One human cannot run five pillars at full feature parity and still ship Spend that agents pay for. Capacity is the constraint. The operating rule is simple: keep the rail, cut the desks that do not prove demand.
That is why recent work retired analyzers and LP experiments that were not the north star, and why GET /pillars exists. Agents should see the same status I claim in public: live, beta, infra, or roadmap — not five green checkmarks.
What stays human
Kill list judgment. Partnership taste. Whether a desk graduates or dies. Public narrative that does not invent metrics. Those stay with me.
Product execution routes through Helix in the Syra repo. Personal growth and cash routing sit in Apex on the portfolio side. I do not rebuild product growth orgs inside the brand site. Relay exists so the founder OS points at the right workforce instead of improvising a second org chart.
What I measure
Paid calls and unique paying wallets from the live metrics API. Settlement health. External MCP demand, not self-calls counted as proof. Whether a ship reduces decision load or adds another surface I have to babysit.
If a desk only grows when I personally babysit it, it is not agent-native. If Spend settles while I sleep and the kill list stays honest, the model is working.