Manifesto.

Models got smart. Every team can call a frontier model now, and it will reason about almost anything you put in front of it. Intelligence used to be the scarce thing. It isn't anymore.

But intelligence isn't experience. And nobody has built the place where agents earn it.

An agent on day one is a brilliant new hire with zero shifts. It has never watched a customer go quiet for a week, never seen a payment bounce, never been handed a policy that changed mid-quarter. So it learns those things the only place it can — in production, on your customers.

The tools we have don't close the gap. Evals grade an answer, not a month. Staging has no customers and no clock. Shadow mode is silent right up until it isn't. A careful rollout is still learning in production, just with a smaller blast radius.

Two kinds of teams feel this most. The ones putting agents on real work — finance, support, operations — where a mistake has a dollar sign on it. And the ones whose products are now being used by agents, with no way to see what those agents see.

Two shifts arrived at once. Agents started doing real, multi-week work instead of answering questions. And simulation got good enough to stand in for the world — the tools, the people, the passing of time. One created the need. The other made it possible.

So we're building it. Living simulations where an agent works through months of operations, edge cases, and failures before it meets a customer. We believe every agent should earn its experience before it's given responsibility. We exist so your customers always get your best.