macOS desktop fly powered by real FlyWire connectome
Original: A 3D fruit fly on macOS desktop powered by the real FlyWire connectome
Why This Matters
Demonstrates real connectome-driven behavioral simulation accessible as consumer desktop software, bridging neuroscience and interactive computing.
Developer DenisSergeevitch released 'desktop-fly,' an open-source macOS app that renders a 3D fruit fly on the desktop driven by a live spiking neural simulation using 23,210 real neuron positions and a 668-neuron circuit from the FlyWire connectome.
GitHub user DenisSergeevitch has published 'desktop-fly,' a macOS application that places an animated 3D fruit fly on the user's desktop. The fly's behavior — walking, grooming, sleeping, and fleeing the cursor — is governed by a live leaky-integrate-and-fire (LIF) simulation running at 1 kHz, built on the real FlyWire v783 connectome dataset. The simulation uses 23,210 actual neuron soma positions out of 139,255 mapped in FlyWire, displayed in a rotating brain window with live spike flashes colored by super-class. A 668-neuron functional circuit with approximately 19,000 real synaptic connections drives all behaviors. Key neurons included are LC4 (104) and LPLC2 (210) looming-detector visual neurons, Giant Fiber escape command neurons (DNp01/GF), steering neurons (DNa01, DNa02), forward walking neuron DNp09, grooming neuron DNg11, and backward walking neuron MDN. Users can click brain regions in the visualization window to directly stimulate circuits. The project is open-source and includes a build script, Swift source files, and a Python ETL script for data preprocessing.