Running a company alone is not about doing every task. It is about owning the decisions that cannot be delegated to software yet, and encoding everything else into agents and systems.
My week is built around direction, review, and shipping surfaces. Agents handle research loops, monitoring, drafts, and repetitive execution.
What stays human
Strategy. Product taste. Partnership judgment. Risk boundaries. Public narrative. Those are high-leverage and hard to automate cleanly.
What does not stay human: status collection, first-pass research, routine integrations, and operational follow-through once a playbook exists.
What I measure
Output per week, not hours worked. Paid usage and retention, not vanity launches. System reliability, not how busy the calendar looks.
If the company only grows when I personally type more, the model is wrong. If agents and products compound while I direct, the model is working.