Expertise lives in the work.
Project files and expert decisions hold knowledge that open-web training alone cannot supply.
THE SPATIAL INTELLIGENCE LAB
Intelligence for
the built world.
An AI research lab exploring how the built world works—and what happens when it changes.
Explore the research
Physical AI starts with spatial intelligence.
01 / THE HARD PROBLEM
Beyond the mesh.
Into the world.
The built world is
a system of systems.
Buildings, streets, infrastructure, and the spaces between them. Change one part, and you can alter how people move, resources flow, and an environment functions.
Generating a shape is one step. Spatial intelligence must understand these relationships—and compute the consequences of change.
WHY A DOMAIN LAB
Model scale is only the beginning.
Frontier models are powerful starting points. Our thesis: built-world intelligence also requires knowledge, tools, and evaluation grounded in the domain.
Geometry needs meaning.
Models must connect shapes to structure, services, regulations, and design intent.
Plausible isn’t enough.
Physical consequences demand geometric checks, engineering analysis, and practitioner judgment.
02 / WHY ARCOL
Research for
the world around us.
Explore ArcolArcol is where knowledge
becomes geometry.
Its authoring platform connects spatial models, live metrics, and the people making design decisions. That gives the lab a rare starting point: an environment where intelligence can learn to do useful work.
Learn from the decisions.
Work with practitioners to understand why designs succeed and where they fail.
Act on real geometry.
Editable models and connected metrics offer a place to take actions and measure results.
Close the research loop.
Test in design workflows. With permission, use expert corrections to improve the next model.
03 / THE LAB
World models.
For the built world.
Our research agenda: connect spatial understanding to computation, so machines can reason about the built world and test the consequences of change.
R.01REPRESENTATIONUnderstand the system.
Represent geometry, components, and their relationships in coherent world models that can update as environments change.
R.02REASONINGAnticipate the consequences.
Trace constraints and dependencies across spaces, structures, and infrastructure. Predict the effects of a change, then test those predictions against domain tools and expert review.
R.03COMPUTATIONMake the result testable.
Use geometric operations and analysis tools to create, edit, and evaluate designs. Measure reliability on unfamiliar environments and tasks.
THE RESEARCH LOOP
Built to learn from the work.
- 01
Capture expertise.
Spatial models and the decisions behind them.
- 02
Train & evaluate.
Build models. Test against domain constraints.
- 03
Learn in practice.
Expert feedback informs the next research cycle.
04 / THE SCALE OF THE AMBITION
Start with Arcol.
Reach beyond design.
Physical AI needs to understand
where it acts.
Spatial understanding gives Physical AI the context to plan, reason, and act across the built world.
If our research succeeds, it could become a foundation for others to build on—across the life of the built world.
From plans to action.
Reason about access, assembly sequences, and changing site conditions.
Understand what can change.
Connect infrastructure and environmental systems to maintenance, energy use, and adaptation.
Test before acting.
Give Physical AI structured environments to explore tasks and consequences.
THE LONG-TERM AMBITION
Intelligence for how we
design, build, and inhabit.
The built world.
The next frontier.
Spatial intelligence for Physical AI.Back to the beginning