The mushroom body is the fly's learning and memory center: in the real fly it learns which odors and cues predict reward or danger, a pattern → value judgment.
Every move is computed by a spiking network wired from the real MaleCNS mushroom-body connectome — 4,510 measured neurons, 1,027,152 neuron-to-neuron connections, excitatory/inhibitory signs from neurotransmitter annotations. On the right, those neurons are drawn at their real measured soma positions and light up as the circuit runs. The network's state is never reset: it carries forward from move to move, so what you see at move 40 descends from move 1. The whole network runs in your browser (a Web Worker) — no server computes anything.
In a real Drosophila this circuit learns odour→reward associations. Here the identical wiring is fed Othello: 138 board features (own discs, opponent discs, side to move, a disc-count bucket) plus the move history as tokens, projected onto the Kenyon cells. The circuit takes one timestep per move, forced passes included; the readout of that timestep is the move. Because the state is carried and never reset, the lateral signal keeps propagating across the whole game.
Othello rules, the board, the fly cartoon and the brain atlas renderer are original code for this project. The spiking network runs in this page as plain JavaScript in a Web Worker and is checked numerically against the reference implementation.