Measured connections
The hemibrain v1.2 MB + CX subcircuit: 2,846 neurons and 177,903 connections. A persistent LIF simulation produces the spike readings.
FEFE LABS · LOADING / Opening the observation room…
ONE BRAIN. A NEW BODY.
A fruit-fly connectome. A little mechanical body. Stay a while. Watch it sense, turn, and take a wrong turn.
THE EXPERIMENT LOG
Same arena, same controller. Change the seed. Then disconnect the neural output.
WHAT YOU ARE WATCHING
This is a first sensory–motor coupling experiment. It has not demonstrated learned scent seeking or a complete fly body simulation.
Try the existing odor-learning experimentThe hemibrain v1.2 MB + CX subcircuit: 2,846 neurons and 177,903 connections. A persistent LIF simulation produces the spike readings.
Virtual feelers sample an odor field through PN inputs. MBON activity is cross-coupled to wheel speeds. New positions change the next sample; contact constrains movement and attenuates the next odor input.
Weights are fixed. The controller and 2D contact constraints are engineered. There is no reward training or natural motor pathway. The disconnected control tests coupling, not intelligence.
HHMI Janelia FlyEM · hemibrain v1.2 · CC BY 4.0. Scheffer et al. (2020) ↗ · Data license & selection
Parameters: dt = 0.2 ms; Wsyn = 0.275 mV; probe = 50 ms × 2; world step = 0.25 s; wheel gain = 0.8 / 8 Hz; seed = 42. Wheel speeds are not biological motor-neuron outputs. Probe order alternates each step.
Run: body-01-v1-42-neural
Circuit SHA-256: 93975312e1003fa9b901bfaa1195e412d848cda6f290d9dbd5e1db9cfaf65d29
Recorded: 2026-09-14T18:31:08.842Z
These are independent trials, not the growth history of one individual. New shared runs are stored in the runner archive.