Foundational AI
WORLD DYNAMICS MODEL
A new class of AI models for real-world systems that learn state and dynamics, predict consequences and decide.
Why now
AI IS LEARNING TO ACT.
IT NOW NEEDS TO MODEL CONSEQUENCES.
When AI acts within real systems, its outputs become interventions. Each action changes the system’s state, constraints and dependencies, shaping what can happen next.
REAL SYSTEMS CARRY HISTORY FORWARD
Software, infrastructure, organizations and physical systems do not reset between prompts. Their current state reflects past events and actions. Every new intervention becomes part of that history.
ACTIONS CREATE DIFFERENT FUTURES
Making a decision requires more than understanding the current state. It requires anticipating how the system could evolve under different actions.
DIFFERENT ACTIONS. DIFFERENT POSSIBLE FUTURES.
The thesis
STATES CHANGE.
UNDERLYING MECHANISMS PERSIST.
Learn how systems evolve across states: DYNAMICS.
The technology
FROM STATES TO DYNAMICS.
MODEL WHAT HAPPENS NEXT.
Understanding a system requires more than a snapshot of its current state. ThreeDots is developing a World Dynamics Model that learns how systems evolve in response to actions and events. It maintains a persistent representation of system state, learns transition dynamics and incorporates new evidence as conditions change.
Maintain a structured representation of the system’s entities, relationships and current state.
Learn how actions and events transform the system from one state to another.
Incorporate new evidence as the environment evolves, without rebuilding the representation from scratch.
The team
Built by people 
who understand complex systems from the inside.
ThreeDots brings operational experience to foundational AI research. We’re bringing together researchers, operators and domain experts to develop the World Dynamics Model and test it against the complexity of real-world systems.
Interested in advising or partnering with us?
Talk to us
