A reality model of your operation.
We are working toward a future where organizations can understand the consequences of a decision before they make it.
Explore our thinkingPanmentis is a research and product company building reality models.
The world has been digitized. Buildings, grids and industrial sites produce more data than ever, yet decisions are still made one part at a time, looking backward. Heating is optimized without regard to the grid connection, investments are sized without regard to how a site actually runs, and reports describe yesterday rather than tomorrow. The data exists. What is missing is a way to compute the whole.
That is what a reality model is for. It is to an operation what an Earth system model is to the planet: one model of the whole, built on proven science, calibrated to the real thing, and honest about what it does not know.
We are founders, operators and executives with decades of combined commercial and technical experience across industries. We have run operations, built products and made the kind of decisions a reality model is meant to inform.
The whole, not the parts
Operations are connected; decisions should be too. Weather drives energy use, prices drive costs, and a change in one system creates a bottleneck in another. Yet most tools answer one question at a time: the best heating schedule, the best solar size, the best charging plan. Each answer can be right on its own and wrong together, the moment all three reach for the same grid capacity in the same hour. A reality model calculates the parts together. When goals compete, the conflict is made visible and the whole is recalculated, so no part wins at the expense of the others without anyone noticing.
Cause and effect, not just patterns. Much of how the physical world behaves is already known. Heat flows, sunlight converts and equipment ages according to well-studied laws. A reality model starts from that knowledge and uses data to calibrate it to the real site, rather than trying to rediscover the laws of physics from a few months of measurements. This is what lets it answer the questions that matter most: what would happen if we did something we have never done before?
Forward-looking. Dashboards describe yesterday. A reality model shows how an operation is likely to develop and compares options before anything is committed, so decisions move from reacting to anticipating.
Alive to change. Buildings are renovated, equipment wears and habits shift. A reality model is recalibrated as reality changes, so it keeps describing the operation as it is, not as it was when someone last configured it.
The ontology: knowing what we know
The world and our knowledge of it are not the same thing. A temperature reading, a technician's note, an estimate and a forecast can all describe the same room, yet they deserve very different trust. Most systems flatten them into one table of numbers, as if every value were equally true. Errors then compound quietly: an assumption becomes an input, the input becomes a result, and the result becomes a decision, with nothing along the way recording that it all began as a guess.
The Panmentis ontology is built on that distinction. It describes what exists in an operation, its spaces, equipment, flows and limits, and it keeps a separate, permanent account of what we know about each of them: when it was true in the world, when we learned it, where it came from, and how certain it is. Every number carries its age, its source and its state. A reality model built on this ontology does not only calculate; it knows the standing of everything it calculates with.
Proposals stay proposals until a person confirms them. Anything that has not been measured or confirmed is treated as a proposal, and whatever is built on a proposal inherits that status. Confidence is earned, never assumed.
Old information is treated as old. Knowledge has a shelf life. A limit confirmed years ago, or a measurement from a sensor that went quiet, is recognized as out of date and kept out of decisions until someone renews it.
Contradictions come to the surface. When new evidence contradicts an earlier claim, the contradiction is shown rather than averaged away. Disagreement between sources is information, not noise.
Every decision can be revisited. Nothing is overwritten; corrections are added, not erased. For each decision, the model keeps a record of exactly what was known at that moment, so any decision can be understood afterward on the terms it was made.
The model's own influence is kept apart. Once a reality model helps run an operation, it changes the very data it learns from. The ontology keeps the operation's natural behavior distinct from the effects of the model's own recommendations, so the model never mistakes its own footprint for a law of nature.
Many futures, not one forecast
Conceptual illustration. No operational data or calculated probabilities.
There is no single future to predict. Prices may spike or fall, a winter may be mild or harsh, a tenant may move in or out. Planning against one forecast hides all of this behind a single line that is almost certainly wrong. A reality model holds many possible futures at once, each with its own likelihood.
The scenarios stay alive. A study run once is out of date the day after. A reality model updates its scenarios continuously as new measurements, prices and decisions arrive, so the picture of what may happen always reflects what has actually happened.
It shows what would change the outcome. Beyond what is likely, a reality model shows why. It identifies the few factors that would most change the future that actually unfolds, so attention goes to what matters: the price that would make an investment fail, the assumption that would make a plan unsafe, the signal worth watching before it is too late.
Decisions that hold up. With the range of futures in view, the question changes from "what is the best plan if the forecast is right?" to "which plan holds up across the futures we may face?"
Honest about certainty
Clear about what can be known. Tomorrow's solar output can be forecast closely. Markets and human behavior cannot. A reality model states how certain each result is, and what it offers follows from that: close forecasts where the science allows, ranges and scenarios where it does not.
Checked against limits before acting. Every recommended action is checked against the limits a site has confirmed, such as equipment ratings, the main fuse, comfort bounds and regulation. Where information is uncertain or out of date, the model holds back rather than guess.
People stay in charge. A reality model supports the people responsible for an operation. Automation is introduced gradually, site by site, as each site earns the trust to take it on.
Starting with buildings
We begin where the science is well understood, the data already exists, and the value can be measured in euros and emissions. For buildings, we are developing:
Predictive HVAC optimization. Heating, ventilation and cooling planned ahead of weather, prices and occupancy, using far more than the building automation system sees: meters, external sensors, weather and market data.
Energy investments, sized and tested before you commit. Solar, batteries, heat pumps and EV charging, sized for the real site and tested against how it actually runs, with the payback and peak-power effect made clear.
Operational autonomy. Routine adjustments handled automatically within confirmed limits, so operations run more efficiently and people can focus on decisions that need them.
Emissions you can trace. The emissions effect of each decision modeled hour by hour, with every figure traceable to the meter or record it came from.
Beyond buildings
The same reality model extends to any operation built from connected systems: industry and infrastructure, logistics and ports, and in time, the systems societies depend on. If your field is one of them, we would like to explore it with you.
Your data stays where you decide
You bring data in by uploading files or by authorizing connections to your own systems. Panmentis never accesses a system on its own, and you can disconnect at any time. Raw data stays in the jurisdiction you choose, and regulated sectors and regions with local rules get their own in-region deployment. Learning across sites is optional and uses anonymous metadata only, never raw data.
The future this makes possible
We believe that a world where decisions are understood before they are made is safer, more prosperous and more resilient than one where we act and hope.
Buildings that look after themselves. Most of the energy a building wastes is wasted because no one can see the whole. With a reality model, buildings anticipate the weather, the price and the people inside them, and quietly do the right thing within the limits their owners have set.
Capital that goes where it works. Every solar array, battery and renovation is sized and tested against the real site before money is committed. Investment follows evidence rather than rules of thumb, and the energy transition moves faster because fewer bets go wrong.
Automation people can trust. When every number carries its age, its source and its state, and every decision can be revisited on the terms it was made, handing routine work to a model stops being a leap of faith. Trust is built one confirmed step at a time.
Resources used where they matter. Energy, capacity, materials and money are wasted when each part of a system is planned on its own. When the whole can be computed, the same resources go further: less spare capacity built just in case, fewer peaks that strain the grid, and less energy spent on buildings that are empty.
Foresight for everyone, not only the largest. Today, only the biggest organizations can afford teams of specialists to model their decisions. A reality model gives a housing cooperative, a municipality or a small business the same foresight, so better decisions are no longer a privilege of scale.
Resilience built in advance. When the futures that could go wrong are visible early, preparation replaces reaction. Shortages, price shocks and extreme weather become scenarios planned for rather than crises lived through.
Knowledge that compounds. Each site teaches the model something true about the physical world. What is learned, honestly labeled for what it is, makes every next site faster and every next decision better.
Societies that test before they act. In time, the same approach can help grids, cities and economies explore the consequences of a policy or a shock before anyone has to live through them.
The horizon
As the world digitizes, we make it computable. We believe this work leads toward causal superintelligence: reality models that let industries, cities and economies see the consequences of a decision before a single action is taken. That is our mission. The reality model for buildings is where it begins.
Let’s understand the operation.
Interested in a reality model for your operation? Tell us about it and we will prepare an inquiry you can send through your preferred channel.