Humanoid hand wraps wontons as Nature Will releases Natura-1, a foundation model trained on 1M hours of human data
CyberRobooo · x · 2026-10-13
A demo shows a humanoid robot's dexterous hands wrapping wontons — harder than dumplings in some ways, since the thinner, softer wrappers tear and leak if force is off, testing model-hardware coordination.
The quoted post covers Natura-1, a general foundation model for the physical world just released by Nature Will, trained on 1 million hours of egocentric human data. The hard part of dexterous manipulation is maintaining control through whole interactions — objects move, contact points shift, and two hands swap roles — not just grasping. Combining dexterous control, zero-shot generalization, and long-horizon execution in one foundation model could meaningfully expand general-purpose humanoid capabilities.
More from Embodied
- WUJI's dexterous robot hand packs 20 active DoF with high-frequency control — Olivier__OG · 2026-10-13
- Natura-1 runs multi-minute autonomous dexterous hand workflows with a single policy — stingning · 2026-10-13
- Natura-1 demo: a single-take 30-minute recording of everyday human-robot interaction — stingning · 2026-10-13
- Collov Labs launches NewEyes AI: camera glasses plus open-ear earphones give agents eyes — CodeByPoonam · 2026-10-13
- Surface Ultra unboxed: NVIDIA RTX Spark N1X chip, 20 cores, 64GB unified memory — TinfoilTricorn · 2026-10-13
- Replicating laundry real2sim in MuJoCo and training RL WBC in 3 hours with Opus 5.5 — ZeYanjie · 2026-10-13