Dyna-2 trained on 1M+ hours of egocentric video; data infra detailed, open-sourced
Scobleizer · x · 2026-08-19
Robotics firm Dyna published how they trained Dyna-2 repeatably on over 1,000,000 hours of egocentric video, noting that what worked at 10k hours broke at scale: ingestion capped at 14,000 episode-hours/week (a million hours would take over a year), training manifests took 48 hours to build, and reading a petabyte from cloud storage exposed GPUs to latency and packet loss.
A YC team resharing the thread called it a gold mine and open-sourced HFlow, a reference implementation of the described infrastructure: Airflow for DAG orchestration, MCAP for episode storage with random access, and DuckDB for analytical queries and dataset curation over columnar metadata. Their takeaway after two months at YC: at scale, AI problems become software engineering problems—fix one bottleneck and it moves elsewhere.
More from Embodied
- Unitree's 'Superman' humanoid robot hits 45.6 km/h, outpacing top athletes — FinanceYF5 · 2026-08-19
- Netizen Jokes SpaceX Booster Leg Collapse, Firebots — teortaxesTex · 2026-08-19
- Physical AI Creates New Jobs: Robot Fleet Operators, Sim Designers — Scobleizer · 2026-08-19
- World Robot Conference showcases various bionic robots — jiqizhixin · 2026-08-19
- Engineer Open Sources Desktop Robot Using On-Device Model to Capture Emotions in Overlap Zones — 面壁智能 · 2026-08-19
- AI Security Tool Finds Hardware Encryption Flaw: Can Decrypt All Devices — CtrlAltDwayne · 2026-08-19