AI for Materials Discovery
via Quantum Game Theory

The Problem: Prediction Outpaced Validation

It can take 20 years and hundreds of millions of dollars to bring a new quantum material from lab to market. Google's GNoME (Graph Networks for Materials Exploration) has predicted 2.2 million new crystal structures. But the bottleneck has shifted — from prediction to validation and synthesis. Current discovery pipelines are fragmented, manual, and lack the infrastructure to handle the sheer volume of AI-generated candidates.

Meanwhile, classical optimization forces trade-offs — improve one property, sacrifice another. And even when bulk crystals deliver all three properties, transitioning to thin films or nanostructures for scaling degrades the very properties you need. The field faces "pick two" at every level.

But trade-offs aren't physics limits — they're mathematical assumptions.

The Insight: Quantum Math Without Qubits

COGNISYN uses quantum game theory on classical compute — the same principle as simulated annealing, which borrows metallurgy's mathematics without molten metal. Full quantum mechanical structures run on classical hardware:

The one thing we leave out is the Born rule — no random numbers, no wavefunction collapse. Selection is argmax over amplitudes. The Hilbert space doesn't know what substrate implements it.

The Solution: Where Cooperation Is the Ground State

COGNISYN turns static crystal predictions into a dynamic, production-ready pipeline — the connective tissue between deep-learning models like GNoME and downstream industrial applications.

The Care Operator (C_λ) reshapes the energy landscape so cooperation is the ground state — not something agents negotiate, but something the mathematics produces. It is embedded directly in the Hamiltonian. Four components — energy-directed effort, homeostatic regulation, support, and goal alignment (which measures synergy across the first three) — map onto host quality, coherence, optical transparency, and multi-property synergy.

Imbalanced materials are high-energy states. Only compounds where all three properties cooperate reach the ground state. When all three score above threshold simultaneously, a Care equilibrium emerges beyond the Pareto frontier. Not a better trade-off — no trade-off at all.

Why Agents?

Screening a known database is the easy problem. These compounds have been in the Materials Project all along — they'd never been evaluated through this lens. The hard problem is what happens when the database runs out.

COGNISYN agents don't just screen — they hypothesize, discover, and remember. The LLM proposes hypotheses freely, exploring correlations across complex multi-dimensional data with a fluency humans can't match. But the LLM never computes the answer. The Hamiltonian returns real eigenvalues. Creativity is unconstrained. Results are mathematically constrained.

Three AI Agents. One Grammar. Real Physics.

Each agent writes rules in a compositional grammar called Baba is Quantum — where tokens ARE mathematical operations:

Each rule triggers real Hamiltonian computation across 7 mathematical frameworks. Creativity lives in the LLM. Truth lives in the eigenvalue. Agents don't communicate with each other — H_care IS the coordination.

The grammar grows with every discovery. Previously discovered rules route instantly — checked before built-in vocabulary or exploration. The platform gets smarter with every compound evaluated.

Click to view live proof of concept app

First Application: Quantum Computing Materials Discovery LIVE

AI designing the substrate for its own future

Finding Yb-171 host crystals where host quality, optical properties, and coherence are ALL high — a problem where materials scientists are stuck with sub-optimal trade-offs.

Why Host Materials Matter: The Full Stack

Trapped ion quantum computing is a 7-layer engineering challenge. At every layer, three properties compete. COGNISYN starts at the foundation.

Each layer has the same "pick two" problem. Same engine, new database adapter. The grammar compounds upward through the stack.

Same Mathematics. Any Domain.

Anywhere "pick two" is accepted as inevitable is a COGNISYN opportunity:

COGNISYN makes multi-objective optimization accessible without needing to understand Hamiltonians.

Technical Infrastructure