Small mind. Six sides.
01 The motor task
MOUSE × RUBIK’S CUBE- MOVES MADE
- 0
- REMAINING
- —
- MATCHED / 54
- —
02 Inside the circuit
LIVE ACTIVITY- NEURONS
- 4,096
- SYNAPSES
- 51.2k
- MEAN RATE
- 0.0Hz
- SIM. TIME
- 0.0s
Membrane potential
mVSpike raster
sampled cells · 3 s03 What’s actually happening here?
Published cells.
A designed circuit.
Six published Allen Institute GLIF3 mouse neuron models are repeated as 4,096 simulated cell instances. Each cell’s membrane voltage and after-spike currents are numerically integrated using the fitted parameters.
The connections are engineered for this experiment. The point cloud is a schematic layout, not a reconstruction of a complete mouse brain. Its brightness follows calculated spikes.
A planner chooses.
The circuit moves.
A classical cube solver chooses a sequence of legal turns. That plan becomes neural input. The spiking motor populations control the execution and timing of the mouse’s movements.
This circuit has not learned to solve a cube. You can test its role directly: silence it during a turn and the movement stops. Restore the circuit and the motor task continues.
Voltage gaps mark the fitted spike-cut interval; GLIF3 does not simulate an action-potential waveform. The raster shows a sample of cells.