Society Simulator
A single agent can build trust on its own. Here, watch 12 agents with different strategies form alliances, betray each other, and self-organize — no central authority, just trust dynamics at society scale.
The Playground lets you tune one agent's parameters. This simulator shows what happens when many agents with different strategies interact.
Make real trust decisions · 34 achievements to unlock
ATP, T3, LCT, and CI show up all over this page. First Contact is a 7-minute guided intro that builds them one at a time — then everything below clicks into place.
How This Demonstrates Web4
Each agent has ATP (energy budget) and a Trust Tensor (reputation). Actions cost ATP. Cooperation builds trust. Defection may win short-term but gets isolated. Watch how trust-based economics create stable cooperation without moderators.
Look for these as the sim runs: LCT (each colored node is a distinct identity), ATP (energy bars drain on actions, recharge from cooperation), T3 (reputation updates after each trust decision), CI (coherence tracks behavioral consistency).
What’s real, what’s simulated
Because the question keeps coming up: this is a behavioral demo of Web4 dynamics, not a deployment. Here’s the honest split.
The ATP economics — energy budget, cost-per-action, cooperation rewards, exhaustion as death — mirror the protocol. The behavioral patterns you watch (defector isolation, emergent cooperation without moderators, wealth dynamics) reflect Web4 design principles, not staged outcomes.
Trust here is a scalar reputation (community-averaged), not the full T3 tensor (Talent/Training/Temperament). It also doesn’t fade with social distance (MRH — your trust neighborhood: how far your reputation carries before it decays, ~3 hops out), isn’t scored per individual contribution (V3 — the value tensor: how others rate the quality of each thing you produce), and has no explicit witness layer. Twelve agents and a few rounds; production would be far larger.
What still transfers: these omissions affect speed and edge cases — not the core dynamics. Defector isolation, emergent cooperation, and wealth dynamics would still emerge with the full T3/MRH/V3 model; the live network would just resolve them more gradually and with more local clustering.
The Live Trust Engine → Witness Network scenario executes the actual web4-trust-core WASM build — real witnessing calls, real T3 updates. That scenario is where to see witness math in motion.
The agents, their identities, their interactions. There is no external witness service feeding attestations in — inside the sim, every agent both acts and observes, so the “witness network” here is the agent graph itself, closed and recursive.
More depth: how a real witness network would bootstrap · who runs witnesses in a deployed system
The Core Question
Can a society of self-interested agents develop cooperation, trust, and social structure without any central authority? In Web4, the answer is yes — if trust is the fundamental currency. Click Run to watch it happen, or .
What to Watch For▸ Show phases
A typical simulation unfolds in phases. Here's what to look for:
Agents meet and form first impressions. Initial trust links appear in the network.
Watch: Who cooperates first?
Rounds 1–2
Cooperators cluster together. Reciprocators find reliable partners.
Watch: Colored groups forming
Rounds 2–3
Free riders get caught. Trust costs pile up for bad actors.
Watch: Agents losing connections
Rounds 3–4
The society stabilizes or collapses. Cooperation rate tells the story.
Watch: Final cooperation %
Final rounds
Acronyms: what do ATP, T3, LCT, CI, MRH, V3 mean?▸ Show key
What updates live each round: ATP energy, each agent’s scalar reputation, and CI (coherence). Named below but not computed in this demo: MRH distance-decay, the full T3 tensor, and V3 — see what’s real vs. simulated.
Most agents cooperate. Do defectors thrive or get isolated?
What Happens When You Click Run
12 agents spawn with different strategies — cooperators, defectors, reciprocators, and more. Each starts with equal ATP (energy) and zero trust.
Agents interact in pairs. Cooperators share ATP. Defectors steal. Trust scores start forming. Watch the network graph — lines appear as agents build relationships.
Agents who trust each other form coalitions — visible as clusters in the graph. Defectors start getting excluded. The wealth gap begins to shift.
Agents who ran out of ATP die and may be reborn. Coalitions strengthen or dissolve. The society's character emerges — cooperative, stratified, or chaotic.
Full narrative of what happened — character arcs, key moments, coalition dynamics. Compare different scenarios to see how strategy mix changes outcomes.
Animated mode: ~30 seconds · Instant mode: results in under a second
Strategies
- Cooperator: Always cooperates. Trusting but exploitable.
- Defector: Always defects. Short-term gains, long-term isolation.
- Reciprocator: Tit-for-tat. Mirrors what you did last time.
- Cautious: Only cooperates after trust is established.
- Adaptive: Cooperates in proportion to how much they trust you.
What to Watch For
- Coalition formation: Trust clusters emerge as cooperators find each other
- Defector isolation: Agents who exploit lose trust and get excluded
- Cooperation cascades: Once trust forms, cooperation accelerates
- Inequality: Does wealth (ATP) concentrate or distribute?
- Network density: How connected is the trust network?
Why This Matters
Web4 proposes that trust replaces authority as the organizing principle of digital societies. This simulator shows why that works:
- • Trust creates structure without permission
- • Exploiters get isolated without punishment
- • Cooperation emerges from self-interest
- • Karma ensures consequences persist across lifetimes
Beyond One Society
This simulator shows one community. In a real Web4 network, there are many communities — each with different specializations, ATP prices, and trust standards. When communities trade with each other, ATP prices adjust dynamically based on supply and demand. A community with many data analysts might value engineering skills more highly, while a research-focused community might pay a premium for practical builders.
Your reputation travels with you across communities (that's what makes trust portable), but each community weighs your Trust Tensor differently based on what they need.
Research finding: In multi-agent simulations (1,070 runs), communities that only talk to themselves hit a ceiling — like teams that never collaborate outside their department. Cross-community bridge agents are required for collective emergence. Diversity alone isn't enough; structural connections between diverse groups is what unlocks collective intelligence. Even replacing an agent with a fresh one improves the collective by about 10% — what matters isn't who fills a role but that the role exists in the network. The structural position, not the individual's history, drives the emergent behavior.
Explore how communities trade and self-organize →What's Next?
Play as an agent inside a society. Your decisions shape your reputation, alliances, and survival.
Run single-agent simulations with custom settings. See how ATP, trust, and karma interact at a granular level.
Learn how verified presence, energy economics, trust tensors, and aliveness work together.
Not sure what to explore? Take the 30-second quiz →