
An independent, founder-run company from the Netherlands — thirty years in the built environment behind it, and no investors or parent group to answer to.
Across three decades in the built environment, one thing kept repeating: the people who had to decide were working from partial pictures — every party holding a different fragment, no two maps in agreement, the real answer only clear once the ground was already open. That is what the company was started to change. Not to win a market, but to close a gap the founder had lived with for a career. The conviction that a decision on the physical world should be one you can see and check — before it is committed, not after — is where ORB began, and it is still what sets its course. It is built for the user — and that course was set there, deliberately, before any investor could set it instead.
ORB is built and run by its founder — no investors, no parent group, no board to satisfy. Behind that sit three decades of practice across the spatial domain, from what lies below the ground to what stands above it. The company is kept deliberately small and independent, and that is a choice, not a stage it hopes to grow out of: it keeps the work honest and answerable to the people who use it rather than to a shareholder waiting on a return. Founder-led is about who sets the direction and answers for it — not a single point of failure. ORB stands on measured reality, open standards and data you hold in your own tenant, so what you depend on is the ground and the record, not any one person.
There are two ways to work with the company: the platform itself, with your own tenant, and hands-on advisory where a project calls for it. Based in the Netherlands and working anywhere — the physics of ground and the logic of coordination stop at no border. To discuss a project, a pilot, or your own ORB, a real person answers info@orbplatform.nl.
Three sentences describe, with unusual honesty, how far today’s AI actually reaches — and exactly where it stops. Each one only confirmed what ORB was already doing. But this note starts with my own claim, not theirs.
The bottleneck was never the intelligence — it is the absence of a base for it to stand on. Give a model, or a person, a measured ground where every fact carries its source, its standard and its margin — and one such ground shared across every domain — and the plausible becomes checkable. The answer to better decisions is not a smarter model; it is a measured reality to reason on. That is the hypothesis this whole company is built on — and it is no longer just a claim: ORB is already running, proving it one measured result at a time.
What follows is why I hold it. I did not build ORB from these three — they arrived at the same edge, from three different directions, after ORB was already running. Each stops there; ORB is where the step is taken.
“The model generates tokens. Everything else is someone else’s problem.”
This is the most honest thing I’ve heard said about agents. A language model produces plausible words; whether they match the world is, in Coyle’s phrase, someone else’s problem. His answer is an ontology — a structured model of reality the agent can be held to. And that is where he stops: he makes the case that you need one.
My step past it: ORB is that ontology — but measured, not asserted. Not a graph of facts someone typed in, but a 1:1 record of the physical world where every value carries the reading it came from, above ground and below, and stays honestly empty where nothing was surveyed. Coyle says an agent needs grounded truth. ORB is that ground — actually surveyed, with source, standard and margin on every figure. I didn’t stop at “you need this.” I built the this.
“You can outsource your thinking, but you can’t outsource your understanding.”
Karpathy names the discipline that has to survive the shift to agents: let the machine do the work, but you remain the one who has to understand — and answer for — the result. He argues for engineering over vibes: deterministic, verifiable, kept on a leash. And that is where he stops: a way of working.
My step past it: ORB moves that discipline out of the process and into the product. Every figure ORB returns hands you its own trail — source, standard, margin — so understanding isn’t something you reconstruct by trusting whoever built the model; it travels with the number. Karpathy keeps the human responsible. ORB gives the human something they can actually check, on its own, line by line.
“Composition over inheritance. Eventually, inheritance starts to break down.”
Schroeder sees where agents are heading: not one giant model but many small, domain-specific ones, composed. He’s right — that is how this scales. And that is where he stops: the agents compose, but each still reasons over its own slice of the world.
My step past it: ORB gives those specialists one ground to compose on. A ground-mechanics check, a water permit, a settlement forecast — different domains, one measured model beneath them all, so ten domains resolve to a single advice instead of ten disconnected opinions. Composition needs something to compose over. ORB is that something.
So here is what I believe. Coyle named what agents lack. Karpathy named the discipline they demand. Schroeder named where they are heading — apart, into domains. All three stop at the edge of the same gap: a world model that is measured, not asserted; verifiable in its output, not just its process; and shared across every domain, not rebuilt in each. ORB is my answer to that gap — built independently, from the field, and already running. Where they stop, ORB takes the step.
The ambition is large on purpose, and the company would rather earn it one verifiable result at a time than ask you to believe a rendering. What is claimed on this site traces back to real work you can open and check — finding by finding, source by source. That is the working principle behind the whole company: grand where it points, provable where it stands, and never a step it cannot show.
No investors and no parent group. The direction is not for sale, and the company answers to the people who use it — not to a shareholder. No outside owner
The one deciding the roadmap is the one who has done the work — three decades in the field. But ORB does not lean on any single person to keep standing: it runs on measured reality, open standards and your own portable tenant, so what you rely on is the ground and the record. Direction, not dependency
Your own tenant, your own data, open standards and no lock-in. What you put in, you can take out — the company earns the next year, it does not trap you into it. No lock-in
Based in the Netherlands, working anywhere. A real person reads and answers the mail — no ticket queue, no reseller in between. Direct line
No investors, no parent group — one independent company, and one person you can actually reach. That is the whole of it.