AI for
_
Unreasonable builds AI that discovers new knowledge and solves the hardest problems across domains, making experts faster and smarter.
Our mission is to build superintelligence for autonomous knowledge creation and discovery. Unreasonable builds AI that harnesses the ever-expanding body of global data to surface novel insights and ideas. By unlocking creativity, boosting research productivity, and accelerating breakthroughs, Unreasonable empowers innovation across scientific and technological frontiers. Our AI is natively multidisciplinary and thinks in deep, cross-cutting abstractions that integrate science, technology, art, and human intuition
Problem
_
Too Much Data, Too Little Insight
AI Incapable of True Invention
Specialized Tools Limit Cross-Domain Discovery
Unreasonable.DISCOVERY
An AI discovery engine that helps you compose new knowledge across disciplines.
How Unreasonable Labs Works
_
Multidisciplinary knowledge organization that connects real-world data from across domains. This transforms fragmented information from literature, simulations, lab data, sensor streams, and enterprise datasets into a connected, multidomain representation of your entire problem landscape.
Unlike early generative AI systems that primarily learn surface-level correlations from text and data, Unreasonable is designed to form deep abstractions - analogous to scientific theories - that generalize across domains. Its unified world model captures mechanisms, constraints, and organizing principles that persist across physics, biology, materials, and markets, enabling transfer, analogy, and prediction beyond the training distribution.
This allows the system to:
Integrate physics, biology, chemistry, materials, and market context
Connect relationships hidden across fragmented sources
Enable cross-domain analogies and mechanism-level insights
Support dynamic, self-updating organizational memory
Our AI performs deep reasoning, hypothesis generation, and insight extraction, supported by the unified reasoning fabrics we built across domains. It creates new knowledge instead of simply retrieving existing facts.
Unreasonable Labs performs deep, multi-step reasoning on top of its unified world model, enabling it to:
Build theory-like abstractions that apply across domains, rather than brittle, domain-specific correlations
Generate novel hypotheses, mechanisms, materials, and designs
Decompose complex problems into smaller scientific workflows
Call tools, simulations, and external models as needed
Iteratively refine insights based on new evidence
Orchestrate agents that behave like teams of scientists and engineers
A collaborative environment where experts guide, critique, and extend AI-driven discoveries. Transparent reasoning traces and interpretable models that let users inspect how insights are formed. This keeps scientists in the loop while letting AI expand the search space beyond human limits.
Unreasonable Labs provides a transparent, collaborative environment where:
Scientists can guide, question, and critique AI-generated insights
Reasoning chains, intermediate steps, and model updates are fully visible
Mechanisms can be validated, modified, or extended by users
Multiple stakeholders work together in shared discovery workspaces
Enterprise data stays secure with sandboxing and robust permissions
An AI-Native Tool Store allows teams to plug in tools, models, simulations, and scientific agents to extend and customize discovery workflows
What this means for you
Straightforward integration of existing tools and workflows
Simple adaptation to new specialized domains
Cross-domain breakthroughs that siloed approaches miss
Realize true creativity across multiple disciplines - science, engineering, art and humanities
Who this is for
_
Unreasonable serves deep-tech R&D teams, scientific research organizations, and corporate innovation groups working on complex, multidisciplinary problems. This allows you to realize true creativity across multiple disciplines - science, engineering, art and humanities. Example use cases:
Academic and government scientists working at the frontiers of fundamental knowledge who seek to uncover new principles, mechanisms, and laws governing complex systems. Unreasonable Labs supports theory formation, cross-scale reasoning, and the discovery of latent structure in data-sparse or open-world regimes where existing models break down.
Industrial R&D and innovation teams in materials, deep tech such as energy, biomedical engineering, biotech, and semiconductors looking to design new materials, processes, and products.
Strategy, innovation, and corporate development groups that need to explore future markets, technologies, and scenarios beyond standard research reports, with integration of complex data sources, models and reasoning strategies.
Scientific organizations and labs seeking an AI collaborator for hypothesis generation, literature synthesis, and experiment design.
Team
_
Former Research Scientist at Google DeepMind, where he worked on advancing the research, development, and real-world deployment of cutting-edge AI systems. His work at Google spanned neural machine translation, state-of-the-art dialogue and multimodal models, and emerging approaches to agentic reasoning, contributing to major efforts across the Gemini model family and their integration into production products such as Google AI Mode. He holds a strong conviction that open-ended knowledge creation is the defining capability required for higher-order machine intelligence, and is dedicated to advancing this frontier - from theoretical foundations to practical, scalable systems - to accelerate scientific discovery and amplify human research capabilities.
Markus J. Buehler is the McAfee Professor of Engineering at MIT and a pioneer in physics-aware artificial intelligence for scientific discovery. His work establishes quantitative links between molecular structure and emergent function across materials and biological systems, integrating first-principles simulation, graph-based modeling, and machine learning. He has advanced graph-native and agent-based frameworks that enable AI systems to reason across scales, compose models, and generate testable hypotheses. His research has led to foundational contributions in bio-inspired materials, multiscale mechanics, and AI-driven design, with broad impact across academia, industry, and government. Buehler is a member of the U.S. National Academy of Engineering, and has received major international honors including the Feynman Prize and the Washington Award. His work bridges science, engineering, and art, reflecting a sustained focus on generative systems that expand the boundaries of discovery.
In his former roles as a Senior Director, Strategic Solutions at C3 AI and a Senior Scientist at SSI, Andrew has both shaped the high-level application of generative AI for Fortune 500 biotech clients and led the deep-level hands-on implementation of AI systems for accelerating government research. During his PhD at MIT, he united AI models, physics-based simulation, and real-world experimentation to close the loop on designing novel material structures with bespoke properties. He is passionate about bridging domains and perspectives to tackle previously impossible problems and driven to harness AI-augmented workflows for expanding human potential.
Member of Technical Staff. Haiqian is a soft-matter physicist developing physics-grounded AI for biological, material, and other complex systems. He received his PhD in 2026 from MIT, where he studied biological self-assembly using machine learning, continuum mechanics, and optical imaging, and published work at the interface of physics, biology, and computation in journals including Nature Methods, Nature Physics, PRX Life, and PNAS. His work is guided by the view that embedding physical structure into learning is essential for building interpretable and robust models for complex systems.
Matt is an experienced software engineer and leader with over 20 years of experience building and scaling products and platforms at companies including Airbnb and Dropbox. He has led engineering teams at multiple companies, specializing in data infrastructure, developer tooling, and full-stack product development. A YC founder and former startup CTO, Matt brings deep expertise in turning complex technical systems into polished, production-ready products. He holds a B.S. in Computer Science from the University of Maryland.
%20(1)%201.png)
With over a decade of experience building scalable product systems across healthcare, fintech and biotech, Jenn works at the intersection of design, product and AI to translate complex technical concepts into clear, production-ready experiences. Her work emphasizes systems thinking, rapid iteration, and close collaboration with engineering and research teams to ship high-quality products. She is driven by making advanced technology practical, usable, and impactful.
Julia is an experienced People Operations leader who is passionate about building people centered organizations where employees feel supported and set up for success. She brings expertise across the full employee lifecycle, including benefits strategy, HRIS implementation, onboarding, and scalable people programs. With experience across multiple industries, Julia partners closely with leadership teams to support growth and change while fostering strong, values driven cultures. She is known for blending structure with empathy and using data informed decisions to create environments where people can do their best work.
Investors
_
Backed by leading AI and deep-tech investors, including top-tier venture firms focused on frontier AI, scientific computing, and industrial transformation.
