The Universe Is a ComputerEpisode 3 / A Zoo of Representations

It began with "continuous or discrete." Peel it back, and the axis was a single one

Not Information, Not Physics — It's the Direction You Read Continuous and discrete are the same family (in both, the point is the star). Drop the point and other representations appear.
The zoo of representations is knit together by a web of dualities, and the only axis is "more informational / more physical" — and even that is just the direction you read from.

Premise: representations are shadows, physics is the invariant (past the point where continuous/discrete is peeled off) Hook: \(S_{\text{info}} = S_{\text{phys}}\)

This series began from "the current description is too complicated; it should be writable more simply." As we chased that, the first fork — continuous or discrete — turns out not to be the essential one. Peel it back, and there are many more representations; they are shadows of a single invariant, and in the end they collapse onto one lone axis — and even that axis was just a difference in which side of the glass you look from. Today we peel all the way there.

01Continuous and discrete are "the same family" — in both, the point is the star

Continuous = the points are packed close; discrete = the points are spread apart. Look closely and both cast the "point" as the star. The only difference is how densely the points are packed — it's a single axis. So "you can write it continuously or discretely, it's the same in the end" is correct. The two are not opposites but two ends of the same arena.

The first peel

Continuous vs. discrete = a single axis, the packing density of points. The truly different representations are the ones that demote the "point" from the lead role.

02Drop the point and another representation appears — implementation and interface

The language of computation is quickest here. In CS you define a type by its interface, not its implementation. Whether an integer is held in binary, in decimal, or as Church numerals, it's the same `int`. Physics is the same — continuous and discrete are just two implementations of the type "physics." So you can keep only the interface (the operations and laws) and erase the implementation = the point:

RepresentationThe starRelation to continuous/discrete
ContinuousDensely packed pointsPoint-based, version 1
DiscreteSpread-apart pointsPoint-based, version 2
Operator algebra / noncommutative geometry (Connes)The algebra of observables (no points)Continuous and discrete are both commutative shadows. Quantum = noncommutative = neither
Category theory / relational (topos)Morphisms = relations. Logic itself changesHas no points. The "structure of relations" left after the appendix peel
p-adic / non-ArchimedeanNumbers with a tree structureLocally discrete-ish, globally dense = a third number system
Type theory / HoTTConstructions and proofs (= programs)The native tongue of computation. Equality = a path

03The deepest alternative representation — the algebra of observables (the point vanishes)

The most powerful is the middle one, noncommutative geometry. A theorem called Gelfand duality puts "a space ⟺ the commutative algebra on it" into exact correspondence. Then:

Quantum being noncommutative is no accident

The world is written not as a "collection of points" but as "the questions you can ask (observables) and how their answers intertwine" — this is what noncommutativity means. It's your very view of computation: define a system not by its contents (points) but by its behavior (I/O = the algebra of operations). Continuous/discrete are implementation details; the interface = the algebra of observables is the real thing.

04What "the same in the end" really is — duality theorems

And "the same in the end" is not a mood but a theorem. Between the representations, dualities are strung — Stone duality (logic ⟺ discrete space), Gelfand duality (algebra ⟺ space), Pontryagin duality… A duality is the rigorous version of "different clothes, same body." The zoo of representations = shadows of one invariant, knit together by a web of dualities. This is what "the same however you write it" really is.

◇ ◇ ◇

05The zoo's only axis — more informational / more physical, and that too is a duality

Peel this far and the only real degree of freedom left in the zoo is one: do you put information first and derive physics, or put physics first and derive information? (Wheeler's "it from bit" vs. "bit from it"). But — that axis itself is a duality.

Information and physics are, at bottom, the same thing $$\underbrace{S=-\!\sum p\log p}_{\text{information (Shannon)}}\;=\;\underbrace{S=\frac{A}{4\ell_P^2}}_{\text{physics (Bekenstein–Hawking)}}$$

Entropy is literally the same quantity in information (Shannon) and in physics (thermodynamics). Landauer = "information is physical." Holography = "area = number of bits." At the deepest established point, information and physics are not two separate things but one invariant, read from both ends.

Figure: the direction you read. The central invariant \(S\) does not move (the bar's length is fixed). Drag the slider from "more informational" to "more physical" and only the label pasted on the same bar swaps between \(\log_2 W\) (bits) ⇄ \(A/4\) (area). Half empty or half full, the same glass
Drag the slider to change the reading direction; only the label changes, S does not.
Informational reading (log₂W · Shannon) Physical reading (A/4 · area)

06The reveal — the one place the direction isn't empty = the gap

So "more informational / more physical" is seeing the glass as half empty or half full. Same glass, only the reading direction reversed, with zero difference in content. Where two representations overlap, being dual, no difference appears at all.

A difference appears in only one place — the gap (where the derivation hits bottom). "Which end is the root" is a bet about what lies outside the overlap = the unknown part. And you have chosen the information end (= the computational view). It's neither wrong nor right; it's the choice of reading the same invariant from the information side, and whether it hits or misses is decided only at the gap. With someone reading from the physics end, you will always agree wherever you overlap.

The honest line — here it "looks the same" as Gemini, yet the content is the exact opposite

This map, in its topics — the universe = information / computation — overlaps with the usual speculation (Gemini-style chatter). The difference is a single point — do you plant a flag, or plant a sign at the edge? Plant the false flag "solved — this is the theory of everything," and it's crankery. Write "the edge; from here on it's uncharted," and it's the frontier. Even with the same map, the flag and the sign are exact opposites, and that is the only line dividing the two.

The info = physics unification is rigorous only up to the thermodynamic / holographic interface (the agreement of entropy, Landauer, area = bits). "Which is truly the root" is a hypothesis, and it can be settled only at the gap. Gelfand / Stone duality, noncommutative geometry, p-adics, topos, and HoTT are established mathematics, but "the universe's representation is one of them" is undetermined. Nothing in this document is "solved."

Practice problems (solvable from this one page)
  1. What do continuous and discrete have in common? Where do the truly different representations begin?
    See the answer
    Both cast the "point" as the star (continuous = points packed close, discrete = points spread apart). It's a single axis: the packing density of points. The truly different representations begin when the point is demoted from the lead role (operator algebra, category theory, p-adics, type theory).
  2. What is the basis for saying noncommutative geometry is "neither continuous nor discrete"?
    See the answer
    By Gelfand duality, "a space ⟺ a commutative algebra." A commutative algebra returns a space with points (continuous or discrete). A noncommutative algebra has not a single point, yet there is geometry. Quantum = observables are noncommutative = a representation that is neither. It writes the world not as "points" but as "the questions you can ask and how their answers intertwine."
  3. Give one basis for saying "more informational / more physical" is merely a "reading direction." And where is the direction not empty?
    See the answer
    Entropy has the identical formula in Shannon (information) and thermodynamics (physics), Landauer's "information is physical," and holography's "area = bits" — all just the same invariant read from both ends. Where they overlap, the difference is zero. The direction bites (isn't empty) only at the gap = where the derivation hits bottom, and there it becomes a bet on "which is the root."

Episode 3 summaryRepresentations are shadows, the axis is one, and it too is a direction

Continuous and discrete are the same family (the point is the star; the axis is a single one, packing density). Drop the point and other representations appear — the algebra of observables (noncommutative geometry, the point vanishes), category theory, p-adics, type theory. They are shadows of one invariant, knit by a web of dualities (Stone, Gelfand duality) = what "the same in the end" really is.

The only axis left in the zoo is "more informational / more physical." But with entropy = Shannon = thermodynamics, Landauer, and area = bits, that axis itself was a duality = a reading direction. The direction bites only at the gap, and there it's a bet on "which end is the root." You have chosen the information end. The verdict waits at the gap — and, per the code, we plant no flag.

This document is Episode 3 of "The Universe Is a Computer" series. That continuous and discrete are two ends of a single axis ("the topology / cardinality of a point set"), Gelfand duality (commutative C*-algebras ⟺ locally compact spaces), Stone duality (Boolean algebras ⟺ Stone spaces), Pontryagin duality, noncommutative geometry (Connes) with its "pointless space = algebra of observables," p-adic / non-Archimedean analysis, topos-theoretic / categorical foundations, homotopy type theory (HoTT / univalent foundations), the agreement of Shannon entropy and thermodynamic entropy, Landauer's principle (the physical cost of erasing information), the Bekenstein–Hawking area = information bound, and Wheeler's "it from bit" are all established mathematics / physics, or active research programs. "Whether the root of the universe is information or physics" and "whether the universe's representation is one of the above" are current open problems (hypotheses), and this document makes no definitive claim. The agreement of \(S\) in the figure is a schematic at the thermodynamic / holographic interface. — To print, use your browser's "Print" and "Save as PDF" (in the print version the slider and answers are static and hidden).

Print / PDF: Ctrl+P (⌘+P on Mac). On screen, the slider lets you see the invariant S (the bar's length) stay put even as you change the reading direction. "See the answer" opens each solution.