Intelligence begins
with an honest model
of the world.

Pale mineral manufacturing terrain beneath a large beige moon

Visual Normality

Visual Normality (V/N) is building the world model for manufacturing: a machine-readable representation of industrial reality where humans, AI and machines can understand state, simulate possible futures, make decisions, execute work, and learn from the outcomes.

Reality deserves a better interface

Factories are among the most consequential systems people have built. They turn ideas into medicine, food, devices, materials and the infrastructure of everyday life. Yet the people responsible for them are asked to make critical decisions from fragments: an old drawing, a dashboard, a spreadsheet, a photograph and the memory of someone who knows the site.

The manufacturing value chain behaves as one connected system. Its information does not. We believe that must change.

Visual Normality builds a system that connects products, demand, engineering, supply, physical assets, operations, capability, capacity, maintenance, quality and projects, so teams can understand what their network can really support, test what could change and govern what must happen next. The system routes work between AI agents, humans and, eventually, robots. And every decision that moves through it leaves behind something rare: an interventional record of what was true, what was predicted, what was decided and what actually happened.

The data makes the system smarter. The system is the only way to collect the data.

The future we are working toward

We imagine manufacturing environments that are understandable before they are altered, simulatable before capital is committed and governable before machines are permitted to act.

We imagine a decision infrastructure that spans the physical operation without pretending to replace the systems that already run it.

We imagine autonomy arriving responsibly: first through understanding, then prediction, recommendation, simulation, preparation, bounded execution and verified learning.

Reality should be visible before it is changed.

Treat reality as a set of states

What was designed is not always what was built. What was built is not always how it is operated. What is proposed is not what has been approved. And what was installed will one day be removed.

These differences are not inconveniences to hide. They are information. We preserve them, make them legible and attach confidence to every consequential claim. A compelling image is not sufficient evidence. A prediction is not a fact. Provenance must travel with the conclusion.

Route the work to whoever does it best

A plant runs on work: monitoring, checking, drafting, verifying, deciding, approving. Today nearly all of it lands on people, and too much of their time goes to assembling context instead of exercising judgement.

We route the work. Agents watch, assemble and draft, with every statement linked to its source. People judge, decide and approve. In time, machines will act, within envelopes precise enough to make action safe.

Intelligence must remain accountable

Industrial AI should make uncertainty clearer, not hide it behind fluent answers.

Understand before predicting. Predict before recommending. Simulate before executing. Act only where action is safe. Learn from every outcome.

Let every outcome compound

The most valuable industrial intelligence is not a volume of sensor data. It is the verified relationship between a state, a constraint, a decision, an action and what happened next.

Every prediction is compared with its outcome. A model that never faces its own predictions never improves. Ours will face every one.

Build the next view
of manufacturing with us.

Product
Location
Amsterdam, The Netherlands

VN SyntaGraph™

Abstract beige manufacturing plateau with process infrastructure and a pale moon

Manufacturing change is usually reviewed across disconnected sources: site plans, engineering models, photographs, equipment files, utility data, documents, and local knowledge. Each source describes part of the operation. None carries the whole decision.

VN SyntaGraph is our working model for bringing those parts into one spatial context. It connects the facility, line, asset, route, constraint, observation, and proposal while preserving where each piece of information came from.

The distinction between current and proposed matters. So does the distinction between documented, observed, measured, and inferred. SyntaGraph is designed to keep those states visible, so a team can inspect the basis of a change before accepting it.

Technical View and RealView are two readings of the same operation. The evidence graph links each claim to its source. SyntaGraph does not replace engineering judgment or existing systems of record. It provides a clearer environment in which that judgment can be made.

Inside
SyntaGraph.

The connected site model, reality registration, and placement checks, shown with concept imagery.

Technical model

The technical model is the structured view of the site. It combines building and line geometry with rooms, equipment, utilities, asset identity, and known dimensions. Teams can measure clearances, inspect dependencies, and place proposed equipment without changing the approved baseline. Current and proposed objects remain visibly distinct throughout the review.

RealView

RealView is a calibrated camera view of the current facility. It helps reviewers confirm what is physically present, inspect access and working conditions, and identify differences between the documented model and the observed site. Selecting a building, line, or asset keeps the same object in context when switching between RealView and the technical model.

Evidence and constraints

Evidence and constraints are attached directly to the site, asset, or location they describe. They can include drawings, measurements, observations, equipment requirements, safety zones, utility capacity, and route restrictions. SyntaGraph records whether information is observed, measured, documented, or inferred, and surfaces missing or contradictory evidence before a decision is made.