People ask me what my AI agent team does. The reality is that it is incredibly horizontal, incredibly open and flexible, and fully driven by…

People ask me what my AI agent team does. The reality is that it is incredibly horizontal, incredibly open and flexible, and fully driven by a flywheel of verbalized, documented, and reviewed goals-context-decisions. See example prompt below. (And though a bit needy, it worked.) In the span of 10m49s, Simon and the crew reviewed my calendar, slack, gmail, and recent fireflies recordings. They completed 11 tasks in parallel (emails, documents, workshop agenda), set up 5 auditors (that don't need me) to track projects in progress, and suggested 2 meetings to move off my calendar. ...and people are still using ChatGPT like Google. My current agent infra: I have one CoS named Simon, an assistant for Simon to document and memorialize, 6 named directors/avengers, 20+ unnamed but pre-specified subagents/sidekicks managed by the directors, and unlimited civilians spawned by Simon or the directors/avengers).

Source: Allie K. Miller (X) | 2026-05-05