MatterSpace · Universal Generation Engine
Generation, not discovery.
Describe what should exist.
The Universal Generation Engine creates novel candidates — materials, drugs, chips, algorithms, biological interventions — that satisfy physical constraints by construction. Built from first principles for your agents.
Not screening. Not filtering. Not predicting. Your agents generate valid candidates from target specifications. Every output is physically realizable.
Case Studies
What MatterSpace has generated
From blind rediscovery of known catalysts to novel battery materials — real generation results across materials science and beyond.
Blind Rediscovery of SAA Catalysts
Rediscovered Re₁@Ni, Ir₁@Ni catalysts
Discovering optimal single-atom alloy catalysts by brute-force screening is prohibitively expensive — the combinatorial space of dopant elements and host lattices is too large. MatterSpace generated 600 candidates across 23 dopant elements and matched known optimal SAA configurations without any prior knowledge of the targets. Structural match within half an angstrom.
Battery Cathode Discovery
Novel cathode candidates generated
High-performance battery cathodes require specific voltage and capacity profiles, but current approaches to cathode design discard over 90% of candidates that violate stability constraints. MatterSpace generated stable lithium-ion cathode materials with constraint enforcement ensuring thermodynamic stability and ionic conductivity at every step of the generation process.
Green Hydrogen Catalyst Optimization
Catalyst candidates for HER/OER
Green hydrogen requires catalysts that balance overpotential, long-term stability, and earth-abundance — competing objectives that make manual optimization intractable. MatterSpace generated catalyst compositions optimized for HER/OER with multi-objective scoring across all three constraints simultaneously.
Drug Candidate Generation
Early Testing
Virtual screening pipelines evaluate millions of molecules but discard the vast majority for failing drug-likeness, solubility, or ADMET checks applied after generation. MatterSpace generates small molecules that satisfy binding affinity, selectivity, and ADMET constraints simultaneously — physics-grounded navigation of chemical space, not brute-force filtering.
Chip Topology Optimization
Early Testing
Semiconductor layout iterations take weeks when design-rule violations are caught only after placement. MatterSpace generates layouts that satisfy thermal, power, and area constraints by construction — design-rule enforcement during generation eliminates costly post-layout iterations.
Matrix Algorithm Discovery
Early Testing
Finding faster matrix multiplication algorithms has been a decades-long manual effort — the search space is combinatorially vast and correctness constraints are strict. MatterSpace navigates the solution landscape to discover novel algorithms that minimize arithmetic complexity while preserving numerical stability.
How It Works
The generation pipeline
From target specification to validated candidates. Every stage is observable, every constraint enforced during generation.
Adaptive Dynamics Controller
Four physics modes fire in real time based on gradient state and exploration history. Local refinement, stochastic exploration, barrier crossing, and rapid stabilization — orchestrated automatically.
Constraint Enforcement
Physical constraints are enforced during generation, not after. Bond lengths, coordination numbers, symmetry groups, and charge neutrality validated at every step. Every output is valid by construction.
Domain Packs
Field-specific force fields, physical constraints, objective functions, and sampling strategies. The core engine is domain-agnostic — domain packs supply the science for each application.
Candidate Validation
Multi-tier validation pipeline. Fast filters eliminate non-viable candidates. Relaxation confirms local stability. Property prediction scores against objectives. High-fidelity verification on top candidates.
Sub-Engines
One architecture. Five domains.
The core engine is domain-agnostic. Domain packs supply the physics, constraints, and objectives for each field. Lattice is ready. Pharma, Algo, Tessera, and Longevity are available for early testing.
Lattice
Materials & Energy
Crystal structures, alloys, catalysts, battery cathodes, superconductors, photovoltaics. 9 domain packs covering the most important classes of functional materials.
ReadyPharma
Drug Discovery
Molecular generation guided by binding energy landscapes. Constraint-aware synthesis ensures drug-likeness, solubility, and ADMET compliance.
Early TestingTessera
Chip Design
Semiconductor architecture, photonic layout, and circuit topology optimization. Design-rule landscapes produce physically valid, manufacturable configurations.
Early TestingAlgo
Algorithm Discovery
Matrix n-rank algorithm search and computational optimization. Novel algorithmic structures discovered by navigating solution landscapes under complexity and correctness constraints.
Early TestingLongevity
Biological Interventions
Partial epigenetic reprogramming target discovery. Navigates the Yamanaka factor landscape to identify safe, reversible rejuvenation interventions grounded in cellular biology constraints.
Early TestingConstraints in. Candidates out.
MatterSpace Lattice is available now for materials discovery. Talk to us about early access to Pharma, Tessera, Algo, and Longevity.