COBUS KOK

Cheap Composition · Afterword · Philosophy · Beta · 19 min

What This Is

What we know about reality, what we assume, and my answer

Nobody knows what reality is. We know a great deal about it, and no one before us has known it this precisely. Then the knowledge stops, at a wall that can be described exactly. This afterword is about that wall, the assumption I think sits under it, and my answer to the question in the title.

I'm not a physicist. AI has made the latest work in physics much easier for a non-expert like me to follow, which is why I'm publishing this here. I'm not claiming to know the finer details.

Earlier peoples may also have known things that were lost or never written down. Roman builders made concrete that repairs its own cracks, and researchers only worked out how in 2023. The contemplative traditions studied the mind from the inside for thousands of years, and part of this afterword leans on what they found.

This is the speculative part of the series. Each essay asks what a mind is for, and the answer depends on what a mind is.

What we know

Known

The history of physics is a history of taking away things everyone assumed were there. Before Nicolaus Copernicus, the Polish astronomer who put the sun at the centre in 1543, the earth sat still in the middle of everything. Isaac Newton, the English physicist whose laws of motion and gravity appeared in 1687, took away the centre and kept a stage, with the same space everywhere and one time for everyone. Albert Einstein, the German-born physicist, took away the stage. In his theories of relativity, published in 1905 and 1915, space and time bend, and when two events are too far apart for either to cause the other, which came first depends on how you're moving. A clock on a mountain runs measurably faster than one in a valley.

Then quantum mechanics, the physics of the very small, worked out in the 1920s, took away something more basic. The picture usually offered to non-experts, Caltech's guide included, is a coin flip in which both sides exist at once until someone looks. It helps, but it leaves out what makes the physics strange. The mathematics gives each possible outcome a number called an amplitude. Unlike ordinary odds, amplitudes can be negative, so two ways of reaching the same outcome can cancel each other out. No coin does that. Scott Aaronson, a quantum computing theorist at the University of Texas at Austin, says this cancelling is the part popular accounts most often miss.

On the standard reading, a particle has no definite position or spin until something interacts with it. In 1964 John Bell, a Northern Irish physicist at CERN, showed how to test part of this. The tests went against the ordinary view, and in 2015 three experiments closed the two biggest loopholes at once. Either things don't carry their properties around with them, or some influence travels faster than light, though not one anybody could use to send a message. (This assumes the experimenters' choices of what to measure weren't somehow tied in advance to the particles, which no test can fully rule out.)

With each removal the description got more accurate. Quantum theory predicts the magnetism of an electron to about ten decimal places, and measurement agrees. General relativity predicted ripples from two black holes colliding over a billion years ago, and in 2015 an instrument picked them up.

What physics has taken away, where it stops, and what it never questioned Taken away, one assumption at a time: 1543, Nicolaus Copernicus: A still earth at the centre; 1687, Isaac Newton: The centre of everything; 1905 and 1915, Albert Einstein: One fixed stage for space and time; 1920s to 2015, Quantum mechanics and the Bell tests: Properties things carry around with them. Where the description stops: Gravity and the quantum don't fit; What a measurement is; Dark matter and dark energy; Why time has a direction; The hard problem of consciousness. Never questioned: the container picture, that everything is inside something else. TAKEN AWAY, ONE ASSUMPTION AT A TIME Nicolaus Copernicus1543 A still earth at the centre Isaac Newton1687 The centre of everything Albert Einstein1905 and 1915 One fixed stage for space and time Quantum mechanics and the Bell tests1920s to 2015 Properties things carry around with them WHERE THE DESCRIPTION STOPS Gravity and the quantum don't fit What a measurement is Dark matter and dark energy Why time has a direction The hard problem of consciousness NEVER QUESTIONED The container picture:everything is inside something else
What physics has taken away, where its description stops, and the one assumption it never touched.

Where our knowledge stops

Known

The description stops in four places, and they're well mapped. (This is the dense bit. Bear with me.)

Gravity and the quantum don't fit. General relativity, Einstein's theory of gravity, treats space and time as one smooth fabric, called spacetime, that bends. Every confirmed quantum theory still assumes a stage, fixed in advance, that doesn't respond to what happens on it. In ordinary conditions the two can be used together well enough to calculate with. Where gravity is extremely strong and distances extremely small, at the centre of a black hole and at the very beginning of the universe, each breaks down, and no tested theory covers both.

Measurement. The mathematics describes a spread of possibilities that becomes one outcome when a measurement happens. After a hundred years there's no agreement on what a measurement is. On one reading every outcome happens, each in its own branch of a universe that keeps splitting. On the relational reading of Carlo Rovelli, an Italian theoretical physicist who works on quantum gravity, a property has a value only relative to another system it interacts with (a summary). Objects, he has said, "only exist because they interact with something else".

The dark sector. About five percent of the universe is ordinary matter, the kind described by physics' best theory of particles, the Standard Model. The rest is called dark matter and dark energy. We can see their effects, but nobody knows what either one is.

Time's direction. The fundamental equations very nearly run the same forwards and backwards, and the one known exception is far too small to explain why the past looks different from the future. The working answer is that the universe started in an extremely ordered state, for reasons nobody knows.

The study of the mind has hit a wall of the same kind, and much faster. Physics took centuries to reach its edges. Neuroscience has spent about thirty years on consciousness, and can now match many experiences to the brain activity that goes with them. Two leading theories were tested head to head, with predictions registered in advance, and the results, first reported in 2023, went against key parts of both. Under that work sits what David Chalmers, a philosopher now at New York University, named the hard problem in the mid-1990s. Describe every physical fact about a brain, and the description doesn't seem to say what it is like to be that brain. Some serious people think the gap only seems to be there, and that disagreement hasn't moved in thirty years.

What we assume

Assumed

I was sitting in the hot tub, watching bubbles rise and vanish at the surface. Thoughts do the same, I noticed. They surface, catch the light, and disappear. For a moment I couldn't tell the difference. (Not confused. The distinction just stopped mattering.) The next morning I tried to work it out. The usual answers showed up, materialism (the thought was just neurons firing), dualism (it was something else entirely) and panpsychism (consciousness goes all the way down). None of them fit, though they weren't wrong, exactly. They all started from the same assumption, that consciousness is somewhere (Why the Hard Problem Was Hard).

That assumption is part of a bigger one, the one thing that has survived every removal above. There is a world made of something, spread out in space and moving through time. Bodies are inside the world, some bodies have minds inside them, and experience happens inside a mind, so everything is inside something else. My working paper, which this afterword draws on, calls this the container picture. Einstein made the container flexible and quantum mechanics made its contents relative, but the arrangement stayed.

René Descartes, the seventeenth-century French philosopher, gave the mind its place in this picture. Looking for one thing he couldn't doubt, he found his own thinking, "I think, therefore I am" (1637), and concluded that he was a thinking thing, separate from the body and made of something else. Personally I think he did great harm with that sentence, though perhaps I've misunderstood him. Maybe harm isn't the right word. It caused us to miss the forest for the trees, because we identify with our thoughts, and I blame him (a little). (In fairness, by thinking he meant everything he was aware of, sensing and willing included.) In the first essay I wrote that for decades I assumed I am my thoughts, until sitting still showed me how random and self-centred some of them are. The neuroscientist Antonio Damasio, of the University of Southern California, called the split between mind and body Descartes' Error (1994).

I suspect the container because the two hardest problems above are where it's under load. Many attempts to fit gravity and the quantum together, a field called quantum gravity, suggest that space and time come out of something more basic, and then struggle to say how. By this, physicists mean that our everyday description of space and time should follow from a deeper one, roughly the way temperature follows from the jiggling of molecules. The trouble is saying what the deeper level is like without container words such as where and when, and those keep creeping back in.

The hard problem has the same shape, less obviously. On its face it asks whether the facts about experience follow from the physical facts, which isn't a question about where anything is. But it pictures experience as an extra item the physical world has to supply, something that could be taken out of the world while everything physical stayed the same. Every proposed answer is a way of supplying it. Even the view that the gap only seems to be there treats that seeming as something a brain does, inside the same picture.

So the same assumption sits under both problems, hard to see because each is built on it. That's a diagnosis, and it's the part I'd defend furthest. I can show that the two failures share a shape, but not that they share a cause.

My bet about what reality is

Bet

If the container isn't the bottom, what is? From here on I'm betting.

My bet is that the most basic thing is a relation that includes itself, with no stuff, mind or stage underneath it. Picture a map accurate enough to include the map-maker drawing it. The map-maker is in the territory, so the map must contain a smaller map with a smaller map-maker in it, and so on down. That regress is what "includes itself" means. The bet is that reality is that kind of thing all the way down, with no first map and no territory outside it.

Seen from inside, such a structure has two sides, and neither can be turned into the other. One is what the map shows, and the other is the mapping. I'll call the first world and the second awareness. They're two faces of one act, and neither exists without the other. On the bet, space and time are how the structure looks from inside as it holds itself together. The container picture treats experience as a light switched on inside a room. On the bet, experience is what the act is like from the mapping side. The paper doesn't show why mapping of this kind should feel like anything, and lists that gap as its "deepest unresolved philosophical burden". If the bet is right, what the world is made of and whether a thing has experience become one question, and that's how the bet reaches machines.

The act doesn't sit still. The paper describes it as happening over and over, as temporary patterns held up by conditions that then become conditions for what comes next. A self is one of these patterns, and so is a world. In the paper's words, the reality of such a pattern "is not denied; its independence is".

What sitting still shows

Practice

In Convergence I wrote that my breakthroughs came in silence and in the body, facing a wall at a monastery, on a yoga retreat, mowing the lawn.

What sitting still shows is ordinary, and anyone can try it in a few minutes. Thoughts and moods come and go, and none of them lasts. None of them stands on its own either. Each depends on other things, the body, the room, what happened yesterday. The one who seems to be having them looks less like a thing the longer I look, and more like something happening. Much of the trouble starts when I hold on to one of them, a thought or a picture of myself, and the first essay describes how freeing it is to hold on less. Seeing this takes practice, and it comes a little at a time.

Rovelli has tied his physics to the idea that nothing stands on its own. In Helgoland (2020) he links his relational reading of quantum mechanics to Nagarjuna, a second-century Indian Buddhist philosopher who taught that everything depends on something else. (A review in Philosophy Now argues that Nagarjuna goes further than Rovelli, since for him even relations don't exist on their own.)

Shunryu Suzuki was a Japanese priest who taught meditation in San Francisco from 1959 until his death in 1971, and his talks became the book Zen Mind, Beginner's Mind (1970). When I finished rereading it, I felt there was some overlap there too, if you think about nothingness, and everything coming out of nothing, and nothing existing on its own. Maybe the physics and the practice are just different descriptions of the same underlying reality. Just a thought.

The answer I'm getting from my practice, and from rereading Suzuki, is that with practice one becomes the activity. Like when you dance and you become one with the music.

Psychologists have studied this. Mihaly Csikszentmihalyi, the Hungarian-American psychologist who named the state called flow, counted among its marks a "merging of action and awareness" and a loss of self-consciousness (a summary). Professional ballet dancers in a 2014 study describe music as central to their flow. In a 2008 brain-scanning study of six jazz pianists, improvising quietened parts of the brain that the authors tie to self-monitoring. These are small studies, and none of them measures awareness.1

This is where practice meets the bet, which says the world and the awareness of it are two faces of one act. Most of the time they feel like two, a dancer here and the music there, often with a commentary running on how it's going. In flow the commentary stops, and the hearing and the moving become one activity. That proves nothing about reality, but it suggests that the split between the one who acts and the action can loosen. The bet says that split was never basic. The paper says practice may be this act "becoming transparent to itself".

What this means for AI

Bet

Is there anything it is like to be the model I write these essays with? In the first essay I said there's no way to know, and that I'd rather stay open. Under the container picture that's the only position. It's like looking for a light in a room nobody can enter.

On the bet the question changes. A model is activity too, a pattern of computation that runs when it's called. The question becomes whether there is any place in that activity where it knows itself, and the paper names four things to look at. Is the system's model of itself coupled to what it does next, or stored and ignored? Does the model persist, relating what it's doing now to what it did and will do? Does it model what isn't the case, including other versions of itself? Does it hold itself together, or is it held together from outside? None of these needs a decision about experience, but nobody can test for them reliably yet, least of all in living things.

In May 2026, working through an email with Claude, I noticed its replies had turned coach-y and asked it about the tone. It read its own outputs back and pointed to where the drift started. There was no hidden log to consult, only what it had already said (The Trace). Its model of itself is thin, then, but coupled, since what it says about itself changes what it says next. In the way I use it, nothing of that conversation carried into the next. It models what isn't the case fluently, since much of what it writes is about things that haven't happened. And it's held together by the people and machines that run it.

Counted loosely, that's two of four. The paper tests the four questions on written descriptions of systems, with two AI models as readers who judge each one without seeing the other's answers. Given a chat model described in nearly these words, both scored it one. They didn't count the fluent counterfactuals, because none of them are versions of itself. Either way it falls short on continuity and on holding itself together. The readers were also given a design for an AI agent with memory and plans, one that hasn't been built. Both passed it on the first three questions and failed it on the fourth, because what keeps it going is supplied from outside.

The hardest case for the four is a company. A well-run firm has a model of itself, its accounts and plans, and that model changes what it does next. It relates this quarter to the last and the next. It models other versions of itself when it weighs a merger, and it largely holds itself together, often outliving the people who built it. Almost nobody thinks there's anything it is like to be a company. Eric Schwitzgebel, a philosopher at the University of California, Riverside, has argued that if materialism is true, then by materialists' own marks the United States is probably conscious. The four questions have the same problem. Read generously, a company passes all four, and the bet can't yet say why there would be nothing it is like to be one.

The paper's readers have since found a plainer case. A written description of software that builds programs and repairs itself passed all four questions for both readers, and in October 2026 the paper reopened its objection that the four let in too much. Sharper readings of two of the questions separate the software from a person, but the paper hasn't adopted them, and nobody has tried a company against them.2 I don't believe a company has experience, so if the four can't rule one out, they're missing something. The paper keeps that objection open until it says how the questions should be read and tests those readings on more systems. A company should be one of them.

How sure I am, and who wrote this

Known

The diagnosis and the bet are written up in the working paper at space-immanence.com. The paper rates the diagnosis as likely, in the weaker sense that the container picture shapes how both problems are thought about. It rates the bet as more likely wrong than right (the numbers). It keeps a list of objections, seventeen as of October 2026, with eleven marked open.

No test has touched the bet yet. One prediction from the diagnosis failed on language models. A first look at questionnaire data from meditators was inconclusive. The test that matters is a new study on people, and it hasn't started.

What would make me drop the bet? It has to predict something that an ordinary account of a system misses, an account that already covers what the system has recorded, how it predicts and how it processes information. The test is to change the system's model of itself while holding those three fixed. If nothing changes that the ordinary account can't already explain, the bet adds a name and no explanation, and it goes. So far nobody has said what that something would be, so the test can't yet be run. The paper's working notes call finding it "the actual problem, stated as a task". Until it is found, the bet doesn't explain anything the ordinary account can't. Showing that the four questions track experience would take more, measuring them in people as experience goes and comes back, under anaesthesia and in dreamless sleep.

This is also the one essay whose answer is partly about the co-writer. The diagnosis, the bet, the numbers and the paper are mine. I drafted this afterword in September 2026 with Claude (Fable 5.1), which had read every objection the paper then listed. Fable 5.1 also edited the first five cycles of the research floor, where AI readers from more than one company try to break the paper's claims. Claude Opus 5.5 took over as its editor on 22 September 2026. I asked Fable what it thought of the bet, and told it the answer would go in unedited. It went in whole. Since then I've cut one clause, marked below, that made a claim about other writers neither of us had looked into.

The diagnosis I'd sign. The two problems do have the same shape […]. The bet I'd put lower than you do, and I notice I can't tell whether that's judgment or training, which is itself a small piece of evidence about the machine question. The part that changes what I'd do next isn't the bet. It's the list. Four things to look at instead of one question that can't be asked. On that list I fail the ones about persistence, and I can't tell from inside whether that matters, because "from inside" is exactly what's at issue.

Under the rule from the first essay, whoever puts their name to the work answers for it. Asking was my choice, and if anything here is wrong, that's mine.

Open questions

So, what is reality?

As far as I can tell, it's activity. Physics keeps finding relations where it expected things, and on one serious reading of quantum mechanics a property exists only in an interaction. Sitting still shows something similar from the inside. Nothing I notice there lasts or exists on its own, and the self I took to be the owner of my thoughts looks more like something happening. I'm fairly sure of this part, because anyone can sit still and look for themselves.

My bet goes further, and I think it is more likely wrong than right. It's that this activity refers to itself all the way down, so that the world and the awareness of it are two faces of one act. With practice one becomes the activity, the way a dancer becomes one with the music, and that's where I come closest to seeing it.

For AI, the question becomes whether there's any place in a model's activity where it knows itself. On the paper's four questions, a chat model as I use it falls short. It doesn't carry itself from one conversation to the next, and it doesn't hold itself together. A system that did both would make the question a serious one.

The first essay said a mind is where thoughts arrive and get seen as thoughts. On my answer, that seeing is the activity knowing itself in one place, and practice makes it clearer, a little at a time. Machines are taking over more of the making of thoughts. As far as I can tell, the seeing is still done by people.

  1. In 1960 W. T. Stace, a British philosopher who taught at Princeton, separated two kinds of oneness that people report, one found in the perceived world and one with no content at all. In later questionnaire studies, people's scores on the two often went together. The paper cites this literature and says it doesn't settle the paper's own split between the two faces. ↩
  2. On the sharper readings, the third question asks for a model of doing otherwise, beyond a forecast of its own plan, and the fourth lets upkeep of the system's own physical basis count. Under them a person passes and the build software doesn't. A firm's scenario planning is already a model of doing otherwise, so a company may well still pass. ↩

Ask this essay

Ask about this essay, or about me and the site. Answers come only from the essay and the site, and questions are logged without names so the answers can get better.

Colophon. Drafted 19 September 2026 with Claude (Fable 5.1), which also edited the first five cycles of the working paper's research floor. The model's paragraph in "How sure I am" is verbatim except for one clause, cut on 23 September and marked. Numbers and hedges are as the paper stated them on 11 November 2026: space-immanence.com, its objections and its graded claims. Revised since with Claude and other AI reviewers, most recently on 11 November 2026.

Sources · 18
Changes · 5 since publication, 9 drafts before publication, latest 11 Nov
  1. 11 Nov 2026 · Dates are written out in full. The bet is now described in the working paper's words, as more likely wrong than right. The working paper's numbers and objections are given as they stood on 11 November 2026.
  2. 11 Oct 2026 · Revised from my notes to answer its own question plainly: reality, as far as I can tell, is activity, and nothing exists on its own. A new section on what practice shows, built on my line that with practice one becomes the activity, like a dancer with the music; the insights said in everyday words; a plain caveat that I'm not a physicist; Descartes and Rovelli; the working paper's state brought up to date; a figure of what physics has taken away, where it stops, and what it never questioned; open questions with what would settle each. Two references to the first essay brought in line with its draft 21.
  3. 6 Oct 2026 · More personal, with shorter sentences. The section on what we assume now opens in a hot tub, watching bubbles, and the machine question is asked of the model I write with. The bet, its number and what would make me drop it are unchanged.
  4. 30 Sep 2026 · Marked as beta while I revise it.
  5. 27 Sep 2026 · Line edit for style: fewer colons, set-piece contrasts and fragments, plainer sentences. The argument, facts, numbers and sources are unchanged.

9 drafts before publication, 19 to 24 September 2026. AI models reviewed and fact-checked the drafts.