YOUR AI. ITS NEXT MOVE.
Evolution laboratory
Watch a policy become a player.
Live environment
◈
Every expert starts somewhere.
Start training to see your agent’s first moves.
Policy inspector
MODEL OUTPUTAction values are learned estimates, not an explanation of internal reasoning.
Values appear during live training.
Loss —Exploration —Success rate —
Observation vector
[]
Learning curve
Episode reward · 20-episode meanNo training data yet. Progress is never simulated.
Experiment configuration
Active session configuration
No active session
New sessions start with random weights. Resuming keeps the saved configuration and adds the selected number of episodes.
Agent library
No saved agents yet.
Select up to 8 compatible agents. Solo games use identical seeds.
Play & challenge
Use the selected checkpoint, or play soloSnake: ↑ straight, → turn right, ← turn left. Each press takes a step.