Founder
Founder, NeoAmorfic Ltd. London and Athens.
“Physics intelligence is a game-changer for any field where ‘close enough’ is catastrophic.”
The next decade belongs to systems that propose, and the one after it to whoever can tell which proposals are true. Language models will draft more and more of what we read and decide; learned policies will act for us in the physical world. Both are extraordinary and both are probabilistic: their confidence is not evidence, and no amount of scale changes that. What is missing is a layer that reads the physical state of the system a proposal concerns — a market, a molecule, an engine, an aircraft — by the laws of physics, and reports it in a form anyone can re-run. That layer is physics intelligence, and building it is what I do.
The first step exists. QEIv18™, our physics engine, was certified on production in July 2026. It reads the structural condition of an evolving system from the system’s own record — entropy, geometry, stability, coupling — and gives the same reading to anyone who re-runs it, to at least twelve significant digits on any machine. Above it we are building the intelligence layer in a laboratory of our own design, where controlled systems with known equations and failure boundaries supply the ground truth and the engine is never allowed to grade itself. The foundational question is whether structure, read only from present and past observations, can recognise approaching instability before ordinary measurements show it. We have built much of the laboratory; we have not yet made the discovery, and we say so.
The frontier is five capabilities in rising order of difficulty: structural measurement, which exists; early warning before conventional alarms; trajectory intelligence, which tells similar-looking presents apart by where they are heading; counterfactual intelligence, which says how a trajectory changes under an alternative action; and intervention intelligence, which selects the permissible action that changes the outcome and proves that it did. The destination is a physics-native decision layer that answers not what is likely to happen but what structural process is occurring, what happens if it continues and which intervention changes it — for jet engines and grids, for spacecraft and robots, for molecules and cells, and as the layer that judges what a learned model proposes before it is allowed to act. That is the work, and it will take longer than a product cycle. I would rather build it properly than announce it early.
I have always been an early adopter, and innovation and technology inspire me. I completed MIT Sloan’s programme on AI strategy in 2021, explored artificial intelligence in an exhibition for a Swiss private foundation in Zurich in 2022, and have used large language models from the first week they were available. I trained in swaps and derivatives at INSEAD and in investment management and financial regulation with the Securities and Investment Institute, and invested in derivatives for years. Throughout my professional life I have been in entrepreneurial mode, founding and running companies in design, brand strategy and renewable energy.
Everything is design. Trained as an industrial and transportation designer in Vienna and at Art Center in Pasadena, I learned to ask how a thing works before deciding how it should look, and I have used the reverse of the rule I was taught ever since: function follows form. It applies wherever the form already exists and the function is still to be discovered — a market, a molecule, an airframe in the wind, and now the invented forms of language models and machine learning. The structure of a thing determines what it can do and how it will fail, and physics is the discipline that reads it. I did not come to physics through a laboratory and I do not present myself as a physicist; I came to it as an inventor with a designer’s question — what is the structure, and what does it determine. NeoAmorfic is that idea, built.