100,000 hours of egocentric skill data open-sourced on Hugging Face, free for all
lukas_m_ziegler · x · 2026-11-17
Hub, together with Hugging Face's LeRobot team, open-sourced HubEgo-100K — billed as the most diverse egocentric dataset ever released: 100,000 hours of skilled real-world tasks from 100,000+ contributors, full resolution with annotations, free for research and commercial use with no application required.
Author lukasmziegler notes commercial workplace data has mostly sat with whoever could pay for it; this first batch of a planned 100,000 hours means "a student with a laptop now starts from the same place," potentially changing who gets to build embodied AI.
More from Embodied
- Embodied AI field debate: why VLA work is a 'influencer field' and pretraining is the steepest slope — ChongZzZhang · 2026-11-17
- Barbell-tracking app built on custom physics engine: LLMs compute forces in 3 steps — wordgrammer · 2026-11-17
- Skild AI and Nvidia's Steel solver hits 13x PhysX throughput for robot learning — lukas_m_ziegler · 2026-11-17
- Runway open-sources Praxis-1, a 27B MoE robot policy topping RoboLab-120 — lukas_m_ziegler · 2026-11-17
- PathRobotics raises $300M at $2.45B as ARR grows 17x in 18 months — lukas_m_ziegler · 2026-11-17
- NVIDIA unveils GR00T 2 robot model, open weights coming in December — lukas_m_ziegler · 2026-11-17