THE ARCADE OBSERVATORY / PLAY & DISCOVER
Snake AI.
A small world. A learned instinct. Watch instinct become strategy.
NEURAL POLICY / SNAKE02 / ARCADE
The trained agentPPO POLICY
AI FOOD COLLECTED0
BOARD FILLED6%2 of 36 cells
SESSION BEST0
Training progress 6 × 6 policy
This policy trained for roughly 76 hours. Four additional danger inputs tell it whether each direction leads to a collision.
Change the world size to see the learning curve for that model.
