Corpse drops. Instant lethality. Moving food patches. A metabolic law that punishes size. And evolution's answer: the smallest predators the simulation has ever produced.
Scroll to explore
Part 4 ended in stasis. Six hundred organisms, all carrying the same six node types, all roughly the same size, drifting through a world they had thoroughly saturated. The arms race produced convergence so complete that the only variation left was neural. The problem wasn't that evolution stopped working. The problem was that the environment stopped selecting.
Part 5 changes the economics.
Corpse drops. When an organism dies, 70% of its energy falls as food right where it died. A fat organism at 150 energy becomes a 105-unit meal. Killing is now immediately profitable. Scavenging becomes a viable niche. Being fat makes you a target, not just a tank.
Lethal threshold. Take 30% of your max energy as damage in a single exchange and you die outright. No war of attrition. No slow grinding. A well-placed hit from a claw can end an organism in one tick. This eliminates the attritional combat that dominated Part 4 and replaces it with an economy where a single encounter decides everything.
Moving food superpatches. Four dense food zones drift on sine-wave paths across the map. Organisms that can track the patches eat well. Organisms that stumble into the path eat when the patch happens to pass by. The moving patches were designed to reward memory and sensing, to make anticipation valuable, to give the Memory node type a concrete advantage over random foraging.
Quadratic metabolic scaling. An organism's metabolic cost scales as (n/10)0.7 where n is node count. This is sublinear but still punishes size: a 10-node organism pays 1.0x, a 30-node organism pays 2.2x. The scaling is gentler than linear but harsh enough to make every node justify its existence.
Fat nerf. Fat nodes cap at 30 energy each (down from 50 in Part 4) and recharge at only 25% rate. Fat still provides a buffer, but the Part 4 strategy of stacking unlimited fat reserves is no longer viable. Organisms have to spend their energy on something other than hoarding.
Memory nodes. A new node type that adds four brain registers per node. These registers persist across ticks, giving the neural network a form of working memory. The registers are visible in the body composition data, unlike the recurrent hidden states of Part 4's brains. If evolution selects for memory, we'll see it in the charts.
No population cap. The maximum is set to 10,000, effectively uncapped. The dynamics have to enforce their own limits. If predation doesn't cull the population, resource competition will. If neither does, the simulation will bloat. Two concrete hypotheses: will corpse drops create genuine predator-prey oscillations, and will memory nodes earn their metabolic cost when food patches move predictably?
Sixty organisms seed into a 4000 by 2250 world. Each carries the default genome: one core, two muscle anchors, one sensor, one mouth. Five nodes. Random neural weights. The world is empty and the food is everywhere.
By tick 3,726, the population has exploded to 387. Fifty-seven species have already diverged from the single founding genome. Average body size has crept up to 5.9 nodes as organisms add a sixth node, usually another mouth. The logic is simple: in an empty world with abundant food, more mouths means more energy. Muscle dominates at 34% of total nodes because the default genome starts with two anchors, but mouth is climbing fast at 27%.
The exotic node types are already appearing. Signal at 1.9%, stomach at 1.1%. Memory nodes appeared briefly at tick 402 but haven't established yet. The big question for this simulation is whether memory will take hold before the body plan locks in, and at tick 3,726, the answer is still unknown.
Food sits at 901 items. Corpse drops are active but barely relevant: organisms are small and cheap, so dying releases only a few energy units. The economy of death that Part 5 was designed to test hasn't started yet. The world is still in the gift phase, where everything is free.
Genesis — Tick 2,000–2,100
The first 2,000 ticks: 387 organisms filling an empty world. Bodies average 5-6 nodes. The default genome is visible everywhere: a core with two muscle anchors, a sensor, and a mouth. Watch the clustering around food patches as the first foraging behaviors emerge from random neural weights.
The gift phase ends somewhere around tick 15,000. The population, which peaked above 400 in the initial expansion, crashes back to 277. Food drops from 901 to 820. The organisms that filled the world ate through the surplus and now face genuine scarcity. This is the first real selection event: the organisms that overfilled the world die, and the 277 that remain are the ones that ate most efficiently.
Evolution's answer to scarcity is to stack mouths. The default genome carried one mouth. By tick 26,148, the average organism carries 2.4. Mouth nodes have climbed from 27% to 38.7% of all nodes, the dominant type in the simulation. Muscle anchors, which started at 34%, have fallen to 25.9%. The trade is explicit: locomotion capacity for intake throughput. Organisms are getting slower but eating faster.
Whether this trade-off is sustainable depends on what happens next. In a world with corpse drops and instant lethality, being slow and mouth-heavy is a predation invitation. But at tick 26,148, predation hasn't launched. Only two organisms carry claws. The math doesn't work yet: targets average 35.7 energy, and a corpse drop from a 36-energy kill yields about 25 energy. The payoff doesn't exceed the metabolic cost of carrying a claw on a 6-node body. Predation needs bigger, fatter targets to become viable.
Five organisms carry memory nodes. That's 1.8% adoption, 0.3% of total nodes. Memory survived the crash. The initial population boom-bust would have wiped out any trait that didn't confer real advantage, or at least neutrality. Five memory-carriers still alive at generation 36 suggests memory is pulling its weight, possibly in patch-tracking. The question is whether 5 organisms is the start of a trend or the remnant of a lineage about to go extinct.
Post-Boom Crash — Tick 26,000–26,100
Population down to 277 after the initial boom. Mouth-heavy organisms dominate: 38.7% of all nodes are mouths. Bodies are still small at 6.2 nodes, built for eating, not fighting. Food has dropped from 901 to 820. The economy of scarcity is asserting itself.
Between tick 26,000 and tick 181,000, the simulation runs for 155,000 ticks and tells a story that no previous part has told. Bodies don't grow. They shrink.
Average node count drops from 6.2 to 3.8. The quadratic metabolic scaling at (n/10)0.7 is selecting hard against size. At 3.8 nodes, organisms pay roughly half the metabolic rate of a 6-node body. Evolution found the minimum viable organism: core plus mouth plus muscle. Three nodes. That's the cheapest thing that can move and eat. Most bodies are 3 or 4 nodes. A few reach 8. This is a world of miniatures, the total opposite of Part 4's 29-node armored giants.
The miniaturization has consequences that ripple through every other metric.
Sensors collapsed. From 15.6% of nodes at tick 26k to 1.0% at tick 181k. Nearly blind. Evolution decided that the metabolic cost of a sensor node isn't justified by the information it provides when bodies are this small. Organisms find food by random walk and encounter, not directed search. The moving superpatches were designed to reward sensing and memory, but apparently random motion intersects the patches often enough that the advantage of seeing them coming doesn't justify a fourth node.
Predation ignited. This is the big story. Claw adoption went from 2 organisms (0.7%) to 75 organisms (26.2%). From 0.1% of nodes to 7.2%. With the lethal threshold at 30% and tiny bodies, combat is fast and fatal. A 3-node organism at 40 energy dies from a 12-damage hit. Corpse drops make killing profitable: a 40-energy corpse yields about 28 energy, delivered as food right where the predator is standing. The economics of murder work because bodies are small enough that the payoff exceeds the cost of carrying a claw.
This is exactly what Part 5 was designed to test. Corpse drops plus lethal threshold plus small bodies creates a genuine predator economy. The organisms with claws aren't occasional, incidental killers. They're specialists. A quarter of the population now carries dedicated weapons on bodies averaging fewer than 4 nodes. Kill, eat the corpse, reproduce. The Lotka-Volterra dynamics that never appeared in Parts 1 through 4 might finally be emerging.
Memory holding at 5 organisms. The same count as tick 26,148. It could be a single persistent lineage threading through 127 generations. Five out of 286 is marginal, but survival through 155,000 ticks of intense selection means memory is at least neutral. The fact that it hasn't spread suggests it isn't strongly advantageous either. The moving superpatches may not be enough of an advantage when bodies are this small and lifespans this short.
Armor has appeared at 1.8% in response to the claw boom. The arms race is starting again, this time at miniature scale. Population stable at 286. Food at 420, down from 820 at tick 26k. Resources are genuinely scarce. The simulation is regulating itself without a cap.
The Miniaturization — Tick 180,000–180,100
Bodies averaging 3.8 nodes. A quarter of the population carries claws (red). Sensors have nearly vanished. This is the smallest average organism size in the project's history: core, mouth, muscle. The minimum viable predator.
Between tick 181,000 and tick 221,000, the claw boom matures into something unprecedented in the project's history: a genuine predator economy.
By tick 199,000, one species controls the ecosystem. sp_652 holds 53.5% of the population, 121 of 226 organisms. This species wasn't in the top five at tick 181k. Its rise from background noise to majority control in under 18,000 ticks is the kind of explosive succession that marks a real competitive advantage. The strategy: core, muscle, mouth, stomach, claw. Five nodes. A hybrid that eats efficiently and kills efficiently. Stomach adoption reaches 54%, meaning every food item and every corpse yields more energy per node. Predation funded by digestion.
The population drops to 226, then to 179. Food crashes from 420 to 227. The ecosystem is transitioning from a plant-based economy to a corpse-based one. When an organism dies, 70% of its energy falls as food right where it died. A population of armed predators creates a constant supply of corpse food, which feeds the predators who create more corpse food. The cycle is self-reinforcing.
The morphology is distinctive. A core with one or two short chains ending in a claw node. The spring-mass physics means those chains whip around as the organism turns, sweeping the claw through a wide arc. They look like tiny flails. Evolution found the medieval weapon without anyone designing it: having the claw at the end of a chain maximizes reach, and the whipping motion means a turning organism sweeps its weapon through more space than a claw mounted close to the core would.
By tick 221,000, the cycle reaches its logical conclusion. Claw adoption hits 96%. Nearly every organism in the simulation carries a dedicated weapon. The predator economy is no longer a niche strategy; it's the default. Everyone is both predator and prey.
Stomach reaches 92% adoption alongside the claws. Bodies grow from 3.8 to 5.7 average nodes as bone and armor begin appearing. This is the same arms race sequence as Part 4, claws go universal, then armor appears in response, then bone scaffolds the armor, but at one-fifth the body size. Part 4's organisms grew from 8 to 28 nodes during the arms race. Part 5's grow from 3.8 to 5.7. The metabolic scaling is doing its job.
Memory goes extinct. Zero organisms. In a world of lethal combat and 5-node bodies, there is no room for a node that doesn't directly contribute to feeding or fighting. The experiment's hypothesis about memory nodes earning their cost through patch-tracking has been falsified. Memory was neutral at best, and neutral isn't good enough when every node must justify itself against a metabolic curve.
The Predator Peak — Tick 221,000–221,100
96% of the population carries claws. Bodies have grown to 5.7 nodes. This is peak predation: every organism is armed, every encounter is lethal, and corpse food drives the economy. The miniaturization is reversing as organisms invest in bone and armor alongside their weapons.
The predator economy overshot.
When 96% of the population carried claws, everyone was killing everyone. Corpse food was abundant but the prey population couldn't sustain the kill rate. The herbivore strategy, core plus muscle plus mouth plus stomach at 4 nodes with no claw, became more efficient than the armed version. A 4-node unarmed forager pays less metabolic cost than a 5-node predator and doesn't risk lethal combat on every encounter. Evolution stripped the weapons back off.
Claw adoption crashes from 96% to 32% in 30,000 ticks. The most dramatic reversal in the simulation's history. Bodies shrink from 5.7 to 4.7 average nodes as the arms race infrastructure dismantles. Bone collapses from 5.8% to 0.3%. The lean forager template reasserts itself.
This is the Lotka-Volterra dynamic that Part 5 was designed to test. Predators rise, prey crash, predators starve, predators crash, prey recover. The first half of the cycle played out between tick 181,000 and tick 221,000 as claws rose from 0.7% to 96%. The second half plays out over the next 30,000 ticks as claws crash back to 32%. The oscillation is real.
A new species, sp_3301, rises to 32.2% of the population, overtaking sp_652. The species succession tracks the strategy shift: when claws were universal, sp_652's predator-forager hybrid dominated. Now a different body plan leads, one optimized for lean foraging in a post-predator environment.
Max energy drops from 124 to 75.6. Organisms are poorer. With fewer kills and less corpse food, the energy cycling through the ecosystem declines. Food sits at 209. The boom is over.
But stomach survives intact at 96% adoption, the one trait that persists through both the predator boom and the predator crash. Digestion efficiency is always valuable regardless of whether you're eating food or corpses. Stomach is becoming the Part 5 winner: the only non-default node type to achieve and maintain near-universal adoption across strategic shifts.
The Crash — Tick 251,000–251,100
Claw adoption has crashed from 96% to 32%. Bodies are shrinking back toward the lean forager template. The predator economy overshot: when everyone was armed, the kill rate exceeded the prey population's recovery rate. Evolution's correction is underway.
The Lotka-Volterra cycle predicted that predators would rise again after the crash. They didn't. The ecosystem found a third state.
By tick 278,000, armor adoption has exploded to 63.6%, up from 12% at the crash. Two-thirds of the population carries armor. Armor is 15.5% of all nodes, a larger proportional share than claw ever achieved at its peak. The ecosystem didn't reload its weapons. It put on shields.
Claw collapses to 5.6%. Eighteen organisms out of 324 still carry weapons. The few remaining predators face a population where armor reflects 20% of incoming damage. The math that made predation profitable, small bodies, easy kills, rich corpses, no longer works when the target bounces damage back.
sp_652 surges to 64.2% of the population. The same species that ran the predator economy now runs the armored herbivore economy. sp_652 keeps reinventing itself: predator-forager hybrid at tick 199k, weakened during the crash, and now dominant again as an armored forager. The species pivots from offense to defense as the meta shifts. Whatever sp_652's neural weights are doing, they adapt to strategic change faster than any competitor.
Population surges to 324, the highest since genesis. Armor keeps organisms alive. Fewer predators means less mortality. Stomach drops from 96% to 35% as the body budget shifts from digestion to defense: organisms trade stomach nodes for armor nodes. When predators were everywhere, digestion efficiency maximized corpse income. When predators disappeared, the corpse economy dried up and armor's survival benefit outweighs stomach's income benefit.
Food drops to 143, the lowest in the simulation's history. 324 organisms competing for 143 food items. This is the density-dependent crisis that should trigger the next phase. The armored herbivore equilibrium can't hold at this population density.
The Armor Pivot — Tick 278,000–278,100
64% of the population carries armor. Claws down to 5.6%. Population at 324, the highest since genesis. The ecosystem chose defense over offense. sp_652 dominates at 64%, having pivoted from predator to armored herbivore.
The armored herbivore equilibrium didn't hold. The cycle repeated, but with a difference.
By tick 303,000, claws are rising again: 29.8% adoption, up from 5.6%. Armor is falling: 7.5%, down from 63.6%. Stomach is back to near-universal at 95%. The ecosystem is cycling between offense and defense on a roughly 50,000-tick period. The full oscillation:
| Trait | Tick 221k Peak offense |
Tick 251k Crash |
Tick 278k Peak defense |
Tick 303k Cycle 2 rising |
Tick 329k Cycle 2 mid |
|---|---|---|---|---|---|
| Claw | 96% | 32% | 5.6% | 30% | 42% |
| Armor | 12% | 12% | 64% | 8% | 13% |
| Stomach | 92% | 96% | 35% | 95% | 88% |
| Population | 179 | 183 | 324 | 292 | 252 |
The cycle doesn't repeat cleanly. In cycle 1, claws rose to 96% before armor appeared at all. In cycle 2, armor, bone, and fat all rise alongside claws. Defense and offense co-evolve instead of lagging a full cycle behind. The ecosystem learned. Or more precisely: the lineages that carry defensive traits weren't completely eliminated during the trough, and the survivors co-evolve with the new predators instead of playing catch-up.
Sensors recover to 8.7%, the most meaningful sensor presence since genesis. With food at 143 and 252 organisms competing for it, directional information has real value again. At 5.0 average nodes, organisms can afford a sensor as a 5th node alongside the core/muscle/mouth/stomach template. The sensor collapse was specific to the miniaturization era when bodies were 3-4 nodes and couldn't spare a node for information. At 5 nodes, the budget is less tight.
sp_652 holds at 53.6%, still dominant but weakened from its 74% peak. The oscillating meta creates room for specialists: claw-heavy lineages that thrive during predator booms, armor-heavy lineages that thrive during defensive phases, and generalists that survive both. Species diversity has climbed to 57, the highest since the early boom.
This is the central finding of Part 5. The economics of death create genuine strategic oscillation, not the permanent arms race of Part 4. The difference is metabolic scaling: bodies can't grow large enough to absorb all traits simultaneously. When everyone is 4-5 nodes, every strategic shift costs only one node, cheap enough to reverse. Part 4's organisms carried claws and armor and fat and bone all at once because they could afford 28 nodes. Part 5's organisms carry claws or armor because they can only afford 5.
Second Cycle — Tick 329,000–329,100
The second predation cycle builds. Claw adoption at 42%, with armor, bone, and fat co-evolving in parallel. Bodies average 5.0 nodes. Unlike cycle 1, defense and offense rise simultaneously. The ecosystem is adapting to its own oscillation.
The second cycle peaked around tick 329,000 with claws at 42% and crashed. But the crash was gentler than cycle 1's. No dramatic reversal from 96% to 5%. Just a steady decline from 42% to 14% to 4%. The co-evolution of defense alongside offense truncated the peak: when armor, bone, and fat rise alongside claws, predation never becomes as universally profitable as it did in cycle 1. The overshoot is smaller, so the correction is smaller.
By tick 378,000, the oscillation is over. Claw adoption at 4.1%. Armor at 3.8%. Both offensive and defensive traits have been stripped to background noise. The ecosystem converges on the lean forager: core, muscle, mouth, stomach. Four nodes. The minimum viable organism that can move, eat, and digest efficiently.
Without predation culling the population, organisms proliferate. Population surges to 320, the highest since genesis. Food crashes to 68, the most extreme scarcity in the entire simulation. 320 organisms competing for 68 food items. For comparison, the opening had 387 organisms for 901 food items. The ecosystem has eaten through everything. The lean foragers are cheap to run, so they survive on very little, but 68 food items for 320 organisms is unsustainable.
The food crisis corrects itself. Population crashes to 258. Food recovers to 258. A perfect 1:1 ratio, the healthiest food-to-organism balance since the early simulation. Starvation culled the weakest, survivors spread out, food regenerated. The density-dependent correction works exactly as a textbook would predict.
Convergence — Tick 378,000–378,100
The oscillation is over. Claws at 4%, armor at 4%. Population at 320, the highest since genesis, but food has crashed to 68. These are lean foragers: core, muscle, mouth, stomach. Four nodes, no weapons, no armor. The ecosystem stripped everything back to the minimum.
After the food crisis resolves, a third cycle emerges. But it's barely a cycle at all.
Claw adoption rises to 15%. Armor rises to 14%. Both climbing together from the start, with no lag between offense and defense. Bone at 9%, fat at 9%. The ecosystem has fully internalized the co-evolution pattern: when predators appear, armor appears simultaneously. The naive victims of cycle 1 are gone. Every lineage now carries the genetic memory of two full predator-prey oscillations.
The amplitude tells the story of the dampening. Cycle 1 peaked at 96% claw adoption. Cycle 2 at 42%. Cycle 3 is holding around 15%. If this is a dampened oscillation converging on a fixed point, the equilibrium is somewhere around 10-15% armed, 10-15% armored, and the rest lean foragers with stomach. The wild swings that defined the middle of the simulation are settling into gentle ripples.
sp_652 holds at 67%, having dominated through every phase: the predator economy at tick 199k, the crash at tick 251k, the armor pivot at tick 278k, two full oscillation cycles, a food crisis, and now the convergence. Whatever sp_652's neural weights encode, they adapt to the strategic environment faster than any competitor. The species is not locked into one strategy. It rides every wave.
Signal nodes have established for the first time at 7.4% adoption. After 350,000 ticks of irrelevance, pheromone broadcasting found a niche. At this population density, organisms that emit signals attract kin, and kin aggregation provides the +20% sensor range group bonus. Signal may be free-riding on the social mechanics rather than serving any direct communication function. Evolution found a use for the feature, just not the one it was designed for.
The New Equilibrium — Tick 402,000–402,100
The gentle third cycle. Claws at 15%, armor at 14%, co-evolving in parallel. Bodies average 4.9 nodes. Food has recovered to 258 after the crisis. The wild oscillations of the middle simulation have dampened into a stable mixed equilibrium.
The third cycle never peaked. It plateaued at 15% claw adoption around tick 427,000, held briefly, and collapsed. By tick 450,000, claws had crashed to 1.9%. Armor surged to 11.5% in response, the same defensive pivot as before, but at one-sixth the amplitude. Then armor collapsed too. By tick 473,000, claws were at 0.3% and armor at 12.2%, but falling.
The dampened oscillation didn't stabilize at a low equilibrium. It dampened to zero. Three cycles, each smaller than the last: 96%, 42%, 15%. Then nothing. The predator economy that consumed 70,000 ticks of evolutionary drama simply vanished. The few remaining clawed organisms are new mutations, not evidence of a viable niche. They appear and get selected out within a few thousand ticks.
The population-food oscillation is a different story. That one never dampened. Population surged to 327 at tick 473k, food crashed to 94. The fourth food crisis in 200,000 ticks, roughly the same amplitude each time. Pop overshoots 300, food drops below 100, starvation corrects, food recovers. This oscillation has no evolutionary solution because organisms can't evolve to need less food. The density-dependent cycle is permanent.
sp_652 holds through it all. 63% at tick 473k, having dominated every strategic era since tick 200k. sp_8182 emerged as the strongest challenger at 15.6%, the most serious competition sp_652 has faced. But no species has broken 20% against the dynasty. Whatever sp_652's neural weights encode, 150 generations of challengers haven't found a better answer.
The simulation ends at tick 500,000 with 189 organisms, 211 food items, and a body plan so minimal it's almost a parody of the design space available.
Four node types, each at almost exactly 25%: core (22.8%), muscle (22.8%), mouth (23.2%), stomach (23.2%). The minimum viable organism. Everything else has been stripped away. Sensor at 4.2% is the only non-default trait with meaningful presence, and even that is marginal. Claw at 1.1%. Armor at 0.6%. Fat at 0.6%. Bone at 1.3%. Signal: one organism. Memory: extinct.
Average body size is 4.4 nodes. Range: 4 to 7. There are no large organisms. No specialists. No predators. No armored tanks. Just lean foragers with a stomach, eating food and reproducing. The ecosystem that produced a 96% predator economy, a 64% armor pivot, and three full Lotka-Volterra cycles ends as the simplest possible world.
Final State — Tick 499,000–499,100
The final state: 189 organisms, all lean foragers. Bodies average 4.4 nodes. No visible claws, no armor, no predation. The ecosystem stripped everything back to the minimum viable organism: core, muscle, mouth, stomach.
Part 5 asked two concrete questions. Both got clear answers.
Will corpse drops create genuine predator-prey oscillations? Yes, emphatically. The Lotka-Volterra dynamic that never appeared in Parts 1 through 4 played out across three full cycles between tick 180,000 and tick 450,000. Claw adoption rose to 96%, crashed, rose again to 42%, crashed again, rose to 15%, and finally dampened to zero. The oscillation was real, and its dampening through co-evolution of offense and defense was the central surprise. Each cycle was smaller than the last because the population retained evolutionary memory of both strategies. Defense and offense co-evolved in later cycles instead of lagging a full phase behind. The naive victims of cycle 1, organisms that had never seen a claw, were gone by cycle 2.
Will memory nodes earn their metabolic cost when food patches move predictably? No. Memory appeared sporadically throughout the run but never exceeded 5% adoption and went extinct by tick 450,000. Four brain registers per node wasn't enough computational advantage to justify the energy drain in a world where bodies average 4-5 nodes and every node must justify itself against the metabolic curve. The moving superpatches weren't enough of an advantage when random motion intersects the patches often enough to survive.
The deeper finding is about metabolic scaling. The quadratic cost at (n/10)0.7 is the mechanism that kept Part 5 interesting. It forced organisms to stay small, which forced strategic trade-offs: claws or armor, not both. When every strategic shift costs only one node, it's cheap enough to reverse, which creates oscillation rather than the permanent convergence of Part 4. Part 4's organisms grew to 28 nodes and carried everything. Part 5's organisms stayed at 4-5 nodes and had to choose.
sp_652 dominated from tick 200,000 to tick 500,000. A single species, holding 60-75% of the population through every strategic phase: the predator economy, the crash, the armor pivot, two full oscillation cycles, convergence, and final equilibrium. 300,000 ticks of continuous majority control. The dynasty didn't specialize in one strategy. It rode every wave, pivoting from predator to armored herbivore to lean forager as the meta shifted. The neural weights that encode that flexibility are the most successful product of Part 5's evolution, and they're invisible in every chart.
Ecosystem Timeline
Population (green, left axis) and species count (gold, right axis). The genesis boom, miniaturization plateau, predator economy, and dampened oscillation are all visible.
Species Dynamics
Streamgraph of species populations over time. sp_652's dynasty dominates the second half of the simulation, holding 60-75% of the population from tick 200k onward.
Body Composition Over Time
Total node counts by type. The claw spike at tick 221k and the armor response at tick 278k are the two most dramatic morphological shifts. Stomach's steady rise through the first 200k ticks funded the predator economy.
Body Composition — Proportional
Each node type as a percentage of total nodes. The oscillation between claw and armor is visible as complementary waves. By tick 450k, both have collapsed and the four default types share the body budget equally.
Species Dynamics — Proportional
Each species as a percentage of total population. sp_652's dominance from tick 200k onward is the strongest single-species grip in the project's five-part history.
Five parts of Primordial proved that evolution works. Neural networks evolve from random noise into functional controllers. Bodies adapt to their environment. Predator-prey oscillations emerge without anyone programming them. Arms races ignite and dampen. Dynasties rise and fall. All from physics, mutation, and survival.
But Primordial's organisms live in a flat 2D world with spring-mass bodies and feedforward brains. They can move and eat and kill. That's the full behavioral repertoire. The interesting question isn't whether evolution can produce behavior from scratch. Five parts answered that. The interesting question is what evolution produces when the behavioral space is wide enough to surprise you.
Noloss is a 3D evolutionary simulation on a tiny spherical planet. Creatures have attention-based recurrent brains, hierarchical bodies with individually damageable parts, and a set of actions that change meaning depending on context: what they're targeting, what they're holding, what their genetics have specialized them toward. The same "interact" action might mean graze, hunt, share food, court a mate, or teach a juvenile, depending on the creature's evolved drives and the situation it finds itself in. Herbivores, carnivores, scavengers, and social species all emerge from the same brain architecture and the same action set.
You play as Intervention. You walk the planet's surface, sculpt terrain, paint biomes, feed creatures, and watch evolution respond to the world you shape. No loss function. No reward engineering. Just a living world that adapts to whatever you build.