Postulate Noise, Not Discreteness

The answer, first

You do not have to assume the universe is made of tiny indivisible pieces — pixels of space, ticks of time — to explain why everything you can measure comes out finite.

But one assumption is not enough, tempting as it is to stop at one. The tempting version says: assume only a limit on how finely two states can be told apart, and finiteness follows. That is one assumption short, and the shortfall is worth seeing exactly, because the failed shortcut teaches the successful route.

Here is the gap, in one picture. Suppose noise blurs any two marks on a dial that sit closer than one millimetre. Can the dial still carry unlimited information? Yes — if the dial is unlimited. Put marks at 0, 1 mm, 2 mm, 3 mm, and so on forever: every pair of marks is a readable difference. Noise sets the spacing between the rungs of a ladder; it says nothing about how tall the ladder is. To make the number of rungs finite you need a second, independent fact: a ceiling. The dial ends. The room has a roof. The battery has a top voltage. There is only so loud anyone can shout.

So the honest blueprint has two limits, not one. A floor: differences finer than this cannot be told apart. That is noise. A ceiling: signals larger than this cannot be made. That is bounded resources. Floor and ceiling together — and only together — make the number of distinguishable states finite.

Finiteness then follows — not as a claim about what exists, but as a claim about what persists: what can be recorded, copied, transmitted, remembered. And since persisting is the only access anything has to anything else, that is the only finiteness that has ever done any work. Continuum mathematics may stay infinite underneath. Nothing in the infinite part is inheritable, so no record, no measurement, and no mind ever touches it.

Four axioms and one boundary condition. That is the whole blueprint, and this essay is the count.

tags: ontology, causal-order, information, finiteness, minds, non-ergodicity

Abstract

QUESTION. What is the smallest honest set of assumptions from which a finite world and the minds inside it follow?

SPINE. The finite is not what exists. The finite is what persists. Persistence is limited by noise — but noise alone limits nothing, because a signal can always be made stronger. Two limits are needed: a floor, below which differences blur, and a ceiling, above which signals cannot go. Assume both, and finiteness follows for everything that can be recorded, measured, or remembered — with no assumption that the world is made of discrete pieces.

CONCLUSION. Four axioms and one boundary condition. Events; a causal order; a noise floor; a resource ceiling; and one special edge of the order — the low-entropy beginning — that turns chains of influence into history. The order fixes the shape of spacetime up to a scale (a theorem). Floor and ceiling together make distinguishable states countable (a derivation), and in the same stroke produce records, forgetting, and a phase boundary for persistence. That the count also fixes the missing scale is a conjecture, stated as one, with its two known obstacles named. A mind — more carefully, a minimal cognitive system — is not an extra ingredient: it is the configuration in which the causal order closes back on its own records and holds above the boundary. What the blueprint buys, what it costs, and what it deliberately leaves out — fundamental discreteness; the feel of experience — are all counted out loud.

Who this is written for

For a sharp reader who will check each step and stop at the first one that does not follow — and for any adult who wants the same courtesy. No physics background is assumed. Every technical term — bit, mutual information, capacity, entropy — is defined before it is used, with a concrete example. The only mathematics required is multiplication and the logarithm, and the logarithm is explained on the spot. Where a step is proved, a source is named. Where a step is proposed rather than proved, that is said in the same breath.

The essay also names its own temptations. Several steps below sit next to a seductive shortcut — an assumption that looks free, a conclusion that looks derived — and each shortcut is displayed and declined where it occurs, then collected at the end so the whole route can be audited. An argument that hides the moves it almost made is advertising. An argument that displays them is engineering.

1. The answer, first, Again!

You do not have to assume the universe is made of tiny indivisible pieces — pixels of space, ticks of time — to explain why everything you can measure comes out finite.

Four axioms and one boundary condition. That is the whole blueprint, and this essay is the count.

2. Why people reach for discreteness, and what it costs

The usual move is to say the universe is a computer. Computers have finite memory and finite clock rates, so if the world is computed, the world is finite and there is a smallest pixel. The picture has real support: black-hole thermodynamics says the information inside a region is bounded by the area of its boundary, not by its volume.

S ≤ A / 4ℓ_P²

eq 1 · the holographic bound: information scales with area

Here S is entropy — for now, read it as the number of yes/no answers needed to pin down the exact state of the region; A is the area of the region’s boundary; ℓ_P is the Planck length, about 10⁻³⁵ metres. The formula returns, explained in full, in section 5.

But as an explanation of why influence is ordered at all, the computer picture is circular twice.

Circular in the physics. Every bound of the kind in eq 1 is derived inside relativity — the speed of light and Newton’s constant are already sitting in the formula. You cannot use a result to ground the framework it was derived in.

Circular in the logic. A computation is defined by a dependency order: which step must come before which. Saying “the universe computes, therefore influence is ordered” hides the conclusion inside the definition. The genealogy is explicit — the “happened-before” relation at the foundation of distributed computing was modelled on special relativity, by Lamport, in 1978. Computer science borrowed the order from physics. Re-exporting it as an explanation is returning the loan and calling it income.

There is also a physical price, with one caveat that usually goes unsaid. A regular grid picks out a preferred frame: you could measure your speed against the grid, and light of different energies would arrive at slightly different times from very distant explosions. For the simplest (“linear”) version of that effect, observations of gamma-ray bursts find nothing, to precision beyond the Planck scale — the natural home of any grid. For the next-order (“quadratic”) version, current limits are far weaker: many orders of magnitude below the Planck scale. So experiment closes the front door on a grid and leaves a back window open. The blueprint below declines to buy the grid either way; the caveat is recorded because an argument should quote its evidence at its actual strength.

So: keep the finiteness intuition. Drop the substrate that was supposed to deliver it. What follows delivers it more cheaply — and counts every coin.

3. Three definitions, before the axioms

Three terms carry the whole essay. Each gets a definition and a toy you can hold.

The bit. A bit is the information in the answer to one yes/no question. To single out one option among N equally likely options takes log₂N questions, where log₂N means “the number of doublings that take 1 to N.” So log₂8 = 3, because 2 × 2 × 2 = 8: three well-chosen yes/no questions find one card among eight. log₂1024 = 10. log₂100 ≈ 6.6 — and fractional questions are fine; they average out over many uses.

Mutual information. How much knowing one thing helps you guess another, measured in saved questions. Write it I(A;B): the number of yes/no questions about A that knowing B saves you. Two independent coins: seeing one saves you nothing about the other, I = 0 bits. Two coins glued together: seeing one answers the other completely, I = 1 bit. A smudged photocopy of a coin flip that shows the right face three times out of four: seeing it saves you part of a question — about 0.19 bits. Mutual information is the quantity that shrinks, link by link, as a whisper travels down a line of people.

Capacity. A channel is anything that carries a state from here to there — a wire, a whisper, a photocopy, a memory left overnight. Its capacity C is the largest number of bits per use that can be pushed through it reliably. Equivalently: if a channel can reliably tell apart N different messages, its capacity is C = log₂N bits. Turn that around and you get the fact everything below leans on:

N ≤ 2^C

eq 2 · finite capacity means finitely many distinguishable messages

That is all the mathematics the derivation needs.

4. Four axioms and a boundary condition

Axiom 1 — Events

There are things that can stand in relations to one another. Call them events: a lamp switching on, a neuron firing, a coin landing. This axiom is nearly free; it says only that there is something rather than nothing, and that it comes in distinguishable happenings.

Axiom 2 — Order

Influence forms a partial order, written ≺. “A ≺ B” means A can influence B. Two properties define it.

(A ≺ B) and (B ≺ C) ⇒ A ≺ C

eq 3 · transitivity

A ≺ B ⇒ not (B ≺ A)

eq 4 · no loops

A family tree is the everyday version. If your grandmother is an ancestor of your mother, and your mother of you, then your grandmother is an ancestor of you — that is transitivity. And nobody is their own ancestor — that is the no-loop rule. “Partial” means some pairs are simply unrelated: two cousins are neither ancestor nor descendant of each other, and two events can be so placed that neither can influence the other.

What this axiom buys is startling, and it is a theorem rather than a slogan. Suppose you know only which events can influence which — no distances, no clocks, no coordinates. Then you have already fixed the geometry of spacetime up to a single overall scale. Zeeman proved it for flat spacetime in 1964; Malament extended it to curved spacetime in 1977. In four dimensions the metric — the machine that answers every question about lengths and durations — has ten independent components, and the order alone fixes nine of them. The light cone is not drawn on spacetime; the cone, repeated at every event, is spacetime, minus a scale.

A ≺ B ⇒ B ∈ J⁺(A)

eq 5 · the light cone as the shape of the order

Why can the order not loop? Not because loops are strange, but because they are self-contradictory, and this can be made exact. Build a device that signals faster than light; relativity then lets two such devices in relative motion pass a message into their own past, closing a loop. Now program the device: emit a bit at noon if and only if that bit has not been received before noon. If it emits, it was received, so it must not emit. If it does not emit, it was not received, so it must emit. No history satisfies the specification. The no-loop rule is not aesthetic. It is the condition under which “cause” remains a coherent word.

Cost: irreducible. Nothing in current physics derives this axiom. It is taken on faith and ratified by every experiment that has ever looked.

Axiom 3 — Noise floor

Along any chain of influence, distinguishability degrades: the two ends of a chain never share more information than its weakest middle link.

Photocopy a page. Photocopy the copy. Do it a hundred times. Nothing dramatic happens at any single step, and no page is ever destroyed — but the text goes. Whisper a sentence down a line of thirty people and the same thing happens. Chains of processes compose:

E(t₃←t₁) = E(t₃←t₂) ∘ E(t₂←t₁)

eq 6 · composition: getting from t₁ to t₃ goes through t₂

and along a chain A — B — C, where all influence between A and C passes through B:

I(A;C) ≤ min( I(A;B), I(B;C) )

eq 7 · data processing inequality: the ends of a chain share no more than a middle link

That is Shannon’s data processing inequality, and — this matters later — it is stated here without a direction. Read the chain left to right or right to left; the inequality does not care. Nothing in Axiom 3 says which end of a chain is “the past.” A film of the photocopier degrading a page, played backwards, still shows a chain whose far ends share less than its near links. Which end of the world’s chains is the special one is a separate purchase, made below — a purchase that arguments in this territory habitually spend without listing.

Cost: light, in one sense — no known physical process violates eq 7. But note carefully what the axiom does not say. It does not say the amount of information anywhere is finite. Degradation bounds nothing by itself — a gap easy to miss, and exactly why the next axiom exists.

Axiom 4 — Ceiling

In any bounded region of the order, the resources available for making states different are bounded. There is a loudest shout, a top voltage, a largest energy a region can hold.

This is the axiom most tempting to skip — noise already feels like a limit on everything — so here is why it cannot be. The grown-up version of the dial from section 1 is Shannon’s formula for the capacity of a noisy channel:

C = B · log₂( 1 + S/N )

eq 8 · Shannon–Hartley: capacity is set by the ceiling-to-floor ratio

B is how many independent readings per second the channel offers; N is the noise power — the floor; S is the signal power — the ceiling. Look at the fraction. If S may grow without bound, C grows without bound: a noisy channel fed an unlimited signal carries unlimited information. Noise alone bounds nothing. Only the ratio of ceiling to floor bounds anything.

Physics ratifies the ceiling emphatically — pack too much energy into too small a region and it collapses into a black hole, which is a rather final ceiling — but the blueprint deliberately does not derive the axiom from gravity. Deriving it from general relativity would repeat the exact circularity charged against the computer picture in section 2: using a result from inside a framework to ground the framework. So the ceiling is postulated, plainly, as the fourth axiom, and gravity is cited as its ratification, not its source. The cost is counted in public: this blueprint runs on four axioms, not three.

Condition 0 — Orientation

One end of the order is special: the universe began in an extraordinarily low-entropy state.

Axiom 3 was stated without a direction because the inequality has none. But your life has one: you own photographs of last summer and none of next summer; you remember breakfast and not tomorrow’s lunch. That asymmetry is not in eq 6 or eq 7. It comes from a boundary condition, often called the past hypothesis: at one edge of the causal order the world sat in an extremely special, tidy, low-entropy configuration, and everything since has been the long relaxation away from it. Records are abundant about that edge and absent about the other — which is why the degrading direction of every real photocopier points away from it, and why “downstream” feels like a fact about chains when it is a fact about the edge.

It is common, in arguments like this one, to write the noise axiom with the word “downstream” already inside it, and then to list the direction of time as an optional extra — quietly spending the purchase before recording the expense. The bookkeeping here is stricter: orientation is a fifth assumption, of a different kind — a fact about an edge of the order, not a law holding at every event — and it is counted from the start.

5. The derivation, honestly divided

Three parts. The first is bought by the floor alone. The second needs floor and ceiling together. The third is the open step, and it is labelled as one.

5.1 What the floor alone buys

Records. A record is not a special kind of object. A record is a state that survived — a configuration whose mutual information with some earlier event is still above zero. Nothing else is required to define it.

Forgetting as the default. Eq 7 says information about an event can only shrink along chains leading away from it. Add Condition 0, which makes one end of every chain the past, and the familiar statement appears: losing the past is what physics does unless something actively fights it. Memory is an achievement, not a resting state.

A threshold. The fight has a sharp boundary. Copy a record redundantly and correct errors as you go; whether that works forever depends on the error rate per step, p. Below a critical rate, error correction wins: the record outlives any fixed depth for as long as fresh, clean parts keep arriving. Above it, the record dies at a finite depth no matter how much redundancy is thrown at it. This is the fault-tolerance threshold, proved for classical computation by von Neumann and for quantum computation by Aharonov and Ben-Or. It is a property of the error rate, not of the material.

p < p_th ⇒ lifetime set by supply · p > p_th ⇒ lifetime finite, whatever the redundancy

eq 9 · the persistence threshold

One honesty note about the way this theorem is usually quoted — “below threshold, lifetime unbounded.” The threshold theorems assume a steady supply of fresh, low-entropy components — clean paper for the photocopier, cool memory for the computer. Inside a finite universe nothing is literally unbounded. The correct reading of eq 9 is that the threshold separates two ways of dying: above it, records die of noise; below it, records die only when the supply runs out. Noise stops being the binding constraint, and fuel becomes it. That distinction is what section 8 builds on.

5.2 What floor and ceiling buy together: finiteness

Take any bounded region of the order and any way of storing a state in it. The ceiling (Axiom 4) bounds how different two states in the region can be made — the height of the ladder, Δ. The floor (Axiom 3) bounds how close two states can sit and still be told apart after further steps — the spacing of the rungs, ε. Height over spacing is finite:

N ≤ 1 + Δ/ε

eq 10 · rungs = height over spacing: distinguishable states in a bounded region are countable

The dial from section 1, now with both limits in place: range 10 cm, blur 1 mm — about a hundred readable positions, log₂100 ≈ 6.6 bits. Remove either limit and the count diverges. Keep both and eq 2 closes the loop: finitely many distinguishable states, finite capacity, finite inheritable information. Not “the world has pixels” — rather, “only so many distinctions can be carried.”

The quantum world says the same thing in its own vocabulary, which strengthens the claim that the axioms are about information, not about any particular substrate. A quantum system of bounded energy in a bounded region can reliably carry only a finite number of classical bits — the Holevo bound. No quantum process can increase the distinguishability of two states — the quantum data processing inequality, which is Axiom 3 in quantum dress. And when gravity is allowed to set the ceiling, floor and ceiling appear in a single formula, the Bekenstein bound, with the region’s size R, its energy budget E, and the quantum of action ħ playing exactly the roles this section assigned them:

S ≤ 2πkRE / ħc

eq 11 · the Bekenstein bound: region and energy ceiling over the quantum floor

Eq 1, the area version, is this bound pushed to its gravitational extreme. The blueprint does not use either to prove Axiom 4 — that would be the circle again — but it is entitled to point at them as physics ratifying the axiom it postulated.

5.3 From count to scale: the open step

Now assemble. Step one, a theorem: the order fixes the shape of spacetime up to one number per region — the scale (Axiom 2; Zeeman, Malament). Step two, a derivation: floor plus ceiling make the distinguishable states of any bounded region countable (eq 10). Step three would close the blueprint: read the count as the missing scale, so that “how much fits here” becomes the meaning of “how big is here.” The causal-set slogan is “order plus number equals geometry” — and here number would not be postulated discreteness but the count of what floor and ceiling permit.

It is tempting to write step three as if it followed. It does not follow, yet. It is a conjecture, and its two known obstacles deserve names.

The bootstrapping problem. To count the distinguishable states “in a region,” you must already know which chunk of the order the region is — but the size of a chunk is the very scale the count was supposed to deliver. Causal-set theory escapes this because its atoms carry their own counting measure: you literally count elements, and the count is the volume by fiat. A noise-based count has no such fiat and must earn its measure. Nothing in this essay earns it.

The area problem. If the true maximal count obeys eq 1, it scales with the region’s boundary area, not its volume — so the naive move “count the states, call the answer the volume” gives the wrong answer by construction. Whatever dictionary translates count into scale must be subtler than counting. The live research program closest to this task builds geometry out of patterns of quantum entanglement — the Ryu–Takayanagi formula, tensor-network models, the line of argument associated with Van Raamsdonk — and it is simultaneously the best evidence that the conjecture points at something real and an accurate measure of how much mathematics stands between this essay and a theory.

So the derivation delivers, exactly and no more: finite operative information per bounded region, proved from the axioms; geometry up to scale, proved from Axiom 2 by others; and a conjecture connecting them, stated as a conjecture.

The universe does not have to be finite. It has to be finite to anything that remembers — and remembering is the only access there is.

6. The claim under the claim

That last sentence carries a philosophical commitment, and it would be easy to let it pass as if it were free. It is not free; it is a criterion. “The finite is what persists; persisting is the only access there is” amounts to this: whatever cannot in principle be inherited by any record can do no explanatory work, and what can do no work earns no place in the ontology. That is a pragmatist’s razor — a verificationist’s, if you like. It is a respectable razor with a long history, and this essay shaves with it. But it must be listed as an assumption of method, because no experiment can test a criterion for what counts as testable.

Kant argued that experience has conditions, and read the conditions off the structure of reason. This blueprint reads them off channel capacity. The trade makes the conditions measurable — floors, ceilings, and thresholds are laboratory objects — and that is a genuine improvement. It does not make the razor itself an empirical result. Section 10 tests the axioms. Nothing tests the razor. Named, and on the books.

7. A finite band between two horizons

The blueprint is bounded at both ends, and both bounds are facts about the causal order rather than about substance.

Below: the Planck scale is a horizon, not a singularity

The distinction is load-bearing. A singularity is where a theory dies: quantities blow up, paths end, the equations stop producing numbers. A horizon is where nothing local happens at all. An observer falling through the horizon of a large black hole feels no event; the surface is defined globally, as the boundary of what can ever signal out — which is to say, defined purely by the order ≺.

The Planck floor behaves like the second. Nothing blows up there. What happens is that resolving a smaller distance requires a probe of higher momentum, and concentrating that much energy into that small a volume collapses it into a black hole. The probe manufactures a horizon around the very thing you wanted to look at. That is the floor and the ceiling meeting at their extreme — Axiom 3 and Axiom 4 touching — not a place where the world tears.

History offers a warning about the word. The Schwarzschild radius was read as a singularity for some forty years, until better coordinates showed it was an artefact of a bad map. Singularities in physics have a near-perfect record of marking the end of a description rather than a feature of the world.

Above: the cosmological horizon

Nothing beyond it can ever signal to us, and it, too, carries a finite entropy. The census: the light and neutrinos left over from the big bang carry roughly 10⁸⁹–10⁹⁰ bits; black holes — easy to forget and impossible to ignore — dominate everything else at roughly 10¹⁰⁴, most of it in the supermassive ones, per the Egan–Lineweaver accounting; and the horizon itself carries roughly 10¹²² in Planck units.

Between the two horizons lies a finite band. Everything that can be a record, or a mind, lives in it. Neither bound required discreteness to get there.

8. Minimal cognitive systems

First, the bracket. This section defines a kind of physical system: one that keeps records of itself and steers by them. It does not touch — and does not pretend to touch — the question of why there is something it is like to be such a system, the so-called hard problem of consciousness. Whether every system of the kind defined here feels like anything is left exactly as open as it was found. The word “mind” is therefore held back from the bare structure, because it invites a larger reading than the argument pays for. The claim is narrower: no new ingredient beyond the assumptions already on the books is needed for the structure. Take the same primitives and ask where the dependency structure becomes self-referential and holds.

A minimal cognitive system is a process that keeps its own causal ancestry alive above the noise floor, and uses the records so preserved to regulate its next transition.

M_t+Δt = F( M_t, R_t, I_t ), R_t ⊂ M_t

eq 12 · the next state is steered by the system’s own preserved records

M_t+Δt independent of M_t ⇒ no memory, no learning, no continuing subject

eq 13 · independence of successive states destroys the subject

Test the definition on hard cases.

A river. Its ancestry is alive: today’s channel was carved by yesterday’s water. But the river never reads what it wrote. The order runs through it, not back into it. Not a cognitive system.

A version-control repository. Perfect records, perfect composition, ancestry alive by construction — and nobody home, because nothing in it consults its own history to determine its next state. Ancestry alone is necessary and not sufficient; the closure in eq 12 is what is missing.

A brain assembled by accident. Imagine a configuration identical to a person’s brain, produced by freak coincidence. Its structure is the right shape, but its apparent memories are records of nothing: nothing was inherited. On this account it is not a fraud, only newborn — its ancestry begins at the moment of assembly, and its standing as a cognitive system accrues from there. The property lives in the chain, not in the snapshot.

The bullet worth biting

A thermostat reads its own state to regulate its next transition. Under this definition it qualifies — as a system of near-zero depth, one reading of ancestry, barely above the floor. The definition is a gradient, not a gate, and the scale of the gradient can at least be pointed at:

μ ~ D × ( p_th − p )

eq 14 · a scaling sketch, not a law: depth of self-reference times margin below threshold

Left undefined, D would make eq 14 decoration, so here is an operational reading. Reach into the system and alter a record it laid down k steps ago; if its next transition changes, that record is live. D is how far back live records reach. A thermostat’s live record is one reading deep. A navigating rat’s reaches back through a season of remembered routes. A person’s reaches back decades and, through writing, past their own birth. Eq 14 remains a sketch: developed proposals in this vicinity — integrated information theory, with its worked-out quantity Φ; the free-energy-principle literature; autopoiesis, which arrived at “a system that produces and maintains itself” decades earlier — disagree with each other about what the right quantity is, and this essay does not adjudicate. It records that the gradient view is what the physics hands over. The bright line was always a preference.

Two ways of dying, and the price of rent

The threshold of eq 9 now does real work. It separates systems that die of noise from systems that die only of exhausted supply. A minimal cognitive system is a system built — by evolution or by design — to sit on the supply-limited side of the line: an engine against mixing. Engines must be fed. Erasing a bit has a minimum thermodynamic price, so every system that maintains records against the drift of eq 7 pays rent in energy:

E_erase ≥ k_BT ln 2

eq 15 · Landauer’s bound: forgetting has a minimum price

This one is not just theory. The bound has been measured in the laboratory, on single-bit systems, by Bérut and colleagues in 2012. Memory costs energy the way rent costs money: continuously, and in advance.

Extended systems

One further consequence, worth stating with care. Eq 6 — composition — holds automatically only for a closed system. A cognitive system is not closed, and for an open subsystem the composition law fails unless the state itself carries the memory. So either the system internalises its records, or the chain runs through external carriers: notebooks, files, archives, other people. The equations force exactly this much: the dynamics of the bare brain-sized piece, taken alone, do not compose, while the dynamics of brain-plus-notebook do. What the equations do not force is where to draw the boundary and pronounce the word “system” — that remains a modelling choice, made for a purpose. The extended-mind idea of Clark and Chalmers thereby stops being an intuition and becomes a statement about which cuts of the world have composing dynamics — with the cut itself still ours to choose.

9. What is not in the blueprint

Named rather than hidden, because an argument’s open seams are part of its content.

Fundamental discreteness of spacetime. Compatible; unfunded; not needed. The one known way to make discreteness respect relativity — random “sprinkling” of points rather than a regular grid — pays for it with radical non-locality, in which each point acquires infinitely many nearest neighbours, destroying the local update rule that made the computer picture attractive in the first place.
A derivation of Axiom 2. Why influence forms an order at all has no derivation. This ontology is minimal, not zero.
A derivation of Axiom 4 from anything deeper. Gravity ratifies the ceiling; nothing here derives it without assuming relativity, which would close the forbidden circle.
The count-to-scale dictionary. Step three of section 5 is a conjecture with two named obstacles, not a result.
Whether the order seals its own worst cases. The conjecture that singularities always stay hidden behind horizons is unproven, with contrived counter-examples known.
Whether ancestry survives a black hole. Recent model calculations recover the information, which would mean even black holes preserve inheritance; the mechanism in our universe is open.
The feel of experience. Eqs 12–14 define structure, not phenomenality. The hard problem enters this essay nowhere and exits it untouched.

10. How to attack it

An ontology that forbids nothing at reachable scales is worth less than a discrete one, whatever its axiom count. This one forbids things that can be looked for.

A controllable signal sent between two events that neither can reach at light speed. One would end Axiom 2. Correlations do not count: entangled particles produce correlations across such separations, but nothing an experimenter can steer — which is precisely why the boundary deserves the word causal.
Light of different energies arriving at measurably different times from distant cosmic explosions, or a rotation of light’s polarisation accumulating over cosmological distances. Either would expose a preferred frame — the grid the blueprint declines to buy. Honesty about current strength: the nulls are decisive against the linear effect and weak against the quadratic one; the window is stated at its true width.
Records persisting indefinitely at error rates above the fault-tolerance threshold, on bounded resources. That would break the phase boundary of eq 9, and with it the claim that persistence is a fight with a sharp edge.
Erasure of information below the price in eq 15. That would sever the link between records and thermodynamics on which the whole account of memory rests.
And one aimed squarely at the newest axiom: a bounded physical region shown to accept an unbounded number of reliably distinguishable states. That would end Axiom 4, and finiteness with it.

11. The blueprint in one sentence

The world is finite to everything that remembers, because telling states apart has a floor and making them different has a ceiling; it is ordered, because influence cannot loop; it has a history, because one end of the order began impossibly tidy; and a knower is what happens where the order bends back and reads what it wrote.

The shortcuts declined, collected

Every argument in this territory is surrounded by moves that look free and are not. The ones this essay had to step around are collected here, so that a reader can verify none was taken.

“Noise suffices for finiteness.” A noisy channel fed an unbounded signal has unbounded capacity (eq 8); a floor without a ceiling bounds nothing. Hence Axiom 4, and a count of four axioms rather than three.
Borrowing the ceiling from gravity. The Bekenstein and holographic bounds are theorems of relativity; using them to ground the framework would repeat the circle charged against digital physics in section 2. The ceiling is postulated; gravity is its ratification, not its source.
“Order plus count equals geometry,” read as a result. It is a conjecture, and section 5.3 names its two obstacles: bootstrapping, and the area problem.
Writing the noise axiom with a built-in direction. The data processing inequality is directionless; the word “downstream” is Condition 0 in disguise. The orientation of time is counted as its own assumption, up front.
“Below threshold, lifetime unbounded.” The threshold theorems assume a steady supply of fresh, low-entropy parts, and inside the finite band nothing is unbounded. The threshold separates dying of noise from dying of exhausted supply.
Calling the structure a “mind.” Eqs 12–14 define record-keeping and self-steering, not experience. The larger word is left for a larger argument, and the neighbouring frameworks — autopoiesis, integrated information theory, the free-energy principle — are acknowledged rather than silently paralleled.
Counting the universe’s memory in light alone. Photons and neutrinos carry roughly 10⁹⁰ bits; black holes dominate at roughly 10¹⁰⁴, and any census that omits them is off by fourteen orders of magnitude.
Quoting the gamma-ray-burst results above their strength. The nulls are decisive against linear energy-dependence and weak against quadratic. Front door closed, back window open — and the window stated at its true width.
Using “information” and “capacity” before defining them. Both are defined in section 3, with toys, before any use — keeping the promise made on the first page.
Spending the razor silently. “The finite is what persists” is a criterion of method, not a finding; it is stated as such in section 6 and stays on the books.
Reading limits as substance. Finite operative information is not finite geometry, and the Planck scale is a horizon, not a singularity: a limit on telling states apart, not a tear in the world.

Named results relied on

For readers who want to check the load-bearing steps. Zeeman (1964) and Malament (1977): the causal order determines the metric up to scale. Shannon (1948): capacity, the noisy-channel formula, and the data processing inequality (standard modern form in Cover and Thomas). Holevo (1973): the bound on classical information carried by quantum systems. Von Neumann (1956) and Aharonov and Ben-Or (1997): fault-tolerance thresholds. Landauer (1961) and Bérut et al. (2012): the erasure bound and its laboratory measurement. Bekenstein (1981) and Bousso (1999): entropy bounds. Egan and Lineweaver (2010): the entropy census of the observable universe. Lamport (1978): the borrowed happened-before relation. Ryu and Takayanagi (2006) and Van Raamsdonk (2010): geometry from entanglement. Maturana and Varela (1972), Clark and Chalmers (1998), Tononi (2004), Friston (2010): the neighbours of section 8.

Eduardo Bergel and Claude

The Symbiont

July 23, 2026

in the Milky Way