EXPERIMENTS IN EMERGENCE / 002

Growing
cells.

A learned local rule grows a picture from one living cell.

Paint damage into the body. Watch the hidden state try to put the pattern back together.

Enter the workspace
01 / THE WORKSPACE

Grow · damage · inspect

LOADING MODEL
Regenerating lizardpublished neural cellular automaton
FRAME0000
Loading the local model…
Preparing WebGL
Drag
Local run recordSeed 137 · no interventions
A completed challenge is required.
02 / WHAT THIS IS

A local rule
with a memory.

Every cell reads a tiny neighborhood, updates a hidden state, and paints the next frame. The rule is shared across the whole canvas.

This lab runs the published Growing Neural Cellular Automata inference model in your browser. The weights are fixed; they are not being retrained here. The colored body is only one view into sixteen channels of state.

PUBLISHED MODELGrowing Neural Cellular Automata

Alexander Mordvintsev · Ettore Randazzo · Eyvind Niklasson · Michael Levin

Read the original Distill article ↗
EXPERIMENTAL LAYERJev chooses; code measures

Jev does not supply the cell weights. It receives a bounded menu and its selection is recorded beside the measured outcome.

This is a visual computation experiment, not a biological model and not evidence of cognition. Damage tests are repeatable browser runs with a fixed seed and explicit interventions.