training environments to raise general intelligence.training environmentsto raise general intelligence.

We're an applied research lab in San Francisco. We build training environments and evaluations for frontier AI labs.

Our environments are real workspaces, mid-flight: a persona with history, work already in progress, the next move not obvious. The model walks in and tries to do the job. The people who actually do that job design the verifiers that score it.

Most environments today are one tool, one reward function. That's enough for math and code. It isn't enough for work that takes hours, spans apps, and changes shape as you go.

Every environment we ship is a contract: train on it, and the model gets meaningfully better at work someone is paid to do. The frontier moves where the work is.