Larger bodies. Richer senses. Sexual reproduction. Scarcer resources. 300,000 ticks of evolution.
Scroll to explore
Part 1 ended with every organism wearing the same body. Three nodes. A core and two mouths. No sensors, no muscles, no bones. Evolution had done exactly what it was supposed to — optimize for survival — and survival meant stripping everything away. The problem wasn't evolution. The problem was the world.
Part 2 rebuilds the world with seven changes:
Bodies you can see. Spread goes from 1.5 to 8 units. Organisms are structures now, not dots. Nodes grow outward from the core, so bodies actually look different from each other.
Eating isn't fighting. Three neural outputs instead of two: eat, attack, reproduce. A creature can graze without biting its neighbor. Herbivores and predators evolve as separate strategies.
Sensors that matter. Six inputs per sensor instead of three. Food angle, food type, organism size. Awareness scales with sensor count — more sensors, wider field of view, longer range. Finally worth the metabolic cost.
Half the food. Plant spawn rate drops from 2.5 to 1.2 per tick. Meat decays slower, making carnivory a real option. When food is scarce, every strategy gets tested.
Sex. Two genetically similar organisms within 50 units can mate. Genome crossover. Traits from both parents. Fresh mutations on top. Evolution stops being a solo act.
Triple lifespan. 15,000 ticks instead of 5,000. Complex strategies need time to pay off. Most organisms still die young — from competition, not old age.
Bigger is better. Force scales with node count. Bones reduce drag. Fat lowers reproduction thresholds. For the first time, there's a mechanical reason to grow.
Seven changes. Same algorithm. Radically different life.
Fifty organisms are dropped into a world with half the food of Part 1. The scarcity hits immediately. Within 1,000 ticks, the population crashes from 50 to just 20. A 60% die-off in the first moments. Only 3 species survive the initial bottleneck. In Part 1, the founding population had enough resources to decline gently. Here, it's a massacre.
But the survivors are different from Part 1's founders. Their bodies are physically larger, 8 units of spread instead of 1.5. They're visible structures in the world, not invisible dots. And the ones that survive have neural networks that, by chance, produce food-seeking behavior strong enough to overcome the scarcity. These three lineages will father hundreds of species.
Genesis — A Harsher World
The founding population faces half the food supply of Part 1. Watch the population crash to 20 within the first thousand ticks. Only 3 species survive.
Recovery is explosive. By tick 2,000, the population rebounds to 73 with 8 species. By tick 3,000, it reaches 145 with 13 species. The three surviving founder lineages radiate rapidly, their offspring testing mutations in a world with empty niches and just enough food to reward the bold.
Sexual reproduction is now active. When two genetically similar organisms come within 50 units of each other, both are reproductively ready, and both signal willingness, their genomes cross over. The offspring inherits traits from both parents plus fresh mutations. This genome recombination accelerates the pace of innovation — instead of tweaking one parent's strategy, evolution can combine the best features of two.
Then the first crash. Tick 4,000: population drops to 67. Tick 5,000: 63. The classic boom-bust, but more violent than anything Part 1 produced. The ecosystem is learning to oscillate.
First Divergence — Boom and Bust
Rapid speciation from the three surviving founder lineages. Watch species proliferate, then crash as resources are consumed. Bodies are visibly larger than Part 1.
At tick 8,000, with only 88 organisms alive, the simulation reaches peak species diversity: 25 species. Nearly one species per 3.5 organisms. An incredibly fragmented ecosystem where most species are represented by just a handful of individuals.
This is Part 2's Cambrian Explosion, and it's fundamentally different from Part 1's. In Part 1, peak diversity came with peak population: lots of organisms, lots of species, but all converging on the same body plan. Here, peak diversity comes at low population. These aren't clones with slightly different weights. The bodies are visible, the node counts vary, and the species occupy genuinely different niches.
Most of these 25 species won't survive. They're evolutionary experiments, most of which will fail. But the ones that succeed will define the ecosystem for the next 100,000 ticks.
The Cambrian Moment — 25 Species
Peak species diversity: 25 species coexist in a population of just 88. One species per 3.5 organisms. Most will go extinct within thousands of ticks.
By tick 66,000, the simulation crosses generation 100. Two-thirds the time it took Part 1 to reach generation 128. The scarcer resources create more selective pressure, forcing constant adaptation. Organisms die from competition, not old age, despite having lifespans three times longer.
The generation counter doesn't climb smoothly. It wobbles. At tick 102,000 it reaches 156, then drops to 147 by tick 103,000, then climbs back to 157 by tick 110,000. The most evolved organisms are dying. The bleeding edge of evolution is being culled. Survivors from earlier generations persist while the newest experiments fail. Evolution isn't a smooth march forward. It's a jagged frontier where innovation is punished as often as it's rewarded.
Between ticks 30,000 and 90,000, the ecosystem consolidates. Winners emerge. Species count settles from the wild 20+ swings of early life to a more stable 12-16. The population finds a carrying capacity of about 125 organisms — roughly half of Part 1's 280. Fewer organisms, but each one has been through more rounds of selection.
Deep Evolution — Generation 230+
Organisms refined across 230+ generations of selection. Neural networks have been optimized far beyond anything Part 1 achieved. Watch for the generation counter wobble — the most evolved lineages sometimes go extinct.
Around tick 126,000, the ecosystem breaks. Population plummets from 125 to 73. A 42% crash. Species collapse from 18 to just 3. Three species out of hundreds that had existed. This isn't the gentle oscillation of the earlier era. It's a genuine mass extinction event.
The cause is likely competitive: a few species became so dominant that they suppressed all others, then their own population outstripped the food supply. With no diversity to buffer the crash, the entire ecosystem nearly collapses. The simulation's anti-extinction safeguards don't trigger (the population never drops below the minimum floor), but it comes dangerously close.
The ecosystem never fully recovers. Pre-extinction carrying capacity was ~125 organisms. Post-extinction, it stabilizes around 95. The population permanently shifts to a lower level. The three surviving species rebuild the world, but they build a smaller one.
From three surviving species, the ecosystem rebuilds. Species count rebounds to 16 by tick 155,000, then crashes back to 4 by tick 168,000. Rebounds to 16 again. Crashes to 4 at tick 203,000. The same cycle, over and over: speciation, brief flourishing, collapse. The ecosystem keeps trying to diversify and keeps failing.
Each recovery produces a different cast of species, but the population never reaches pre-extinction levels. The three survivors are too well-adapted. They've had tens of thousands of ticks to optimize, and newcomers can barely compete. New species fill empty niches for a few thousand ticks, then the incumbents expand and crowd them out.
Biologists have a name for this: punctuated equilibrium. Long periods of stasis interrupted by brief bursts of change. The simulation found it on its own.
Late Radiation — Recovery After Extinction
New species emerge from the post-extinction landscape. Watch diversity spike briefly before dominant species reassert control.
By tick 263,000, the simulation crosses generation 400. These organisms have been through 400 rounds of selection, crossover, and mutation. For context: Part 1's entire 100,000-tick run produced only 128 generations. Part 2 has 3.6 times the evolutionary depth. The generation rate holds steady at 1.6 generations per 1,000 ticks — no slowdown, no plateau. Evolution hasn't finished.
The late game settles into a clear rhythm. Three to six dominant species occupy most of the population. Periodic bursts of speciation create 10-16 species. The newcomers get outcompeted within 10-15 thousand ticks. The cycle repeats. The incumbents are just too good — their neural networks are the product of hundreds of generations of refinement.
At tick 300,000, the simulation ends. 102 organisms. 11 species. Generation 457. The ecosystem is still cycling, still producing new species, still killing them. Given another 300,000 ticks, it would keep going. There is no equilibrium. There is no final state. Just an endless oscillation between creation and extinction — the same pattern that real paleontology finds in the fossil record, discovered from scratch by a few thousand lines of Python.
Deep Time — Generation 457
The final moments. Organisms with neural networks refined across 457 generations. Bodies shaped by mass extinctions, recoveries, and relentless competition. These are the survivors.
Part 1 ran for 100,000 ticks and produced a stable, boring ecosystem. Population locked at 280. Every organism identical. 128 generations of refinement that led to the same minimal body. Part 2 ran three times longer and never settled. Population hovering around 95 — a third of Part 1's — because each organism costs more to run. Bigger bodies, richer brains, higher metabolic overhead. Fewer creatures, but far more interesting ones.
The generation counter tells the story. In Part 1 it climbed steadily — each generation slightly more refined than the last, a smooth march toward the same destination. In Part 2 it wobbles. The most evolved lineages go extinct regularly, replaced by younger species with better strategies. Being highly evolved doesn't mean being safe. It means you're at the frontier, and the frontier is where things die.
Part 1's organisms all converged. Part 2's diverged. Sensors survived because six inputs made them worth keeping. Bones persisted because they made organisms faster. Fat mattered because it lowered reproduction thresholds. The rules rewarded complexity, and evolution responded. Different world, different life.
The narrative above tells the story. The charts below show the numbers. Population, species count, and generation depth — tracked every 100 ticks across the full run. The colored phases correspond to the chapters: initial struggle, Cambrian bloom, long consolidation, mass extinction, and the late-era cycles that never settle.
Ecosystem Timeline
Population (green, left axis) and species count (gold, right axis). Hover for exact values at any tick. Phase colors show: Struggle, Bloom, Consolidation, Extinction, Late Life.
The streamgraph below shows every species that ever lived as a colored band. Width = population. Watch the early explosion of color, the long dominance of a few survivors, the extinction event that wipes the palette nearly clean, and the tentative new colors that appear in the aftermath.
Species Dynamics
Streamgraph of species populations over time. Each colored band is one species. Wider = more organisms. Hover for species ID and peak population.
Body composition tells the morphological story. Part 1 collapsed to pure mouths. Part 2 has a longer arc: bones, sensors, and muscle all persist through the early and middle eras. Watch the first 100,000 ticks. Genuine morphological diversity. Multiple node types competing for metabolic bandwidth. Then watch the second half. The palette simplifies. Mouths expand. Everything else contracts.
Body Composition Over Time
Stacked area chart of total node types across the population. Core (cream), bone (grey), muscle (red), sensor (blue), mouth (gold), fat (green), armor (purple). Hover for details.
The absolute counts above show what the population is building. The normalized views below strip away population size and show proportions: what fraction of all nodes are muscle? What share of the ecosystem does each species control? These reveal convergence patterns that absolute counts can hide.
Body Composition — Proportional
Each node type as a percentage of total nodes across the population. Shows compositional shifts independent of population size. Hover for exact percentages.
Species Dynamics — Proportional
Each species as a percentage of total population. Shows dominance shifts and competitive displacement independent of population booms and crashes.
The species dynamics are real. That streamgraph shows genuine ecological drama: radiations, extinctions, fragile recoveries that collapse before they stabilize. The mass extinction at tick 130,000 is a textbook example of competitive exclusion followed by adaptive radiation, the same pattern paleontologists find in the fossil record. Separating eat from attack created actual behavioral diversity. Some organisms became dedicated herbivores. Others stumbled into predatory strategies that their ancestors could never have attempted. The longer lifespan gave complex strategies time to mature across hundreds of generations.
So the ecosystem works. The bodies don't.
Stare at the body composition chart long enough and Part 1's lesson reappears wearing a different outfit. Bones, sensors, and muscle survive through the early eras while mutation is still exploring randomly. But by tick 200,000, the composition simplifies to mouths and cores. The same convergence, just slower. Seven rule changes bought 200,000 extra ticks of diversity, and then evolution arrived at the same answer anyway.
The economics explain why. A muscle node costs 0.06 energy per tick but barely affects how fast you move. A bone costs 0.01 and reduces drag by a fraction. An armor plate costs 0.02 and protects against attacks that rarely kill anyone. Mouths cost 0.04 and directly convert food into survival. Evolution does the math in every generation. The answer keeps coming back the same.
But the problem goes deeper than metabolic accounting. These organisms don't have limbs. They have nodes floating around a core, connected by springs, drifting in roughly circular arrangements. A bone doesn't support anything because there's nothing to support. Muscle doesn't power locomotion because locomotion comes from Brownian motion and whole-body spring forces. Armor doesn't save lives because most organisms starve long before anything eats them. Every node type except mouth is an answer to a question the world never asks.
Part 2 proved that evolution will build complex ecosystems under the right pressure. The species dynamics, the extinction cycles, the generation depth are all genuine. What it also proved is that body diversity requires more than cheaper parts. The world has to demand bodies that do different things. Speed that only muscle can provide. Reach that only bone can extend. Danger that only armor can survive. Until the physics rewards having legs, evolution will keep building spheres.
Part 3 attacks the convergence problem from both sides. The organisms get new capabilities. The environment gets new demands. Neither change works alone.
Articulated bodies. Nodes can now attach to other nodes in chains, forming limb-like structures: bone-bone-mouth arms that reach further than any core-attached mouth. Muscle nodes at joint positions let the neural network control limb movement directly. Bodies stop being circles and start being architectures.
Muscle means speed. An organism's top speed scales with its muscle-to-mass ratio. No muscle, slow organism. Lots of muscle, fast enough to catch prey or escape predators. Sprint bursts burn extra energy but double velocity for a few ticks. For the first time, locomotion has a body plan cost.
Bone means reach. Bone chains extending outward from the core increase foraging radius. A mouth at the end of a three-bone arm eats from three times further than a mouth pressed against the core. Structure becomes a competitive advantage.
Predation gets real. Attack damage doubles. Armor reflects damage back at attackers. Energy transfer from kills jumps from 50% to 70%. When predation becomes a genuine cause of death, armor stops being dead weight and starts being the difference between surviving an encounter and feeding someone else's offspring.
Kin recognition. Organisms evolve a tolerance gene that controls whether they attack members of their own species. Some lineages discover cooperation. Others stay solitary. Parents stop eating their children (unless evolution decides that's still the winning strategy).
The world shifts. Seasonal food cycles. Sudden resource crashes that hit without warning. Spatial gradients that make some regions lush and others barren. A world that never sits still, demanding organisms that can adapt or move.
The question from Part 2 was whether better rules produce better bodies. The answer was: only up to a point. Part 3 asks a sharper question. What happens when the world punishes simplicity?