The Charlot Lab.
The Charlot Lab works on the foundations of trustworthy embodied AI: systems that are provable, physically grounded, and cheap enough to run on the device. Two threads: Interface Engineering and Swap-2C Constrained AI.
Led by Dr. Charlot.
Publications
Every paper from this lab is a living one. Read it, run its model inline, ask it questions, call its method as a tool from Claude, navigate its place in the corpus, or let the Institute host speak it — a presence assembled from the paper's own concepts.
◆ Energy First Architecture — the on-device verification layer: the certificate that vetoes an action before it commits◆ The living corpusThe graphIn space · XR◆ Run as tools · MCP◆ Ferromotion · drive it◆ The Ferromotion Textbook — 16 interactive chapters◆ Textbook · ch.1 · the body is the controller◆ Textbook · ch.2 · agreement & λ₂◆ Textbook · ch.3 · the command it will not obey◆ Textbook · ch.4 · when touching becomes holding◆ Textbook · ch.5 · show it once◆ Textbook · ch.6 · the estimator that stays honest◆ Textbook · ch.7 · as fast as the motors allow◆ Textbook · ch.8 · where to put your foot◆ Textbook · ch.9 · make it linear◆ Textbook · ch.10 · do everything at once◆ Textbook · ch.11 · the robot that bends◆ Textbook · ch.12 · revising the past◆ Textbook · ch.13 · held by cables◆ Textbook · ch.14 · planning through contact◆ Textbook · ch.15 · landing a rocket◆ Textbook · ch.16 · turning to fitThe Rust library
26 topics.
Each is a live instrument, a paper, and — where it exists — a course. Open one to run it.
Building the energy compute future
A novel machine has to answer two questions at once: how it does on the work we already do, and what it does that the old machine cannot do at all.
Open the topic →The energy-first turn
Electricity, not transistors, is now the binding constraint on computing. What the escape looks like, who is building it, and what would count as proof.
Open the topic →MathGround: joules, not tokens
Every decision priced in joules, carrying a signed receipt — the substrate the rest of the lab runs on.
Open the topic →Provable by construction
A controller that ships its own stability proof, checked on the device in milliseconds.
Open the topic →OmniSense: perception as a volume
A mesh of nodes resolves a space into occupied, empty, and — through RF — the unknown behind walls.
Open the topic →The contact layer
Vision-based touch: perception where line-of-sight ends.
Open the topic →Touching sound
Contact audio: the sense that works through occlusion and in the dark, and reveals what light and touch can't.
Open the topic →Ternary Physical AI
A policy with no multiplies, on silicon anyone can make.
Open the topic →Thermodynamic Physical AI
When the noise is the computer: sampling as the compute.
Open the topic →Thermodynamic compute in gaming and media
Probabilistic computing has never arrived with a workload waiting for it. A playable world and a codec whose decoder is the sampler, measured on silicon people already own.
Open the topic →The open field of computing
Every way to compute, mapped — and how old each one really is.
Open the topic →Quantum information at the edge
What quantum information science actually offers on-device Physical AI — mostly quantum-inspired structure, run classically, held with honest skepticism.
Open the topic →Spatial AI
The world held as a living, predictive field of points.
Open the topic →The Computable World Model
Geometry an agent can read, simulate, and manufacture from.
Open the topic →Graph of the World
The world as a queryable graph of grounded relations.
Open the topic →Spatial RF
Seeing through walls with the radio the room already has.
Open the topic →The surface that pays twice
A material that harvests energy and computes at once.
Open the topic →The printed body
A humanoid whose body — and its actuators — you can print.
Open the topic →VLI: built to interact
Vision-language interaction that reads intent, not just commands.
Open the topic →Space logistics & transportation
Physical AI across the whole transport chain, from launch to on-orbit capture.
Open the topic →Physical AI in space
Above the Kármán line there is no labor pool, so Physical AI is the space economy's labor supply rather than a substitute for it. The market, the unit economics, and who is building it.
Open the topic →Siting the computation
Where should the compute that serves an embodied machine physically sit? Eleven hypotheses about the alternatives to centralised facilities, priced economically and environmentally, and the physical limits that bound the most distributed end.
Open the topic →Physical AI on the farm
US agriculture is counted well at the top and thinly where a machine has to work. Eleven hypotheses about what Physical AI can take on across the estate, the labour market and the logistics layer that moves the crop, each resolved to a trajectory position and the constraint that binds it.
Open the topic →Physical AI in the field
Telecom operators run one of the largest distributed physical estates in any industry and hold an approximate record of its contents. Ten hypotheses about what Physical AI carries there, each resolved to a trajectory position, a binding constraint and a threshold.
Open the topic →The matching principle
One boundary condition, from a 1939 antenna to a robot's touch.
Open the topic →The adaptive unit
Learning without forgetting — the dendrite as a small network.
Open the topic →Two threads.
What the lab works on.
Clean seams between parts
The engineering of the interfaces between the pieces of an embodied system (hardware, software, sensors, and subsystems) so modular systems compose without bespoke glue.
Provable, grounded, cheap to run
AI grounded in physics and math rather than language: closed-form primitives, controllers traced to a Lyapunov function, and a picojoule energy receipt on every call. Behavior is provable and the power cost is known before deployment. It runs in a WebGPU browser in under 400 KB.
Open Interface Engineering.
One thesis runs through the topics, the threads, and the tracks: the lab is Open Interface Engineering in practice. The hard failures in complex systems happen at the interfaces between parts, and the binding cost of AI is energy.
Where systems fail
The hard failures in complex systems happen at the seams between parts. The lab engineers those interfaces so systems compose and stay verifiable, instead of fusing into bespoke glue.
Physics, not vibes
Behavior is grounded in physics and math and traced to a certificate, so what a system will do is known before it runs, not discovered after.
Account in energy
Every computation has a thermodynamic cost. The lab measures and minimizes it in joules, the honest unit for AI that acts in the physical world.
Five tracks, each a live site.
The lab's research runs in the open. Each track has its own thesis, a fancy visual, and a working site you can use.
MathGround
The Swap-2C runtime: physics- and math-grounded primitives, closed-form control, and a picojoule energy receipt on every call.
Open the research track →Pattern-Lang
Recognizing patterns by naming, composing, and verifying them over a small finite lexicon, instead of learning them from billions of examples.
Open the research track →Play Dimension
A strategy game whose mechanics map directly to ideas from AI and neuroscience.
Open the research track →openIE CAD
CAD/ECAD that AI agents can actually drive: geometry, PCB, optics, simulation, and manufacturing through one MCP endpoint, in a WebGPU viewer.
Open the research track →OpenIE Compute
The Periodic Stack of Computation: every computational primitive mapped by consistent axes, including its thermodynamic cost in joules.
Open the research track →Build this with us.
The lab takes on students through internships, and works with investigators across the Institute.