Design a spaceship's sensors and train it by playing: your moves become training examples for a neural network that learns to dodge bombs and collect stars.

How does a machine learn to play with nobody teaching it the strategy? Boxes of beads that reward and punish moves discover perfect play on their own — from the stone game to Hexapawn to a teamwork game with no single solution.

Why does cooperation emerge among selfish players? From the one-shot prisoner's dilemma to a tournament of strategies, to the evolution that rewards the best scorers. Rebuild Axelrod's experiment and discover why winning every duel is not winning the tournament.

Build, train and watch artificial neurons: weights, sigmoid, gradient descent, backprop. No manual calculation — everything is automatic, you design.