Seedling

Magnolia × soulangeana

DAY 000 germinating
how it grows ↓

How it grows

The crown exists before the tree does

Every seed on the shelf carries the shape of its grown crown as a cloud of light. Each day, every point of light pulls on the nearest branch tip, and that tip grows one step toward the average pull. When a branch comes close enough, the light is spent. Nothing tells a branch where to go. The shape is decided by where the light is.

0

Thickness

A branch is as thick as what it feeds

There is no rule for a trunk. Each branch is drawn as thick as the number of twigs above it, so the trunk is simply the branch that feeds everything. Botanists call this the pipe model: the cross-section below a fork equals the sum of the cross-sections above it. Move over the tree to see what each branch carries.

Hover a branch

Chaos to order

Random points, random steps, always a tree

The light is scattered at random. Every growth step has a small random wobble. And yet the result always reads as a tree, and as the right kind of tree. The order is not in the parts. It is in the rule. Replant, and the particulars change while the species stays the same.

The shelf

Five seeds, five habits

Neural blossoms

A flower that grows itself

Every blossom on the shelf can also be grown by a neural cellular automaton: a grid of cells where each cell runs the same tiny neural network, sees only its eight neighbours, and decides how to change. Nobody draws the flower. The network is trained by gradient descent until a single seed cell unfolds into the species' bloom, and because no cell ever sees the whole picture, the flower also heals when you tear a piece out. Train it here and watch it learn, or start from weights we trained, then hang the grown flower on the tree.

Target, 32 × 32
step 0 · click the flower to wound it
The twelve hidden channels
Training loss

loading