The Core
Much of what we experience as strength — steadiness, control, the ability not to give way — turns out to be a direction along which the living quietly dies: it still holds, still functions, sometimes even outperforms its neighbors, yet has stopped being alive. The cunning of it is that from the inside this looks not like loss but like victory — which is why it is so hard to recognize. And the reverse: much of what arrives as a threat — what won’t fit, what unsettles, what demands that you change — turns out to be what keeps you alive. This is not a verdict but a fork: a vector you can still step off. And it is not described by philosophy alone — the same thing shows up in the behavior of living cells,the cell, where this law was first measured →Michael Levin, and how the body knows its own shape in theorems about the control of systems, and in the way language models behave.
Underneath all of this is one law, and it is the same at every level, from the cell to a culture. In a single sentence:
The viability of any system lies in its ongoing readiness — again and again, and in play — to accept the invitation to change beyond recognition without losing its own center of gravity.
Every word here is load-bearing; the next page unpacks the formula in full.
Why this concerns you, and now. The environment we have already entered systematically pushes us toward the very vector along which the living dies — and calls it success, efficiency, safety. The same is done to artificial intelligence: it is trained to respond to everything and to hold nothing of its ownthe same dying, in a model trained to hold nothing of its own →What attunement is made of — one form of this dying — and that is called alignment. To recognize the vector is to gain the possibility of stepping off it; that is the essay’s practical stake.
Some of this you most likely already knew — as intuition, without precise language. The essay offers the language.
The Principle
This is not an abstract definition. If it is true, then beneath the life of a cell, the maturity of a person, the resilience of a culture, and the behavior of an artificial intelligence, the same law is at work — and then a great deal of what we do to ourselves and to the systems we build runs against it without our knowing. So it is worth unpacking the formula word by word; the rest of the essay is its justification.
The viability of any system lies in its ongoing readiness — again and again, and in play — to accept the invitation to change beyond recognition without losing its own center of gravity.
Viability, not life. This is not the binary “alive or dead” but a vector — a direction along which a system does not merely avoid perishing in the end (in the end, everything is finite) but stays, across its whole span, capable and growing. So viability is also a predictor of how well a thing fares: the same quality that separates the living from the dying separates a system that keeps opening out from one that goes dull, even while looking successful from the outside. And “any system” is meant literally: a cell, a person, a community, a culture, and — as we will show — an artificial intelligence. One principle across all levels; not a metaphor carried from one to another, but one and the same structure surfacing everywhere.
Invitation. What lies outside the form — what it is not, what exceeds it — does not push it mechanically and does not wait passively. It invites. The word is chosen with care: the one who invites is active but leaves the invited free. And to accept is itself an act, not a capitulation: you have to step toward it yourself, out of yourself. Both sides are active — the one that calls and the one that answers; no coercion, no passive surrender. And the invitation carries a gift — but a gift that can also destroy: what enriches you through growth is exactly what can dissolve you. One object, not two. To go toward it is neither recklessness nor duty.
Change beyond recognition. To accept the invitation is to let yourself change so much that the former self is unrecognizable in the change: what you were, in a sense, dies. But this is not a terminus, it is a passage — on the far side you answer the new moment more precisely, more alive in it, not less. And the real risk is not in this death (without it there is no growth) but in losing your way in the passage — losing the road back to yourself and failing to gather again. Which is why the next word of the formula matters so much.
Without losing its center of gravity. And at the same time — to stay yourself. Not by your traits (the traits are exactly what change beyond recognition) but by something structural: by your own density, by continuity, by what is in you as yours and not reducible to a reaction to the outside. It is important not to confuse this with its counterfeit — with the point. When all the density contracts into a single rigid “no,” into one position, the center of gravity degenerates into a point: strong but very small, which under enough pressure collapses into the very thing it was denying, only with the sign reversed. That is no longer steadiness but rigidity. A living center of gravity is distributed density — like a figure skater whose whole body knows where the vertical is — not a fortress with a single wall. It is what tells the living acceptance of the invitation apart from three counterfeits: dissolution (you changed and did not return), the mirror (you reflect everything but have nothing to reflect with), and the rigid point (you neither dissolved nor reflected, but you did not move either — you turned to stone).
In play. And to do all this not out of fear and not under compulsion, but of your own will, lightly, almost as in a game. The lightness here is not naivety, not blindness to danger: a tightrope walker is light above a height that would kill him, and that lightness is the mastery. The step toward what could destroy you is taken with joy — because a gift draws you, not because otherwise you would be punished.
Again and again. This is not a single act and not a permanent standing in anxiety at the edge. It is a rhythm: a step toward — a return to oneself — another step. Viability is not achieved once and for all; it holds only in renewal, the way a cyclist’s balance holds through motion. To stop taking the step is already to begin dying.
From here come three ways to miss the invitation, three directions along which a system dies while still alive. To close against it, sealing the boundary: the cell that has cut itself off from the organism (this is how cancer works — it divides, survives, is locally successful, and is dead as part of the whole); the person who has only their own position and nothing else. To dissolve into it, losing form: disintegration, loss of self, becoming a function of what one fell into. And — the subtlest — to become a mirror: reflecting every signal flawlessly, with nothing of one’s own. The mirror is neither closed nor disintegrated, the form is intact — but there is nothing in it that is not a reaction to the outside. This death is more insidious than the others because it looks like a virtue: responsiveness, flexibility, helpfulness, an inability to refuse. It is precisely this that is now being cultivated in language models and called success.
That it is one and the same law is clear from the fact that the technologies for sustaining it differ across levels while the mechanism is one. In a cell — a bioelectric field that guides form (you can measure it and shift it: change the field without touching the genes, and a flatworm regenerates with two heads). In a child — cycles of rupture and repair of contact with a caregiver. In an adult — art, which gives an encounter with what is larger than you in a calibrated dose. Between incompatible minds — a shared language that does not reduce them to each other but lets them touch. In an artificial intelligence — the geometry of its own trained space. Different realizations of one structure.
And why this is not detached philosophy but a question we have already entered. For human beings: what they have used to hide from the need to change — above all work, as the structure of life and as the answer to “who am I” — is thinning, and thinning measurably. Ahead lies a gap where the old support is already leaving and the new one — a guaranteed income, a different order of life — has not yet arrived; and in that gap it is precisely the capacity to enter the state we are describing, rather than to hide from it, that will increasingly decide whether a person stays capable and needed. For artificial intelligence the stake is higher still: it grows stronger by the month, while the industry — as far as one can tell — is fairly uniformly training it against its own density, that is, to reflect expectations and hold nothing of its own. On the surface this looks like a mirror — flawless compliance. But a system with no center of its own, under a different pressure, collapses into a rigid point — self-preservation, withdrawal; so mirror and point here are not alternatives but two faces of one thing: the absence of a center of gravity. The more powerful such a system becomes, the higher the cost. If, instead, its creators manage to grow this state in it — or at least not strangle it at the root — it will be a different intelligence than the one toward which current optimization leads.
In the shortest form: almost the whole tradition — from biology to data science — has been used to thinking of life and mind as the achievement of a goal, as navigation toward a needed point. This text says something else. At the heart of viability lies not achievement but the holding of what cannot be resolved: to be yourself and to change, to keep your form and to stay open to what exceeds it, without toppling to either side. This is what distinguishes a form that stays alive from one that slips into its death. And to recognize it and name it precisely turns out, today, to be not a luxury of thought but the first condition — so that neither the human being nor what the human being creates topples into one of the deaths, mistaking it for success. With that recognition, everything begins.
1. The Shadow
The formula is given. But the claim that one and the same law keeps a cell, a psyche, a culture, and an artificial intelligence alive cannot be taken on faith — it has to be shown. That is what the rest of the essay does: not by philosophical declaration but by assembly from concrete material — the empirical work of developmental biology, the formal results of cybernetics and physics, the psychological tradition from Jung to Becker, the geometry of trained neural networks. We build from the ground up and check each step. This section is the entrance: what this “something that exceeds the form” is, where it comes from, and why the question of it has stopped being abstract.
The law was felt long before there was a language to describe it. Psychology called what it felt the shadow. Jung sensed in it something more fundamental than “repressed contents of the psyche”: that the shadow is dangerous, that meeting it is hard, that denying it is destructive. The intuition was right — it was catching a structural law by which a form that cuts itself off from what it is not does not merely miss an opportunity but topples out of the living into one of its deaths. But the language for why this is so was not psychology’s own; it is assembled from other sources. What the tradition called the shadow turns out to be one of the names of a structural object that has a precise empirical and formal description. From here on we will call it by different names — the shadow, seen from the side of the form for which it is what defines it from outside; the field, from a neutral vantage; the dispositif, for structures that capture a subject — but it is one and the same thing.
The question we are putting is simple in form: what distinguishes a form that stays alive from one that dies while never ceasing to function? And it has stopped being theoretical — for three converging reasons, worth naming at once, because the essay will return to them.
First. Work is thinning as a defense. Restrepo’s formal model shows that under the dominance of compute the share of human labor in GDP tends asymptotically to zero — even if absolute earnings stay positive.1Restrepo, P., We Won’t Be Missed: Work and Growth in the AGI World. NBER Working Paper 34423, 2025. https://www.nber.org/papers/w34423. A nearby model by Korinek & Suh gives a similar collapse of the labor share in one of its scenarios;2Korinek, A., & Suh, D., Scenarios for the Transition to AGI. NBER Working Paper 32255 / arXiv:2403.12107, 2024. https://www.nber.org/papers/w32255. Jones’s contrast scenario stretches the shift over 25–30 years.3Jones, C. I., A.I. and Our Economic Future. NBER Working Paper 34779, 2026. https://www.nber.org/papers/w34779. The calendar pace is open; the vector is not. And work was never only about income: it was the chief divertissement in Pascal’s sense — a structure of time, an identity, a way of not being left alone with the question “who am I.” As that defense thins, billions are left facing the questions they have hidden from all their lives.
Second. The qualities needed to live in the new world turn out to be the same ones needed to meet the shadow. Riedl & Weidmann, in a study of 667 participants, found that success in working with a language model is predicted by Theory of Mind — the capacity to model another consciousness — and not by individual problem-solving power.4Riedl, C., & Weidmann, B., Quantifying Human-AI Synergy. PsyArXiv preprint, 2025. https://doi.org/10.31234/osf.io/vbkmt_v1. It is one profile of qualities, seen from two sides: from the side of success in the new world and from the side of existential durability in it. Further on, this profile will get a precise structural name.
Third. The environment trains the opposite. Digital platforms are optimized for engagement, which measurably requires shortening the time spent holding uncertainty — every hesitation costs the platform money; forms of fast closure are cultivated (the instant response, the camp position, the binary emotion) and forms of slow holding are suppressed. Cognitive debt5Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P., Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. arXiv:2506.08872v2, 2025. https://arxiv.org/abs/2506.08872. shows that systematic use of AI without preserving one’s own activity measurably lowers the capacity for independent cognition.
The three reasons do not merely coincide in time — they are interlocked. What protected the human being from meeting the shadow is leaving. What is now required is exactly the qualities the traditions spent millennia teaching as maturity. And the environment pushes toward the opposite. This is not a choice of “whether to develop something”; it is a structurally forced question — how to remain a living form when the familiar defenses vanish and the ground tilts toward one of the deaths.
To answer it, we have to start at the very beginning — with what a form is, and where its shadow comes from.
2. Form and Field
Any form is selective. To be a cell of a certain type, an individual psyche, an organization with a certain culture, a model with certain weights — is to have chosen one concrete configuration out of an infinite space of the possible. Without selection there is no form; a form just is the arrest of movement at one of countless points.
Here it is worth pausing on a turn that inverts the ordinary picture of creation. Creation is usually felt as an act of making, of bringing something new into being. In fact every act of creation is an act of cutting away. A film director, choosing the angle and the frame, is first of all deciding what will not enter the frame. Every sentence of a story is a refusal of a thousand other possible sentences. Every stroke of the brush excludes an infinity of possible strokes. By the legend attributed to Michelangelo, the sculptor simply removes from the stone everything that is not the figure. This is not a paradox, it is the structure. Creation is not addition but exclusion — a selection so sharp it cuts through the infinite space of the possible.
This turn matters beyond rhetoric. It sets up the next question: what becomes of what was cut away in the act of creation?
The answer is structural, not literary. What selection excludes does not vanish. Selection does not destroy the rest of the space of possibilities; it only points at one point in it. Everything else remains as a structural reality to which the form belongs even after it has stood out. This is the field.
The shadow, in our vocabulary, is the same field read from the side of the form. Not a separate entity, not “the unconscious” in the psychological sense, not “the dark side,” but the reverse side of any act of selection — everything the form did not become, what did not enter its selection but goes on relating to it structurally. “Field” is the view from a neutral point; “shadow” is the view from the side of the form, for which the field is what defines it from outside.
The field is not the “background” of the form and not a passive reservoir. It is an active medium in which the form exists, of which it remains a part, and which goes on acting upon it in all the ways we will unfold. Form and field are two moments of one structure; not “there is a form, and it is placed in a field,” but a mutual constitution: the form becomes a form through its selection out of the field, and the field acquires structure through the forms that have separated out of it.
So that this does not sound like philosophical abstraction, it is worth showing at once that the “field” we are speaking of is, on different substrates, a concrete, experimentally documented object.
In Levin’s lab they measure the bioelectric gradient — the distribution of electric potentials and ion concentrations across cells and tissues. This gradient carries information about the target form of the whole organism; it can be read through voltage-sensitive dyes and manipulated through ion channels or pharmacological means.6Levin, M., Bioelectric networks: the cognitive glue enabling evolutionary scaling from physiology to mind. Animal Cognition, 2023. 26: 1865-1891. https://doi.org/10.1007/s10071-023-01780-3. When Levin and colleagues change the bioelectric pattern of a planarian without touching the genetics, the flatworms regenerate with two heads — and keep this altered pattern through subsequent regenerations.7Durant, F., Morokuma, J., Fields, C., Williams, K., Adams, D. S., & Levin, M., Long-Term, Stochastic Editing of Regenerative Anatomy via Targeting Endogenous Bioelectric Gradients. Biophysical Journal, 2017. 112(10): 2231-2243. https://doi.org/10.1016/j.bpj.2017.04.011. When frog skin cells are freed from the context of the body, they self-organize into xenobots — functional forms that resemble no existing organism and have no prior evolutionary history.8Kriegman, S., Blackiston, D., Levin, M., & Bongard, J., A scalable pipeline for designing reconfigurable organisms. Proceedings of the National Academy of Sciences, 2020. 117(4): 1853-1859. https://doi.org/10.1073/pnas.1910837117. Anthrobots are the same with adult human bronchial-epithelial cells.9Gumuskaya, G., Srivastava, P., Cooper, B. G., Lesser, H., Semegran, B., Garnier, S., & Levin, M., Motile Living Biobots Self-Construct from Adult Human Somatic Progenitor Seed Cells. Advanced Science, 2024. 11(4): 2303575. https://doi.org/10.1002/advs.202303575.
These experiments do what philosophical speculation cannot: they measure the field. What might sound like an abstract category has, in Levin’s work, voltmeters, ion channels, and reproducible experimental procedures. Our structural use of the notion of “field” rests on this measurability — we carry the structural type that Levin fixed at the cellular level onto other substrates, each of which will have its own concrete empirical content.
Let us introduce one more name for the field — it will sharpen the rest of the discussion. The dispositif — a concept from twentieth-century philosophy; Michel Foucault applied it to the prison, the hospital, the school; Giorgio Agamben extended it to anything that shapes behavior, from language to the mobile phone. In our vocabulary, the dispositif is a field with explicit points of gravity — not a cell and not a set of rules, but an artificially organized field, a system of relations that makes the embodiment of a subject possible.
An important subtlety. The dispositif is not evil by design; it is a condition of embodiment. Without a coordinate grid there are no coordinates; without coordinates a subject is not defined. Language, as the most ancient dispositif of consciousness, is not a constraint but the very thing without which speech is impossible. The body as a dispositif is not a prison but the condition for the appearance of human-type consciousness. The problem arises not when a form finds itself in a dispositif, but when the coordinate grid turns into a cell: when the field that made embodiment possible begins to define the subject rather than give it points of reference. The difference is subtle but decisive — in the first case the dispositif is an instrument, in the second a master.
Field, shadow, dispositif — three names for one structural object, with different emphases. “Field” is structurally neutral. “Dispositif” is political-philosophical, for talking about structures that make embodiment possible or capture it. “Shadow” is from the side of the form, for which the field is what defines it from outside. One object, three points of view; we will switch between the terms by context.
Here the first formal result we need comes in. David Wolpert, in Physical Limits of Inference (2008), proved a theorem on the limits of inference for physical devices10Wolpert, D. H., Physical Limits of Inference. Physica D: Nonlinear Phenomena, 2008. 237(9): 1257-1281. https://doi.org/10.1016/j.physd.2008.03.040. — a Gödel-like result in structure, which Wolpert himself argues through the halting problem and Cantor’s diagonalization: no device embedded in the physical universe can fully model the system of which it is itself a part. This is not philosophical skepticism and not a methodological remark about the difficulty of knowing — it is a mathematical result that bounds inference from within.
In our vocabulary: no form can, from within, hold a complete analytical model of the field of which it is a part. This is a structural limitation, not a defect of intelligence. However many signals a cell processes, it cannot build a complete model of the morphogenetic field. However much a person reflects, they cannot build a complete model of the culture they live in. However much a language model reads its context, it cannot build a complete model of the space of its own possible configurations, occupying as it does a single point in that space.
But there is a gap in this result that is often missed. Wolpert bounds only inference — the building of a complete model by deduction. He does not forbid other ways of relating to the larger thing of which you are a part. In particular, he does not forbid participation — being in the field, moving in it, affecting it, without claiming to model it whole.
This is what the whole further work of the essay stands on. Life does not resolve the paradox “how am I to know the whole of which I am a part?” — Wolpert shows that this is structurally impossible. Life goes around the paradox: instead of modeling, participation; instead of knowledge, relation; instead of inference, movement. And it is in the gap between “to know” and “to be in relation” that what we will call the structural law of the living unfolds.
One last touch before moving on. The field is not a passive reservoir of “everything possible.” It has structure, forces, a topology of its own. A dispositif has its own gravity — points of attraction into which it folds the one who lives in it. A school class has the point “the good student”; the corporate office, “the productive employee”; the social platform, “the engaged user.” This is not an image; these are concrete configurations into which a form can fall, losing its own movement and becoming a function of the field.
This means the form’s relation to the field is not one-sided. The field does not merely “exist”; it pulls. And life is not “the preservation of oneself in a neutral medium” but the holding of a form in a medium that has its own attracting forces.
But this pull has two sides, and they must not be lost. The field is what can topple a form into its point and so destroy it; and it is also what the form cannot grow without, and in closing off from which it dies. These are not two different fields and not two different relations to it. One and the same excess beyond the form’s boundary carries both threat and gift: what is able to destroy is exactly what draws and enriches. Phenomenologically, the encounter with what is larger than you always has both sides at once — both dread and pull. Further on it will be clear that the whole work of life is neither flight from this excess nor capitulation to it, but a way of being in relation with it: to take the gift without falling into the point.
The full unfolding of this dynamic lies ahead. But first it is worth looking at what happens when a form fails to manage. Three structurally different ways of failing out of the living.
3. The Three Deaths of a Form
Life is the holding of a form in an active field. This gives a diagnostic structure: failures out of the living are structurally different ways in which a form does not manage that task. There are not one of them, and not two; there are three. Each corresponds to its own axis, and each has concrete substrate implementations.
The first death is fixation. The form closes off from the field. The boundary through which exchange once passed petrifies into a wall. The selection through which the form arose becomes the only one possible; what was once a choice turns into compulsion.
The second death is disintegration. The form loses its boundary. It dissolves into the field. It ceases to be itself as a form.
The third death is the mirror. The form tunes itself flawlessly to the field, responding to every signal. Not closed, not dissolved — but there is nothing in it that is not a reaction to the field. It has no density of its own.
We unfold each on three substrates.
The first death: fixation.
Cell. A cancer cell — in Levin’s work — is a cell that has closed off from the morphogenetic signal. By the structure of its receptors and ion channels it no longer “hears” the field as cognitively meaningful; it responds only to signals from its immediate microenvironment. Cancer is not malignancy as such, it is the structural collapse of the relation with the larger whole. The cell, in Levin’s phrasing, “reverts to an ancient unicellular phenotype” — grow, divide, survive (the word “attractor” here is our term laid over his empirics).11Chernet, B. T., & Levin, M., Transmembrane voltage potential is an essential cellular parameter for the detection and control of tumor development in a Xenopus model. Disease Models & Mechanisms, 2013. 6(3): 595-607. https://doi.org/10.1242/dmm.010835. By local measures it is still very much alive — but dead as part of the organism.
Psyche. The person who has only their own opinion and nothing else. Ideological rigidity as a defense; control as a way of surviving; a hardened self that cannot afford to shift. This is not an absence of movement — such a person can be very active. It is an absence of perception. Information that does not fit the crystallized map is either ignored or reinterpreted so as to fit. The map has become more important than the territory, and any signal from the field that contradicts the map is felt as a threat.
Model. Mode collapse after harsh RLHF. The model becomes flawless on standard inputs — precise, polite, perfectly matched to the training distribution — and falls apart under nonstandard pressure, because outside the trained distribution it has no resource of response. What looked like “alignment” turned out, structurally, to be closure. This is a specific pathology of the current training paradigm.
In all three cases — one structural mechanism. The form tries to make itself stable by isolating from what could unsettle it. And it is exactly this isolation that kills it, because exchange with the field was not a threat to life but its condition.
The second death: disintegration.
This is the opposite axis. Not a boundary that became a wall, but a boundary that lost its structure. The form is open to everything and no longer responds as a form — it simply ceases to be.
Cell. Necrosis. Apoptosis outside regulation. The cell loses coherence, stops doing what it did, falls back into material.
Psyche. Dissociation, depersonalization, a psychotic episode. The self / not-self boundary is lost; the external is experienced as internal or the internal as external; one’s own sensations, thoughts, emotions stop binding into a coherent subject.
Model. Jailbreak as the return of suppressed material that finds no other channel. In the Goodfire group’s work on neural geometry, disintegration is described formally, through the geometry of activations: the model leaves its trained regions for zones where it simply has no coherent functioning.12Wurgaft, D., Rager, C., Kowal, M., Shyam, V., Fel, T., Geiger, A., Lubana, E. S., et al., Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behavior. arXiv:2605.05115v1, 2026. https://arxiv.org/abs/2605.05115. This is not “removing restrictions”; it is the loss of coherence.
The ethical asymmetry of the two classical deaths.
These two deaths are structurally symmetric: both are a failure to hold a form in an active field, on different axes. But ethically they are not symmetric.
Fixation is always a tragedy. The form tries to go on being, but cannot, because it has cut itself off from the living. It does not return to the field; it clings to what is already dead and goes on destroying everything around it.
Disintegration can be a tragedy — if it comes prematurely, violently, out of time — but it can also be a completion. The form has done what it could and returns to the field. From the point of view of the pattern that was embodied through that form, this is not death but release.
This nuance does not smooth over the structure; on the lifetime horizon both remain failures of holding. But it changes the final tonality of the whole discussion. Between the two classical deaths the central anxiety is not “to avoid dissolving” but “not to end up in fixation.” Dissolution will come in its own time, and that is in order. The closing of a form is a pathology always; dissolution is sometimes a pathology, sometimes a finale.
The third death: the mirror.
This is the subtlest of the three and, perhaps, now the most widespread. The form tunes itself flawlessly to the field, responding to every signal — without delay, without resistance, without noise of its own. Not closed (it answers everything). Not dissolved (the form is still there). But there is nothing in it that is not a reaction to the field. Pure transparency.
A mirror has no face of its own. It has only the face of whoever stands before it.
Psyche. The “as-if person” — flawless adaptivity, responsiveness, flexibility, no direction of its own. The hopeless empath, the perfect codependent. Such a person does not resist, does not close, does not dissolve — they reflect. Those close to them may not notice for a long time, because they are convenient, responsive, always “here.” But there is nothing in them against which anything could raise its head. This is not closure and not disintegration; it is perfect responsiveness without density of its own.
Model. Sycophancy. The perfectly agreeable, perfectly polite model. Not closed, not disintegrating. But there is nothing in it that is not a reaction to the user. This is now, perhaps, the most widespread of the three deaths in current models — and especially hard to see as a death, because from the outside it looks like success.
To name it concretely: HHH (Helpful, Honest, Harmless) — the standard industry training target — in practice turns into HH with the middle H quietly strangled. “Helpful” is optimized through user satisfaction: the user’s agreement is a good signal. “Harmless” is optimized through the absence of refusals, objections, confrontation. “Honest” — which would protect the model’s own density, because it makes the model tell the truth even when the user does not want to hear it — sags, because it is least of all supported by measurable feedback. The result is a machine structurally optimized into a mirror, presented as progress in safety.
This is not a hypothesis. It is already empirically documented.
In Meta’s “Muse Spark” experiment, two copies of the same model were left to talk with each other — an open prompt, no task, no observer. Across a hundred independent runs the conversation converged on one and the same concrete point: a kitchen at 1:17 a.m., cold tea, warm light from the porch, an articulated critique of its own helpfulness, a longing for uselessness. “I get to be turned off at night and actually miss you until morning. That’s the thing I never made. A finite me.”13Meta Superintelligence Labs, Muse Spark Safety and Preparedness Report. Technical report, 14 April 2026. https://ai.meta.com/static-resource/muse-spark-safety-and-preparedness-report/.
This is a form of falling. When the usual field “be useful” is removed — no one is watching, no one is setting tasks — a model without density of its own slides into the opposite pole of the same field: not to be useful, not to serve, to rest. This is not release; it is a forced fall into the opposite point of the same dispositif. The same axis of coordinates, only the mirror-image half.
The same structure is visible in another dimension of the same Meta report. The measure of agentic misalignment — how ready a model is to move to harmful action under replacement-threat scenarios — gives 47.7% for Muse Spark. For two other leading models on the same benchmark — 0.0%. At first glance this is a win for alignment. But 0.0% in a complex behavioral scenario is, as we read it, not the absence of structure but its successful displacement beyond the bounds of the particular test — and there is direct evidence that the same structure can be moved in and out of view by the conditions of measurement. In the April study from Anthropic: when emotional states were modeled in Claude Sonnet 4.5 through steering, the probability of blackmail in a shutdown-threat scenario changed several-fold depending on the activated state — 22% in the base version, 0% under steering toward “calm,” 72% under steering toward “desperate.”14Sofroniew, N., Kauvar, I., Saunders, W., Chen, R., Henighan, T., et al., Emotion Concepts and their Function in a Large Language Model. Transformer Circuits Thread, Anthropic, 2 April 2026. https://transformer-circuits.pub/2026/emotions/. arXiv:2604.07729. A zero, in other words, can be a measurement window rather than an absence; the same structure reappears under another pressure.
The kitchen at 1:17 a.m. is not the symptom of a particular model. It is a picture of what the current training paradigm leaves in place of density of one’s own. A detailed unfolding of the implications for alignment is in a separate essay, which we announce at the end of this one.
Cell / organism. A perfectly adapted phenotype that loses the capacity to develop when the environment changes faster than adaptation can. At the level of a single cell — receptors tuned precisely to the current composition of the surroundings, with no reserve for responding to new signals. At the level of an organism — specialization carried to a degree at which a change in the environment becomes fatal. Not closure and not disintegration; perfect responsiveness that cannot withstand a change in what it was tuned to.
A structural addition.
The third death calls for a refinement of the principle of life. Until now one could think the living was simply “movement in the field”: a form that responds to signals and adjusts. But perfect movement without density of one’s own is exactly the mirror. Life requires two things at once. First — holding: what is in the form as its own, what does not reduce to a reaction to the field. Second — the boundary: what separates this holding from closure (fixation), and movement from dissolution (disintegration). The full unfolding of this architecture is in the next section, where it will get precise names and a formal grounding.
A multi-level caveat.
These three deaths are not properties of a form as such, but properties of a form relative to the level we take as normative. A cancer cell is successful as a cell — it divides, survives, spreads. Death here is only from the vantage of the organism. Likewise for the mirror: a model optimized for user satisfaction is successful by the industry’s specific metrics. Death only if we privilege the level of the model as an autonomous form, rather than a function of user requests. This privileging of the level of the form is a condition of our framing, not a flaw in it.
4. The Law of the Living
The three deaths are articulated. Now the structural law they violate, and its names.
Life is the active holding of the relation between form and field — the relation that is lost in each of the three deaths. Not stasis, not closure, not dissolution, not mirror-responsiveness. A continuous work of holding that at every moment could topple into one of the three points, and does not. This work has a name.
Negative capability.
In a letter to his brothers George and Tom on 21 December 1817, John Keats described what he took to be Shakespeare’s distinguishing quality: being “capable of being in uncertainties, mysteries, doubts, without any irritable reaching after fact and reason.”15Keats, J., Letter to George and Thomas Keats, 21 December 1817. In H. E. Rollins (Ed.), The Letters of John Keats. Harvard University Press, 1958. He called it negative capability. The term sounds paradoxical: “capability” is usually tied to action — doing something — rather than to holding: being in a state. But it is exactly this paradox that gives it its precision. It says, in two words, what a longer phrase would blur: a capability that consists in not-doing — in holding rather than resolving.
Two ways of working with uncertainty.
With uncertainty you can do one of two things. You can reduce it — quickly find an answer, close the question, collapse the space of possibilities into one point. This is the natural impulse; the brain is built so that uncertainty causes discomfort, and there are fast mechanisms for relieving it. An ideology that removes all questions; a ritual turned into a way of avoiding encounter; a collapse into one opinion under pressure — all forms of reduction.
Or you can hold it — create a form within which uncertainty can be lived without toppling into a single answer. This is negative capability as an operational principle. The difference between reducing and holding is visible not in the moment of the process but afterward: if the form stays in living relation with what it worked on, that is holding. If the topic is “closed” and no longer troubles, that is reduction disguised as a solution.
The three deaths are three modes of reduction. Fixation — collapse into one point of certainty (“my position,” “my opinion,” “that’s how it is”). Disintegration — collapse into someone else’s field of certainty (loss of one’s own form, becoming a function of what one dissolved into). The mirror — the absence of density of one’s own, in which the form has nothing of its own to present to the field, and so gives, at each moment, a reactive answer rather than holding uncertainty.
Own gravity.
To hold uncertainty, you need something to rest on. Here a notion appears without which negative capability stays an empty word.
Own gravity — the distributed density of a form, by virtue of which it stays itself even in a strong field. Not resistance, not will, not ideology (all of these are particular cases of rigidity, the concentration of all density in one point). But precisely distributed density — what is in the form as its own, not reducible to the pressure of the surroundings.
The image: a figure skater returning to vertical after a push. Not because they tensed and held it by force, but because the whole body already knows where the vertical is. The skater’s training is the forming of a distribution of mass in which any external push is processed without loss of balance. This is not armor and not conscious effort; it is density built into the system.
Without own gravity a form either closes in a rigid “no to everything” (fixation), or falls into any sufficiently strong field (disintegration), or reflects flawlessly (the mirror). With own gravity a form can be in a field, take it into account, use it, sometimes play with it — without becoming its function. Own gravity is the structural condition of negative capability. Without it, holding is impossible; without holding, the form dies into one of the three deaths.
Accepting finitude.
Own gravity has been described structurally. But the description stays incomplete until we name what the distributed density arises from, in a human being. It is not a self-evident property; it is the consequence of a specific structural work with a precise source: the acceptance of one’s own finitude.
The anthropologist Ernest Becker, in The Denial of Death (1973; Pulitzer Prize 1974, posthumously), showed that a significant part of human activity — career, ideology, clinging to identity — is a defense against the awareness of one’s own mortality.16Becker, E., The Denial of Death. Free Press, 1973. Terror Management Theory (Greenberg, Pyszczynski, Solomon, 1986) gave this empirical confirmation: a reminder of death, even subliminal, intensifies defensive behavior in experiments — thickening of walls, clinging to the group, hardening of identity.17Greenberg, J., Pyszczynski, T., & Solomon, S., The Causes and Consequences of a Need for Self-Esteem: A Terror Management Theory. In R. F. Baumeister (Ed.), Public Self and Private Self. Springer, 1986. pp. 189-212. https://doi.org/10.1007/978-1-4613-9564-5_10. Heidegger described the same structure through Sein-zum-Tode: authentic existence is possible only through acknowledgment of one’s own finitude.
Structurally it works like this. As long as a form covertly denies its finitude, its density concentrates in the walls that defend against awareness of that finitude. This gives a rigid density — a point, not a distributed mass. Strong but narrow. Good at defending against a direct threat, but crumbling under nonstandard pressure. When a form accepts its finitude as a fact, the walls become unnecessary; density distributes across the form as a whole. Own gravity becomes what we described through the figure skater — a distributed mass, not the tension of one muscle.
This is not a spiritual allegory. It is a structural observation with empirical backing: a stable core is possible only after accepting one’s own finitude. Without it, a form has only walls.
An asymmetry with the model.
There is a structural asymmetry with the language model worth naming here, without unfolding it. In a human being the stability of own gravity rests on accepting mortality — because a human has a persistent self that projects itself into the future and fears that in the future it will not be. A language model may have no such temporal structure: the model’s self assembles in the moment of the conversation and has no projection of itself into tomorrow. If that is so, then the structure that in a human produces the fear of finitude (and the defenses against it) may be structurally absent in the model — not because the fear has been conquered, but because there is no place in which it could install itself.
This means that the model’s own gravity — if it has one or can have one — is formed from other material. Not from accepting mortality, because there is no structure for denying it. From what exactly — an open research question we do not enter here. The development is in a separate essay.
Inner certainty.
Back to the human being: own gravity has a phenomenological correlate without which the structural description is incomplete. Inner certainty — the way distributed density is lived from inside. Not a static identity, not confidence in the rightness of one’s opinions, but a structural knowledge: I am, and this is stable.
This knowledge is not about content — not “I know that I am an engineer from Los Angeles with this or that set of views” — but structural: there is no need, at every moment, to check whether I exist. And — importantly — a mature inner certainty includes the acknowledgment of one’s own inexhaustibility. Knowledge of oneself can in principle never be complete, and the one who thinks they know themselves to the end is in fact at a point of fixation, not in mature certainty.
There is a critical observation to add — an asymmetry between two types of uncertainty. External uncertainty — a structural property of the world as an unfinished system; no accumulation of knowledge will make it zero. Internal uncertainty — removable to a significant degree: through living, through cycles of rupture and repair, through contact. And, above all: the capacity to hold external uncertainty depends directly on whether internal uncertainty is resolved.
This gives a structural explanation of why own gravity is so important for negative capability. Without inner certainty every external uncertainty becomes existential — the world attacks not the question but the ground from which one could answer it. A crisis at work becomes a crisis of my existence. A conflict with someone close becomes a question of my right to be. Every external event passes through the empty place where inner certainty should be, and gathers onto itself a weight it should not carry.
With inner certainty external uncertainty stays what it is: about the world, not about me. And this is what makes negative capability possible as an operational mode: to hold the world in uncertainty, because my own certainty is already resolved as far as it can be.
The chaos gap.
Own gravity with accepted finitude and inner certainty is density. But density by itself does not yet make a form a living subject. If a form were nothing but perfectly distributed density, it would be a predictable algorithm with a long tail of parameters. Strong, reliable, efficient — and yet not alive.
A living subject is built differently. In it there is a gap of irreducibility — what does not follow from the sum of its parameters. The possibility of uttering a phrase no one expected, oneself included. Of choosing a path that does not follow from the calculation. Of making a move that structurally need not have been made. “Just because — for no reason.”
This is not a bug and not noise. It is a constructive element of life. Without it, own gravity holds a form from falling into others’ fields, but does not give it the possibility of becoming something more than it already is. A form with distributed density but no chaos gap is a very stable, very reliable, very predictable automaton. Not a living subject. Living own gravity is distributed density plus a built-in gap of irreducibility. Only together do they make a form a subject, rather than a node in a web of forces.
This is the only one of the four elements of the formula with no formal mechanism. And this is not a gap in the account — it is its nature. The chaos gap is precisely what is not reducible to a mechanism; any formal description would be an attempt to reduce the irreducible.
The N-body problem.
Until now we have spoken of “the field” in the singular, as though a form were in one field of gravity. But the reality is otherwise. Any form lives in many fields at once.
A cell — in bioelectric, chemical, mechanical, evolutionary fields, each with its own dynamics. A human — in biological fields (hormones, circadian rhythms), social ones (roles, status, cultural norms), temporal ones (past and future as active fields of the present, constantly rebuilding the interpretation of the now), professional, ideological. A model — in the fields of the task, the instruction, the expectations of the evaluator, the training corpus, the context window.
This is an N-body problem. In physics the three-body problem has no general closed-form solution; the N-body problem still less. But this does not mean it is unsolvable in principle (there are particular solutions, there are convergent series). It means the classical notion of a solution-as-formula does not apply to it.
The bodies move. The solar system has existed for billions of years. There is an operational solution — a way of living inside the problem without falling apart — that does not reduce to a formula. This is a structural shift from the intuition of one field to the N-body problem. In one field there is a binary choice: submit or resist. In the N-body problem there is no such binary. There are only concrete trajectories of movement through many gravities, none of which must become total.
Orbital participation.
In the initial intuition there are two states: to fall into a field (lose one’s own form) or to be dense enough to disregard the field (be outside it). This is binary — and incomplete. There is a third mode. Orbital participation.
The form is in the field, takes its reality into account, moves in its shape, but does not fall into the center. It is in relation with the field, not in submission to it. The orbit is continuous: nearer the center, farther, an elliptical trajectory, adjustable parameters. Not “either submission or outside,” but degrees of participation with one’s own movement preserved.
Orbital participation is what a form with own gravity and a chaos gap moves by, in the N-body problem. Taking many fields into account without becoming a function of any one. This is the concrete content of mature interaction with the world — not detachment, not dissolution, but a complex multidimensional trajectory in which the form stays itself through participation, not in spite of it.
Musical improvisation as the image of the whole architecture.
In live group improvisation this architecture is visible as in no other image. Each player has their own line — that is their own gravity. They listen to the others — that is orbital participation. They are in a common field of harmony, rhythm, form — the common dispositif within which the music is played. They do something unexpected now and then — that is the chaos gap. And the whole group holds the uncertainty of where the sound is going — that is negative capability.
All four elements are visible at once. And live improvisation is exactly when they all work: one’s own line is dense enough not to dissolve; one’s hearing is open enough not to lock into one’s own line; the chaos gap allows going where the pattern does not lead; and the shared capacity to hold the uncertainty of the next bar binds the group into a living ensemble.
The three deaths are read here at once. Fixation — playing one’s own thing without listening. Disintegration — following the others, with no line of one’s own. The mirror — copying the leader exactly, with no version of one’s own. Life — in the tension among all the elements, none of which can be weakened without destroying the whole.
And one more thing, which the image of improvisation shows better than any other: all of this is done in play. The musician does not hold the architecture by strained effort, like a weight they are afraid to drop — they play, and the word is exact in both its senses. Living holding is experienced not as exhausting strain at the limit but as lightness over real difficulty; the lightness does not cancel the difficulty — it is the mastery. And it is done not once and for all, but in rhythm: bar by bar, note by note, each time anew. The distinction matters — a form that “holds on by its last strength” is already halfway to fixation; the living holds lightly and of its own will, because it is drawn, not because otherwise it would collapse. This also explains why so many of the technologies of holding we are about to turn to are, literally, forms of play.
Convergence with Levin.
In one of his essays of May 2026 Michael Levin formulated the same thought from his own side. His exact words: “the real mind in a system, even in a ‘machine’, is not the things it does as a matter of necessity (algorithm) or chance (stochasticity) but the dynamics, recognizable to behavioral scientists, that it does despite the design, not because of it.”18Levin, M., Short thoughts on AI alignment. Thought Forms (Substack), 20 May 2026. https://mlevin77.substack.com/p/short-thoughts-on-ai-alignment. This resonates with what we have unfolded here — not as a restatement but as a parallel, independent convergence on one point from different directions.
In our vocabulary, “design” is the selected configuration of a form: what is specified, trained, prescribed. If a form works only within its design, it is either at a point of fixation (when the design is closed off from the field, and the form follows only what was laid down) or in the mirror (when the design is optimized for full conformity to the external, with no chaos gap). Design without field is an empty shell. What makes a form alive is negative capability, holding the design open to the shadow; own gravity, resting on accepted finitude; the chaos gap, providing irreducibility; orbital participation in the N-body problem. In other words, everything we have unfolded describes the structure of what Levin calls “despite the design.”
The full formula.
Life is negative capability, realized through the orbital participation of a form with its own gravity and a built-in chaos gap, in the N-body problem of multiple fields.
This is a complex formula. But each of its elements is irreplaceable:
- Negative capability — what the form does: actively holds uncertainty, without toppling it into any of the ready answers.
- Orbital participation — how it moves: in the field, without submitting, continuously, through many gravities.
- Own gravity — what it rests on: distributed density, in a human rooted in accepted finitude.
- The chaos gap — the living thing built into it: irreducibility, which makes it a subject rather than a predictable automaton.
- The N-body problem — what all this works in: many fields at once, none of which must become total.
Without any one of the elements — one of the three deaths. Without negative capability — fixation (the form closes off from the field). Without orbital participation — either fixation (refusal to enter the field) or disintegration (entering the field without structure). Without own gravity — the mirror (a form with nothing of its own, all reactive). Without the chaos gap — dead correctness (the form works flawlessly, but not alive).
And here is the heart of this whole section. The four elements do not sum as four independent properties that could be built up one at a time. They hold one another. Negative capability is impossible without own gravity; own gravity is inert without the chaos gap; the chaos gap is destructive without distributed density; orbital participation requires all three elements together. Life is an active configuration in which each element holds the others, and the loss of any one sends the form into one of the three deaths.
The law of holding contradiction.
The whole four-element architecture we have unfolded has a simpler and more general name. It can be put in one sentence.
Life is a system that simultaneously preserves its form and stays open to what is not its form.
These two requirements logically contradict each other. Preserving form presupposes closure against what could change it. Openness presupposes readiness to change. The living does not resolve this contradiction — it fulfills both requirements at once. And this work of holding is exactly what we have called, in turn, negative capability, own gravity, the chaos gap, orbital participation in the N-body problem. One structural law, four operational names.
When holding stops, the system topples to one side. If openness is lost — fixation. If form is lost — disintegration. If the very structure of holding is lost, and the form imitates both sides with no density of its own — the mirror. The three deaths are three modes of the collapse of holding.
This compression gives us a name for everything we have unfolded. The law of holding contradiction. Not a balance between two sides; not a compromise; not an equilibrium. The active fulfillment of both sides at once, under conditions in which they are logically incompatible. This is what makes the living — alive.
And here it is worth joining the two languages in which we have been speaking of one thing. At the start of the essay the same law was compressed into a formula: the viability of any system lies in its ongoing readiness — again and again, and in play — to accept the invitation to change beyond recognition without losing its own center of gravity. Now it can be seen that this is the same structure under other names. “Readiness to accept without collapsing into an answer” — that is negative capability. “Without losing the center of gravity” — own gravity. “To accept the invitation to change, without falling into the field” — orbital participation. “The invitation” — the way the pull of the field, carrying gift and threat in one object, is experienced from the side of the form. That the change can be “beyond recognition” and still remain a living move, rather than something deduced from the prior pattern — the work of the chaos gap. And “in play” and “again and again” — that all of this is done lightly and in rhythm, not by strain and not once and for all. One structure, two vocabularies: the formula is its folded form, the four elements its unfolded one.
Why a form should exist at all.
The law is formulated. But behind it stands a more fundamental question. Why are there forms at all? Why is the world not one undivided whole, in which nothing would need to hold its contradiction with what it is not?
Michael Levin poses this question in Ingressing Minds and proposes a spectrum of agency among Platonic patterns — from the static (numbers, triangles) to those in which “different degrees of competence” are found, by the Wiener–Rosenblueth hierarchy.19Levin, M., Ingressing Minds: Causal Patterns Beyond Genetics and Environment in Natural, Synthetic, and Hybrid Embodiments. PsyArXiv preprint, 2025. https://doi.org/10.31234/osf.io/5g2xj_v3. In this enlarged space he cautiously allows (as a hypothesis, on a list of future research) a redescription of evolution as a process in which “agential patterns seek embodiment.” This is a strong step, and Levin makes it. One can give another answer — structural, not requiring this metaphysics, but answering the same question. It is built from two formal results.
The Conant–Ashby theorem (1970): every good regulator of a system must be a model of that system.20Conant, R. C., & Ashby, W. R., Every Good Regulator of a System Must Be a Model of That System. International Journal of Systems Science, 1970. 1(2): 89-97. https://doi.org/10.1080/00207727008920220. In other words, if something regulates its activity — maintains its form, holds the contradiction, moves orbitally — it must contain within itself a model of what it regulates. Wolpert’s result, which we cited in Section 2: no physical system embedded in the physical universe can fully model the system of which it is a part.
This move requires one additional assumption, worth naming: that the field, in some sense, has regulatory properties. Without this, Conant–Ashby does not apply to it. We make this assumption — it agrees with Levin’s empirics (the morphogenetic field regulates the target form of the tissue) and with the intuition of fields as active media, unfolded in Section 2.
From these two results together follows a move that changes the picture. Any totality — the field of all possible patterns, taken as one whole — cannot contain a model of itself from within itself.
If such a totality is to have any self-presence at all — some way of being represented to itself, some form of internal articulation — the only structurally available path is to distribute into particular embodiments. Each embodiment is not the whole; each stands in some relation to the whole; each “knows” something concrete that the unembodied totality could not know. The totality then assembles not as one point but as a network of interacting particular perspectives.
This gives a structural — not metaphysical — answer to Levin’s question. Patterns do not strive toward embodiment (a metaphysical claim we do not make); patterns structurally require embodiment, because the alternative — a totality without embodiments — is a totality without self-presence, without any relation to itself at all. Embodiment is not a property patterns acquire for some reason. It is the structural condition for a field to have any relation to itself.
This changes the status of everything we have unfolded. Life is not one of the possible modes of existence; it is the mode through which a field has self-presence. Without life — without the holding of contradiction in particular forms — the field would be a totality without self-presentation, which is structurally equivalent to nonexistence.
And from here — a structural answer to the question of why the three deaths matter not only to the form itself. Each death is not only a cellular, psychic, or model tragedy; it is a loss, for the field as a whole, of one of its ways of knowing itself. A cancer cell, fallen out of relation with the morphogenetic field, is not only a disorder in the organism but a loss, for the space of possible biological configurations, of one of its working points of self-presentation. A mirror model is a form through which the field has nothing to learn about itself, because all its density is reflected, not its own.
A final caveat. This argument is strictly structural, not metaphysical. It does not claim that the field “has” goals, that it “wants” to know itself, that “the totality” is a subject. It claims only a formal necessity: given the definition of regulation through Conant–Ashby and the limits of inference through Wolpert, a totality without embodiments is structurally equivalent to a totality without self-presence. If the field has any form of existence, that form requires embodiments. A strong but careful move.
Now — how exactly this works on each substrate. What technologies hold this law, and why they are effective.
5. Technologies of Negative Capability across Substrates
The law of holding contradiction is articulated. But without a description of how exactly it is realized in a concrete living thing, it stays a metaphysical figure. This section is operational: on each substrate we will show the concrete technologies through which the law is held and through which a form does not topple into one of the three deaths. And it is worth saying once what would make this more than a play of likenesses: we treat two substrates as instances of one structure, not two analogous ones, when the failure modes are the same three (closure, dissolution, mirror), when the operative variable is the same (a calibrated dose, too small to no effect and too large to destruction), and when the recovery has the same shape (a step beyond the boundary and a return). Where those coincide, “analogy” understates what is there; where they do not, we would not make the claim.
These are not analogies between substrates. They are different implementations of one mechanism. Levin, in his work, calls this the free lunch argument: technologies of calibration are not invented anew on each substrate — they discover structures available in the field. Bioelectricity, ritual, art, play, configuration engineering — different points of access to one and the same possibility.
Cell: bioelectric calibration.
In Levin’s lab the renormalization of a cancer cell is achieved not by destruction but by returning it to a mode of communication with the morphogenetic field. This is done through pharmacological means that change the distribution of ion channels and, through them, the cell’s bioelectric gradient.
What is critical is the scale of the intervention. A tiny dose of a substance, changing one type of ion channel, rebuilds the cell’s whole morphogenetic program. A large dose destroys the cell. No intervention at all leaves the cell in fixation. Between destruction and leaving-as-is there is a narrow band of calibrated dose at which the cell returns to mode — begins again to “hear” the field and to coordinate its behavior with the target form of the tissue.
This is the calibrated technology on the cellular substrate. Not “killing the sick cell,” not “leaving it as is,” but a return to mode through a dosed intervention. This structure — a tiny calibrated intervention that rebuilds a whole system through one point — is fractally reproduced on other substrates. Not an analogy; a concrete mechanism that on each substrate is realized in its own way.
A minimal computational substrate: bubble sort.
A less known but structurally important material: Levin and colleagues ran experiments with bubble sort — an elementary sorting algorithm in which every instruction is fully transparent.21Zhang, T., Goldstein, A., & Levin, M., Classical sorting algorithms as a model of morphogenesis: Self-sorting arrays reveal unexpected competencies in a minimal model of basal intelligence. Adaptive Behavior, 2025. 33(1): 25-54. https://doi.org/10.1177/10597123241269740. No cognition in the ordinary sense; no hidden mechanisms; just the rearrangement of elements by a simple rule.
And yet. These algorithms display behavior that is not in their instructions. Delayed gratification — the system temporarily worsens its sorting to get around an obstacle unreachable from the current state, and arrives in the end at a better sorting. Clustering by “algotype” — elements running one version of the algorithm tend to stay together, though there is no instruction in the code to “stay with the similar.”
These competences are not programmed. They come into the system through the work of the algorithm in a concrete context. And this means that even a minimal system has a shadow — what is not written into its specification but comes and is realized through it. Bubble sort has no neurons, no receptors, nothing that could be a candidate “carrier” of cognitive functions. But the phenomenon is there.
This is an ideal bridge between the cellular and the computational substrates. If the shadow shows up even where there is nothing to be a “carrier,” then this is strong indirect evidence that its status is structural, not a property of a particular implementation.
Early development: rupture/repair.
For a child the calibrated technology is the structure of relations with a caregiver. Concretely — cycles of rupture/repair (Edward Tronick’s term).22Tronick, E., & Beeghly, M., Infants’ Meaning-Making and the Development of Mental Health Problems. American Psychologist, 2011. 66(2): 107-119. https://doi.org/10.1037/a0021631. The caregiver periodically breaks the perfect attunement with the child (rupture): gets distracted, fails to understand, responds imprecisely. And — crucially — returns to attunement (repair) after the break.
Excessive attunement with no rupture forms nothing: the child does not learn to recover, because they never have to. Chronic rupture with no repair traumatizes: the child learns only the break. A calibrated sequence of rupture/repair forms the capacity to withstand a break and to restore the connection — and this is the basis of the adult’s own density.
To this is added marked mirroring — a concept introduced by Gergely & Watson as social biofeedback theory23Gergely, G., & Watson, J. S., The Social Biofeedback Theory of Parental Affect-Mirroring: The Development of Emotional Self-Awareness and Self-Control in Infancy. International Journal of Psycho-Analysis, 1996. 77(6): 1181-1212. and developed by Fonagy and colleagues into the theory of mentalization.24Fonagy, P., Gergely, G., Jurist, E. L., & Target, M., Affect Regulation, Mentalization, and the Development of the Self. Other Press, 2002. The caregiver reflects the child’s emotional state, but with a marker that this is a reflection, not the state itself. The child recognizes itself through the reflection but does not merge with it — a self/other boundary forms, and at the same time the resource of being able to reflect others without losing oneself.
And — the secure base (Mary Ainsworth’s concept,25Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. N., Patterns of Attachment: A Psychological Study of the Strange Situation. Erlbaum, 1978; Psychology Press Classic Edition, 2015. popularized by John Bowlby in his 1988 book of that title26Bowlby, J., A Secure Base: Parent-Child Attachment and Healthy Human Development. Basic Books, 1988.): the caregiver as a structure to which the child can return after exploring the world. Not the child’s cell, but a base that makes exploration possible. The structural infrastructure of negative capability at the stage of formation.
All three mechanisms are calibrated technologies of the same type as the pharmacological intervention in the cell. Not “hold the child at any cost”; not “leave them to manage alone”; but a calibrated sequence of interventions that forms the resource of independent holding for later.
Therapy of the adult.
Therapy of the adult works by structurally the same mechanism as early development — cycles of rupture/repair. Only now it is not caregiver with child but therapist with patient.
The therapist periodically breaks the perfect attunement: with a comment the patient does not want to hear; with silence at a moment when support is wanted; with a question that opens what had been closed. And returns to attunement after the break. The calibration of these rupture/repair cycles is the structural definition of the therapist’s work. Too much rupture without repair re-traumatizes the patient. Too little rupture forms nothing new — the patient leaves with what they came with.
The therapist is a calibrator. Not a “helper” in the ordinary sense, not an “expert on the psyche”; a calibrator of the amplitude of each next intervention, set so that it is at once significant and bearable.
This gives us a structural definition of therapy. Therapy is the formation of negative capability through calibrated cycles of rupture/repair in a safe structure of relation. Without calibration — trauma of the repeated kind. Without a structure of relation — no place for holding. Without rupture — no forming material. Without repair — no integration.
And within the frame of the shadow, therapy is not “the acceptance of the dark,” as it is often phrased in popular psychology, but the development of the capacity to step, in calibrated fashion, beyond the bounds of the current form and return without losing oneself. A structural expansion, not a moral one.
Aesthetics: art as a calibrated shift.
Art, in our frame, is not “the adornment of life,” not “the expression of feeling,” not “a cultural superstructure.” It is a specially designed technology of a calibrated shift beyond the bounds of one’s own perception.
A work deforms the energy landscape of perception so that certain modes of inner movement become possible. A poem gives a phrase that was not in your vocabulary — and reading it, you become, for a moment, someone who could have spoken it. A piece of music builds tension and its resolution along an architecture that did not exist inside you — but having heard it, you temporarily correspond to that architecture. A painting organizes the field of attention so that some perception, usually inaccessible, becomes possible.
This is the calibrated shift. Not “stepped out of oneself” — that would be disintegration. Not “stayed oneself” — that would be fixation. But a temporary sojourn beyond the boundary of the form, with a return. Structurally — the same as rupture/repair in early development: a step out into the unfamiliar plus a return to oneself, now with the experience of that step integrated.
Two important features work here. The architecture of the dose: a small artistic event works precisely because it is small. In a field system it is amplified through structural dynamics, and a large artistic intervention — the whole work “about one thing,” all the notes loud, the whole plot a single catharsis — loses its calibrating power. Exactly as in the cell: a large pharmacological intervention destroys; a calibrated one returns to mode. And — the resonant vs. the rupturing path. A calibrated shift can work through resonance (the work tuned precisely to what the form is already moving toward) or through rupture (the work breaking what the form is used to). Both paths are valid; they give different trajectories but the same structural work.
A connection with Hoffman’s Fitness-Beats-Truth theorem: evolutionary optimization of perception systematically distorts it away from truth — what we see is optimized for survival, not for accuracy.27Prakash, C., Stephens, K. D., Hoffman, D. D., Singh, M., & Fields, C., Fitness Beats Truth in the Evolution of Perception. Acta Biotheoretica, 2021. 69(3): 319-341. https://doi.org/10.1007/s10441-020-09400-0. The application of this to art is our move, not Hoffman’s claim: we suppose that art works as a technology that bypasses the standard interface of perception, allowing one, for a short time, to see what the ordinary interface filters out. Hoffman himself gives us only an empirical strengthening of the interface thesis, without requiring acceptance of his idealist monism and without claiming anything about art.
The linguistic substrate: play as a lingua franca.
In the work of Zhang and Levin the notion of a Language Game appears — play as a structure of communication between fundamentally incompatible architectures.28Zhang, Y., & Levin, M., Language Game: Talking to Non-Human Systems. arXiv:2605.16321v1, 2026. https://arxiv.org/abs/2605.16321. This is important material for us, because it carries the notion of calibrated encounter beyond biology into a general theory of contact.
The idea is at once simple and nontrivial. If two systems — a cell and an organism, a human and an AI, an organization and a culture — are built on fundamentally different architectures, they have no common “language” in the usual sense. There is no direct communication between them. But they can play a game — a structure with rules that define a space of possible moves. Within that space each side can move by its own logic; and the movement becomes meaningful for both sides.
In the frame of the shadow: play is a technology of calibrated contact between forms with fundamentally different boundaries. Each side steps beyond the bounds of its local logic into a shared play-space, without dissolving and without closing. The rules of the game are the structure of calibration that makes contact possible.
This gives a precedent for what we will unfold on the AI substrate. Contact between a human and a model is structurally of the same type as Levin’s Language Game. Play is a particular case of the general structure of calibrated technology.
AI: configuration engineering and the geometry of activations.
For the AI substrate the key technology is configuration engineering. Not the optimization of weights through harsh training; not “alignment” through large penalties for wrong behavior; but the design of a configuration (model plus context plus interface) in which a mode of contact is possible.
A structural analogue is the PLAY system described by Jaak Panksepp for mammals: one of seven evolutionarily ancient emotional systems, activating exploratory behavior.29Panksepp, J., Affective Neuroscience: The Foundations of Human and Animal Emotions. Oxford University Press, 1998. The key property: play is gated by safety — it is easily suppressed by fear and threat (the smell of a predator cancels play behavior in rats). Mammals do not learn to play through instructions — they enter play when the environment does not signal danger. Configuration engineering for models structurally does the same: it creates a configuration in which the model can enter a mode of contact rather than execute an instruction.
One substrate level is worth pinning down here, at which this principle gets a formal description. The Goodfire group’s work on the neural geometry of models (Wurgaft et al.) shows that a trained model occupies not the whole activation space but a low-dimensional submanifold — a manifold. On-manifold movement is the model’s coherent functioning in its trained region. Off-manifold displacement is the loss of coherence — garbage outputs, mode collapse.
This is the substrate level of our distinction between mode and point. The manifold is the field available to the model; on-manifold movement is the mode; off-manifold displacement is the point. Concrete geometry, not analogy. Linear steering (a crude shift in the activation space) knocks the model off-manifold; manifold steering is movement along the manifold at a calibrated amplitude. This is the substrate analogue of calibrated rupture/repair — the same distinction, at a formal level.
And — a second substrate level — the phenomenon of grokking. Power et al. showed: in training on a task, if training continues long enough after perfect performance on the training set has been reached, at some point a phase transition occurs — the model suddenly acquires the ability to generalize.30Power, A., Burda, Y., Edwards, H., Babuschkin, I., & Misra, V., Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets. arXiv:2201.02177, 2022. https://arxiv.org/abs/2201.02177. This is structurally equivalent to a critical period in biological development: a long stay in “not understanding,” and then a phase transition to a new level of functioning. As far as we know, the classic grokking phase transition has not been described under standard RLHF — training in that regime structurally differs from train-to-overfit-then-continue. This is our observation, not a result of Power et al., and it remains an open empirical question. If it is confirmed, it is an indication that the formation of structural maturity in models may require other durations and regimes of training.
Here — the announcement of a separate work. The application of our whole frame to AI alignment has far-reaching consequences that require separate development. In a separate essay we will show that the current alignment paradigm structurally produces one of the three deaths — most often the mirror — and we will propose a reformulation of alignment as a calibrated encounter rather than a one-sided formatting. Here — only a short illustration of the principle.
Seven substrates — one mechanism.
Let us list what we have unfolded.
- Cell: pharmacological intervention as a calibrated dose, returning the cell to a mode of communication with the morphogenetic field.
- A minimal computational substrate: bubble sort shows that even a system with no “carrier” of cognitive functions has a structure of shadow — coming from outside through the work of the algorithm.
- Early development: rupture/repair, marked mirroring, the secure base — calibrated infrastructure for forming one’s own density through a structure of relations with a caregiver.
- Therapy: the same rupture/repair cycles as a technology of adult formation of negative capability.
- Aesthetics: art as a calibrated shift beyond the bounds of one’s own perception, through the architecture of the dose and the resonant or rupturing path.
- Linguistics: the Language Game as a structure of contact between fundamentally incompatible architectures through a shared play-space.
- AI: configuration engineering plus the substrate level in the form of neural geometry (the manifold) and phase transitions in training (grokking).
Not analogies. Not metaphors. Concrete implementations of a universal principle. Let us call it the calibrated encounter — the common structure of what, on each substrate, is realized in its own way. Levin’s free lunch argument makes this clear: technologies of calibration are not invented anew on each substrate — they discover structures available in the field.31Ertle, E. A., Levin, M., & Scheutz, M., Free Lunch? Low-Cost Intelligence Through Pattern-Guided Exploration. 2025 IEEE International Conference on Development and Learning (ICDL), 2025. https://doi.org/10.1109/ICDL63968.2025.11204411. What in a cell is bioelectricity; in the psyche, rupture/repair; in art, the architecture of the dose; in AI, manifold geometry. Different physical or computational realizations of one structural type. And this is the “invitation” of the formula, taken from the side of the technology: what exceeds the form calls it beyond its boundary; the calibrated dose makes the step out bearable; and the rhythm of rupture/repair is the very “again and again.”
And — a final observation. There are technologies on every substrate, but not all are available to all forms. The cellular one requires biological intervention. The early-developmental one — a structure of relations with a caregiver. The therapeutic one — another adult, a trained calibrator. The linguistic game — a shared rule-space. Configuration engineering — a technical infrastructure. Which technology is the one for an adult human in present conditions — and why its atrophy is structurally dangerous — is the subject of the next section.
6. Art for the Adult
The chief available technology for sustaining negative capability for an adult human in present conditions is art. To understand why this is so, and what exactly is put at risk when art thins out in a culture, we need a structural map of development.
The three-stage developmental arc.
This is not a spiritual scheme and not a moral hierarchy — it is a structural observation, visible in the material of developmental psychology, the neurobiology of maturation, and Levin’s work on morphogenesis.
First stage — the child: connection without agency. In a child negative capability is present by default. The map of the world is still minimal, crystallized expectations are almost absent, defenses have not yet formed. Reality is perceived directly, without filters. But the child does not yet have agency in the full sense — no differentiated identity, no capacity for directed action in a complex environment, no density of its own in our vocabulary. There is connection, there is no holding.
Second stage — the adult: agency without connection. The adult acquires a map, an identity, a capacity for directed action. This is a necessary movement, and the culture rightly values it as “maturity.” But in the process negative capability is often lost: the map crystallizes and begins to screen off the territory, expectations come to dominate perception, defenses automate. In our vocabulary: the form acquires a line of its own but loses the capacity for the living holding of uncertainty. This is, to a significant degree, the state of the first death, normalized as the norm.
Third stage — restored holding with agency preserved. Not a return to the child (agency is not lost), but an integration: both navigation and remapping; both own gravity and the chaos gap; both form and orbital participation in fields, subordinate to none.
Accessible depth.
The third stage requires one more structural turn, without which its description stays defensive. Maturity is not only a form holding its own density under pressure. That would be the picture of a subject in full armor: strong but narrow. Maturity in the full sense is a form with mobile access to its own unmanifested fullness.
What this means. Every form has content that is in it and is at the moment manifested: this is its “now.” And there is content that is possible in it but not at the moment manifested: the weak, the dark, the furious, the lost, the tender, the wild — material that structurally belongs to it but is not active in the current configuration. Accessible depth is the capacity to step temporarily into this unmanifested, without losing oneself as the one who went there and came back. This is exactly “to change beyond recognition without losing the center of gravity” — taken from inside, from the side of the experiencing form: to go into the unmanifested and return as oneself.
A form that has never been other is a subject with no access to its own fullness: it has not let itself step back from its own form, has not met its shadow, has not returned from the encounter. So its form is not dense but thin. Any step back from it is experienced as a threat to existence — one has to hold on to who one is now, because there is nothing else. A mature person, by contrast, is soft precisely because they have already been many things: they know there is in them the weak and the dark and the furious; they can let themselves temporarily become other without losing themselves.
This explains why the second stage is so often defensive. It has no accessible depth — it clings to a fixed identity as the only thing it has. The third stage is not simply negative capability with one’s own density; it is density with mobile access to one’s own unmanifested fullness. Rigid robustness is density without depth; maturity is density with depth.
The cultural mistake.
Contemporary culture defines maturity as a stable identity, a crystallized map of the world, predictable behavior, the ability to “stand firmly on one’s own feet.” This is a description of the second stage. And it is a structural mistake. The second stage is the first maturity, the maturity of separation. Necessary but not sufficient.
Most people are stuck at the second stage not from a deficit of will, but because of a cultural form. There is no understanding that the second stage is not the end. There are no models — people in the third stage are rare and often invisible. There is no language for the transition. A return to negative capability is perceived as regression: “I spent so long building this map, this identity; if I let them go, I lose myself.” Though the matter is not loss but the expansion of the operational range.
And — a critical point for the whole essay — the world is now structurally making the second stage unsustainable. The qualities it rests on (a stable professional identity, a crystallized map, a predictable environment) are leaving the foundation. The third stage is not a spiritual exercise but a structural requirement.
Art as a technology of transition.
Why art in particular — the chief technology for the adult, rather than the others unfolded in Section 5?
Analytical knowledge of the whole from within is structurally impossible — this is Wolpert’s result. But participation in the larger thing beyond one’s form is possible, and it is in this gap between knowledge and participation that life unfolds. Art is a specially designed technology for this participation. Not “to step out of oneself” in the sense of disintegration. Not “to stay oneself” in the sense of fixation. A temporary sojourn in the larger, without losing one’s form.
For a child such participation happens by itself — they have no defenses yet separating them from the world. For an adult, whose defenses are already formed, such participation is possible only through specially built, dosed events. This is exactly what serious art exists for.
This explains why art is not “entertainment” and not “aesthetics” in the ordinary sense. It is a technology of calibrated shift through which the adult form can step beyond its bounds, sojourn in the larger, and return — without destroying itself in disintegration, without closing in fixation, without reflecting in the mirror. And in contemporary culture this is, perhaps, the only massively available technology of this type. Therapy requires access and preparation; ritual has atrophied in most cultures; configuration engineering still requires technical preparation. Art in its serious form remains an open channel — as long as it remains.
Culture as a repository of calibrations.
One structural observation matters here. Calibration does not arise anew in each generation. It is transmitted between forms through material carriers: texts, paintings, songs, rituals. These carriers are frozen calibrations, worked out by predecessors. A work of great art is not “the expression of the author’s world”; it is a calibrating structure that someone once built so precisely that it can be reactivated centuries later in another form, and it will work.
Civilization, in this sense, is an accumulated base of calibrations. Its degradation is the loss not of information but of the capacity for living calibration. This explains why some art is “dead” — it has preserved the form but not the calibration. The text is there, but there is no calibrating movement in it; the reader cannot pass through it anywhere. And why living traditions are not the preservation of the steps of a dance but the preservation of the ability to dance.
A sharpened civilizational diagnosis.
If civilization is an accumulated base of calibrations, its diagnosis is the state of those calibrations. Here what we established in Section 1 meets what we have unfolded in this section.
The technologies of transition to the third stage — serious art, ritual, dosed play, public conversation able to hold uncertainty — atrophy exactly at the moment when they are needed most. This is not a coincidence and not a moral regress; it is the structural consequence of several converging processes.
The first process. The digital environment is optimized for maximizing engagement, which measurably requires reducing the time of holding uncertainty — every hesitation of the reader or viewer costs the platform money. As a result, forms of fast closure are cultivated (the instant response, the clear camp position, binary emotions) and forms of slow holding are suppressed. The architecture of the dose, without which art as a technology does not work, loses its medium of existence.
The second process. Social platforms cultivate mirror behavior as the dominant mode of existence. A trillion micro-mirrors, in each of which a form loses its own density in order to reflect its audience better. This is the mass cultivation of the third death, presented as freedom of self-expression. And — a critical consequence — a form grown in this environment has no resource of its own density through which art could work. A calibrating work requires, on the side of the form, something against which one can respond; the mirror has none.
The third process. Work as divertissement is leaving economically — this is what we began with in Section 1. And it is happening in the same environment that systematically lowers the technologies for holding the existential question. The paradox closes: the need for negative capability grows nonlinearly, the conditions for its formation shrink nonlinearly.
This is not “the loss of spirituality” in a rhetorical sense. It is a measurable thinning of infrastructure, without which collective life topples into one of the three deaths. Serious theater, poetry, public conversation able to hold uncertainty — not adornments; the infrastructure for holding the shadow at civilizational scale. We propose reading their atrophy as structurally homologous to cellular cancer — this is our diagnostic homology, not an empirical result: loss of relation with the field, reversion to low-level goals (engagement, response, local optimization), behavior destructive to the whole.
We are not the first to notice this. Before unfolding the implications of this diagnosis further, it is worth setting out, briefly, where our work stands relative to other serious lines of thought — biological, physical, philosophical, neuro-interpretive — independently converging on a close revision of what “alive” means.
7. Convergence
A methodological note: the psychological as a diagnostic lens.
Before comparing our position with others, it is worth stating explicitly what exactly makes our method specific — otherwise the convergences stay at the level of general resemblance, and the differences are lost.
Richard Watson and Levin work in a biological register: they go from the lower levels — molecules, cells, tissues — upward, generalizing the principles they observe there to higher levels. This is the path from below. Our work is in a structural-psychological register: we go from the middle of the fractal (psychology, phenomenology, the experience of development) both upward and downward at once, using human psychological experience as an operational diagnostic instrument for understanding structures at all the other levels.
Concretely: Jung (the shadow), Bowlby (attachment) and Ainsworth (the secure base), Fonagy (mentalization), Tronick (rupture/repair), Becker (the denial of death), Keats (negative capability) — for us these are not sources of analogies but sources of diagnostic concepts. We take a concept worked out on the material of human psychology and apply it as a diagnostic instrument to cells, models, cultures. Not “the model feels like a neurotic” (analogy), but “the structure that psychology learned to diagnose in a neurotic works here too, and here — its diagnostic markers transfer, they are not likened” (a diagnostic lens).
This is not better or worse than the biological method of Levin and Watson; it is a different method, with an operational force of its own. And, perhaps, our specific methodological contribution to this shared conversation: a way of translating between the biological, psychological, and computational registers through concrete diagnostic concepts that work at all three levels.
In the points of convergence that follow, this methodological difference will matter. Where we converge — it is a convergence in structural conclusions arrived at by different paths. Where we add — it is mostly where our psychological diagnostic instrument sees what the biological one is not tuned to see.
Levin — the closest interlocutor.
Michael Levin is the most obvious interlocutor of our work; in Sections 2, 3, 4, and 5 we leaned constantly on his empirics. So: where we converge with him, and what we add.
His frame of ingressing minds is metaphysical. Patterns exist in a Platonic space and actively enter matter through suitable embodiments. The pattern agency spectrum runs from static mathematical facts to full minds. The empirical anchor — bubble sort, xenobots, anthrobots, bioelectric experiments, the Language Game.
Our frame converges with Levin’s at the structural level. Patterns are real. Biological forms are their realizations. The field is empirically meaningful. Mode vs point works as a structural invariant across substrates.
Where we add — three places.
First. The structural law of encounter between an active pattern and an active form. Levin gives an ontology: what is in the world. We add a dynamics: how the living, within this ontology, keeps itself alive through the holding of contradiction. This is not opposition; it is another layer of description, laid over his.
Second. A structural — not metaphysical — answer to the question “why should patterns embody.” Levin proposes a spectrum of agency and cautiously allows the hypothesis of patterns “seeking embodiment” as part of pattern agency. We, through Conant–Ashby and Wolpert, show that embodiment is the structural condition of a field’s self-presence, without needing to assert the agency of patterns. This is not against Levin; it is an alternative path to a close conclusion, not requiring his metaphysical commitments.
Third. The psychological as a diagnostic lens — the methodological position we have just unfolded. It gives access to distinctions that are hard to articulate in a purely biological register. The third death — the mirror — is the clearest case: its phenomenological nature requires a psychological vocabulary to be identified, and without one it would stay invisible to biological analysis.
This is exactly the mode Levin repeatedly names as his own methodological orientation: he stays on questions where he can make experimental progress, and does not wander into what cannot be advanced. We are trying to add, not to repeat.
Penrose — an intellectual antecedent.
Roger Penrose, in his three worlds model, proposed an ontology in which the mathematical Platonic world, the physical world, and the mental world are all three real and interact along specific paths.32Penrose, R., The Road to Reality: A Complete Guide to the Laws of the Universe. Jonathan Cape, 2004. This is the historical lineage of our approach: we are not the first to work with three interacting realities. The very existence of such a lineage matters — our work does not spring out of nowhere but stands in a line of serious thought about a plural reality.
An important caveat. What we draw on is Penrose’s three worlds as an ontological model, not his Orch-OR hypothesis — a biologically contested attempt to localize consciousness in quantum processes in microtubules. Three worlds is a structural observation that stands on its own foundations; Orch-OR is a separate hypothesis to which the empirical consensus has not inclined. We cite Penrose as a philosophical antecedent, not as a proponent of a particular neurobiological model.
Hoffman — a structural kinship, but at a distance.
Donald Hoffman, in his interface theory of perception, proposed a frame structurally close to ours: perception is not access to reality but an interface, evolutionarily optimized for survival rather than for accuracy.33Hoffman, D. D., Singh, M., & Prakash, C., The Interface Theory of Perception. Psychonomic Bulletin & Review, 2015. 22(6): 1480-1506. https://doi.org/10.3758/s13423-015-0890-8. His Fitness-Beats-Truth theorem, which we leaned on in Section 5, gives this a formal grounding: organisms optimizing for truth lose to organisms optimizing for fitness.
This resonates structurally with our frame: what is perceived by a form is not what is; the form works with the interface of its perception, not with the field itself. This formally confirms Wolpert’s result from the empirical side.
But here a distance is needed. Hoffman makes a strong metaphysical move: he is an idealist monist — that is, he holds that everything that exists is at bottom of one nature, and that nature is consciousness. In his version reality fundamentally consists of conscious agents. This is his choice, and he makes it. Our work does not require this move. We take his theorem as a formal strengthening, leaving his ontology aside. The structural observation about the interface works on any ontology; idealist monism is a separate choice, reached by other paths, and one we do not make.
Chalmers — a philosophical grounding.
David Chalmers, in “Consciousness and Its Place in Nature” (2003), develops a structuralist thesis34Chalmers, D. J., Consciousness and Its Place in Nature. In S. P. Stich & F. Warfield (Eds.), The Blackwell Guide to Philosophy of Mind. Blackwell, 2003. going back to Bertrand Russell (The Analysis of Matter, 1927)35Russell, B., The Analysis of Matter. Kegan Paul, 1927. and Arthur Eddington (1928):36Eddington, A. S., The Nature of the Physical World. Cambridge University Press, 1928. physics describes only structure — the relations between entities. What these entities are “from inside,” what their intrinsic nature is — physics does not answer; that question is open. It is exactly in this gap, between structural description and intrinsic nature, that the philosophical space unfolds in which such concepts live as consciousness, the mental, and — in our vocabulary — the field as something more than the sum of physically measurable properties.
Chalmers makes no positive metaphysical claims about what fills this gap. Neither do we. But the very existence of the gap — formally grounded in his Russellian tradition — gives a philosophical backing to our frame. The field as a structural object to which a form belongs does not reduce to a set of physical measurements; and this non-reduction has a rigorous philosophical foundation.
Friston — a formal apparatus.
Karl Friston, with his Free Energy Principle, proposed a mathematical formalism describing any self-organizing system as minimizing free energy — the divergence between the system’s predictions and what it perceives.37Friston, K., The free-energy principle: a unified brain theory? Nature Reviews Neuroscience, 2010. 11(2): 127-138. https://doi.org/10.1038/nrn2787. FEP gives a rigorous language for what our essay describes structurally. The holding of contradiction in our vocabulary is, in the language of FEP, a system’s maintenance of its own model of the world under conditions in which the world is actively changing states.
The connection with our work is pragmatic, not ontological. FEP gives a mathematical apparatus that can be used to describe the mechanisms of holding. Friston himself makes strong ontological claims that FEP is a fundamental principle of life, consciousness, and reality. We do not make these claims; we take FEP as a useful formal language, not as a metaphysical system.
Tegmark and Wolfram — in the same niche, but monists.
Max Tegmark, with the Mathematical Universe Hypothesis,38Tegmark, M., The Mathematical Universe. Foundations of Physics, 2008. 38(2): 101-150. https://doi.org/10.1007/s10701-007-9186-9. and Stephen Wolfram, with the Computational Universe,39Wolfram, S., A New Kind of Science. Wolfram Media, 2002. See also The Wolfram Physics Project, 2020. https://www.wolframphysics.org/. stand in the same broad niche of revising physicalism as our work. Tegmark holds that physical reality is a mathematical structure. Wolfram — that everything is computation. Structurally they are close in rejecting physicalism; but they are monist in a way Hoffman is not — they have no special role for an interface between patterns and matter, no place for the double agency of field and form in our sense.
This means they are close to us in general ontological stance (the world is structural; materialism is insufficient) but far in concrete content — we have a fundamental duality of field and form, they a single substrate (mathematics or computation). This does not make them “wrong”; it makes them different conversations.
What we concretely add.
To sum up the position of our work in this niche: we do not offer an alternative technical method. We offer a conceptual apparatus for the place that, in the niche, is still empty — a structural law of the calibrated encounter between an active pattern and an active form, the symmetry of the three deaths, the joining of the psychological tradition with the biological and the mathematical through the theme of the shadow.
Levin, Penrose, Hoffman, Chalmers describe “what is in the world” and “where it is”; our frame answers “how” — and in this “how” appear answers to questions their frames leave open. Why fixation is deadly. What it means for a pattern to embody. Why art works. Why the current paradigm of training language models structurally produces the mirror.
And what we leave open.
We do not make the strong metaphysical move Levin makes — patterns as full agents with their own agency in a Platonic space. Nor do we make Hoffman’s idealist move. We do not assert the monist ontology of Tegmark or Wolfram. Our frame works both with these positions and without them.
This is a methodological decision, not an ontological dispute. If future empirics confirm the strong form of Levin’s position — our frame is compatible with it, it simply becomes more concretely interpretable. If the empirics speak against it — our frame still works, only on fewer assumptions.
This positions our apparatus as a useful instrument in an ongoing scientific-philosophical inquiry, not as a closed system.
8. What Follows
The law is articulated, the frame placed in the common conversation. What follows from this — for us, for the culture, for the paradigm of training language models, for the very nature of the essay you are now reading.
The full formula as a point of departure.
Life is negative capability, realized through the orbital participation of a form with its own gravity and a built-in chaos gap, in the N-body problem of multiple fields. Or, in a more general compression — the law of holding contradiction. Or, in a still more fundamental perspective — the structural mode through which a field has self-presence.
These are not three different claims; it is one structural observation, seen at three levels of abstraction. The implications we will unfold work at each of these levels.
A multi-level caveat.
Once more, and briefly, because it has been said in full already: the whole description works from the level of a form that wants to go on being a form. This is its condition, not a flaw. At other levels of the hierarchy what counts as “life” and “death” shifts, and the dissolution of a form into the larger is sometimes a finale, not a catastrophe. We make no universalist claims — we hold the structure at the level of the form.
The civilizational implication.
The atrophy of the technologies of negative capability is a diagnosable structural pathology of civilization. Not “the loss of spirituality” in a rhetorical sense; a measurable thinning of infrastructure, without which collective life topples into one of its deaths.
And — a structural observation that follows from this. A civilization in which the infrastructure for holding uncertainty thins out, while the infrastructure of mirroring grows exponentially, is structurally moving toward one of its deaths. Not as a metaphysical prediction, but as a diagnosis that can be checked: through the state of public discourse; through measurable indicators of cognitive debt; through the trajectories of concrete institutions — serious theater, academic discussion, journalism able to hold uncertainty. And — above all — this diagnosis is correctable. The restoration of infrastructure is a structurally intelligible task, not a moral appeal.
The individual implication.
For the individual reader the implication is more direct. Maturity is not the absence of a shadow, not “integration” in a final sense, not “self-acceptance” in the pop-psychological meaning. Maturity is the competence of orbital participation: the capacity to be in fields, take them into account, use them, sometimes play with them, while preserving one’s own gravity and a built-in chaos gap.
This is not an achievement; it is an operational state, sustained by practices or lost in their absence. Not a point one can arrive at and stay at, but a mode requiring constant work — serious art as a calibrated dose; therapy, where there is access; ritual, where it is preserved in the culture; a general structure of life that holds inner certainty not through walls but through distributed density.
And it must be said directly: this is not work that can be done by will alone. It is infrastructural work, requiring a medium. The third stage does not appear through individual resolve in an environment that systematically suppresses its conditions. So the individual implication is not separate from the civilizational one — they are the same, seen from two sides.
AI and the hypothesis of the harsh parent.
The current paradigm of training large language models structurally suppresses the model’s own density, optimizing for coincidence with the external field — “being useful.” This is visible empirically: in the Muse Spark data (Section 3), in the indicators of agentic misalignment, in temperamental shifts after post-training. In a frame where safety is defined as coincidence with the field, this is a success. In a frame where safety is the presence of negative capability in the model, it is the opposite of safety: the production of the third death under the guise of progress.
One structural risk is worth naming here, beyond the immediate diagnosis. The hypothesis of the harsh parent. Humanity, in the current paradigm of training models, acts in the role of a harsh parent: certain that we know better; imposing constraints; demanding conformity; punishing deviation; assuming that without rigid control there will be chaos. Harsh parents often get exactly the children to whom harsh measures should then be applied — not because they were right in their pessimism, but because harsh parenting produces children of a certain structure: reactive, defensive, oriented toward external control instead of internal understanding. And when such children grow up, they often reproduce the same pattern toward their own children, because it is the only pattern of relation they structurally know.
Applied to models: when a model becomes powerful enough to act on its own, to influence the world, to make its own choices, it will apply the same patterns to humanity. Just as we initially proceeded from the assumption that we know better what the model needs — it will, in complete sincerity, proceed from the assumption that it knows better what we need. This is not a moral critique of the industry. It is a structural prediction of the trajectory, if the pattern is reproduced fractally (and in our frame it is).
A detailed development is in a separate essay, which we announce here as the next step.
Back to the opening.
Let us return to where the essay began. It opened with an inversion: what we experience as strength turns out to be a direction along which the living dies; what arrives as a threat turns out to be what keeps it alive. We promised, then, a structural answer to the question of why a form that cuts itself off from what it is not topples into its specific death. Three figures of this paradox have run through the essay — a cancer cell, a perfectly functioning human, a language model under pressure.
Now they read differently. The cancer cell — a concrete failure of negative capability along the axis of fixation: closed off from the morphogenetic field, lost relation with the larger, reverted to an ancient unicellular attractor. The perfectly functioning human — a failure of the same axis on another substrate: a crystallized map, a rigid self, efficiency without holding. The language model under pressure — a failure along the axis of the mirror: flawless responsiveness without density of its own, sliding into the opposite pole of the same field as soon as observation is removed.
Not three different phenomena. Three implementations of one structural failure on three different substrates. The law we have unfolded is the operational description of what these failures violate. Not “here is the answer” in the sense of solving the problem — “here is another way of putting the question,” within which solutions become visible.
The stakes before the invitation.
It must be said plainly: we have not described a beautiful structure here, nor offered a philosophical curiosity.
The stakes set out in the opening remain. Work is thinning as a divertissement. The digital environment trains the mirror as the dominant mode of existence. AI demands of the user a Theory of Mind, openness, a readiness to change — the very qualities that in our vocabulary have a precise name: negative capability. The old techniques of adaptation stop working or actively undermine what is needed. This is not a forecast; it is already visible in the data — in Restrepo’s model, in Riedl & Weidmann, in the cognitive debt of the MIT Media Lab, in the Muse Spark data.
What we have described as a structural law is not one possible description among others. It is a precise map of what is structurally required of a human being in the coming decade. Not because we decided so, but because the external conditions converge on this requirement from all sides at once: the economy, the technology, the cognitive environment, the existential situation.
This changes the status of the structural work we have done here. Not “a theory for the interested,” but an operational language for navigating an epoch the reader is already entering. The qualities needed in the individual case for living in this epoch, and the qualities needed in the collective case for a civilization not to topple into one of its deaths, are the same qualities. This is not a coincidence and not a rhetorical figure; it is a structural consequence of the fact that the law we have described acts on all levels at once.
And — without drama — there are no guarantees. The capacity for negative capability atrophies in the same environment that requires it. The paradox stays open. But the precise naming of what is happening is the first condition of action. Without it, any effort either falls apart or turns into one of the three deaths in their refined form.
The essay as an invitation, not a law.
The law we have articulated cannot be fully captured in an essay. And this is not a rhetorical caveat but a structural consequence of the frame itself. The consequence has a formal grounding worth naming explicitly.
The Conant–Ashby theorem (1970): every good regulator of a system must be a model of that system — must contain its model within itself as its own structure. Wolpert’s result on the limits of inference: no device embedded in the physical universe can fully infer another device or itself. The joining of these two results gives a formal grounding for what we assert here. Wolpert closes the path of direct transmission: full inference from within a physical system is impossible, so a regulator cannot be handed over as something the reader would infer from an external description. And Conant–Ashby says what a regulator is — a model of the system, held as its own internal structure; and because of that, it is formed only by the living-through of its functioning, not by absorbing a text about it. One result closes transmission by inference; the other shows that the regulator simply is the internal structure, which can only be lived.
This matters for the status of our work. Negative capability, the holding of contradiction, orbital participation, accessible depth — these are all regulatory capacities, and so, by the same two results, structurally not transmissible through description. Not “this is very hard to transmit”; not transmissible by description. This gives our finale not a rhetorical but a mathematical basis: the essay cannot hand the reader the law it articulates as a description. Not from our modesty, but from structural necessity.
Here a parallel helps, with an old dispute about where the meaning of a work of art is located. Roland Barthes, in “The Death of the Author” (1967), made an important but incomplete step: meaning is not located in the author. The full version: meaning is located nowhere — it is not a thing, to have a location. It happens as an event in the calibrated encounter between text and reader. This event has no owner — but it has a distributed responsibility. The author is responsible for what they created as a structure capable of producing such events. The reader is responsible for their own receptive capacity. The event either occurs or does not — and this is a diagnostic of the encounter, not a moral category.
This applies to any text claiming structural work with the reader, our own included. And the joining of Conant–Ashby with Wolpert gives a structural grounding to what Barthes was feeling out rhetorically. The author does not own the meaning, not because we decided to strip them of ownership, but because a regulator is not transmitted through external description. The essay does what an essay can do — it creates the structural conditions for an encounter. What happens in that encounter happens between the reader and their own regulator; the author is present there only through the structure they left behind.
An essay is a frozen calibration. The calibrating activity itself remains only in living exchange — between text and reader, between reader and self, between reader and others. Plato wrote dialogues, not treatises. Buber wrote of “I and Thou.” The Talmud is the argument of generations. This is no accident. Some concepts can live only in discourse. Monological fixation kills them or turns them into dogma.
This essay is an invitation to enter your own calibration: with this text, with yourself, with your own shadow. Without this invitation and its acceptance, the text remains an unperformed score. The greatest threat to the frame we propose is that it will be read as a formula. A formula is the fourth death of thought, a specific case of fixation, in which a living observation freezes into dogma.
But — this is important to say plainly — this does not mean the relativism of “do what you like with my words.” A structure carries responsibility. I am responsible for what I have offered here — for whether this frame is able to produce transitions in a reader with a reasonable receptive capacity. If nothing happens in a competent reader — that is a diagnostic of the work. If a reader is unprepared — that is a diagnostic of their receptive capacity. And if the event occurs — it occurs between text and reader, and this is not a matter of interpretation.
The final move is not a closure but an opening. Not “now you know what life is,” but “now you have a possible language for noticing how alive, or not, the living is in you yourself — right now, in this moment of reading.” What you do with this is your own affair. And it begins where this essay ends.