Skild AI founder explains why robotics data needs four sources, each flawed
deepakpathak · x · 2026-09-12
Skild AI founder Deepak Pathak laid out a framework for how a $100M ARR robotics company thinks about data scale, arguing no single source can solve the robotics data problem.
- Robots collecting their own data: best-targeted, slowest to scale since the physical world can't run faster than real time
- Teleoperation: industry default, but almost zero diversity
- Simulation: far faster than real time, but every new task needs a hand-built environment
- Human video: most abundant, furthest from the robot
He evaluates data quality along three axes — scalability, diversity, and closeness to the robot — and argues only a combination of sources will solve robotics data.
More from Embodied
- GPT-6 Astra Rebuilds Office Scan in Blender, Drops Humanoid Gym Into Newton Sim — liuziwei7 · 2026-09-12
- Building a physical Codex Pet for OpenAI Dev Day — gabrielchua · 2026-09-12
- Edinburgh Spinout Singular Photonics Launches Litavis, a SPAD Sensor That Processes Photons In-Pixel — Scobleizer · 2026-09-12
- Open-source robot arms now sell for $999 with 3.3 lb payload and 0.2mm accuracy — Scobleizer · 2026-09-12
- Tesla Europe recaps a year of FSD Supervised testing across Swedish streets and Nordic weather — yunta_tsai · 2026-09-12
- FADA (CoRL 2026) open-sourced: humanoid robots adapt to new conditions from 2 minutes of experience — GuanyaShi · 2026-09-12