FlyHamlet

The live session computes a fresh run of the full FlyWire v783 connectome: 139,255 neurons and 15,091,983 connected pairs. Neurons use the repository’s leaky integrate-and-fire model at 0.1 ms steps. Letters are assigned to the 90 × 30 mm arena’s 3 × 9 grid by a seeded permutation. Saved 300-second recordings remain available in the session menu.

The brain model supplies descending-neuron motor commands. Giant-fiber escape activity can start the downstream wing motor; an asynchronous 175–225 Hz wing oscillator then generates repeated strokes, and stroke velocity produces the lift and thrust that change x, y and z. DNa02 + DNa01 bias left/right wing force for turning. On the keyboard, DNp09/MDN locomotor drive sets an alternating modified-tripod leg oscillator; stance legs generate ground traction while swing legs reposition. The downstream plant represents VNC, muscle and body mechanics missing from the brain connectome and never selects a key or destination.

Randomness comes from injected Poisson sensory inputs. The connectome turns that activity into motor commands; deterministic wing and leg mechanics turn those commands into movement. A key is typed only on keyboard contact. The renderer uses the simulated wing phase, gait phase, altitude and attitude directly, including exposure-style wing blur at flight frequencies too fast for a display to resolve frame by frame.