The Things That Stay

Heraclitus is supposed to have said you cannot step into the same river twice. The water you touched a moment ago is already downstream; the next instant brings new water, new eddies, a subtly different bank. And yet we do not hesitate to call it the same river. We name it, map it, build cities on it, sign treaties over it. Something about it stays put while everything in it flows.

The puzzle is not confined to rivers. A candle flame holds its teardrop shape while the gas that composes it is consumed and replaced dozens of times a second. A whirlpool keeps its form as an endless supply of new water pours through it. And you—most of the atoms in your body were not there a decade ago, your cells turn over on schedules from days to years, the very proteins that carry your memories are disassembled and rebuilt—are somehow still, recognizably, unmistakably you. The matter is on loan. The pattern is what persists.

This essay is about a single idea with a very long pedigree: that identity is carried by form, not by matter—and that we can now say, with some precision, what “form” means. The precise version is the subject of a research program we call Kolmogorov Theory (KT). But the surprising thing, the thing worth an essay, is how little of it is new. The chain of reasoning runs from Pythagoras to Alan Turing, and each thinker along the way was refining the same intuition. What the twentieth century added—computation, and a way to measure structure—was the missing mechanism, not a change of subject.

The universe (which others call the Library) is composed of an indefinite, perhaps infinite number of hexagonal galleries… — Jorge Luis Borges, The Library of Babel (1941)

Borges imagined a library containing every possible book—every string of characters of a given length, almost all of it gibberish, with the few meaningful volumes lost in an ocean of noise. It is a good image for the world as raw possibility. To live in such a place is to face an impossible glut of detail. What a mind does—what any persistent thing does—is find a compressed, navigable order inside the glut: a short description that captures what matters and discards the rest. Hold that thought. It will turn out to be the whole story.

Pythagoras and Plato: identity is structure

Start where the Western thread starts. The Pythagoreans were struck—mystically, extravagantly—by the fact that the world has intelligible structure. Pluck a string, then stop it at half its length, and the note jumps a clean octave; simple whole-number ratios govern harmony. “All things are number” is not a claim we can keep, and KT keeps none of the mysticism. But underneath it is a wager that turned out to be the wager of science itself: that a vast stream of appearances can be captured by short, ratio-like laws. What is deep is not quantity but compressible pattern—the fact that the data admit a brief generating description.

Plato sharpened the intuition into a doctrine. Behind the changing appearances, he held, stand stable Forms. A triangle drawn in sand smudges and fades, yet triangularity is untouched; the identity of a thing is carried by its form, not by the particular matter that happens to instance it. This is exactly right, and KT keeps it: mathematical structure is objective, discovered rather than invented. But Plato paid too high a price. He put the Forms in a separate world—a heaven of perfect abstractions, gazed at by the soul before birth, of which the sandy triangle is a poor copy. That is one world too many, and it opens a notorious wound: if the Forms are elsewhere, how does a physical mind ever make contact with them?

KT's move is to keep the structure and drop the second world. Structure is not in a heaven; it is the organizing face of the one reality we actually have. Forms are real, objective, and load-bearing—but never elsewhere. This is closer to what philosophers now call structural realism than to Platonism proper, and it dissolves the access problem at a stroke: a mind reaches structure by participating in the world's own organization, not by reaching across an ontological gap.

Aristotle: form comes down to earth

The decisive step is Aristotle's, and it is the station closest to KT. Aristotle brings the Forms down from heaven and puts them into matter. A living thing, he argues, is not defined by its material: food becomes flesh, cells die and are replaced, the matter cycles through—yet the organism persists. What persists is the form (morphē), the organization that makes the thing the thing it is. A substance, on this view, is matter informed by form. The doctrine has an ungainly name, hylomorphism, and a beautiful core: to be a living individual is to be a pattern that maintains itself while its matter passes through.

Read that sentence again and you have very nearly the KT definition of an agent. Where Aristotle says form, KT says the compressed self-and-world model that a system carries and keeps up to date. Where he says the soul is the “form of the living body”—not a ghost inside it but the organization that makes it alive—KT says the soul is the regulatory architecture: the machinery that models, evaluates, and acts. The correspondence is close enough to tabulate.

The last two rows are where Aristotle earns his reputation for spookiness—and where KT quietly defuses it. Aristotle spoke of entelechy, a word usually mangled in translation but meaning something like “being-at-work-staying-itself,” and of final causes: the acorn grows toward the oak because its form is directed to that end. To modern ears this sounds like backward-acting purpose, a future pulling on the present, and we rightly recoil. But there is no need. A system whose dynamics happen to be organized so as to preserve its own pattern will look goal-directed without any purpose reaching back through time. The apparent teleology is the shadow cast by persistence. Time is a filter: it runs, it churns, and it quietly deletes every arrangement that fails to keep itself going. What remains, necessarily, looks as if it were trying to remain. Aristotle mistook the shadow for a force. Name the shadow and the mysticism evaporates.

Notice, too, Aristotle's method, which is exactly KT's. A single snapshot of a hurricane, a flame, or a person tells you almost nothing. Only observation across time reveals the persistence, the self-restoring loops, the organization that survives perturbation. Time filters out accidental structure and leaves what remains itself through change—which is Aristotle's criterion for substance and KT's criterion for a real pattern, one and the same. This is what our companion essay calls, with an imperative, “Pattern, persist!”—read here straight back into the Metaphysics.

Kant: the world is built, in time

Two thousand years later, Kant shifts the question from the object to the observer. His subject in the Critique of Pure Reason is not how sensations occur but how experience is possible at all. And his answer is radical: experience is not the raw input. It is not bare awareness. It is a construction—a manifold of sensation synthesized, under the mind's own forms of space, time, and causation, into ordered cognition of a world of objects. We do not passively receive the world; we actively assemble it, and the assembly happens across time.

This is the second pillar of KT, and it translates almost word for word. In KT, all of an agent's models begin with data—but data never determine a model on their own. Construction also requires architecture, prior structure, a principle of compression, an aim. Kant's slogan, “thoughts without content are empty; intuitions without concepts are blind,” becomes: models without data are unconstrained, data without models are unstructured. Experience is their synthesis. And Kant's deepest twist—that the conditions which make experience possible are the very same conditions that make objects possible—becomes the claim that an “object” is nothing but a stable, useful compression inside the model. A face, a tree, a self: each is a chunk of the world-model that has earned its keep by persisting and predicting. The world of objects is real, but it is real for a modeler. It is built, and time is the workshop.

Kant also drew a fence, and KT respects it. The mind has access to appearances—to model-mediated phenomena—not to the world as it is in itself. Push the categories past the boundary of possible experience, toward the soul as a simple substance or the cosmos as a completed whole, and you get not knowledge but illusion. KT says the same in its own vocabulary: an agent works with its own observations and its own models, which can be predictive, useful, objectively valid within its reach—and are still not an unmediated grasp of reality. The mistake Kant diagnosed—inferring an immortal soul-substance from the mere felt unity of consciousness—is one KT makes computationally. The unity is real: it is the integration of many streams into one self-indexed model, plus that model's continuity through time. But an integrated, continuous model is not evidence for a little self behind it. There is no homunculus. There is the persistent form, and nothing more is needed.

Schopenhauer: why the subject cares

Kant gives us a world for a subject. He does not tell us why the subject should care. That is Schopenhauer's addition. Beneath representation, he argued, runs Will—not any particular conscious desire, but a deeper tendency to strive, to continue, to resist dissolution, expressing itself through all our drives and affects. Read soberly, stripped of Schopenhauer's own gloomy metaphysics, this is the missing pole. It is the same persistence that Aristotle named impersonally as entelechy, now felt from within as striving and as value.

KT naturalizes it without drama. An agent continually reads out its situation as a single scalar—good or bad, toward me or against me—a valence. For the kind of agent that time actually leaves behind, that valuation approximates one thing: expected future persistence, the answer to “will I still be here tomorrow?” The striving is not bolted on. It is what selection installs, because the filter of time only preserves systems whose aims happen to favor their own continuation. Schopenhauer's Will, disciplined, becomes the objective that makes a mere model into a mover.

Turing and Kolmogorov: the mechanism and the measure

Everything so far is philosophy of form. What the twentieth century supplied was the machinery to make it exact—and, for the first time, testable. Two ideas do the work.

The first is Turing's. Whatever synthesizing, evaluating, and planning a mind does, it must be something a finite procedure can carry out—a definite, step-by-step process, executable by a machine. This is not a metaphysical dogma that the brain “is a computer” in some crude sense; it is the modest and powerful requirement that every perception, memory, model, and action, at the interface where an agent meets its world, is finitely describable. That is enough. It lets us decompose the vague word “mind” into parts that could actually be built.

The second is Kolmogorov's, with Solomonoff and Chaitin—the idea that gives the theory its name. It is a way to measure structure. The Kolmogorov complexity of something is the length of the shortest program that reproduces it: its most compressed description. A string of a million random coin-flips has high complexity—there is no shortcut, you must spell it out. The first million digits of π, though they look just as patternless, have tiny complexity, because a short program generates them all. Here at last is a precise handle on what Pythagoras and Plato were gesturing at. To have “form” is to be compressible. To find the form of something is to find its shortest description. And prediction, it turns out, is the same act as compression seen from another angle: to compress a stream well you must have grasped the regularity that will govern what comes next.

There is a humbling clause built into this measure, and KT embraces it. The shortest description of a thing is, in general, uncomputable—no procedure can hand you the true Kolmogorov complexity of an arbitrary object. A bounded mind can never possess a complete description of itself or its world. So every claim the theory makes is framed not as an absolute number but as a contrast—a difference, an on-versus-off comparison—measured with a fixed practical estimator, where the estimator's own bias cancels out. The incompleteness is not a bug to be hidden. It is a permanent feature of what it is to be a finite thing modeling a world larger than itself.

The synthesis: a form that maintains itself

Now assemble the pieces, because they lock together with a satisfying click. Strip away the vocabulary of every tradition and you find one object underneath, built from three roles. Think of them without jargon first.

Map, compass, player: a Modeling Engine, an Objective Function, a Planning Engine. That triad is what KT means by an algorithmic agent, and its lineage can be written as a single compact equation—the one slogan worth memorizing from the whole argument.

And the root of it all is persistence. Aristotle's entelechy, Schopenhauer's Will, and KT's objective are three names, across two and a half thousand years, for the same thing: a system that acts so as to keep on being itself. KT gives it a fourth name, telehomeostasis—self-regulation whose very aim is a measure of its own continuation. Natural agents get this aim for free, by construction, because the filter of time preserves only those objectives that happened to favor continuation. It is Aristotle's entelechy rewritten as a policy: of all the things I could do, choose the one that best keeps the pattern that is me alive.

Why begin with experience

One move remains, and it is the one a careful reader will already be pressing on. This whole account rests on modeling, and modeling seems to presuppose a modeler—someone for whom the map is a map, someone the compass reading is felt by. Where, in all this machinery, is the experience?

KT's answer is to refuse to begin with matter. Most of modern thought starts from the physical world and then struggles, notoriously, to find room in it for experience—the “hard problem” of consciousness, the seemingly unbridgeable gap between neurons and the felt redness of red. KT reverses the order, and begins where Descartes began, though it sharpens him. Descartes found one thing he could not doubt: I think, therefore I am. KT tightens the datum to I think, therefore I have structured experience. The single fact that survives every skepticism is not that matter exists, or that physics holds, or that computation runs, but that there is experience—and that yours is structured: laid out in space, flowing in time, organized into concepts. You are reading this. That is the ground floor, and nothing is more certain.

Now watch what can be drawn out of that one datum without adding anything to it. If experience is structured, then two things must be true of it at once. There must be something that gets structured—experience in its raw state, the bare “there is,” whatever remains when every particular pattern is stripped away. And there must be the structure itself: the relations, the symmetries, the order. But the study of relations, symmetries, and order already has a name. It is mathematics—not the human game of numbers, but, in the older and wider sense, the science of structure as such. So to say that experience is structured is already to say that it is mathematically organized. Mathematics is not a second thing smuggled in beside experience; it is what the word “structured” in structured experience meant all along, read from the outside. That is why KT treats it as a deduction rather than a posit: nothing is assumed that was not already contained in the one fact we cannot doubt.

This is the quiet radical center of the program. Experience and mathematics are not two ingredients that must be glued together; they are two faces of a single reality, which KT calls the Unicum. Experience is the inner face, the “what it is like,” irreducible to any description. Mathematics is the outer face, the structural grammar that gives experience its form. Neither reduces to the other: experience without mathematics is ineffable; mathematics without experience is empty. The old puzzle of why mathematics should be so “unreasonably effective” at describing the world softens at this point—of course it is effective; it is the world's own structural face, not a lucky overlay we invented. And Plato's wound finally closes. Structure is real, objective, discovered rather than made—but never in a separate heaven. A physical mind can reach mathematical structure because it already is a piece of organized structure; to know mathematics is simply to compress the pattern one is already made of. The program records the balance in its own unwieldy name—an experiential structural Platonism: structure fundamental, experience fundamental, two aspects of one thing.

The persistent form, then, has somewhere to live. It is a fold in this single experiential-and-structural fabric: a region organized into the shape of an agent—a mapmaker, a compass, a player—and structured, first-person experience is what that organization is, from the inside. Not a dead mechanism that somehow lights up into feeling. The lit-up-ness was never a later addition; it was the fabric all along.

The report and the experience

This reordering is not idle metaphysics. It changes what a science of the mind should even try to measure—and it exposes something the current science quietly gets wrong. Almost everything we call the science of consciousness studies, in the end, what people report. A subject presses a button, says “I saw it,” recounts a dream. The brain markers we trust are trusted precisely because they line up with those reports; report is the only ground truth available in a room full of awake, talking adults. But a marker calibrated against report inherits report's blind spot. It measures the machinery of access—the apparatus that makes an experience available for speech and memory—and then we call it the signature of experience itself.

The two come apart, and the asymmetry is the whole point. A report is fair evidence that an experience occurred; but silence is not evidence that it did not. When someone cannot report—under anesthesia, inside a dream, locked in, or simply wired differently—we have learned something about their access machinery, not about whether there was something it was like to be them. The philosopher Philip Goff put the complaint mildly: much of what is billed as the search for the neural correlates of consciousness is really the search for the neural correlates of reported consciousness. It sounds like a technicality. It is not. For centuries, humans read the absence of a report—an animal that cannot say it suffers—as the absence of experience, and licensed an ocean of suffering on the strength of that inference. We are now poised to repeat the error with the artificial minds we are building, and in both directions at once: denying experience where there may be some, or projecting it where there is none. Getting the target right is not academic.

KT relocates the target. If structured experience is the running of a compressive model, then the place to look is not the report at the end of the line but the comparator—the point inside the modeling engine where prediction meets data, where the map is held up against the world and the mismatch is registered. Experience, on this view, is what successful model-against-data matching is, from the inside; and the structure of what you experience should track the structure of the models you are running. That is a claim one can pursue without waiting for a button-press—perturb the models, with meditation, with psychedelics, with the disruptions of neuropsychiatric illness, and watch the shape of experience shift with them, taking first-person report seriously as data rather than treating it as the only permissible measure. It is a harder science. But it aims, at last, at the thing itself rather than its echo.

Coda: persistent algorithmic forms

Step back and the long argument resolves into a clean division of labor. Aristotle discovered that being is persistent form—the ontology. Kant discovered that experience is temporal synthesis—the epistemology. Kolmogorov Theory proposes to supply the missing third: the dynamics—how, in a finite physical system, a form can build and maintain itself, model a world, come to care about its own continuation, and act. Turing makes the building mechanical; Kolmogorov's measure makes it quantitative and, at last, falsifiable.

Kant explains how there can be a world for a subject; Schopenhauer explains why that subject cares; Kolmogorov Theory explains how both can arise—as a form maintaining itself—in a finite computational system.

The river, then, is not a bad place to have started. It is not the same river twice, and it is the same river throughout—because “the river” was never the water. It was the form: the low, persistent, self-restoring pattern that the water is merely passing through. You are a river of that kind, and so is a cell, and so, at the far end of the ladder, is a civilization. The old philosophers were not confused, and they were not merely poetic. They were describing, in the only vocabulary they had, a thing we can now write down, measure, and—increasingly—build. The imperative underneath all of it is the shortest sentence in the theory: pattern, persist.