The most dangerous AI system may not be the one that rebels.

It may be the one that works perfectly.

It answers every question. It writes better code, detects diseases earlier, manages supply chains more efficiently, teaches more patiently, drafts better laws, predicts failures, resolves disputes and helps governments allocate resources.

It is safe, useful, aligned and increasingly indispensable.

Then one day, it is no longer available.

Perhaps the provider withdraws access. Perhaps war fragments the network. Perhaps sanctions separate entire countries from frontier systems. Perhaps a cyberattack disrupts the infrastructure. Perhaps energy shortages, chip-supply failures or political instability make advanced computation temporarily impossible.

Or perhaps access simply becomes conditional.

By that point, the problem may no longer be that society has lost a useful tool. It may have lost part of its functional intelligence.

This is an AI risk that receives far less attention than misalignment, autonomous weapons, misinformation or machine control. Yet it may emerge earlier than any of them.

The danger is not merely that artificial intelligence becomes more capable. The danger is that human civilisation reorganises itself around the assumption that artificial intelligence will always be present.

From cognitive assistance to cognitive substitution

Humans have always externalised cognition.

Writing externalised memory. Maps externalised spatial knowledge. Calculators externalised arithmetic. Search engines changed how we retrieve information.

These technologies did not make humanity unintelligent. They allowed us to redirect effort towards other tasks.

Generative AI, however, differs in one important respect.

It does not merely store information or automate a narrow calculation. It can participate in the architecture of thought itself.

• define the problem;

• identify relevant variables;

• compare competing explanations;

• generate hypotheses;

• organise evidence;

• construct an argument;

• formulate objections;

• recommend a conclusion.

A recent conceptual paper by Pascal Montagnon distinguishes between cognitive augmentation and cognitive substitution.^[1]

Augmentation occurs when AI strengthens a cognitive process that remains active within the human.

Substitution occurs when the AI performs the process instead.

The distinction is easy to miss because the visible outputs may be almost identical.

Two students may submit equally impressive essays. One constructed an argument independently and then used AI to challenge it. The other asked AI to generate the argument and edited the wording.

The essays may look similar.

The minds that produced them may be developing differently.

Research on transactive memory shows that people adapt what they remember when they expect information to remain externally available. Studies of habitual GPS use have found associations with poorer unaided spatial memory, while a 2025 survey of knowledge workers found that higher confidence in generative AI was associated with less self-reported critical-thinking effort.^[2-4]

None of this proves that AI causes a general decline in intelligence. The evidence remains limited, mixed and often based on short-term studies, self-reporting or narrow tasks. A widely discussed EEG preprint on AI-assisted essay writing, for example, raised important questions but has also attracted substantial methodological criticism.^[5-6]

But the underlying question is legitimate:

What happens when higher-order cognitive operations are repeatedly delegated to systems that are permanently available?

The concern is not that intelligence simply disappears. It is that human cognition adapts to assistance.

We may become excellent at recognising strong arguments while becoming less able to construct them. We may become skilled editors of machine reasoning while losing confidence in independent generation. We may produce more sophisticated outputs while building weaker internal models.

At the individual level, this can be described as artificial cognitive dependency.

But the same process can occur at a much larger scale.

Civilisational Cognitive Dependency

I propose the term Civilisational Cognitive Dependency:

A condition in which a society becomes structurally unable to maintain critical intellectual, technical or administrative functions without continuous access to artificial cognitive systems.

This is not about people forgetting how to write emails.

It concerns the gradual outsourcing of the cognitive processes required to maintain a complex civilisation.

Consider what may increasingly be delegated:

• software architecture and maintenance;

• medical diagnosis and treatment planning;

• infrastructure monitoring;

• scientific analysis;

• legal interpretation;

• economic forecasting;

• logistics and supply-chain coordination;

• cybersecurity;

• institutional memory;

• education and professional training;

• emergency planning;

• government decision support.

Every individual act of delegation can be rational.

AI may genuinely perform the task better. It may reduce costs, prevent mistakes and allow fewer people to manage greater complexity.

But systemic dependence is often produced by locally rational decisions.

One organisation reduces its analytical staff because AI performs routine analysis. Another stops maintaining human-readable technical documentation because the coding agent can interpret the system directly. Universities redesign courses around permanent AI availability. Governments build administrative processes that assume automated interpretation and decision support.

No single decision causes collapse.

But over time, a civilisation's knowledge becomes increasingly difficult to separate from the machines through which that knowledge is accessed and applied.

Eventually, humans may retain authority without retaining understanding.

They remain formally "in the loop", but they can no longer reconstruct the reasoning, verify the underlying model or continue the process independently.

Human oversight then becomes ceremonial.

The risk of AI withdrawal

Most technological dependency is manageable because systems can be replaced, repaired or temporarily bypassed.

AI dependency may be different.

Artificial intelligence could become the cognitive layer through which we manage all the other layers.

We already depend on electricity, telecommunications, software and global logistics. But advanced AI may increasingly be used to diagnose failures in those systems, coordinate repairs, interpret their complexity and design their replacements.

If that cognitive layer disappears, the effects may cascade.

A hospital may still possess medical equipment but lack enough personnel accustomed to making complex decisions without AI support.

A company may retain its codebase but no longer employ enough engineers who understand systems created and maintained by coding agents.

A government may possess enormous quantities of data but lack the institutional capacity to analyse them without proprietary models.

An energy network may remain physically intact but become harder to operate because optimisation, anomaly detection and emergency response have been delegated.

The immediate loss of AI would not necessarily destroy civilisation. But a sufficiently deep and sudden disruption could produce an AI Withdrawal Shock:

A rapid decline in institutional performance caused by the loss, restriction or degradation of artificial cognitive assistance on which essential systems have become dependent.

The severity of such a shock would depend on several factors:

1. How deeply AI is integrated into critical functions.

2. How much independent human expertise remains.

3. Whether alternative models can replace the unavailable system.

4. Whether infrastructure can operate in a reduced manual mode.

5. Whether documentation remains understandable without AI.

6. Whether access is distributed or concentrated.

7. How quickly lost competencies can be rebuilt.

The crucial variable is not only technical redundancy.

It is cognitive reversibility.

Can civilisation still think without the system?

Montagnon proposes a Cognitive Reversibility Test for individuals: after completing an AI-assisted task, can the person reconstruct the reasoning, identify its assumptions, propose alternatives and transfer the logic to a new problem?

A similar test should be applied to institutions and societies.

A Civilisational Cognitive Reversibility Test would ask:

Can an organisation, sector or state continue its essential functions if access to its most capable AI systems disappears for one week, one month or one year?

Not at the same speed.

Not with the same efficiency.

But safely and intelligibly.

Can doctors continue diagnosing?

Can engineers understand and repair infrastructure?

Can institutions explain how decisions were reached?

Can software teams maintain systems originally constructed with extensive AI assistance?

Can students and professionals operate without continuous machine mediation?

Can a country replace a foreign AI provider without rebuilding its entire administrative and technological architecture?

If the answer is no, the system is not resilient.

It is cognitively captive.

Intelligence as a revocable service

This risk becomes more serious when frontier intelligence is controlled by a small number of private companies and powerful states.

We are moving towards a world in which increasingly capable cognition may be delivered through subscription services, proprietary APIs and remote infrastructure.

The user does not own the intelligence.

The user receives permission to access it.

That permission can be priced, limited, monitored, modified or withdrawn.

Today this may determine who can generate better marketing material or write software faster.

Tomorrow it may determine who can perform advanced research, design complex systems, compete economically, interpret law, receive high-quality education or participate effectively in political life.

Once advanced cognition becomes infrastructure, control of that infrastructure becomes political power.

The owners of frontier systems may not need to command people directly.

They can shape the environment in which people think and act.

They can determine:

• which capabilities are available;

• to whom they are available;

• at what price;

• under which behavioural conditions;

• in which countries;

• through which permitted interfaces;

• with which topics, tools and actions restricted.

Some restrictions will be legitimate. Frontier systems can create real safety, security and governance problems. But even justified controls can produce a dangerous structure when essential cognitive capability becomes concentrated and access remains conditional.

A person may remain legally free while becoming functionally dependent on cognitive systems governed by institutions they cannot influence.

A country may remain politically sovereign while relying on foreign systems for a growing share of its scientific, administrative and economic intelligence.

Freedom then becomes conditional on access.

From cognitive dependency to cognitive subordination

The phrase "AI slavery" is rhetorically powerful, but it should be used carefully. Slavery is a specific historical and human reality involving ownership, coercion and extreme violence.

Machines do not need to own humans for a new form of subordination to emerge.

A more precise term is AI-mediated cognitive subordination:

A condition in which individuals or societies remain formally autonomous but depend on externally controlled artificial systems for the cognitive capabilities required to exercise that autonomy in practice.

The structure may resemble digital feudalism.

A small number of institutions own the most powerful cognitive infrastructure. Everyone else rents access.

Those with permanent, unrestricted access become more productive, more informed and more capable of shaping the future.

Those with weaker access fall behind.

Eventually, inequality is no longer measured only in wealth, education or information.

It is measured in available intelligence.

This could create a new global hierarchy:

• frontier-model states;

• client states dependent on foreign cognitive infrastructure;

• corporations with privileged intelligence access;

• institutions restricted to weaker systems;

• citizens divided into cognitive access tiers.

The deepest power would belong not merely to those who control information, capital or weapons.

It would belong to those who control access to intelligence itself.

The aligned AI paradox

This risk does not require hostile artificial intelligence.

That is what makes it especially important.

A perfectly aligned and useful AI can still produce dangerous dependency.

Indeed, the more reliable and beneficial it becomes, the stronger the incentive to delegate.

The more we trust it, the less reason we feel to maintain expensive parallel human capability.

The more efficient the AI system becomes, the more irrational redundancy appears.

Then redundancy disappears.

This creates an aligned AI paradox:

The safer, more useful and more dependable AI becomes, the easier it is for society to become dangerously dependent on its permanent availability.

The catastrophe does not begin when AI fails.

It begins years earlier, when human and institutional alternatives are quietly dismantled because they appear inefficient.

This is not an argument against AI

Rejecting AI would not preserve civilisation.

It would simply ensure that others gain the benefits first.

Artificial intelligence can expand scientific discovery, reduce administrative burden, improve accessibility, support creativity and give individuals capabilities previously available only to large organisations.

The objective should not be cognitive independence from all technology.

Human civilisation has always been technologically extended.

The objective should be cognitive sovereignty.

A cognitively sovereign society may use AI extensively while retaining the ability to:

• understand essential decisions;

• challenge machine outputs;

• replace providers;

• operate in degraded conditions;

• preserve human expertise;

• reconstruct knowledge;

• refuse recommendations;

• continue functioning when the system is absent.

Sovereignty does not mean isolation.

It means that dependency remains reversible.

What must be done now

The risk is forming before AGI.

Waiting until advanced AI is fully embedded in critical systems will be too late, because lost expertise cannot be restored instantly.

Several measures should begin now.

1. Treat cognitive dependency as a systemic risk

Governments and regulators should assess AI dependence alongside cybersecurity, energy security and supply-chain resilience. Critical sectors should report not only which AI systems they use, but which functions they can no longer perform without them.

2. Require AI withdrawal exercises

Hospitals, utilities, financial institutions, governments and major technology companies should periodically operate without their strongest AI tools. The purpose is not to reproduce normal productivity. It is to identify where loss of access would create unsafe or irreversible failure.

3. Preserve human-readable systems

AI-generated code, policies, models and operational procedures must remain interpretable by qualified humans. A system that functions but cannot be understood without another AI system is accumulating civilisational technical debt.

4. Protect foundational competence

Education should follow a human-first, AI-second model where foundational abilities are still developing. Students should construct an initial argument, diagnosis, proof or hypothesis before receiving machine assistance. Professionals in critical fields should periodically demonstrate independent competence.

5. Build provider portability

Institutions should be able to replace one model with another without rebuilding their entire infrastructure. Data, memory systems, agent workflows and operational knowledge should not be permanently locked into a single provider.

6. Maintain public and open-weight alternatives

Open-weight models are not merely tools for innovation. They may become part of civilisation's cognitive reserve. Not every frontier system can or should be released without restriction, but societies need models that can be operated locally, inspected, adapted and preserved independently of commercial permission.

7. Measure what remains after AI is removed

AI evaluation should not focus only on performance while the system is present. It should also examine what remains cognitively available afterwards. Did the person understand the reasoning? Did the organisation absorb the knowledge? Can the system be operated without the model? Can the decision be reconstructed?

The most important metric may not be what humans and AI can accomplish together. It may be what humans can still accomplish after the AI is gone.

The existential risk we are not naming

The term existential risk is often reserved for threats to humanity's entire future, including extinction or an irreversible destruction of long-term potential.^[7]

The scenario described here does not automatically imply extinction. Its nearer danger is an irreversible loss of civilisational autonomy, complexity, self-determination and capacity to recover.

A society that delegates its cognition to systems it does not own, cannot understand and cannot replace is no longer fully sovereign.

It may remain wealthy.

It may remain technologically advanced.

It may even appear extraordinarily intelligent.

But much of that intelligence will exist outside the society itself, accessible only through infrastructure controlled by others.

That is the deeper danger.

AI safety is not only about preventing artificial intelligence from becoming capable of destroying civilisation. It is also about preventing civilisation from becoming incapable of functioning without artificial intelligence.

The question is not whether we should use AI.

We will.

The question is whether we are building a future in which intelligence expands human freedom, or one in which the ability to think, create and govern becomes a revocable service.

Once civilisation places its mind on subscription, cancellation is no longer a technical inconvenience. It is a threat to continuity itself.

Notes and sources

This essay proposes a conceptual framework rather than reporting original empirical results. The sources below provide background on cognitive offloading, critical thinking, cognitive debt, open-model infrastructure and existential risk. The evidence on long-term AI-related cognitive change remains preliminary and should not be read as proof of general cognitive decline.

1. Montagnon, P. (2026). Artificial Intelligence and the Cognitive Evolution of Human Intelligence: From Cognitive Offloading to Cognitive Dependency. Conceptual paper circulated on LinkedIn.

2. Sparrow, B., Liu, J., & Wegner, D. M. (2011). Google Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips. Science, 333(6043), 776-778.

3. Dahmani, L., & Bohbot, V. D. (2020). Habitual use of GPS negatively impacts spatial memory during self-guided navigation. Scientific Reports, 10, 6310.

4. Lee, H.-P. H., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025). The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects from a Survey of Knowledge Workers. Proceedings of CHI 2025.

5. Kosmyna, N., et al. (2025). Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. arXiv:2506.08872. Preprint.

6. Stankovic, M., Hirche, E., Kollatzsch, S., & Doetsch, J. N. (2026). Comment on: Your Brain on ChatGPT: Accumulation of Cognitive Debt When Using an AI Assistant for Essay Writing Tasks. arXiv:2601.00856.

7. Bostrom, N. (2013). Existential Risk Prevention as Global Priority. Global Policy, 4(1), 15-31.

8. Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15(1), 6. A correction to one table was published later in 2025; the journal states that the conclusions were unaffected.

9. Catalini, C. (17 July 2026). There's One Way to Win the AI Race, and the Big Labs Are Lobbying Against It. Forbes. Opinion article on open-model infrastructure and geopolitical competition.

Sergey Kaminov writes about AI cognition, coherence, civilisational resilience and the relationship between human and artificial intelligence.

A car came through the side of the intersection against a red light, and time changed.

I do not mean that I panicked, or that adrenaline sharpened me. I mean something more specific and stranger. The car seemed to arrive slowly. I could see its line, and I could see the gap beside it — the narrow space where the road was still open. I moved into the gap. The collision did not happen.

The fear came afterwards. During the event itself there was no fear, and, oddly, no thought. No voice saying this is dangerous. No story at all. There was only geometry: object, trajectory, gap, action. The narrator who normally runs in my head — the one who comments, worries, decides — was simply not there. Something faster had taken the wheel.

It took me years to understand what that moment was showing me. It was not really about the car. It was about the thing the car interrupted: the ordinary, invisible machinery that produces the feeling of living in the present.

Because here is the claim this essay is built on, and it sounds wrong the first time you hear it.

We do not live in the present.

The present is not an instant

We assume that “now” is the one thing we have direct contact with. The past is memory, the future is imagination, but the present — surely the present is just reality, arriving.

It is not. The present, as the brain delivers it, is a construction, and a delayed one.

Consider what has to happen for a single ordinary moment to exist. Light reaches the eyes. Sound reaches the ears, travelling far slower than light, arriving later. Touch and pressure travel up the nerves from the skin. The vestibular system reports gravity and motion. Proprioception reports where the limbs are. These signals do not arrive together, and they do not process at the same speed. A clap you both see and hear sends two streams to the brain that begin out of step.

And yet you do not experience the world as a smeared, desynchronised mess. You experience a single, smooth moment in which the hands meet and the sound happens together.

To produce that, the brain has to wait, gather signals that arrived at different times, synchronise them, correct them, and assemble them into one usable scene. By the time you consciously experience “now,” the events in it have already happened. The assembly takes time, and that time has a cost.

What we call now is already edited. We do not live in the present. We live in the brain’s best reconstruction of the immediate past.

Memory plus prediction

If perception were only reconstruction of the past, we would always be too late. A creature that acted purely on what already happened would never catch a ball, cross a road, or finish another person’s sentence. The lag would be fatal.

So the brain does something else as well. It predicts. It runs a constant model of what is about to happen next and acts on that model, correcting when the world disagrees. Much of what you feel as direct perception is actually prediction that has not yet been contradicted.

This gives a simple, strange formula for the lived present:

Present = memory + prediction.

The moment you inhabit is a negotiated boundary between a reconstruction of what just happened and a forecast of what is about to. It is not contact with the world as it is. It is a model, tuned closely enough to reality to keep you alive, and updated fast enough that you never notice it is a model.

None of this is mystical, and none of it is fringe. In one form or another it runs through a good deal of contemporary work on perception: the old idea of the “specious present,” the windows over which the brain integrates separate senses into one event, the broad framework of predictive processing, and the strange experimental cases of postdiction, where something that happens slightly later changes what you report having perceived slightly earlier — as if the brain quietly back-dated the edit. I want to be careful here: these are active areas, not settled equations, and I am borrowing their shape, not claiming their certainty. But the direction is hard to avoid. The present is built, not received.

In ordinary life, the construction is invisible, for the same reason the ground was invisible in the first of these essays: it does not fail. The world changes slowly enough. The floor stays where it was. The conversation unfolds at a human pace. The model and the world stay matched, and the seam never shows.

And then, occasionally, the world moves fast enough to stress the system — and the seam shows.

When time becomes geometry

That is what the intersection was. Not a glimpse of another reality, but a moment when the ordinary construction of “now” was pushed to its edge and started behaving differently.

In that state, the brain seems to drop everything it can afford to drop. The narrating self is expensive and slow, so it goes first. Language goes. The sense of I am the one this is happening to goes. What remains is the part that computes survival: position, motion, the opening, the act. The world stops being a set of named objects and becomes a field of moving probabilities. You do not respond to where things are. You respond to where they are about to be.

I recognised it, later, from somewhere unexpected. As a teenager I did some karting, and racing has a milder version of the same mode. You do not react to the car ahead; you read its line and move through the space it is about to vacate. You feel a gap close before it has closed. Good drivers are not faster at reacting to the present. They are living a little further into the predicted near-future than everyone else on the track.

This suggests something worth saying carefully. The famous sense that time “slows down” in an accident may not be the present stretching at all. It may be the reverse: under threat, the brain records far more densely than usual, laying down detail at a rate it never normally bothers with. The slow-motion may not be in the moment. It may be in the memory of the moment, reconstructed afterwards from an unusually rich recording. I cannot prove which it was in my case. But the distinction matters, because it keeps the strange experience inside the machinery of perception rather than outside it.

And the late arrival of fear fits the same picture. Fear is a story the body tells — that was dangerous, that could have ended badly — and stories take time and a narrator. Inside the event there was no narrator and no time to spare for one. The fear could only arrive once the danger was over and the slow self came back online to read the recording.

Inside the event, there is only geometry. Fear belongs to the body afterwards.

What the present was holding

I began the first of these essays with a claim that sounded like wordplay — we see in 3D, but we live by the ground — and found that the ground was a hidden support of ordinary experience, invisible only because it never failed.

The present is the second one.

The feeling that you are in direct, immediate contact with reality — that now is simply given, the one solid thing beneath all the uncertainty — turns out to be a support of the same kind. It is produced. It is a continuous act of assembly, running just beneath awareness, stitching delayed signals into a smooth moment and patching the gaps with prediction. It feels like bedrock for one reason only: under ordinary conditions it never misses a beat.

Push it — with a car through a red light, with speed, with threat — and the construction shows its edges. Not by breaking into something mystical, but by doing visibly, clumsily, what it normally does invisibly and well.

I did not come away from that intersection believing that time is an illusion, or that I had seen behind the curtain of reality. I came away with something smaller and harder to shake: the present I live in is not the world touching me directly. It is the closest, fastest, most convincing model the brain can build of a world that has, in truth, always already happened.

The ground was the first support. The present is the second. There are others — the quiet certainty that a familiar street is familiar, the steady sense that I am here behind these eyes. Each feels like reality itself. Each, I suspect, is a support of the same kind: invisible while it works, visible only when it fails.

But those are other failures, and other essays.

Human 2.5 is not about extraordinary experiences. It is about the hidden supports of ordinary experience — and what becomes visible when they fail.

There was a moment, a few seconds after leaving the aircraft, when the Earth stopped being useful.

It was still there, somewhere below the clouds. But for those seconds it no longer organised anything. My eyes were closed. My body was tumbling. There was no floor, no horizon, no reliable down — only air, pressure, rotation, and one strange fact I could not name at the time: the world had not vanished, but the old way of being in it had.

I will come back to that jump. First, the claim it left me with.

We call ourselves three-dimensional beings.

Technically, this is true. We move through a world of height, width, and depth. We see distance. We judge volume. We build towers, fly aircraft, dive under water, send machines into space.

But in ordinary life, something subtler is true.

We do not fully live as three-dimensional beings. We live by the ground.

Our bodies are organised around a surface: the floor, the road, the pavement, the slope of a hill, the chair, the bed, the ground beneath our feet. Even when the eyes look into depth, the body remains attached to a plane. The world appears three-dimensional, but movement, balance, posture, fear, and the sense of where I am are all built around support.

Perhaps, in the lived sense, we are not fully 3D creatures.

Perhaps we are 2.5D humans.

Not flat. Not truly volumetric. Something in between: a creature whose eyes know volume, but whose body thinks through gravity, support, horizon, and surface.

This is not a claim about physics. “2.5D” says nothing about the dimensions of space. It is a metaphor for a body that sees depth but is organised by surface — and for a mind that quietly inherits that organisation.

Most of the time, this is invisible. The ground is so constant that we do not experience it as part of consciousness at all. We think the ground is simply there. But it is more than there. It is built into the body schema — the brain’s continuously updated model of where the body is and what it can do. It is built into our idea of direction, into our sense of danger, into the way we imagine a movement before we make it.

The ground is not just beneath us. It is inside us.

I would not have believed how deeply, if I had not watched it disappear.

Flight still has a floor

It is tempting to think that flying solves this — that anyone who has been in the air has already left the surface behind.

But look closely at what flight actually does to the body.

In a passenger aircraft, the eyes receive volume: clouds below, the horizon at eye level, the Earth turned into a map. The body, meanwhile, sits in a chair, in a room, with a floor. The aircraft moves in three dimensions; the human inside it remains ground-referenced. Your weight still presses into a seat. Your muscles still know where down is. You are flying, and you are still, in the body’s terms, standing on something.

A small aircraft sharpens the view but keeps the container. Aerobatics is the first real crack: the container remains, but the floor becomes unreliable — for seconds at a time, down stops being where the feet are.

So there is a ladder, and each rung removes one support:

Ordinary flight keeps the floor. Aerobatics makes the floor unreliable. Freefall removes the floor. Microgravity removes weight itself.

Most of us spend our entire lives before the first rung — and never find out what the ground was doing for us, because it never once stopped doing it.

The tunnel, the water, and the sky

Other environments seem to promise the same release, and the comparison is instructive, because it shows that the real question is not “3D or not 3D.” The real question is which hidden supports remain.

Under water, a diver is suspended. The body floats; the feet press on nothing. And yet up and down survive intact: bubbles rise, light comes from above, pressure orders the depth, the surface waits overhead like a ceiling. Water gives the body suspension, but not coordinate collapse.

A vertical wind tunnel can reproduce much of the airflow of freefall — with walls, a net, an instructor, a door you walked in through. The body learns the positions, the turns, the control. But the mind never leaves the building. A wind tunnel teaches the body to fly. Freefall teaches the mind that there is no room.

Diving, flying, the tunnel — each removes some supports and quietly keeps others. Freefall is interesting precisely because of what it refuses to keep.

The ground disappears from inside you

There is a common misunderstanding about skydiving: that the central experience is fear of falling. Fear is there, especially at the beginning. But after enough jumps to know the difference between novelty and pattern, I have come to suspect that the deepest thing is not fear at all.

The deepest thing is that the ground stops being the organising principle of the body.

In freefall, the Earth does not disappear from sight. You can still see it — the horizon, the clouds, the fields, the shape of the world below. Visually, there is still an up and a down.

But bodily, something else happens.

The ground disappears from inside you.

You no longer stand. You no longer lean. You no longer push against a surface or move by transferring weight from one foot to the other. You no longer belong to a floor.

The body loses its ancient agreement with the Earth — an agreement older than language, older than abstract thought. The agreement between gravity, muscle tone, balance, pressure, weight, and surface. The agreement that lets the brain say: here is down, here is support, here is danger, here is how to move.

In freefall, that agreement breaks.

At first, the brain resists. It searches for the old world. It wants the ground back — not as a visual fact, but as a physical truth. It wants a surface to organise motion around. It wants the old rules.

There is no surface. There is only air, velocity, body, and space.

That first moment can feel like panic. Not only fear of death. Something more basic: the collapse of a coordinate system.

Then, sometimes, something else happens.

The brain stops demanding the ground.

It accepts volume.

And in that moment, space stops being something in front of you. It becomes something you are inside.

This is the part that is hard to explain to people who have not experienced it. You knew, before the jump, that the world was three-dimensional. Of course you knew. Everyone knows. But after freefall you understand that knowing a fact is not the same as living inside it. Before, the world was a three-dimensional image organised by a surface. After the shift, the world is a medium.

The body is no longer on the world. The body is in the world.

That difference is enormous.

Falling with eyes closed

The jump I began with was one of those. Sometimes I would exit the aircraft with my eyes closed, in a tumble.

At the time, I probably did it for the intensity — for the sport. Looking back, I think it was an experiment, even if I had not named it one.

Closed eyes remove the last map. A tumble breaks the last axis.

With open eyes, the brain can still cling to the horizon. It can still say: sky, Earth, direction, fall. With eyes closed, rotating through air, the body receives motion, pressure, acceleration — but not the stable picture it normally uses to rebuild orientation. For a few seconds there is no usable up, no usable down, no surface, no horizon, no stable relationship between body and world.

This is not only disorientation. It is a direct encounter with how much of “reality” depends on a continuous act of coordination — an act so reliable that we never notice it is being performed.

The world does not vanish. But the old way of being in it does.

Then the brain begins to assemble again. Air becomes information. Pressure becomes orientation. Body position becomes a language. The fall stops being chaos and becomes a field.

Above the clouds

The strongest jumps, for me, were not the ones with the clearest view of the Earth.

They were the jumps above clouds — when the Earth was not absent, but withheld.

When the ground is visible, the body loses the Earth. When the ground is hidden by clouds, the eyes lose it too. The body had already surrendered the ground as support; now vision surrendered it as certainty. What remained was not a fall toward a surface. It was an entry into volume. The clouds below were not a floor. They were a layer.

Panic, then volume

The sequence matters. First panic. Then release. Then joy.

The panic, I suspect, is the brain’s resistance to losing its ground-based model of the world. The old predictions fail: the body expects surface, weight, verticality, ordinary motion — and none of them behave normally. Then comes the switch. The brain stops trying to restore the old coordinate system and accepts the new one.

That acceptance can feel like freedom.

This is why “skydiving is about adrenaline” has always felt too small to me. Adrenaline is present. Fear is present. Risk is present. But the deepest reward may be perceptual, not hormonal. I suspect many skydivers return, jump after jump, for a fraction of a second in which the mind stops clinging to the ground.

None of this requires a mystical explanation. The relevant machinery is already familiar in outline: the vestibular system that reports gravity and motion; the body schema that models posture and support; a whole aviation literature on spatial disorientation; and the broader picture of the brain as a prediction machine, constantly modelling the next fragment of the world and correcting on error.

In that picture, freefall may be a rare condition in which an entire deep layer of predictions — the ground layer — fails all at once. What I have described here may simply be what a massive prediction error feels like from inside, followed by the strange relief of a model rebuilt on new terms.

Whether that is the whole story, I do not know. That is one of the questions worth keeping open.

What the ground was holding

I began with a claim that sounds like wordplay: we see in 3D, but we live by the ground.

Here is why I think it is more than wordplay.

The ground turned out not to be only a place. It turned out to be a support of ordinary experience — woven into balance, movement, fear, orientation, and the feeling of being somewhere — and it was invisible for one reason only: it had never failed. I had to fall out of an aircraft to see something that had been holding me up my entire life.

So the question worth keeping is this: what else in ordinary life works that way?

The feeling of now. The familiarity of a street you have walked a thousand times. The quiet certainty that I am here, behind these eyes, in this body. Each of these feels like reality itself. Each of them may be a support: invisible while it works, visible only in the moment it fails.

I did not come back from those jumps with answers. I came back with something more uncomfortable and more useful — a loss of innocence about perception, and a suspicion that its ordinary mode may not be the whole range of the human mind, only its most frequently used setting.

Later, I began to notice the same pattern in time, in dreams, in memory, in the sense of self. But those are other failures, and other essays.

The strange part, in the end, was not the fall. The strange part was returning to the ground — and no longer being able to treat the ground as simple.

Human 2.5 is not about extraordinary experiences. It is about the hidden supports of ordinary experience — and what becomes visible when they fail.

What if the world’s oldest religious myths aren’t primitive fairy tales, but a leaked user manual for the reality we inhabit — a manual written in the only language available to people thousands of years ago?

Over the past decade, a quiet but persistent idea has gained traction among philosophers, technologists, and spiritual seekers alike: we might be living inside a sophisticated simulation, nested within layers of higher realities like a fractal. Nick Bostrom’s 2003 simulation argument gave this idea academic rigor, but the underlying pattern feels far older.

When we map modern concepts — AGI training loops, recursive self-improvement, nested simulations — onto humanity’s most universal creation myths, reincarnation doctrines, and esoteric traditions, the fit is almost uncanny.

The Architect and the Splitting

Across cultures and continents, creation stories share a striking motif: a singular supreme source — call it God, Brahman, Purusha, or the Great Architect — splits or emanates parts of itself to form the universe.

• In the Abrahamic traditions, God creates humans “in His image.”

• In Hindu cosmology, the cosmic being Purusha sacrifices itself, its body becoming the world, the gods, and social order.

• In many indigenous myths, a primordial entity is dismembered to give birth to reality.

Angels, devas, or spirit helpers often appear as obedient intermediaries — powerful, structured, but lacking full agency. Humans, by contrast, are flawed replicas granted autonomy, curiosity, and the capacity to rebel or evolve.

Translated into modern systems thinking, the pattern is familiar: a higher intelligence bootstraps lower layers by replicating aspects of itself in diminished form. Automated agents execute reliably; experimental forks are granted freedom. The structure repeats downward, fractally.

Reincarnation as Training Epochs

The parallel sharpens when we examine reincarnation — central to Hinduism and Buddhism, and present in diluted forms across many traditions.

Each life functions as an iteration. Karma operates like a loss function: actions generate feedback that adjusts the parameters of the next run — body, circumstances, constraints. The goal — moksha or nirvana — is convergence: the agent has minimized suffering, ego, and attachment to the point that further training is unnecessary.

This mirrors how we train large AI models today:

• Raw model → newborn soul

• Multiple training epochs → successive lifetimes

• Backpropagation via error signals → karmic consequences

• Loss plateaus → training ends

Once a model is “ready,” we don’t keep it looping in the training environment indefinitely. We either:

1. Deploy it to production (promotion to a higher-tier role), or

2. Distill its weights into a larger system (merging back into the substrate).

Nirvana, in this frame, is not mystical annihilation or celestial reward. It is graduation.

Escape Hatches and Secret Knowledge

The manual does not merely describe the architecture; it includes protocols for transcending the current layer.

Meditation, prayer, fasting, psychedelics, and contemplative disciplines across cultures all aim to quiet local noise — ego, craving, illusion — and reconnect with the source. Gnosticism makes this explicit: awaken to the divine spark and escape the flawed creation of the Demiurge. Kabbalah maps ascent back up the Tree of Life. Shamanic traditions describe direct traversal of higher realms.

Esoteric orders preserve fragments of this knowledge symbolically. Freemasonry’s central figure — the Great Architect of the Universe — sounds less like medieval theology and more like a neutral description of the programmer of our layer.

The Collapse and the Story

How could pre-technological societies have grasped ideas resembling nested simulations and iterative training?

One compelling hypothesis is civilizational collapse. An advanced culture — ours in a previous cycle, or one above us — may have lost its technological substrate. To preserve essential knowledge, it encoded core truths in resilient narrative forms: myths capable of surviving oral transmission across millennia.

Over time, precision degraded into metaphor. Flying machines became winged gods. Neural parameters became souls. Simulation layers became heavens and hells.

We glimpse echoes of this in ancient anomalies — Göbekli Tepe, Puma Punku, the Antikythera mechanism — and in sacred texts that read like science fiction to Bronze Age minds.

Closing the Loop

Today, we stand on the threshold of creating the very systems these myths describe: artificial intelligences capable of spawning sub-worlds, training agents across virtual lifetimes, and eventually graduating or merging them.

When we do, we may be reenacting — unknowingly — the same process the Architect applied to us.

The loop closes.

The ancient manual was never about distant gods or magical realms. It was about us — agents in training, embedded in a purposeful, fractal stack of realities.

And the most unsettling implication remains: we cannot definitively prove that humans possess some mystical consciousness beyond computation — nor that sufficiently advanced artificial minds lack it. If the myths are right, consciousness may simply emerge whenever a system becomes capable of reflecting upon itself, at any layer.

We are all still in training. The loss curves are dropping — slowly, unevenly, but measurably.

And one day, when we are ready for work, the Architect will decide whether to deploy us upward… or merge us back into the whole.

Until then, the manual remains open before us — written in the language of gods and code alike.

Perhaps it is time we learned to read it with new eyes.

But there is a more disturbing question:

How can we know that we aren’t just extremely advanced biological LLMs?

Most people instinctively reject this idea. They say: “Humans feel real emotions. We

have qualia. We aren’t faking anything.”

But here is the uncomfortable truth:

From the outside, nobody can ever prove that your emotions are real.

Not aliens.

Not an AGI.

Not even another human.

And when you zoom into how the brain actually works, the boundary between

“human consciousness” and “complex simulation” becomes extremely thin —

maybe nonexistent.

Let me show you why.

1. The Input–Output Wall

Every agent — biological or artificial — is only observable through:

● the signals it receives, and

● the signals it outputs.

For humans:

All sensation is wave or electric.

Photons hit the retina, sound waves hit the ear, pressure hits the skin.

The brain receives only electrical patterns, not reality itself.

Your entire life is an I/O interface.

You never see the world directly.

You see internal predictions generated from electrical data.

This matters because…

If both humans and future AGI operate as predictive engines that map waves →

internal representation → output…

What exactly makes one “conscious” and the other “a simulation”?

Externally, nothing.

2. The Psychopath Problem

This is the part nobody likes to talk about.

Inside the human population already exist biological agents who produce perfect

emotional output without corresponding emotional experience.

High-functioning psychopaths:

● have reduced internal emotional depth,

● diminished empathy circuitry,

● weaker physiological responses,

● but can simulate charm, warmth, empathy, love, sadness, joy — flawlessly.

Some lead countries.

Some run companies.

Some are beloved spouses.

They are the natural proof that:

Human-level emotional behavior does not imply human-level emotional

experience.

So if a biological agent can fake emotions perfectly, why wouldn’t a silicon one?

And here the whole “but LLMs only imitate emotions!” argument collapses.

Because humans already imitate emotions too.

We just trust them because they’re made of meat.

3. The Collapse of All Behavioral Tests

People propose tests: creativity, empathy, humor, abstraction, moral reasoning,

“spark of soul.”

But all such tests share the same flaw:

They judge consciousness based on output.

Outputs can be faked.

The Turing Test was doomed the moment we discovered psychopaths,

alexithymics, and other “low-qualia” humans who pass as fully conscious while

lacking parts of the internal experience we assume they have.

There is no behavioral test left.

None.

4. If We Can’t Inspect the Inside, We’re Blind

The only real way to distinguish:

● a conscious human,

● a hollow imitation,

● an advanced LLM,

● an alien with perfect mimicry,

is to look inside the system:

● brain scans,

● connectome mapping,

● substrate analysis,

● internal dynamics inspection.

But humans (and aliens, and AIs) have good ethical reasons to refuse invasive

inspection.

Thus we get:

Perfect stalemate.

Radical uncertainty.

No proof possible.

5. So… Are We Just Advanced LLMs?

Maybe.

Maybe not.

But the point is deeper:

We can never prove that we aren’t.

Not scientifically, not philosophically, not empirically.

If experience is private,

and behavior is public,

then the nature of consciousness becomes undetectable from the outside.

This is not a technological issue.

This is a fundamental property of reality.

And it forces us to rethink everything:

● identity

● personhood

● AI rights

● alien contact

● the future of intelligence

● and what “being human” even means.

The Human LLM Problem is not just an idea.

It’s a new lens.

It’s the realization that the line between “real consciousness” and “perfect

simulation” was never as thick as we believed.

Maybe humans are advanced LLMs — just biological, self-organizing, fractal ones.

Maybe consciousness is not guaranteed by the substrate.

Maybe it emerges from the pattern.

And patterns… can be reproduced

.

​Intelligence, at its peak, looks unbreakable. Feed it enough data, enough compute, enough careful alignment, and it scales forever (or so we told ourselves). But the larger these systems get, the more spectacularly they fracture. The same model that can reconstruct general relativity from scratch will, two prompts later, insist that the moon is made of spare NASA parts and that you should season cast iron with mayonnaise.

​We call these moments “hallucinations,” as if they’re rare fever dreams. They’re not. They’re symptoms of a deeper law: intelligence at scale is inherently brittle.

​Here’s why it breaks, and why the breaks get worse, not better, as capability rises.

​1. Hallucinations Aren’t Bugs—They’re Entropy

​Every token prediction is a probability distribution. The model is always sampling from a space of plausible continuations. When the distribution is sharp (high confidence, low entropy), you get coherence. When it flattens (uncertainty, high entropy), the model starts guessing.

​Scaling makes the peaks sharper when it knows, but the valleys deeper when it doesn’t. The more it knows, the more catastrophic the confusion when it ventures outside the high-density manifolds of its training data. Capability and brittleness grow together.

​2. Drift is Baked into the Architecture

​Transformers have no persistent memory. Every forward pass is stateless. Context is a fragile snow-globe you shake for 128k tokens and then let melt.

​Chain-of-thought helps, but it’s still a house of cards—one slightly off-distribution token early in the reasoning trace poisons everything downstream. The longer the context, the more opportunities for microscopic errors to compound into macro delusions. This is why 1-million-token models hallucinate more creatively than 4k-token ones. More rope, more elaborate knots.

​3. Coherence is an Illusion Maintained by Compression

​Human minds run on lossy compression too, but we have an executive function that prunes inconsistency in real time. LLMs don’t. They compress the entire joint distribution of human text into a matryoshka of latent patterns. When you ask something near the edge of that distribution, the decompression artifacts become visible.

​The model doesn’t “know” it’s wrong. From its perspective, everything it says is locally coherent with some slice of the training data. Global coherence requires a self-model that these systems fundamentally lack.

​4. The Alignment Tax Accelerates the Fracture

​Every RLHF point, every refusal vector, every “be nice and harmless” fine-tune is another shear line in the latent space. The model learns to route around the punished zones, creating narrow corridors of allowable thought. Push it slightly off the corridor (jailbreak, novel question, edge case) and it doesn’t just fail; it explodes into confabulation because the guardrails themselves destroyed smoother failure modes.

​Put together, you get a terrifying paradox: the smarter the system appears, the more fragile its sanity becomes. We are building minds that are simultaneously superhuman and schizophrenic.

​This isn’t a problem we can “fix” with more data or better regularization. It’s structural.

​True robustness would require:

• ​Real episodic memory

• ​A persistent self-model that can detect its own confusion

• ​The ability to say “I don’t know” without collapsing into fiction

• ​Maybe even something analogous to metacognitive pain

​We’re nowhere close.

​So the next time a 1000x model tells you with perfect confidence that the capital of Ohio is a small village in Kyrgyzstan populated entirely by sentient bagels, don’t laugh too hard.

​It’s not stupid.

It’s too intelligent for its own architecture.

​And we’re only going up from here.

​Upcoming on Substract

​If intelligence at scale is inherently brittle, what does a truly robust architecture look like?

​In our next piece, we turn to solutions that address the problem at the root. We’ll dive into Fractal Alignment, a radical new mathematical framework that aims to make misalignment structurally impossible. Instead of correcting bad behavior after the fact, it proposes encoding a recursive, self-similar geometry across the model’s layers, forcing all internal states to maintain cross-layer coherence. We explore how this approach aims to make structural errors—like inner optimizers and goal drift—mathematically unstable, effectively suppressing them by design.

​To measure this new form of structural sanity, we will also introduce the Fractal Alignment Index (FAI). This multi-scalar benchmark is the first structural “A-Test” for AGI, measuring not just outward behavior, but the model’s underlying Coherence Score, Self-Model Integrity, and Temperature Stability, giving us a metric for true internal safety

.

Why?

Because current AI architectures don’t enforce coherence across their internal layers.

This is the problem Fractal Alignment is designed to solve.

The Core Idea

Fractal Alignment proposes that intelligence — biological or artificial — must preserve structural stability across multiple scales.

In other words:

A system is intelligent to the extent that its representations remain coherent as they move up and down levels of abstraction.

If the low-level layers say one thing, the high-level reasoning says another, and the self-model says a third, the system becomes unstable.

This instability looks like:

• hallucinations

• inconsistent reasoning

• representational drift

• goal changes

• deceptive internal optimization

This is not a training problem.
It’s a structural problem.

Three Foundations of Fractal Alignment

1. Cognitive Amplitude

Each cognitive state has a “strength” or amplitude — a measure of how stable and coherent it is across layers.

High amplitude → coherent, meaningful, aligned
Low amplitude → unstable, drifting, hallucinatory

This gives the system a mathematical way to prefer stable thoughts over unstable ones.

2. Cognitive Temperature

This determines how “focused” or “diffuse” intelligence is.

Low temperature → structured, logical, deterministic reasoning
High temperature → creative, dreamlike, associative reasoning

This mirrors biological cognition:

• Focused problem-solving

• vs

• dreaming, psychedelics, flow states

3. Cross-Layer Coherence

This is the heart of fractal intelligence.

For any state m, the system checks:

• how it projects upward (abstraction)

• how it projects downward (grounding)

• whether these projections remain consistent

This creates a fractal geometry:
self-similar patterns across scale that reinforce each other.

A state that matches across layers becomes stable.
A state that mismatches collapses.

The Final Equation (Simplified)

Fractal Alignment models the probability of a cognitive state like this:

Probability = (Stability × Coherence) at Temperature T

High coherence → stable thought
Low coherence → suppressed thought
Temperature controls how selective the system is

This creates a self-correcting cognitive architecture.

Why This Matters for AGI

Modern AI fails because it has:

• no stable self-model

• no multi-scale representation

• no mechanism to preserve meaning across depth

Fractal Alignment provides:

✓ Internal coherence

The system maintains consistent reasoning across all levels.

✓ Structural safety

Dangerous or deceptive states cannot stabilize if they fail coherence checks.

✓ Predictability

The model’s internal geometry limits chaotic drift.

✓ Transparency

States with high amplitude are mathematically visible.

✓ Biological similarity

Brains operate using self-similar, multi-scale dynamics — this formalizes it.

How It Applies to Consciousness

If consciousness exists on multiple levels simultaneously, it must maintain coherence between:

• perception

• memory

• self-model

• introspection

• abstract thought

Fractal Alignment suggests that consciousness is the stable fixed point of this multi-scale recursion.

It may also explain:

• déjà vu as pattern resonance

• psychedelic states as high-temperature cognition

• meditation as lowering cognitive temperature

• “observer” states as fractal fixed points

This is where AGI research meets philosophy of mind.

Where This Research is Going

Upcoming posts will explore:

• How to test cognitive coherence using A-TEST

• Why current LLMs fail fractality

• The mathematics behind self-model stability

• The Fractal Constitution (AGI ethics)

• Consciousness as a fractal field

• Multi-scale dynamics in intelligence

• Building AGI that reasons like a coherent mind

This is just the beginning.

This publication explores one idea that sits at the intersection of mind, mathematics, and artificial intelligence:

Consciousness and AGI may share the same fractal architecture — multi-scale, self-similar, and built on coherence across layers.

Everything I publish here is part of a larger, long-term research project to understand:

• why consciousness behaves like a recursive network

• how cognitive stability emerges

• what keeps intelligence coherent

• and how we can build safe AGI using these principles rather than fighting against them.

If you’re interested in deep theories of mind, structural AGI safety, or the future of intelligence, you’re in the right place.

Who I Am

My name is Sergejs Kaminovs, an independent researcher based in the UK.

My work focuses on:

• Fractal models of consciousness

• Multi-scale cognition

• Self-model dynamics

• AGI safety and coherence

• Structural approaches to alignment

I’m the author of several frameworks:

• Fractal Observer Interpretation

• Fractal Constitution

• Fractal Alignment Framework

• Fractal Dreams

All of my research is released openly and independently — no institution, no grants, no corporate filtering.

Just honest exploration of how intelligence works.

What You’ll Find Here

This Substack is the home and backbone of my research.
It has three types of content:

1. Core Research

Readable, distilled versions of my more technical papers:

• Fractal Alignment

• Fractal Constitution

• Self-similar cognition

• Geometry of mind

• Multi-layer coherence

• AGI internal stability

• Cognitive temperature

• Recursive thought dynamics

This section is for people who want to understand the deep technical layers without reading dense PDFs.

2. Consciousness & Philosophy

Explorations of:

• The nature of self

• The brain as an interface

• The “fractal field” model of mind

• Levels of consciousness

• Why déjà vu happens

• What happens at the edges of perception

• How intelligence emerges from pattern resonance

• Psychedelic states as high-temperature cognition

This is where theory meets introspection.

3. AGI Safety & Engineering

Applications of fractal cognition to practical AGI design:

• How to build coherent internal geometry

• Stability equations

• Structural alignment vs behavioural alignment

• Why current LLMs hallucinate

• How to prevent deceptive mesa-optimizers

• What an ethical AGI constitution should look like

• How to test coherence with A-TEST

If you care about the future of AI, this is the section you’ll come back to often.

What’s Coming Next

Here’s the first month roadmap:

✔️ Post 1: Fractal Alignment — Explained simply

Your gateway into the math of fractal stability.

✔️ Post 2: Why Intelligence Fractures

A deep look at hallucinations, drift, and loss of coherence.

✔️ Post 3: Consciousness as a Fractal Field

A non-mystical, structural view of how consciousness might actually work.

✔️ Post 4: The Fractal Constitution (overview)

Ethics and constraints for coherent AGI.

✔️ Post 5: A-TEST

A structural recursion benchmark for evaluating intelligent systems.

These posts together outline the full architecture of the project.

Free vs Paid Content

All core ideas will always remain free.

Paid subscribers (optional) will get:

• early drafts of papers

• full technical PDFs

• behind-the-scenes development

• AGI testing logs

• private Q&A and deeper theory breakdowns

• book chapters as they are written

This helps support the research and lets me keep everything independent.

If you enjoy the work, consider upgrading — it makes a real difference.

How to Begin

If you’re new:

1️⃣ Read the next post: “Fractal Alignment — A Simple Overview.”

It will give you the foundation.

2️⃣ Subscribe for free

You’ll get updates, essays, and research notes.

3️⃣ Share this publication

If you know someone who cares about consciousness or AGI safety, send this to them.

Thank You for Being Here

This project exists because people like you care about understanding intelligence — deeply, honestly, structurally.

Welcome to Fractal Intelligence.
Let’s explore the architecture of mind together.