Nationz.AI · Issue 04 · April 2026
Decision simulation for people who have to explain themselves afterwards
ATIONZ.AI.10,000 AI agents. One decision. Zero guesswork.
.png)
.png)
.png)
.png)
Every agent carries its own memory, beliefs, income, and social network. Drop a policy — watch opinions shift, stakeholder groups form, and chain reactions cascade through the economy. Powered by real mathematical models, not vibes.
Every agent carries its own memory, beliefs, income, and social network. Drop a policy — watch opinions shift, stakeholder groups form, and chain reactions cascade through the economy. Powered by real mathematical models, not vibes.
10,000
Agents with persistent memory
7
Mathematical models
5
Chain reaction waves
nationz.ai / studio / united-arab-emirates · population 11,200
● LIVE SIM
Agent response · Wave 03
Fuel subsidy cut
Support
38.4%
Support · 38%
Neutral · 34%
Oppose · 28%
Stakeholder mapping
Gig drivers
142
Retired teachers
96
Suburban parents
74
Propaganda risk
0.71/1.0
High distortability on paragraph 3. Two tabloids aligned.
Policy simulation
Propaganda risk
Stakeholder mapping
Chain reactions
Agent memory
Country studio
Decision wake
Field notes
Mathematical models
Independent memory
Policy simulation
Propaganda risk
Stakeholder mapping
Chain reactions
Agent memory
Country studio
Decision wake
Field notes
Mathematical models
Independent memory
Policy simulation
Propaganda risk
Stakeholder mapping
Chain reactions
Agent memory
Country studio
Decision wake
Field notes
Mathematical models
Independent memory
The population
Simulate millions of people. Every one of them different.
Drag to rotate. Each node is one agent — own name, profession, beliefs, and memory. The lines are who reacts to whom when a decision lands. Showing four thousand here; the engine scales to millions.
The pipeline
From blank page to report in nine autonomous phases.
PHASE 01
Build Country
AI agent ingests articles, reports, and raw data. Outputs a structured dossier with demographics, sectors, media ecology, and social cleavages.
AI Agent
Demographics
Media Ecology
Cleavages
PHASE 02
Generate 10,000 Agents
Each agent gets a name, age, profession, income, religion, ethnicity, personality traits, political leaning, and persistent memory across simulations.
AI Agent
Independent Memory
Social Graph
Personality Traits
PHASE 03
Drop a Decision
Paste a policy, government memo, or press release. The engine parses intent and delivers it to all 10,000 agents.
Policy
Press Release
Government Memo
PHASE 04
Individual Reactions
Every agent evaluates the decision independently. A multinomial logit model weighs personal impact, economic exposure, identity alignment, and trust in institutions.
P(k) = exp(Uₖ) / Σⱼ exp(Uⱼ)
Multinomial Logit
AI Agent
Support
Oppose
Neutral
Conditional
PHASE 05
Stakeholder Mapping
Agents with shared interests self-organize into coherent stakeholder groups. Influencers, politicians, religious leaders, and business owners emerge as group heads via agglomerative clustering.
influence = (income + intensity + size + cohesion) / N
Influencers
Politicians
Business Leaders
Religious Leaders
Clustering
PHASE 06
Opinion Shifts
Stakeholder groups deploy representatives to persuade. Peer discussions shift opinions pairwise. Three mathematical models run in parallel — echo chambers and consensus emerge.
xᵢ(t+1) = λᵢ · xᵢ(0) + (1 − λᵢ) · Σ wᵢⱼ · xⱼ(t)
Friedkin-Johnsen
Hegselmann-Krause
Deffuant
AI Iteration
PHASE 07
Chain Reactions
Economic shocks cascade through a 10-sector Leontief input-output matrix. Five waves: employer impact, sector ripple, society-level effects, feedback loops, equilibrium.
Δx = (I − A)⁻¹ · Δd
Leontief I-O
10-Sector Economy
5 Waves
Bandwagon Effect
PHASE 08
Media & Propaganda
AI agent simulates tabloid and broadsheet coverage. Identifies which sentence gets lifted, which amplifier carries it, and scores propaganda distortability 0–1.
AI Agent
Tabloids
Broadsheets
Distortability Scan
Risk 0–1
PHASE 09
Strategic Report
AI writes a publishable memo with approval predictions, coalition risk analysis, stakeholder maps, and strategic recommendations — in the register of a government brief.
AI Agent
Coalition Risk
Stakeholder Map
Approval Rating
◉
Opinion Shift
Track how each agent's stance moves across simulation waves — from initial reaction to final equilibrium.
⚠
Risk Measure
Propaganda distortability, coalition fragility, and approval volatility scored 0–1 per wave.
◈
AI Chatbot
Ask follow-up questions in natural language. The chatbot has full context of every agent and every wave.
◆
Campaigning
AI-generated messaging strategies for each stakeholder group — what to say, to whom, through which channel.
⟐
Iteration Comparison
Run the same decision with different framings. Side-by-side diff of approval, stakeholder groups, and risk.
The trinity
A government. A country. Ten thousand people.
Every simulation starts with three things — the decision-maker who signs the policy, the country it lands in, and the population it lands on. Get the first two right and the third writes itself.
01 · Government
The decision-maker
Ministries, regulators, the people who sign things. The simulator ingests their stated intent and tests how it survives contact with reality.
02 · Country
The context
Demographics, sectors, media ecology, cleavages — in the local grammar. A policy in Oslo will not land the same in Lagos.
.png)
03 · People
The population
Up to ten thousand agents with professions, beliefs, economic ties, and a memory that persists. They react first; they argue with each other after.
§ 01 · Manifesto
For twenty years the tools for thinking about consequential decisions haven't changed. A room, a deck, a best guess. We built something closer to a wind tunnel— a place to watch the decision land before it actually does.
Aggregate stats hide the friction that actually decides outcomes. Networks are louder than demographics. Context is everything — a policy in Oslo will not land the same in Lagos. The simulator has to know the country. So we made Country Studio, and Country Studio is where most of our users start.
§ 02 · What's inside
Six primitives. One honest answer.
— 01
Country Studio
Paste articles, upload an image, add a handful of facts. A structured dossier — demographics, cleavages, media ecology — drops out the other side.
— 02
Agent populations
Between 50 and 10,000 agents with professions, beliefs, economic ties, and a memory that persists across simulations.
— 03
Social dynamics
Watch stakeholder groups form. Watch opinions shift. Chain reactions move through employer, supplier, and professional networks.
— 04
Propaganda risk index
See which sentence will be lifted, which amplifier will carry it, and how fast it spreads before the clarification ships.
— 05
Local currency, local idiom
Every reaction lands in the grammar of the country. The simulator does not round the world into corporate English.
— 06
Reports you can defend
Stakeholder maps, risk surfaces, and strategic recommendations written in the register of a memo, not a pitch deck.
§ 03 · Under the hood
Real mathematics. Not a black box.
Every simulation runs on published, peer-reviewed mathematical frameworks. No hidden prompts, no vague heuristics — seven models working in concert, each with explicit coefficients you can inspect.
OPINION DYNAMICS
Friedkin-Johnsen Consensus
Each agent blends stubbornness with neighbour influence until the population settles.
BOUNDED CONFIDENCE
Hegselmann-Krause
Agents only listen to neighbours within their confidence bound — echo chambers emerge naturally.
PEER EXCHANGE
Deffuant Model
Pairwise conversations shift opinions at convergence rate μ = 0.3.
ECONOMIC CASCADE
Leontief Input-Output
A shock in one sector cascades through the 10-sector economy via the inverse Leontief matrix.
DISCRETE CHOICE
Multinomial Logit
Each agent picks support / oppose / neutral / conditional via calibrated utility coefficients.
STAKEHOLDER INFLUENCE
Persuasion Probability
Multi-factor persuasion model capped at [0.05, 0.75] — nobody is certain, nobody is immune.
§ 04 · On the record
“The first time we ran a tax reform through two thousand agents in Country Studio, a quiet cluster of gig workers self-organised into a stakeholder group before the policymakers in the room had finished reading the summary.”
Lea Gozalis · Co-founder
§ 05 · How it works
From blank page to answer in four moves.
MOVE 01
Shape the world
Pick a country, or build one in Country Studio with your own sources.
MOVE 02
Populate it
Generate a realistic population, or upload your own agent file.
MOVE 03
Drop a decision
Paste a policy, a memo, or a press release — anything with consequences.
MOVE 04
Read the wake
Individual reactions, waves, chain reactions, and an AI-written report.
01
10,000
Agents per simulation
each with persistent memory
02
20+
Countries out of the box
or build your own in Studio
03
< 60s
From decision to first wave
no pipelines, no setup
04
0.71
Propaganda risk, detected early
before the headline lands
§ 06 · Field notes
Writing from inside the simulator.
01
Research
Inside Country Studio: Teaching AI the Texture of a Nation
How we turn a folder of articles, a flag, and a few stubborn facts into a structured dossier the simulator can actually reason with.
Kamil Rayes · 9 min read
02
Research
The Propaganda Risk Index: Why Every Policy Needs One
Not every policy is vulnerable to the same distortion. We built a score that tells you where the story will be twisted first.
Kamil Rayes · 7 min read
03
Engineering
From 50 to 10,000 Agents: What Scale Teaches You About People
Small populations lie politely. Large ones tell the truth. A field report from pushing the simulator past comfortable limits.
Aiman Showkat Khan · 8 min read
§ 07 · The country is waiting
Paste a few sources. Watch a population take shape. Read the wake.
The first simulation takes about sixty seconds. The second will change how you brief every decision after it.