an artificial life laboratory
Free exploration. Tune parameters and watch evolution unfold.
10 evolution puzzles across 3 difficulty tiers.
AI-powered hypothesis testing with statistical analysis.
Each dot on your screen has a neural network brain. Nobody programmed their behavior. They perceive the world through vision rays, make decisions through 366 evolved weights, and pass their brains to offspring with small mutations. Natural selection is the only teacher.
21 inputs: what they see in three directions, their own energy and speed, pheromone gradients, ambient light, nearby signals, terrain, and two memory cells they control. 12 hidden neurons with tanh activation. 6 outputs: turn, accelerate, attack, deposit pheromone, broadcast signal, share energy. Inherited through neuron-level crossover from two parents, mutated by Gaussian noise on each weight.
Day and night cycles modulate food availability. Seasons shift resource patches across the landscape. Catastrophic droughts and blooms test population resilience. Organisms leave pheromone trails that decay over time and recycle nutrients into the soil when they die.
Predators and prey specialize through diet evolution. Organisms mark territory with pheromones. Camouflage adapts to local terrain. Disease spreads through proximity; immunity follows recovery. Some lineages evolve cooperative energy sharing. Mate preferences drive speciation. None of this was designed. All of it was evolved.
Describe a hypothesis in plain language. An AI designs the experiment: control group, treatment group, measurement window, replicate count. The simulation runs parallel trials at maximum speed. Results are analyzed with Wilcoxon signed-rank tests, Holm-Bonferroni correction for multiple comparisons, and Cliff's delta effect sizes.
Simulation engine in Rust, compiled to WebAssembly for near-native performance in the browser. Canvas2D rendering with level-of-detail scaling.