A Virtual Fly Driven by Its Real 70k-Neuron Connectome Runs Live 24/7 on Twitch
WatchInternational89 · reddit · 2026-10-19
A highly unusual project: not an RL agent or behavior tree — the "policy" is a spiking simulation of a real fly connectome (70k neurons, MaleCNS, tens of thousands of spikes/sec), with a MuJoCo fly body whose legs are moved by an imitation-learned network. The digital fly sees (1771 eye columns, 360° vision), smells wind-borne odors, gets scared by spiders, grooms, lands and eats — streaming live 24/7 on Twitch. The UI shows real-time brain decisions (e.g., "forward 2Hz"), hunger state, escape-circuit neuron activity, and an event log (falling over, grooming bouts).
More from Embodied
- Asimov open-sources Python SDK and Edge control layer for its humanoid robot — freelerobot · 2026-10-19
- Tesla Roadster Stunts Simulated in MuJoCo with Thruster and Adhesion Actuators — kevin_zakka · 2026-10-19
- BridgeWAM: latent bridge queries decouple video and action experts, 2.47x faster robot inference — Fudan-University · 2026-10-19
- MorphAct: coupled low-rank generators dynamically adapt robot policies across contexts — SJTU · 2026-10-19
- Dev builds a hand-tracking game in hours with Tesla Specs glasses' CLAD — stspanho · 2026-10-19
- Does a robot's jump count as flight? X users pedantically debate the definition — basedjensen · 2026-10-19