+ AN OPEN-ENDED SYSTEM
Small decisions. Complex worlds.
Observe the system. Change a condition. Follow the consequences.
SIMULATION TIME
0000 ticks
PAUSED01
Your first observation starts here.
Press Run, then select an agent on the map. Save a checkpoint in Experiments to compare the same world under different conditions.
LIVE WORLD
The commons / 0042
32 × 22 CELLSSEASONAL ECOLOGY
Food-skilled Wood-skilled Water barrier
ENVIRONMENTAL STOCK—
SPATIAL GROUPS—
FOOD / WOOD PRICE—
INVENTORY INEQUALITY—
Historical signals
Measured at every tick · last 2,000 ticks retainedRun the simulation to build a history.
Field notes
Recorded transitions, not generated narrativesCounterfactual bench
CONTROLLED COMPARISONOne checkpoint. Two possible futures.
Clone the complete world, change one condition, and advance both branches for the same number of ticks. The comparison does not advance or replace the live world.
1
Establish a baseline
No checkpoint yet. Capture the current world to begin.
2
Choose an intervention
3
Observe the response
Branch comparison
Equal starting state. Equal elapsed time.A controlled question is a useful question.
Capture a checkpoint, choose a condition, and compare the consequences. Differences are measurements from this model, not evidence about real civilizations.
Baseline Intervention
A single paired run is descriptive. Random streams can diverge after the intervention changes actions. Repeat with different seeds before generalizing; use the repository's experiment runner for repeated trials.
Scientific guardrails
READ THE MODEL, THEN TEST ITWhat we designed
Local utility decisions, resource regeneration, barter, trust, reproduction and soil damage are explicit rules. Observing these actions does not establish emergence.
What we can observe
Population, trade, spatial concentrations, inequality and specialization are computed from actual state. A cluster is a spatial statistic, not a political institution.
What needs evidence
Feedback and collective patterns need baselines, repeated seeds and alternative explanations. Model behavior is not a prediction of human society.
Executed research
REPOSITORY EVIDENCEA model you can inspect
GENESIS / 1.0Collective behavior starts with consequential local choices.
Agents can see nearby opportunities, hold food and wood, harvest, make tools, move, and exchange goods. Different skills and inventory needs create different incentives. Every unit of food and wood has a source, an owner, or a recorded sink.
The causal loop
Local resources→Individual decisions→Harvest & exchange→Soil & population↩
Designed mechanisms
- Ecology
- Seasonal regeneration replenishes spatial stocks. Extraction damages soil; soil recovers slowly. Water cells are impassable.
- Economy
- Harvest moves stock from land to inventory. Barter exchanges owned goods. Food supports survival; wood supplies upkeep and tools. Tools improve extraction and depreciate.
- Demography
- Health responds to consumption. Resource-funded births add agents; starvation and aging remove them. Population is reconciled against births and deaths.
- Social interaction
- Successful local exchanges update bounded trust. Reciprocal aid can transfer food to a neighbor in need. These are programmed mechanisms, not discovered institutions.
- Movement
- Agents evaluate nearby productive opportunities. Moving consumes food, and impassable water changes accessible routes.
Reproducibility contract
The seed, random-number state, scheduler-relevant state, inventories, environment, agents and active interventions travel with every save. The same engine version and conditions reproduce the same outcomes. Floating-point identity across different JavaScript engines is not promised.
Scope and limits
This is an abstract computational model. Agents have no language, beliefs, political institutions, warfare or general intelligence. A tick is a model step, not a calibrated day. Spatial groups have no shared decision process. Parameter choices are assumptions, and scientific claims require controlled experiments.
Inspect the model equations, module contract, and reproduction instructions in the repository.