Demos
Trained agents you can run and play against in the browser.
-
▶ Mapox
Multi-agent, partially-observable gridworlds with a trained policy running live in the browser via WebAssembly and WebGPU.
-
▶ Self-play tic-tac-toe
Play against an actor-critic agent that learned to match minimax with under a minute of self play on a single GPU.
