At Matterspace, we believe AI should advance bits to master atoms by turning computational intelligence into real-world progress
At Matterspace, we believe AI should advance bits to master atoms by turning computational intelligence into real-world progress. Founded by Faruk Guney, MatterSpace brings together over 20 years of engineering experience and years of independent AI research. We're building two distinct, complementary product lines: adaptive intelligence for agents and specialized discovery engines for science and engineering. Duplex is our standalone agent intelligence platform, powered by proprietary Adaptive Continual Intelligence (ACI), inspired by Complementary Learning Systems Theory. It combines runtime learning, memory, decision-making, planning, reusable skills, and governed execution without requiring retraining of the underlying language model. Duplex turns verified experience into updates to learned state, decision policies, and procedures. RAG, knowledge graphs, caching, and written knowledge can support this architecture as tools. Our focus is on how experience improves an agent's future decisions and actions, with the ambition of enabling verified recursive improvement. MS Lattice is a separate product powered by our proprietary Adaptive Learning, Search and Optimization (ALSO) framework. It combines adaptive search, physics-based evaluation, and structured validation for materials discovery and inverse design in atomic-scale spaces. Its four modes are Open Discovery, Prototype-Guided Optimization, Guided Rediscovery, and Blind Rediscovery Benchmark. Duplex develops the agent's working methods; Lattice searches for materials with desired properties. They can work together while retaining distinct roles and approaches. Lattice is our first discovery engine, with more planned for engineering and scientific research. Our mission is to help companies, engineers, and researchers turn computational possibilities into practical outcomes.