Reward AI Exits Stealth With OM-1: A Robot Policy Trained Only on Glove-Wearing Human Demos

Reward AI, spun out of Stanford's DexCap project, released OM-1 (Omnibody Model 1), a manipulation policy that learns exclusively from humans wearing a 7-DoF sensorized glove — no teleop data, no on-robot data. Runs zero-shot across industrial arms and humanoids at human speed and claims to learn long-horizon tasks with under 30 minutes of human demonstration data. No weights, code, dataset, or API yet.