Irvin (in Japan 🇯🇵)(@irvinxyz):Impressive. It’s 1X speed. Not a VLA model btw. Training described as “learning robot actions directly from human motion, no teleop and no on-robot data”. So closer to a scaled, embodiment-agnostic imitation policy (visuo-tactile-proprioceptive) than to OpenVLA / π0 VLAs.

Irvin (in Japan 🇯🇵)

@irvinxyz

Robotics researcher & builder working on physical AI, robot identity, and cultural/economic infra for autonomous robots. Prev NASA, Goldman Sachs, Mysten Labs

参加 March 2012

3.9K フォロー中 17.3K ファン

Irvin (in Japan 🇯🇵)@irvinxyz

2026.09.14 19:19

Impressive. It’s 1X speed. Not a VLA model btw. Training described as “learning robot actions directly from human motion, no teleop and no on-robot data”. So closer to a scaled, embodiment-agnostic imitation policy (visuo-tactile-proprioceptive) than to OpenVLA / π0 VLAs.

Reward AI@RewardAI_

2026.09.14 17:40

Introducing OM-1, our first robot foundation model, zero-shot generalizing to any robot: table-top arms, industrial arms and humanoids. - learned directly from human manipulation data - no teleop/robot data - close to human-level dexterity and efficiency - multi-robot collab