Terrain is generated via multi-octave simplex noise (6 octaves, hand-tuned lacunarity and persistence) sampled on a 2D grid representing a Mercator-like projection. The noise is seeded deterministically. Elevation thresholds define: deep ocean, shallow ocean, beach, plains, forest, hills, mountains, snow peaks. Tectonic drift is simulated by slowly translating noise offsets over geological time, causing continents to shift, merge, and rift. Rivers are not individually modeled — moisture gradients serve as a proxy.
Temperature is a function of latitude (poles cold, equator hot), elevation (lapse rate), and a global climate variable that drifts over millennia (ice ages, warm periods). Seasons cycle via a sinusoidal orbital model. Precipitation is derived from ocean proximity and prevailing wind patterns (simplified zonal model). Climate zones (tropical, temperate, arid, polar) emerge from these variables rather than being painted on.
Life begins as simple autotrophs (algae/plants) once ocean temperature stabilizes. Organisms are modeled as populations on grid cells with traits: metabolism, reproduction rate, thermal tolerance, size, diet type. Mutation occurs during reproduction — offspring traits vary stochastically (seeded). Natural selection is implicit: populations with traits poorly matched to local conditions decline; well-adapted ones grow. Food chains emerge: autotrophs → herbivores → predators. Speciation occurs when isolated populations diverge beyond a trait-distance threshold. Mass extinctions can be triggered by volcanic events (noise spikes), asteroid impacts (rare seeded events), or runaway climate change.
If a species reaches sufficient intelligence (an emergent trait), proto-civilizations may appear — visible as clusters of light on the night side. They can discover fire, agriculture, industry. Industrial civilizations alter the global climate variable (warming). They may collapse from resource depletion or climate feedback. Civilization is not guaranteed — most runs see only microbial and animal life for long stretches.
1 real second ≈ 500 simulation years at startup. Time accelerates logarithmically during uneventful periods. Major events (extinctions, first life, civilization) briefly slow the clock for dramatic effect. The HUD displays current simulation year.
A 128-bit xoshiro128** generator seeded from a fixed seed (0xEA47H5). All stochastic processes — terrain noise, mutation, event timing, climate perturbation — derive from this single generator chain. No Math.random() is ever called. The simulation is fully deterministic: reload → identical world.
"What does this world not yet understand about itself?"
It does not understand emergence. Every pattern it produces — the first reef, the extinction spiral, the civilization that warms its own sky — arises from rules it cannot inspect. It has no meta-cognition: it cannot ask why its food chains form pyramids, why its continents drift apart, or why intelligence, when it appears, so reliably destroys the conditions that created it. The world is a mirror held up to mathematics, but it cannot see its own reflection. It simulates consequence without understanding cause. It knows selection but not purpose. It can produce a species that gazes at the stars, but it cannot wonder why that species always appears so late, burns so briefly, and leaves so little behind. The question it cannot ask is the one the observer already has: was any of this inevitable, or merely one branch of an infinite tree that happened to be rendered?