MoralWorld Live
What happens to 178 countries when AI arrives. Choose an AI path, policies and assumptions; watch the world evolve quarter by quarter; see the odds of 28 endings.
178 countriesCARI-drivenOdds of 28 endingsLead a nationBacktested
Lab
Tools and methods for the AI transition.
Everything here serves the two questions: what could go wrong, and what AI does to us. Three tools are live; the next ones are below.
CARI Live
The Civilizational AI Risk Index: where AI risk stands today, dimension by dimension, with anchors and history.
Six dimensionsAnchored answersHistory since 2019Daily
Consensus Room Live
Several AI models argue the same question in rounds, vote, and produce a consensus report with the minority view kept: for research questions where one model answer is too fragile.
Round-based debateModel comparisonMinority reportsConsensus report
Methods
How the lab thinks.
When every model can produce a fluent answer, the scarce thing is method: how questions are framed, how uncertainty is structured, and how disagreement is kept honest.
Scenario discipline
Turn uncertainty into structured paths: drivers, shocks, transmission channels, adaptive responses, early indicators and decision points. Never one story; always a distribution.
Structured disagreement
Let different models, schools of thought and adversarial roles challenge each other before any synthesis, and keep the minority report.
Thought experiments
Test ideas we cannot yet measure through counterfactuals, constraints and moral trade-offs: what changes if one assumption is reversed?
Long-horizon reasoning
Reason about institutions, technology absorption and social resilience over decades, where most of the consequences of AI will play out.
Next
What we are building.
Ideas at different stages. Each starts as a method, becomes a tool, and is published openly.
Concept
Scenario Atlas
Run the published scenarios. Each major AI scenario (AI 2027, the 2028 crisis, AI 2040, the gradual-disempowerment story) translated into a MoralWorld setup, so anyone can run them side by side, change one assumption, and see which conclusions survive.
In development
Human Agency Monitor
A companion to CARI. CARI measures AI. This measures us: how much human work, decision-making and attention is being handed to AI systems, country by country. The gradual-disempowerment question, made measurable.
In development
Thought Experiment Studio
The human question, debated. Questions with no data yet (a world where no one needs to work, AI companions, machine rights, what remains distinctively human) explored through roles, counterfactuals and multi-model debate, with the disagreements kept on the record.
In MoralWorld desktop
Organisation Lab
AI inside a single institution. How AI changes one firm, bank, hospital or ministry: who adapts, who resists, who stays silent, and how that adds up to what happens to a country.
Concept
Daily Transition Brief
One page a day. What moved in CARI today, what it changes in the simulated world, and the signals worth watching, written for people who have five minutes.