Infinite Randomness

Boltzmann brains and the mathematics of everything

A freak fluctuation in empty space assembles, atom by atom, a functioning human brain — complete with a lifetime of memories, a mood, a headache, the vivid impression of sitting somewhere reading a book. An instant later it dissolves back into the void. This is the classic Boltzmann brain, and the standard reaction to it is a shudder followed by a dismissal: assembling a macroscopic organ by thermal accident is so fantastically improbable, so violent an offense against thermodynamic intuition, that the scenario is filed under cautionary tales for cosmologists and forgotten.

I want to go the other way. The Boltzmann brain is not too strange to take seriously; it is not strange enough. Its weakness is an unexamined assumption — that a conscious moment needs a brain, a particular arrangement of matter floating in space. Drop that assumption and the scenario stops being a freak-show improbability and becomes something close to a mathematical inevitability. This chapter pushes the thought experiment to its limit, shows that the mathematics underwriting it is surprisingly ordinary, and ends somewhere unexpected: at an ontology so minimal it contains almost nothing, yet is empirically indistinguishable from physics. It is the opening move of everything that follows in this part of the book.

Consciousness as Software

Suppose consciousness is fundamentally informational — a state of computation rather than a state of matter. This is not a wild premise; it is the working assumption of most cognitive science, and it is the same move made in A Gigaplex of Parallel Lives, where you are a pattern picked out wherever it occurs, not a particular clump of atoms. If the premise holds, then a Boltzmann brain does not need a brain. It needs a state.

Now imagine an infinite computational substrate — a vast memory, a medium of any kind — cycling randomly and endlessly through its possible configurations. Given infinite random cycling, every finite informational pattern occurs. Not “might occur”: occurs, and occurs infinitely often, regardless of its complexity. There is no thermodynamic miracle to pay for, because nothing macroscopic ever has to assemble. Conscious experiences become software states, spontaneously instantiated as the substrate churns.

The immediate objection is continuity. A momentary state is not a life; whatever flickers into existence for one configuration of the substrate seems to lack a past and a future. But look at what a conscious state actually contains. This present moment of yours includes its own history — coherent memories, a sense of how you got here, anticipations of what comes next. Each such state logically implies predecessor and successor states: the state whose memories are one moment shorter, the state whose memories include this one. And given infinite randomness, those implied states also occur. Entire subjective lifetimes exist in the substrate as internally coherent, logically connected sequences of states — whether or not any causal process ever runs them in order.

This is the radical step, so let me state it plainly: subjective continuity does not require external causal history. The felt connectedness of a life is a fact about the internal logical coherence of the states, not about how the states were produced. A consciousness instantiated this way would perceive a world history — evolution, cosmology, this morning’s breakfast — indistinguishable from one generated by conventional causation. From the inside, there is no experiment that separates a history that was caused from a history that is merely coherent.

Two things follow. First, external causation becomes epistemically optional: nothing in your evidence establishes that your past happened rather than merely cohering. Second, the simulation hypothesis loses its simulator. There is no need for an intentional engineer running ancestor simulations; infinite randomness alone generates every coherent simulated reality, ours included, for free. It is modal realism with a mechanism: every logically consistent possibility instantiated, not by decree, but by exhaustion.

Whether these spontaneous instantiations count — whether a counterfeit of your evidence, however internally coherent, carries any weight when you reason about where and what you are — is a separate question, and it is the hardest one in this territory. It has an answer, and the answer is not “count the copies”: it turns on which realizations of your evidence are admissible, genuinely grounding the structures that make your evidence evidence. That argument belongs to anthropics and is made in You’re Not a Random Sample. This chapter’s job is prior to it: to establish just how much the substrate contains, and how cheaply.

Your Life Is in π

“Every finite pattern occurs infinitely often” sounds like the kind of claim that should cost something — some exotic metaphysical machinery, some leap of faith about infinity. It costs almost nothing. The mathematics that delivers it is a standard, well-studied property of ordinary numbers, and you can hold examples of it in your hand.

Start with the size of the thing to be contained. Take a complete digital record of a human life — one hundred years of continuous high-definition video, efficiently compressed. Call it two petabytes: roughly \(1.8 \times 10^{16}\) bits, eighteen quadrillion. That is your life, every moment and every subtle detail, as one finite bitstring.

A real number is normal in a given base if every finite digit sequence appears in its expansion with exactly the limiting frequency you would expect from chance — each digit a tenth of the time in base 10, each pair a hundredth, and so on for every finite combination. Normality is believed, though not yet proven, to hold for π. If it does, then your eighteen-quadrillion-bit life record appears in the binary expansion of π. Not once — infinitely many times. So does every variant of your life: every road not taken, every alternative timeline, every version of you that never picked up this book. So does every book ever written, every genome, every connectome, every finite pattern conceivable. And since descriptions of π’s own digits are themselves finite strings, π contains infinitely nested descriptions of itself — self-reference embedded recursively in pure numeric form.

Two caveats keep this honest. First, the existence is implicit, not practical: locating your life’s bitstring in π is computationally intractable in a way that shades into impossibility, so this is a fact about what π contains, not a retrieval scheme. Second, π’s normality is a conjecture. Émile Borel proved in 1909 that almost all real numbers — in the precise measure-theoretic sense — are normal in every base; pick a real at random and it contains everything finite with probability one. Yet not a single familiar constant — π, \(e\), \(\sqrt{2}\) — has been proven normal in any base. Normality is simultaneously generic and unexhibited: nearly every number has the property, and we can barely point to one that provably does.

The Vulcan Number

Mathematicians handle that embarrassment by construction: if nature won’t hand over a certified normal number, build one. Champernowne’s constant, \(0.123456789101112\ldots\), concatenates the integers in base 10 and is provably normal there. The Copeland–Erdős constant does the same with the primes. These numbers are deliberately artificial — illustrations rather than discoveries — and that artificiality is exactly their virtue: the everything-container property holds by construction, no conjecture required.

In 2019 I contributed my own to the Online Encyclopedia of Integer Sequences: sequence A308705, the Vulcan Number. The goal was to make the containment property not just provable but obvious. The construction: concatenate every finite binary string in shortlex order — shortest strings first, lexicographic within each length:

0, 1, 00, 01, 10, 11, 000, 001, 010, 011, 100, 101, 110, 111, 0000, ...

Prefix a binary point and you have an infinite binary constant:

0.0100011011000001...

Every finite bit pattern appears, exactly on schedule, because the construction walks through all of them by brute enumeration. No conjecture stands between the definition and the property. And unlike some celebrated constructions in this neighborhood (Chaitin’s Ω is a normal number you provably cannot compute), the Vulcan Number is explicit, deterministic, and simple enough for an introductory programming course:

from decimal import Decimal, getcontext
from itertools import product

# Concatenate all binary strings in shortlex order (lengths 1..8 shown)
bits = "".join(
    "".join(pattern)
    for length in range(1, 9)
    for pattern in product("01", repeat=length)
)

# Evaluate the binary expansion 0.<bits> as a decimal
getcontext().prec = 70
vulcan = sum(Decimal(2) ** -(i + 1) for i, b in enumerate(bits) if b == "1")
print(vulcan)

Numerically, the construction converges on:

0.276387117279486523734198676211901230555089988160685506143676819115...

The uniform enumeration strongly suggests normality in base 2, though full normality in other bases remains unproven. The number was modestly intended, but it earns its place in this argument. It shows that “contains every finite pattern” is not a mystical property that infinity bestows on special numbers in special circumstances. It is a property you can engineer in an afternoon, with a definition that fits in a sentence. Your life’s complete bitstring — and every variant of it — sits at a definite, computable offset inside a number I can write down. Between π, where everything is conjectured to live, and the Vulcan Number, where everything provably lives, the lesson is the same: containing all finite information is cheap. An infinite substrate does not have to try.

And the random substrate of the previous section is just this lesson set in motion. A normal number contains every finite pattern statically, laid out along its digits; an infinite randomizing memory instantiates every finite pattern dynamically, cycling through them in time. Same mathematics, same inevitability. Once you grant an infinite random process, every finite conscious state — coherent memories included — is not an improbable visitor but a standing resident.

Coherence as Physics

One objection remains, and it is the serious one. A random substrate instantiates every finite state, yes — but overwhelmingly it instantiates noise. The coherent lifetimes are there, but so is an unimaginably larger ocean of static. Our experience is not static: it is lawful, quantitative, relentlessly regular. It obeys quantum field theory to eleven decimal places. Surely that is where the minimal ontology dies — randomness cannot fake the most precisely confirmed theory in science.

It can, and the way it does so is instructive. The scenario as stated has a free parameter I have been using informally: coherence — the criterion that picks out, from the ocean of noise, which state-sequences hang together as experience. Nothing forces that criterion to be vague. So make it exact: define logical coherence as the mathematical structure of quantum field theory. A sequence of informational states counts as coherent precisely when it satisfies the constraints QFT imposes — the field configurations, the symmetries and gauge invariances, the amplitudes assigned by the path integral.

With that single stipulation, the minimal ontology and the physical one become observationally equivalent. Every prediction QFT makes, the random substrate makes, because the only state-sequences that support experience are the QFT-coherent ones. Every experiment ever run confirms both theories equally, and no conceivable observation separates them. On one side: spacetime, quantum fields, particles as excitations, the whole furniture of physical realism. On the other: random informational states and one coherence constraint. Identical empirical content.

I am not announcing a discovery about which ontology is true; I am pointing at what the equivalence proves. Observational data cannot distinguish between radically different ontologies that share a logical structure — the choice between “quantum fields exist” and “randomness filtered by field-theoretic coherence” is underdetermined by every experiment that will ever be performed. And of the two, the random substrate is strictly the more parsimonious: it posits no fields, no spacetime, no matter — only states, randomness, and coherence. The most empirically successful theory in the history of physics turns out to run, without loss, on an ontology of almost nothing.

Almost nothing — but not quite nothing. The scenario still has two load-bearing components: an infinite random substrate, and a coherence criterion doing the selecting. Neither has been explained, only invoked. Why randomness as the ground floor rather than something more structured? The answer — that pure chaos is the one foundation that presupposes nothing, and is therefore the prior state of all priors — is the argument of Chaos as Foundation. And what coherence criteria are, where they come from, and how they build worlds is the business of the chapters after that. The Boltzmann brain, pushed to its limit, does not end in a shudder. It ends at the bottom of the stack, where the rest of this book begins.