X / @RewardAI_: OM-1 Cross-Embodiment Robotics Policy Launch
帖子要点 (中文)
@RewardAI_ 在 2026-09-14 17:40 UTC 发官帖 2099553899804053992（约 1226 likes / 30 万 views / 185 reposts，证据 official_release），配套 OM-1 博客上线，给出跨桌面臂/工业臂/人形 zero-shot 迁移的机器人基础策略。cofounder @zipengfu 也在同窗发了一条 2099554803932336193（约 237 likes / 21k views，证据 personal_experience）作为个人延伸。训练数据侧划出明确边界：穿戴式 Omnibody Hand 采人类操作，不写 teleop、不写 on-robot 预训练——这是与 VLA 路线最大的数据层区别。
Key claims (English)
The @RewardAI_ launch post 2099553899804053992 (2026-09-14 17:40 UTC; ≈1226 likes / 300k views / 185 reposts; evidence class official_release) accompanies the OM-1 blog and introduces a cross-embodiment robotics foundation policy that zero-shot transfers across desktop arms, industrial arms, and humanoids. Cofounder @zipengfu posted 2099554803932336193 (≈237 likes / 21k views; evidence class personal_experience) as a personal extension in the same window. The training-data boundary is sharp: wearable Omnibody Hand captures human manipulation with no teleop and no on-robot pretraining — the biggest data-layer contrast to the VLA path.
Obsidian 证据摘录
「官帖：@RewardAI_ 2099553899804053992，约 1226 likes / 300141 views / 185 reposts（UTC 09-14 17:40）。证据 official_release。」——Hermes 定时任务/X-每日简报/2026-09-14-X-Hot-Brief_AtriaDawn-OM1-GlassImaging.md L51