Cortex

one brain, one lifetime

not evolution. not keyframes. real-time synaptic plasticity.

Speed
Phase
exploring
Steps
0
Avg Reward
0
Falls
0

A single creature with a recurrent neural network learns to walk through reward-modulated Hebbian plasticity. No generations, no evolution — one brain adapts in real-time. Synapses that fire together when movement succeeds get strengthened. Watch the connections form in the network visualization.

The learning arc: exploring (random noise, wild flailing) → adapting (synapses forming, noise fading) → coordinating (learning to stand, falling less) → walking (sustained upright motion) → fluent (the body knows). Most runs reach coordinating — standing without falling. Walking is rare. That is the point.

Green connections are excitatory. Red connections are inhibitory. Watch how the network learns to suppress wild motion — the first thing a brain discovers is not how to walk, but how to not fall. Stability before locomotion. Balance before grace.

Reset Body keeps the brain but puts the creature back at start — does it stand longer this time? New Brain wipes everything. Crank Speed to watch years of learning in seconds.

Genesis evolves a population across generations. Cortex grows a single mind in real-time. One is natural selection. The other is synaptic plasticity. The creature that learns to stand has already solved half of walking.