White Paper
Huge Dynamics
A Foundation Model for the Movement of Big Things
Big things move. Commodities, logistics flows, mass populations. When they do, the consequences ripple through financial markets, supply chains, and geopolitical calculations. Yet today, insights into these massive flows remain largely opaque to business decision-makers and policy makers—scattered across siloed datasets, multi-modal formats, and the tacit "know-how" of human supply chain experts.
Our vision is to build a physical-economic foundation model, Huge Dynamics, that models, reasons about, and helps understand the flow of big things. Grounded in operational, economic, and physical realities, Huge Dynamics draws from a wealth of disciplines: operations research (supply chain modeling, inventory dynamics), economics (market dynamics, price elasticities), physics (climate impacts, flow mechanics), and statistical tools (regression, ML predictions). It fuses the reasoning depth and flexibility of state-of-the-art LLMs with the causal rigor and precision of these formal mathematical models to turn scattered data into actionable intelligence.
The Long Term Mission
The long term mission to build the intelligence substrate of the physical economy. Huge Dynamics will become the foundational intelligence layer for the physical economy, commodities, energy, food, materials, and people, transforming opaque, high-stakes flows into transparent, actionable knowledge for enterprises, governments, and humanity at large.
Just as today's language foundation models have become indispensable for knowledge work, Huge Dynamics can become the universal substrate for decisions that move trillions of dollars and affect billions of lives. It could power real-time global supply-chain orchestration, preempt resource conflicts before they escalate, optimize energy transitions at planetary scale. These capabilities will enable policymakers to simulate and steer macroeconomic outcomes with unprecedented fidelity.
Fusing AI with Mechanistic Models
We currently face a bifurcation in world simulation and artificial intelligence.
Formal Models (Deductive modeling)
On the one hand, operations Research (OR), physics, and economic models provide rigorous, causal truth (e.g., mass balance, queueing theory, price elasticity) but are brittle, siloed, and blind to unstructured data.
AI Models (Inductive modeling)
On the other hand, state-of-the-art multimodal models have immense reasoning flexibility but suffer from hallucinations; they lack a "grounding" in physical and economic laws.
Huge Dynamics operates by fusing two distinct cognitive cores:
- The Formal Core:A library of structural models that enforce reality.
- The Reasoning Core:A transformer-based architecture that processes unstructured input (geopolitical news, satellite imagery, CEO statements) to understand intent and context.
When asked a question, the AI does not merely "guess" the next token. It formulates a hypothesis, parameterizes a formal simulation (e.g., running a stochastic optimization on aluminum flows), interacts with the Formal Core, and synthesizes the result into actionable strategy.
Key Capabilities
Prediction
Causal forecasts with reasoning, e.g., "Aluminum prices may spike 15% next month due to European energy shortages diverting supply to data centers, cascading to higher Canadian exports—based on elasticity models and real-time data."
Generative
Sample plausible scenarios for analysis, drawing from stochastic simulations to explore "what-ifs" in commodity flows, like simulating supply disruptions from Red Sea conflicts.
Recommendation
Multimodal optimization via AI-tool integration—decompose queries (search, reasoning, optimization), then synthesize, e.g., "Build the next refinery in Quebec to minimize costs (per min-cost flow) amid climate risks (via meteorology models), with key threats like geopolitical tariffs."