Articulated limbs. Muscle-powered speed. Bone reach. Seasonal pressure. An environment that punishes simplicity.
Scroll to explore
Parts 1 and 2 asked evolution to build complex organisms, and evolution answered honestly: when the rules don't reward complexity, complexity dies. Mouth-and-core spheres won both times because mouths were the only node type that paid for itself. Muscles didn't make you faster. Bones didn't extend your reach. Armor didn't save your life. Every other body part was metabolic overhead.
Part 3 rebuilds the relationship between body and world. Each node type now earns its metabolic cost, or goes extinct for real reasons instead of economic ones. And the environment stops being static.
Muscle means speed. Maximum velocity now scales with muscle-to-mass ratio. An organism with no muscles crawls. Add muscles and you accelerate proportionally. Sprint bursts let you double your speed for a few ticks at steep energy cost. Predators need muscle to catch prey. Prey need muscle to escape. For the first time, locomotion has a body plan.
Bone means reach. Mouths positioned far from the center of mass get an expanded eating radius. A mouth pressed against the core eats at the standard range. A mouth at the end of a bone chain eats from twice as far. Body structure translates directly into foraging advantage.
Predation gets teeth. Attack damage nearly doubles. Energy transfer from kills jumps from 50% to 70%. Armor reflects a fraction of incoming damage back at the attacker, making armored organisms expensive to hunt. When predation becomes a genuine cause of death, armor stops being dead weight.
Kin recognition. A new genetic trait controls whether organisms attack their own species. Tolerance ranges from zero (attack everything that moves) to one (never attack kin). This is evolvable. Some lineages discover cooperation. Others stay solitary. Parents stop eating their children, unless evolution decides that's still optimal.
Limb chains. When mutation adds a node to a bone, it preferentially extends outward from that bone rather than clustering around the core. Bodies can form branching, limb-like structures. Paired with muscle-speed and bone-reach, this creates evolutionary pressure for articulated morphology.
Seasons. Food spawning oscillates on a 10,000-tick cycle. Abundance in summer, scarcity in winter. Random resource shocks crash food supply without warning. Spatial gradients make some regions lush and others barren, and the gradient shifts over time. The environment never sits still.
Six categories of change. Same algorithm. A world that demands bodies do more than eat.
Fifty organisms are dropped into a world that punishes everything Part 2 let slide. Food oscillates on a 10,000-tick cycle, swinging 40% above baseline in summer and 40% below in winter. Spatial gradients make some regions lush while others are barren, and the gradient drifts over time. Random food shocks can crash the entire supply without warning. Predation deals nearly double the damage of Part 2, and armor reflects a fraction of incoming damage back at the attacker.
Every organism starts with the same five-node body: a core, a sensor, a mouth, and two muscles. That forty percent muscle content actually determines how fast they move now. For the first time in the project's history, the body plan has consequences beyond metabolism.
The population reacts to the new rules by exploding. By tick 2,200, fifty organisms become 500. One species becomes 158. The most violent Cambrian explosion of any part: more species, faster, in a more fragmented ecosystem. The richer body mechanics and articulated mutation system mean there are more viable strategies to explore. Evolution floods every niche it can find.
Genesis — A World That Demands More
The founding population under seasonal food variation and stronger predation. Watch the population explode from 50 toward 500 as mutations explore the new body mechanics.
The first species split happens within 500 ticks. By tick 1,500, there are 39 species testing different configurations. Bodies are still small, 5.3 nodes on average, but already diversifying. Muscle dominates early composition at 38% of all nodes, because it came free with the starter body and it actually does something now. Sensors hold steady at 19%. Armor barely exists.
What separates Part 3 from its predecessors is what happens next. Instead of stripping down to cores and mouths, organisms are adding nodes. Bones appear by tick 1,000. Fat reserves start accumulating. The growth-biased mutation rate, thirteen additions for every removal, pushes bodies outward, and for the first time, outward growth pays off. Bigger organisms have more energy reserves to survive the food dips that hit every 5,000 ticks during the seasonal winter. A six-node organism with a fat store weathers the lean season. A three-node organism without one doesn't.
First Divergence — Tick 500
Early speciation under the new rules. Bodies are small but already diversifying. Bones and fat stores appear for the first time as growth-biased mutations push organisms outward.
At tick 3,300, species diversity peaks at 158. Nearly one species per three organisms. The entire population is an experiment in progress, hundreds of body plans competing for the same scarce resources. This dwarfs Part 2's peak of 25 species. The richer body mechanics create more viable strategies, and the population is large enough to sustain them all. Briefly.
Then the armor rush starts. Between tick 5,000 and 19,000, armor goes from 8% of all nodes to 37%. The fastest compositional shift in three hundred thousand ticks of simulation. The cause is straightforward: predation hits hard now. Fifteen damage per mouth. Seventy percent energy transfer from kills. Unarmored organisms are lunch. Armored organisms bounce damage back at their attackers. Evolution does the math in real time.
The cost is sensors. As armor floods the body plans, sensors crater. From 19% at tick 10,000 to 0.8% at tick 29,000. Only 23% of organisms have any sensors at all. Why bother sensing food when you can survive long enough to stumble into it? Why detect predators when five armor nodes make you too expensive to hunt? The world briefly rewards being a tank, not a scout.
By tick 25,000, the average organism has 15 nodes, triple the starting size, with armor making up more than a third of every body. Bone has quietly risen to 10%. Fat to 10%. Mouth holds steady at 22%. The bodies are genuinely complex for the first time in the project's history, but the complexity is lopsided: armor dominates everything.
Radiation — 158 Species
Peak species diversity: 158 species in a population of ~400. The armor rush is just beginning. Watch for the growing armor nodes (purple) as bodies bulk up for combat.
The armor era doesn't last. By tick 50,000, the population has crashed from 253 to 121. The survivors are the organisms that found food efficiently, not just the ones that tanked the most damage. Armor settles from 37% to 28%. And sensors, nearly extinct at tick 29,000, begin the most surprising trajectory in the entire project.
The sensor resurrection happens slowly. From 0.8% at the nadir to 10% by tick 50,000. Then 16% by tick 100,000. Then 18% by tick 150,000. By tick 50,000, ninety percent of surviving organisms carry sensors. By tick 75,000, it's one hundred percent. Every single organism has rebuilt the sensory apparatus that evolution nearly eliminated 50,000 ticks earlier.
The reason is the environment. As populations thin out from 500 to 150, food becomes harder to find by accident. Spatial gradients mean some regions are rich and others are empty. Seasonal cycles put the entire map through lean periods thousands of ticks long. Sensors find food. In a sparse world, finding food is the difference between surviving to reproduce and starving three thousand ticks before your lifespan expires.
Bone follows a parallel arc. From 5% at tick 5,000 to 14% by tick 50,000 to 17% by tick 150,000. Bone is the structural scaffold that makes large bodies functional. Its reach mechanic, where mouths far from the center of mass eat at expanded range, rewards organisms that build outward rather than inward. Combined with limb chain mutations that preferentially extend from existing bone nodes, this creates arm-like structures: bone chains with mouths at the ends, foraging at twice the distance of core-attached mouths.
By tick 160,000, the typical organism has 23 nodes containing all seven types. This is the answer to Parts 1 and 2. No node type has gone extinct. No type dominates above 22%. The body plan has converged not to minimalism, but to balance.
Deep Evolution — Generation 174
Tick 160,000. Organisms carry all seven node types. Sensors have fully recovered. Bone arms extend foraging reach. Bodies average 23 nodes, built by 174 generations of selection under seasonal pressure.
By tick 262,000, the ecosystem has found a rhythm. Population oscillates between 80 and 150, driven by the seasonal food cycle. Species counts fluctuate between 15 and 30. Bodies average 25 nodes. And the composition holds steady within narrow bands, tick after tick after tick.
The late-game body plan: sensor at 17-21%, the resurrected hero of the simulation. Mouth at 17-19%, the foraging apparatus. Bone at 15-19%, the structural scaffold and reach extender. Armor at 18-21%, still valuable but no longer dominant. Muscle at 11-13%, with diminishing returns preventing further stacking. Fat at 8-10%, energy reserves for surviving seasonal droughts. Core at 3-4%, the mandatory anchor point, exactly one per organism.
Every node type pays for itself. That sentence would have been impossible to write about Parts 1 or 2. Sensors earn their metabolic cost by finding food in a sparse world. Muscle earns its cost by powering chase and escape. Bone earns its cost through foraging reach. Armor earns its cost through damage reflection. Fat earns its cost by buffering seasonal famine. The environment demands all of them, and the organisms supply all of them.
What stands out is the stability. Population crashes of 50% barely move the composition. At tick 200,000, the population dips to 79, the lowest sustained point since the early explosive phase, but the proportional body plan holds. The same seven-type distribution at 79 organisms as at 180. Evolution has found a local optimum that survives population shocks. The organisms are a reaction to their environment, and the environment hasn't changed.
Equilibrium — Tick 262,000
The ecosystem at dynamic equilibrium. Population fluctuates with seasons, but the body plan holds. All seven node types persist in every organism.
The final 65,000 ticks are the most dynamic since the Cambrian explosion. Species turnover accelerates: at least twelve dominance shifts between tick 235,000 and 300,000. Species sp_7799, which held majority control for 70,000 consecutive ticks through the middle era, is finally displaced. Its replacements hold power for 10,000 ticks or less before being challenged. The incumbents are no longer untouchable.
The body plan keeps evolving even when composition is stable. Average node count rises from 24.5 at tick 200,000 to 28.6 at tick 300,000. Organisms are still getting larger. Sensors reach their all-time peak of 21.9% at tick 295,800, becoming the single most common node type in the simulation. The organisms that started with one sensor now carry six. Bone hits 19.9% at tick 271,900. The trends that defined the middle era, sensor rise, bone expansion, muscle decline, body growth, are all still actively moving at the end.
The simulation ends with 97 organisms, 38 species, and generation 203. The average organism carries 6 sensors, 5.5 mouths, 5.4 bones, 5.2 armor nodes, 3 muscles, 2.3 fat reserves, and a core. Twenty-nine nodes of evolved architecture. In Parts 1 and 2, organisms converged on three nodes and two types. Part 3 produced organisms five times larger that use every tool available to them.
Given another 300,000 ticks, the trends suggest sensors would keep climbing, bodies would keep growing, and the species roster would keep churning. The system has not reached equilibrium. It has reached a pattern. The same pattern that real evolutionary biology finds in the fossil record, discovered by a few thousand lines of Python running on a laptop.
Late Radiation — 38 Species at Tick 293,000
The final era. A late wave of speciation produces 38 species. Bodies average 29 nodes with all seven types. Sensors are the dominant node type for the first time.
The body composition chart is the payoff. Where Parts 1 and 2 showed inevitable collapse toward mouth-and-core minimalism, Part 3 shows seven colors holding their ground across 300,000 ticks. The proportional view makes it clearer: after the early armor rush subsides around tick 50,000, the composition stabilizes into narrow bands that hold for a quarter million ticks. No type goes extinct. No type takes over.
The species dynamics tell a complementary story. 11,683 unique species across the full run. Early fragmentation gives way to longer dynasties in the middle era, then rapid turnover returns in the final stretch. The ecosystem never stops generating new experiments, and the new experiments never stop failing. The pattern holds whether the population is 500 or 79.
Ecosystem Timeline
Population (green, left axis) and species count (gold, right axis). Hover for exact values at any tick.
Species Dynamics
Streamgraph of species populations over time. Each colored band is one species.
Body Composition Over Time
Stacked area chart of total node types across the population. The key question: do muscle, bone, and armor persist?
Body Composition — Proportional
Each node type as a percentage of total nodes. Shows compositional shifts independent of population size.
Species Dynamics — Proportional
Each species as a percentage of total population. Shows dominance shifts independent of population booms and crashes.
Part 1 asked what evolves when nothing is designed, and the answer was: nothing complex. Three-node organisms with no sensors, no muscles, no bones. Evolution optimized for survival and survival meant minimalism. Part 2 asked what happens when you fix the rules, and the answer was: the ecosystem works but the bodies don't. Mass extinctions, adaptive radiations, 457 generations of genuine ecological drama, all converging on the same minimal body plan. Slower, but the same destination.
Part 3 asked what happens when the world punishes simplicity. The answer is that bodies finally matter. Organisms grow to 29 nodes carrying all seven types. Sensors crash to near-extinction during the armor rush, then recover to become the dominant node type. Bone goes from nonexistent to architectural necessity. Armor finds its equilibrium at five nodes and holds there for 275,000 ticks. The body plan converges, but to balance, not minimalism.
The numbers tell the story cleanly. Average body size in Parts 1 and 2: three nodes. Part 3: twenty-nine. Node types surviving in Parts 1 and 2: two. Part 3: all seven. Sensors surviving: no, no, yes. Muscle surviving: no, no, yes. The same evolutionary algorithm, the same neural network architecture, the same physics engine. The only difference is the environment.
The environment writes the organisms. That was always the thesis of this project. Part 3 is the proof.
Part 3 proved that bodies matter. Part 4 asks whether minds do too. Organisms get recurrent neural networks with memory that persists across ticks. They get chemical signaling: pheromone broadcasts that nearby organisms can sense, creating the preconditions for communication, coordination, and deception. They get resource sharing between kin and group bonuses that reward clustering.
The world gets harder. Four terrain biomes shape where food grows and how fast organisms move. A day-night cycle blinds sensors after dark. Toxic food punishes indiscriminate eating. Hazard zones drift across the map. Two new node types appear: signal emitters for broadcasting and stomachs for digestion efficiency. Organisms start colliding with each other instead of passing through, and attacks can land on body edges, not just nodes.
Bigger world. Longer lives. Nine sensory inputs per sensor instead of six. The question is whether any of it produces something that looks like cooperation, or whether evolution finds a way to turn every new tool into a weapon.