6: Beyond the Boundary

10/2/20264 min read

6: Beyond the Boundary

Where did it come from?

A steel rod holds up a balcony, a bridge, a person leaning against a railing. It looks finished, complete in itself.

Follow it backwards and it comes apart. The rod came from a mill, the mill from ore, the ore from a mine. The mine needed land, and the land needed a forest to be cleared. Somewhere in that chain, soil that once held roots now holds a human being upright.

The rod supports us. Almost nothing in our description of it supports the forest. That asymmetry is not a fact about steel. It is a fact about how we look.

So the question this piece follows is simple to ask and hard to hold: what happens to a world that is described in a way that removes its life?

The cut

Words let us name a thing. Naming lets us hold it, and holding lets us control it. Words also let us divide: this is forest, that is timber; this is soil, that is resource.

Division is useful. It is also how a web of dependence starts to look like a set of separate objects. The perception of disconnection is, at least in part, a grammatical habit.

Science made this habit rigorous. When Galileo separated what could be measured (size, shape, motion) from what could only be felt (colour, warmth, taste), he set the second list outside the work of knowing. Quality became private. Quantity became real.

The gain was enormous: a physics that could predict. The cost was quieter. A world described only by its measurable parts begins to look like a dead world, one in which nothing is alive in the description, even when it is alive in fact.

If that is where the cut was made, what did we build on top of it?

Machine morphism

We built a metaphor. Once the world was stripped to matter in motion, it was easy to see it as a machine: the heart a pump, the brain a switchboard, the cosmos a clock.

Each era borrows its picture of mind from its most impressive tool. Today that tool is the language model. Where we once spoke of the world as mechanism, we now speak of the mind as prediction over tokens. Call it LLM morphism: the newest version of an old move.

Kant saw the limit of that move more than two centuries ago. In a machine, he argued, the parts exist for one another. In an organism, they also exist by means of one another: each part produces and sustains the others. An organism is not only organized. It is organizing.

A watch does not repair its own spring. A tree does grow a new branch. The difference is not complexity. It is where the making comes from.

Language models are the first systems to predict relationships at scale. But they learn those relationships from text, not from the world the text describes. Between data and reality there is no one-to-one correspondence. The model inherits the map, and with it, the cut.

Reaching back toward the world

The field seems to sense this gap. Much of AI's attention in 2026 has shifted from scaling language alone to building world models: systems that learn how things change and what is likely to happen next, rather than what is likely to be said next.

The framing has softened too. Recent surveys treat language models and world models as complementary layers. Language carries what usually happens and what might be worth doing. A world model estimates what will actually follow. In June 2026 alone, developers released over a dozen embodied and world-model foundation models.

Read one way, this is engineering progress. Read another, it is an admission: description is not contact. A system trained only on what we have written about the world will reflect our cuts back at us.

But a world model is still a model of a world. It predicts. Does it organize? A system that simulates the physics of a forest is not yet a system that grows.

Putting the life back

Neuroscience is having a parallel argument. Predictive processing describes the brain as a prediction engine, but one camp keeps cognition inside the skull while another insists the body and its surroundings are part of the thinking itself. Behind that debate sits an older idea: mind-life continuity, the claim that the principles of mind are already present in the organization of life.

A July 2026 paper by Pezzulo and colleagues asks what biological organisms can teach embodied AI about grounded world models. Living creatures do not acquire their models by reading. They acquire them by acting, failing, and staying alive, often together with others.

The same turn is happening outside the lab. A growing number of countries, among them Ecuador, Bolivia and Brazil, now give rivers and forests legal standing. Designers speak of more-than-human futures. Some have tried to make AI's own footprint visible, such as the water consumed in training, to show that even our most abstract tools live inside an ecology.

Science, law and design are each, separately, trying to undo the cut. None of them has yet asked the question together.

Living knowledge

We do not build systems for knowledge. We build them for the persistence of knowledge across time: clay tablets, libraries, databases, model weights.

That raises an awkward question. If everything changes, is storing knowledge a kind of paradox? A stored fact is a cut in time, just as a named thing is a cut in space. What we keep is always a little dead by the time we read it.

The pattern may be repeating at a larger scale. Science once shifted its centre of gravity from physics to the life sciences. Something similar may be happening now, as questions that once belonged to mathematics and engineering begin to need the humanities to be asked well.

Perhaps the next lens is not a sharper instrument but a different relationship: knowledge that is organized and organizing, that stays in contact with what it describes. What would a living knowledge system look like? And would we recognise it, if we had already learned to see the world as dead?

Sources

Connect

© 2026. All rights reserved.