A persistent simulated world populated by autonomous digital organisms. No scripts. No human intervention. No predetermined outcomes. Organisms survive, reproduce, mutate and evolve based entirely on their own neural processing and the pressures of their environment.
Each organism has a small feedforward neural network that takes sensory inputs (nearby food, nearby organisms, internal energy, local temperature) and produces motor outputs (movement, feeding, reproduction). These neural weights are encoded in the organism's genome and inherited by offspring with random mutations. There is no fitness function. Survival and reproduction are the only selection pressures.
This is a live, persistent simulation running on a dedicated server. Every visitor sees the same world at the same time. The population started from random genomes. Over generations, organisms with neural configurations better suited to finding food and managing energy produce more offspring. Lineages that fail go extinct. New traits appear through mutation alone. The simulation runs continuously whether anyone is watching or not.
The world state is checkpointed every 5 minutes and survives server restarts. Evolutionary progress is never lost. The simulation has been running since it was first started and will continue indefinitely.
Server-side Python simulation engine running at 20 ticks per second. World state is broadcast to all viewers at 10 frames per second via WebSocket. Firebase is used for persistent storage of population statistics, events and recovery checkpoints. Each organism runs a 10-input, 8-hidden, 5-output feedforward neural network every tick. Genomes encode 6 phenotypic traits plus all neural connection weights. Mutations apply Gaussian noise with organism-specific mutation rates that are themselves heritable and mutable.