress or retrain it on new inputs. Part of working memory stays with the task we just left, so we arrive at the next one only partly present. Cycling between five or six assistants makes that residue the default state, and it leaves room only for shallow reactive work: fixing formatting, tweaking prompts, reviewing outputs. Long-range strategy and complex synthesis need what constant switching removes: uninterrupted attention on a single problem.