You are using a model right now. To read this sentence, you are predicting the next word before your eyes reach it. To plan your afternoon, you are running a simulation of the future faster than real time. To trust or distrust this page, you are applying a model of what counts as credible.
Models are not optional. Every organism that acts on the future rather than merely reacting to the past is, by definition, running an internal model. The bacterium swimming up a nutrient gradient. The oak shedding its leaves before winter arrives. The city planner zoning land for a population that does not yet exist. The difference between them is not whether they model, but how well they understand the model they are using.
What is a model?
Not a diagram. Not a spreadsheet. Not a computer simulation — though any of these might contain one.
A model is a relationship between something in the world and something in your description of it. You observe a system — its properties, its patterns, the way one thing leads to another. You encode those observations into a form where you can reason about them: equations, narratives, maps, intuitions, business plans, policies, theories. Then you draw conclusions within that form and decode them back into expectations about the world.
When those expectations match what actually happens, you have a model. When they stop matching, you have learned something — not about the world, but about the limits of your encoding of it.
Models are everywhere
They operate at every scale, in every domain, at every level of formality.
Feedforward activation in a biosynthetic pathway
An enzyme early in the chain activates an enzyme further down — before the substrate arrives. The cell is modelling its own metabolic future and acting on the prediction. The model is encoded in the molecular structure itself.
A bird migrating south in September
Day length is shortening. The bird has no experience of the coming winter — it may be in its first year. But it carries a model, encoded genetically and hormonally, that links present light conditions to future survival conditions and changes its behaviour now.
A driver braking before a curve
The road ahead is straight. Nothing requires braking yet. But the driver's internal model — running faster than real time — predicts that at current speed, the curve will be dangerous. Present action follows from a modelled future, not from present conditions.
A company's five-year strategy
The strategy encodes a model of the market, the competition, the technology trajectory, and the organisation's own capabilities. It draws conclusions — invest here, divest there, hire these skills — and decodes them into operational decisions. Every assumption is a linkage in the model. Every omission is a hidden variable.
A government's climate policy
The policy rests on models of atmospheric physics, economic behaviour, political feasibility, and technological development — none of which can be reduced to any other. The policy is only as good as the models it encodes, and every model has a finite horizon beyond which its predictions become unreliable.
A sentence you speak to a friend
You model what they know, what they feel, what they will understand from your words, and what they will misunderstand. You choose words based on that model. If you are surprised by their response, it is because your model of them was incomplete — not because they behaved incorrectly.
A large language model generating text
Claude produces the statistically most probable next token given its training. Is this a model of language, or a simulation of it? Does it preserve causal structure — the reasons why certain words follow others — or does it merely reproduce the outputs? This is not an academic question. It determines what you can and cannot trust it to do.
The questions that matter
Every model — at every level — is subject to the same questions. These are not technical questions. They are the discipline of anyone who acts on the future.
These questions come from the work of Robert Rosen, a mathematical biologist who spent his career developing a formal theory of the modelling relation — the relationship between any system and any description of it. His insight was that the modelling relation is not just a method of science. It is the foundation of all anticipatory behaviour, from molecular to civilisational.
What this place is for
This is not a course on Rosen's theory. It is not a platform for learning concepts, though concepts are here if you want them.
This is a place to bring your models and examine them. A plan you are making. A strategy you are evaluating. A document that encodes assumptions about the future. A system you are trying to understand. A description you suspect may be a simulation dressed up as understanding.
The discipline is simple, though not easy: make your encoding visible. See what it includes and what it omits. Find where the horizon is. Look for the non-equivalent descriptions that your single encoding cannot hold. Notice when you have stopped observing and started classifying.
This is a practice, not a theory. The theory exists — Rosen developed it rigorously — and it is available here for those who want it. But the practice stands on its own: the habit of asking, before every decision, what model am I using, and where does it end?