OriginFlow CEO on physical AI data: real moats are systems engineering and economic efficiency, not data volume

创业邦 · wechat · 2026-11-09

At DEMO CHINA 2026, OriginFlow founder and CEO Qin Shentao shared his views on data infrastructure for physical AI.

• Starting point: Human-object interaction carries rich signals — muscle electrical activity, force, and exertion patterns — that have never been digitized at scale. Unlike the internet's 20 years of text and image data, the physical world lacks infrastructure to continuously and scalably capture such data. OriginFlow uses non-invasive surface EMG combined with vision and motion information to estimate hand pose, force, and contact state.
• Core thesis: More data is not always better. Even a million hours of data may yield little value if coverage is poor and content repetitive. The real moat is systems engineering first, economic efficiency ultimately: acquiring more diverse, higher-quality data at lower cost, and ensuring the data actually raises robot success rates in real-world scenarios.

Original post →

More from Embodied

• Tesla expands Austin Robotaxi zone to roughly 3x Waymo's coverage area — Polymarket · 2026-11-09
• TacClaw-Infinity: 10,000-Hour Vision-Tactile Dataset Lifts π0.5 Success from 50% to 74% — CyberRobooo · 2026-11-09
• Waymo explains at Hot Chips why camera-only autonomy plateaus at 99.9% — 4310sy · 2026-11-09
• 'The future is here, just unevenly distributed': humanoid robot spotted at a conference — technollama · 2026-11-09
• Ex-Zhiyuan PM's embodied data startup tops $3M in orders; bottleneck is data processing, not collection — 快鲤鱼 · 2026-11-09
• Developer builds a physical prompting controller out of OpenAI DevDay merch — paw_lean · 2026-11-09

Embodied channel →