ORB survey vehicle with LiDAR scanning a valley at sunset, terrain rendered as point clouds
Observable Reality Base

Measured. Not assumed.

Every figure ORB returns carries its source, its standard and its margin — above ground or below.

See the solutions Why ORB exists
WHAT ORB DOES
The standard should be simple: every fact a decision rests on, checkable to its source. ORB is that standard.

ORB is the measured ground truth of the physical world. One 1:1 model of a place, built from the data and standards already in use — an ontology of what is actually there: objects, relations and states, each value carrying its source, its standard and its margin. Not a sharper guess about the world, but the world itself, measured.

So bring anything you are working on, above ground or below. On that one model you can assess it, design on it, plan and programme the work, or watch it change over time — and every answer traces back to the reading, the standard and the phase it came from. A whole screening at once, or one figure checked to its source — the same ground underneath.

ORB doesn’t decide for you and doesn’t replace the specialist — it gives the whole field one checkable ground to decide on, honest about what is unknown and traceable end to end. Deciding on the physical world any other way is guessing with confidence. ORB makes the measured way the default.

THE NEXT PHASE
The last decade taught software to reason. The next is about what it reasons over.
“All they can do is give us the next word with a high probability.”
Frank Coyle
Why Agentic Systems Need Ontologies · UC Berkeley

Software moved from code, to weights, to prompts and agents — each phase relocating the intelligence. What none of them moved is the ground. A language model is powerful enough to be dangerous the moment it reasons over information it cannot check. The next step isn’t a smarter agent — it’s a measured world for one to stand on. And for ORB that ground is literal, not just figurative: the measured reality itself, above the surface and below it.

“You can outsource your thinking, but you can’t outsource your understanding.”
Andrej Karpathy
From Vibe Coding to Agentic Engineering · Software 3.0

ORB is that ground — an ontology of what is actually there, every object, relation and figure defined, sourced and bounded by its margin. It is built the way Karpathy describes engineering over improvising: the understanding is never handed off. ORB doesn’t ask an agent — or a person — to trust the model. It hands over the trail.

“Composition over inheritance. Eventually, inheritance starts to break down.”
Justin Schroeder
The Future Is Domain-Specific Agents · Standard Agents

The frontier is fragmenting: not one monolithic agent but many domain specialists, composed — Justin Schroeder’s case. But a ground-mechanics agent, a hydrology agent and a permit agent only reach one answer if they read one reality. ORB is that shared ground across domains: one measured ontology, every domain a lens on it. Ten domains, one advice — because they stand on the same measured facts.

This is not a thesis waiting to be built.

ORB was developed independently, and it runs — on real ground, on real projects. What Coyle and Karpathy describe in principle, ORB already is in practice: a working, measured ontology of the physical world. They arrived at the case for it from theory; ORB arrived from the field. The same conclusion, reached from both ends.

Three arguments, one gap. Each names the need; none builds it. Their case ends at “you need this.” ORB begins where that sentence lands.

Quotes verbatim. Frank Coyle, Why Agentic Systems Need Ontologies (AI Engineer · UC Berkeley); Andrej Karpathy, From Vibe Coding to Agentic Engineering (Software 3.0); Justin Schroeder, The Future Is Domain-Specific Agents (Standard Agents). ORB is independently developed — the reference is convergence, not derivation.

WHAT IT WORKS ON
Something you can observe. Something you can measure. Something with a status.

That is the whole of the living environment — above ground or below, technical or social. Whatever your initiative, it has a line in here.

ObservedMeasuredStatus
Worn markings on a street
condition 7.2 · CROW
re-paving · 2027
A façade above soft ground
57 mm / 30 yr · NEN 9997
attention
A dark corner of a square
4 lux after 22:00 · lighting
flagged unsafe
A turbine beside a hamlet
312 m to nearest home · 41 dB
within setback
A new tower’s shadow
+38 min · midwinter · 12 homes
under review
An avenue tree
canopy Ø 9 m · VTA
retain
A cable under the verge
depth 0.8 m · KLIC
in service
A substation at winter peak
load 87%
congestion
A primary flood defence
crest NAP +4.1 m · AHN
meets norm
A canal crossing permit
6 weeks term left
granted
A cadastral parcel
1,240 m² · Kadaster
owner: municipality
A coastline
−1.4 m / yr · 2010–2025
monitored
An expectation zone
not yet surveyed
high · archaeology
A newly reclaimed plot
no reading yet
unknown

Where a reading is missing, the line stays open — ORB says so, rather than filling it with a guess. Every value carries the source it came from: the difference between a fact an agent can act on and a number it merely repeats.

ONE ONTOLOGY · A LENS FOR EVERY PURPOSE
One ground — a lens for the work in front of you.

Every domain reads the same model; these are lenses on it, not separate products. Pick the one that matches your work.

Engineering

From spatial problem to verified design.

Urban Development

Plan cities as connected systems.

Monitoring

Detect change before it becomes impact.

Operations

Turn plans into coordinated execution.

Governance

Make every decision traceable.

Resilience

Prepare places for what is changing.

Environment

Measure what a plan does to its surroundings.
PROVEN ON REAL PROJECTS
Not a mockup — worked on real ground.

Two screenings that share nothing but the ontology beneath them: a question below the ground, and a question above it over time — each finding traceable to its source and its phase. Not a demo of what could be true — a record of what was measured.

Below ground · a crossing

A bore under a navigable canal.

Ten domains screened at once — advice: may proceed, with noted attention points. Every finding back to its source and phase; location anonymised.

See the case →
Above ground · over time

A street read across fifteen years.

Imagery from 2010, 2018 and 2025 scored against NEN 2767 and CROW — condition carried through to a dated, budgeted plan.

See the case →

Bring what you are working on.

A crossing, a street, a permit, a coastline — above ground or below. ORB reads what is already there, reasons over it, and gives you an answer you can follow back to its source. Measured, and open to check. What to do with it stays yours — and whatever reasons over it next, human or agent, reasons from the same measured ground.

Explore ORB Talk to us