🧬 Paradigm Arena

Six machine-learning paradigms compete in one arena. They all see the same input (nearest 2 food, nearest threat, nearest prey, own size) and use the same 8-direction action space — the only difference is how each policy was learned. Eat food to grow, swallow smaller agents, and the first to reach level 10 wins. The policies are trained in Python and run in your browser with a tiny neural network.