Organism state
Each organism owns position and velocity, heading, energy, age, generation, an eight-trait genome, stamina, alert state and short-lived feeding, capture and death timers.
Truuts experiment
Neverending
A live artificial ecosystem where individual organisms inherit eight behavioral traits, make local decisions and evolve under changing resource and environmental pressure.
CREATURES0
GENERATION0
FOOD0
WORLD AGE0d 00h 00m 00s
BIRTHS0
DEATHS0
CURRENT EVENTCONNECTING TO WORLD…
NEXT EVENTCALCULATING NEXT EVENT…IN —:—
REAL-TIME POPULATION DNA · LIVE AVERAGE · GENERATION 0
LIVE MUTATIONAWAITING WORLD STATE…
SPD0.90
MET1.00
SOC0.20
AGR0.12
PIG0.98
WND0.40
CAN0.00
PER1.00
Genome and mutation
- Speed / SPDThe maximum movement velocity. Fast creatures reach distant food sooner but may separate from resource-rich clusters.
- Metabolism / METThe rate at which stored energy is consumed. Lower metabolism supports longevity; environmental heat can multiply its cost.
- Sociality / SOCThe tendency to move toward nearby creatures. Positive values create clusters; negative values favor solitary exploration.
- Aggression / AGRA heritable competitive tendency distinct from diet. Aggression contributes to lineage behavior and color, but does not determine its food source.
- Pigmentation / PIGAn inherited visual marker rendered directly into the pixel sprite, making related lineages easier to recognize.
- Wander / WNDThe amount of random change applied to a creature's heading. High wander explores unpredictably; low wander moves consistently.
- Cannibalism / CANA heritable value from 0 to 1. Above the hunting threshold, a creature becomes redder, ignores apples and hunts other creatures, favoring isolated or weakened prey.
- Perception / PERThe radius for detecting food sources, nearby organisms and danger. Greater perception supports earlier decisions but slightly increases continuous energy use.
- Inheritance and mutationChildren copy a parent genome with bounded variation. Genetic-instability events temporarily triple mutation strength.
- Population genomeThe DNA panel reports population averages, not one selected creature. Live Mutation identifies the trait changing most between samples.
Quick facts
- Creator
- Eugene Truuts
- Project type
- Artificial life and evolution simulation
- Core model
- Inherited eight-trait genomes with bounded mutation
- Initial world
- 34 organisms, shared resources and 15 weighted events
- Runtime
- Browser · React · HTML Canvas
- Status
- Active experiment · Updated August 2026
About the experiment
Neverending is a continuously running artificial ecosystem. Thirty-four creatures begin with different bodies and eight-trait behavioral genomes, then move through the same rectangular habitat looking for food, forming loose groups and reacting to nearby life.
Speed, metabolism, sociality, aggression, pigmentation, wander, cannibalism and perception vary between organisms. Together they shape how quickly a creature moves, how much energy it spends, what it notices and whether it explores alone, joins a cluster or adopts a predatory feeding strategy.
Every action has an energy cost. Movement consumes energy, sprinting consumes finite stamina and wider perception carries a small metabolic cost. A creature that gathers enough energy and survives long enough may reproduce; its child inherits all eight traits with small mutations. Environmental events keep changing which combinations survive long enough to spread.
There is no score, player character or ideal species. The interesting part is the changing balance between efficient foragers, social clusters, solitary lineages, resource scarcity, predation and sudden world events. Leave the page open and the population becomes its own history.
How it works
The simulation advances continuously while it is visible. Each creature makes local decisions from information available around it rather than from a global plan.
- 1. WanderA genome-controlled heading introduces exploration and individual movement patterns.
- 2. SensePerception determines how far a creature can detect food, nearby organisms and danger.
- 3. Respond locallySocial attraction, separation, resource seeking and warning signals combine into a new steering direction.
- 4. MoveGenome and current behavior determine speed. Faster responses consume finite stamina, preventing endless sprinting.
- 5. ConsumeContact with a valid food source starts a timed feeding action and transfers energy. Depending on its genome, that source can be fruit or another organism.
- 6. ReproduceOlder, high-energy creatures split their stored energy and create a descendant with bounded mutations across all eight genes.
- 7. DieStarvation, age, predation, epidemics and catastrophic events remove creatures from the ecosystem.
Engineering notes
The ecosystem is an agent-based model with an authoritative world state. The renderer receives read-only snapshots; it does not decide outcomes.
- World state
- Local perception
- Steering decision
- Movement & contact
- Energy & environment
- Birth or death
- Population snapshot
- Next simulation tick
Decision model
On each update, an organism samples its local neighborhood and combines wandering, food or prey attraction, separation, social grouping, danger alerts and world forces into a steering direction. There is no population-wide action planner.
Lifecycle and inheritance
Movement and metabolism spend energy; food or prey restore it. Eligible organisms reproduce probabilistically, transferring part of their energy to a nearby child. All eight traits are copied with independently bounded variation, and death removes exhausted or over-age agents.
Emergence and measurement
Clusters, feeding patterns and lineage shifts are consequences of repeated local interactions. Statistics are calculated from living organisms: the population genome is the arithmetic mean of each trait, generation is the highest surviving generation, and Live Mutation tracks the average trait changing most between samples.
Runtime model
The standalone world advances on a 33 ms server interval and publishes snapshots every 100 ms. Atomic disk snapshots preserve its history. The API deployment uses the same engine, catches elapsed time up in bounded steps, and coordinates persistent state with a short-lived lock.
Rendering boundary
The browser polls read-only state every 100 ms, rejects stale snapshots and interpolates entity positions on animation frames. Simulation timing therefore remains separate from display refresh and temporary network jitter.
- Genome
- 8 traits
- World
- 960 × 540
- Initial state
- 34 agents · 72 food
- Population ceiling
- 160 agents
- World events
- 15 types
- State delivery
- 100 ms
World event system
The backend chooses the next event in advance and exposes its name and countdown above the genome panel. When the timer reaches zero, the selected event changes the shared world for every visitor. Selection is weighted: common environmental changes occur more often, while history-changing catastrophes remain rare.
- Temperature phase shiftChanges the energy cost of metabolism to 70% or 140% of normal, favoring different survival strategies.
- Gravitational driftAdds a directional force to every organism, reshaping routes, clusters and access to food.
- Fertile zoneCreates a resource-rich area where new food is concentrated, attracting organisms into local competition.
- Environmental turbulenceMultiplies heading noise, making movement less predictable and weakening established routes.
- Soil depletionPermanently reduces the global food regeneration rate by 20%, increasing long-term scarcity.
- Meteorite impactRemoves organisms and food inside a randomly positioned impact radius, leaving survivors outside the zone.
- Genetic instabilityTriples mutation strength during reproduction, accelerating divergence between descendants and parents.
- Accelerated agingLowers maximum organism lifespan from 4,800 to 1,800 simulation-age units.
- Toxic contaminationGives consumed food a 35% chance to remove energy instead of restoring it.
- Metabolic contagionPeriodically damages vulnerable organisms whose stored energy has fallen below 60.
- Universal constant shiftPermanently changes either baseline metabolism or the nutritional value of all future food.
- Mass extinction eventReduces the existing population to one or two randomly selected survivors; recovery begins from their genomes.
- The system is being observedReduces ongoing energy loss to 70%, briefly making survival easier while observation is active.
- Temporal memory resetResets world age, birth and death counters without deleting organisms, food or inherited genomes.
- World warIntroduces up to 30 new organisms, rapidly mixing genes and disrupting the established ecological balance.
Temporary vs permanent: temporary pressure ends when its timer expires. Permanent changes remain encoded in the shared snapshot and survive page reloads and deployments. “Permanent history” means the event itself ends immediately, but the population it removed or introduced is not restored.
Neverending FAQ
What is artificial life?
Artificial life is the study of life-like processes created with computation. Instead of scripting every outcome, a simulation defines organisms, inheritance, resources and environmental rules, then observes the behavior that emerges.
Is this the same as a biological evolution model?
No. This is an interactive conceptual simulation, not a scientifically calibrated population-genetics model. It demonstrates inheritance, mutation, selection pressure and emergence in an accessible visual system.
How do the digital organisms evolve?
Successful creatures reproduce after reaching an energy and age threshold. Their descendants inherit eight bounded traits that influence movement, energy use, grouping, perception, feeding strategy and survival.
How do traits change behavior?
Traits do not prescribe a fixed script. They adjust costs, sensing ranges and steering tendencies used in each local decision. Small inherited differences can therefore accumulate into visibly different lineages over many generations.
Why does the shared world keep changing?
Movement, feeding, social behavior, reproduction and weighted world events are probabilistic. The server stores one shared history, so small differences compound across generations and remain after visitors leave or the site is deployed again.