Dino RL Lab
A tiny runner with a learning loop
Gemini action actor · Q-learning learner
Reinforcement learning playground
Watch a dino learn to run.
Gemini chooses JUMP or WAIT when an obstacle approaches; the game pauses until its answer arrives. Q-learning uses game rewards and Gemini’s post-run critique to improve later decisions.
The runner
Ready to play
Play yourself: tap the game or Jump. On a keyboard, use Space or ↑; press R to restart. Training lets Gemini choose actions and review runs.
Secure critic endpoint
The Worker URL is saved on this device. Your Gemini key stays on the Worker.