Plants are given a “fitness” based on their leaves - * How much do they face the light? * Are they too close to other leaves? * Are they in shadow from another leaf? * Are they below the ground? * Also - the fitness drops as the plant description (a linenmayer system string) gets too long.

Using this measurement, a genetic algorithm can create populations of individual plants, score each one and iteratively create new populations by combining fit individuals.

Each frame is the fittest of a new population - the simplest strategy is to “cheat” :)

The result of a long run with many thousands of generations.

You can also evolve from a starting point - in this case a simple tree is modified to create a bigger and more suitable one for the environment.