Skild AI trains humanoid to play soccer via 140 simulated years of self-play
deepakpathak · x · 2026-09-24
Robotics startup Skild AI, inspired by the World Cup, used self-play to train a humanoid robot — the "Messinator" — to play soccer:
- 140+ simulated years of play, with the robot facing increasingly capable versions of itself starting from S1
- The only objective was scoring goals: no new demonstrations, no hand-crafted rewards for dribbling, shielding, tackling, shooting or fall recovery
- Those behaviors emerged because they helped it win, and transferred to a real robot
The team frames self-play as a scalable way for robots to create their own curriculum and improve beyond human-provided data.
More from Embodied
- Guava distills GPT-5.4 into a 4B robot agent with just 2,000 simulated trajectories — jbhuang0604 · 2026-11-09
- Dex-One2Many learns dexterous manipulation from one human video, 95%+ success across 4 hands — jbhuang0604 · 2026-11-09
- Fly-by-Code: coding agents pilot drones via active visual and physical feedback loops — jbhuang0604 · 2026-11-09
- μ0 world model learns robot physics from video only, beats π0 by 5 points on RoboCasa365 — jbhuang0604 · 2026-11-09
- NVIDIA unveils GR00T 2 open-world robot action model, hitting 95% success with 5 demos — CyberRobooo · 2026-11-09
- CoRL keynote to explore open-source data and hardware for robotics — micoolcho · 2026-11-09