Notes from the edge of knowing

zoenkol

A home for philosophical musings on computer science, mathematics, artificial intelligence, and physics — and the strange patterns that appear when these worlds overlap.

computer science
mathematics
artificial intelligence
physics

zoenkol.system / 0.2

The archive is still becoming.

Two transmissions have arrived. The archive around them is still being assembled, questioned, rearranged, and occasionally allowed to contradict itself.

read signal 002

state: archive forming ∞ / 00

Signal 001 · 02 August 2026

The finite ocean

Every possible image, every possible film, and what it really means to say that an artificial intelligence can navigate a space too large to search.

Somewhere inside the space of all possible pixel arrangements is every photograph that has been taken, every image that will be made, and an incomprehensible number of scenes that no mind will ever have reason to imagine.

Begin with an ordinary digital image. Under the familiar 24-bit RGB convention, each pixel has three colour channels—red, green, and blue—and each channel can take one of 256 integer values. That gives a pixel 256³, or 16,777,216, possible colours. This is not a rule of nature; modern formats may use different colour depths, alpha channels, or high-dynamic-range representations. It is simply a clean set of constraints for the thought experiment.

A picture as a number

Fix the width and height of the image. Once the number of pixels and the number of values available to each pixel are fixed, the number of possible images is also fixed. For an image containing W × H pixels, the count is:

possible images = (256³)^W×H = 2^24WH

A 1920 × 1080 image therefore has 2^49,766,400 possible raw RGB arrangements—approximately 10^14,981,179.

The number is finite, but “finite” does almost no practical work here. It is so large that enumerating the images one by one is not merely inconvenient. It is beyond any plausible physical computation. The overwhelming majority would look like unstructured noise. A tiny and difficult-to-define fraction would contain recognisable objects, readable text, diagrams, faces, landscapes, or works of art.

Still, each possible image can be placed somewhere in an ordering. Read the pixels from left to right, encode their channel values as bits, and the complete image becomes one very large binary integer. Increment that integer and another image appears. Continue long enough—in the purely mathematical sense—and every image of that exact size and colour format will occur once.

Possible is not the same as present

This is where the language becomes philosophical. It is tempting to say that all of these images already “exist” in a Platonic space. Mathematics supports the more modest claim that every valid image corresponds to a possible configuration, and that every configuration can be assigned a location in a finite set. It does not follow that the images are stored somewhere, waiting to be retrieved, or that they exist physically before a machine renders them.

The distinction matters. A coordinate can be defined without anyone visiting it. A sequence can be possible without being instantiated. The space is real as a mathematical object under the chosen rules; the status of its unseen members as “existing things” is a metaphysical interpretation rather than a computational fact.

Now let the images move

A movie adds a time dimension, but it does not change the logic. Fix the resolution, colour depth, frame rate, and duration. A two-hour film at 24 frames per second contains 172,800 frames. If each frame is a 1920 × 1080, 24-bit RGB image, the number of possible silent films is:

possible films = 2^{24 × 1920 × 1080 × 172,800} = 2^{8,599,633,920,000}

This is approximately 10^{2,588,747,761,649} possible frame sequences. Adding fixed-format audio would multiply the count by another enormous, but still finite, factor.

Inside that set are every two-hour arrangement of those frames: static noise, blank screens, almost identical films differing in one pixel for one frame, every physically accurate recording the format could represent, every impossible visual effect, every coherent story, and every incoherent sequence between them.

The set only remains finite because the constraints remain fixed. Remove the limit on duration, resolution, colour precision, or audio length, and the question changes. “All possible movies” without a bounded format is not the same finite collection.

AI does not search the ocean

This vast configuration space gives us a useful metaphor for generative AI, but the metaphor can easily become misleading. A modern image or video model does not iterate through possible outputs and inspect them until it finds one that matches a prompt. No useful system could perform that search. The space is too large, and there is no simple test that identifies the single correct image because prompts usually admit many acceptable answers.

Instead, the model learns statistical structure from examples. Natural images are not arbitrary pixel arrays. Nearby pixels are correlated. Edges continue. Lighting changes in patterned ways. Faces have recurring geometry. Language associates words with visual features and relationships. Video adds regularities of motion, persistence, causality, and viewpoint. These constraints make the meaningful parts of the full space highly non-uniform.

AI is not opening a catalogue of every possible movie. It is learning where human-like movies are likely to be.

Different model families implement this in different ways. A diffusion model begins with noise and repeatedly transforms it toward an image or sequence that is probable under its learned patterns and compatible with the prompt. An autoregressive model predicts successive tokens or patches from what has already been generated. In both cases, the system uses a learned probability distribution rather than exhaustive enumeration.

“Navigation” is therefore a metaphor for guided sampling. The prompt, the training data, the architecture, the model parameters, and the random seed constrain which regions can be reached easily. They do not provide a literal map with a named coordinate for every possible film. Nor is a particular model guaranteed to be able to produce every member of the full pixel space. Its reachable outputs may form only a minute, structured subset.

The real achievement is compression

Random selection from the complete space would almost always produce noise. A useful generator succeeds because it has absorbed enough regularity to avoid almost all of that space. In this sense, intelligence looks less like raw exploration and more like compression: the discovery of rules, correlations, and abstractions that make some configurations vastly more plausible than others.

A prompt such as “a violinist playing beneath a red moon” does not identify one pre-existing bitmap. It imposes constraints. The model then constructs one of many configurations that satisfy those constraints according to its learned distribution. The result may be new as an exact arrangement of pixels even though its possibility was already contained in the mathematics of the format and its structure is indebted to patterns learned from existing work.

That leaves a useful tension. The space of outputs is finite, but effectively untraversable. Every bounded digital artefact is representable as a number, but its meaning is not contained in that number alone. AI can produce remarkable points in the space, but only because human culture, physical regularities, and prior examples have given the space shape.

Perhaps the mystery is not that every possible image can be numbered. The mystery is that, from an ocean dominated by noise, minds and machines can repeatedly find configurations that mean something.

Boundary of the thought experiment: all counts assume fixed, discrete digital formats. “Platonic existence” and “navigation” are philosophical metaphors; the combinatorial counts and the distinction between enumeration and learned sampling are computational claims.

Signal 002 · 02 August 2026

Beneath the interface

A folder is not made of pixels, a bit is not a tiny switch, and looking deeper into matter may require us to abandon the language of things altogether.

On a computer desktop, a folder appears as a small coloured object with a name beneath it. Enlarge the icon and its curved edges become staircases. Enlarge it further and its colours separate into pixels. But no amount of magnification will reveal the files inside.

The reason is not that the magnifying glass is too weak. The picture on the screen is an interface: a visible sign produced by a chain of systems beneath it. The icon represents a folder, but it is not a miniature folder, and the folder is not composed of the pixels that depict it.

The folder that is not on the screen

On many displays, a visible pixel is produced from red, green, and blue subpixels. OLED subpixels emit light; LCD subpixels regulate light passing through colour filters. Looking closely at the screen can therefore reveal something about the display. It cannot, by itself, reveal what the folder means or how the computer represents it.

Behind the icon are several descriptive layers. The graphical interface draws shapes into a frame buffer. The operating system associates the icon with a file-system object. A directory records names and references to other objects. The file system maps those objects onto logical blocks or extents on a storage device. Each layer translates between a vocabulary useful at one scale and a different vocabulary useful below it.

The arrows do not mean that each layer resembles the layer beneath it. They indicate rules that connect different forms of representation.

folder icon OS object file-system metadata binary encoding physical state

Even the final two labels need care. A bit is not a microscopic bead marked zero or one. It is an abstract binary value. Engineers construct systems in which two distinguishable physical conditions stand for those values. In digital logic, ranges of electrical voltage are interpreted as low and high while transistor networks control how signals propagate. “Off” and “on” are useful shorthand, but the circuit itself remains an analogue physical system with noise, tolerances, capacitance, leakage, and intermediate voltages.

Storage may use different distinctions. DRAM commonly stores charge in capacitors; SRAM uses bistable transistor circuits; flash memory distinguishes threshold-voltage ranges; magnetic disks encode states in magnetised regions. Some flash cells use several voltage ranges to encode more than one bit. Binary is therefore not a special substance at the bottom of the machine. It is an interpretation that can be realised in many materials.

A folder is not made of pixels. A bit is not made of zero and one. Each layer is connected to the next without needing to resemble it.

Zooming into matter

A similar pattern appears when we turn from computers to physical objects. A table looks continuous and solid. Chemistry describes it as arrangements of molecules and atoms. Atomic physics describes atoms through nuclei and electron quantum states. Nuclear physics resolves most nuclei into protons and neutrons. Particle physics describes protons and neutrons as composite systems involving quarks and gluons.

The phrase “look closer” becomes increasingly metaphorical as the scale shrinks. An optical microscope forms an image from light. An electron microscope uses electrons with much shorter wavelengths. Particle accelerators do something less like ordinary seeing: they collide high-energy particles and measure tracks, energies, angles, and decay products. Internal structure is inferred from regularities in how systems scatter and transform.

The discoveries are not simply miniature versions of the objects above them. An atom is not a tiny classical solar system. A proton is not a rigid container holding three little balls. Quantum systems are described through states, probabilities, fields, conserved quantities, and interaction rules that do not map cleanly onto familiar pictures.

Where current physics stops saying “made of”

In the Standard Model, electrons and quarks are treated as elementary: experiments have not revealed smaller constituent parts within them. Protons and neutrons are composite; electrons are not currently known to be. “Elementary” means that the best-tested theory represents an entity without internal structure. It does not prove that no deeper account could ever be discovered.

Modern particle physics is formulated as quantum field theory. Within that framework, particles can be represented as quantised excitations of fields. This already changes the question. Asking what an electron is “made of” may presuppose the wrong form of answer, just as asking which pigment a software folder is painted with confuses an interface object with its implementation.

At one scale we ask, “What smaller objects are inside this object?” At a deeper scale, the productive questions may become: what state is present, which field is excited, what symmetries constrain it, how does it interact, and what remains invariant?

This does not establish that fields are the ultimate substance of reality. The Standard Model is extraordinarily successful but incomplete: it does not incorporate gravity in the same framework, and it does not explain every observed feature of the universe. Fields might be fundamental, emergent, or part of a deeper theory not yet known.

Where the analogy works

The analogy is strongest as a warning about categories. The visible qualities of an icon do not reveal the logical organisation of a file system. Likewise, the colour, texture, and solidity of an everyday object are not qualities we should expect its smallest describable constituents to possess in the same form. Higher-level objects can remain real and useful even when their vocabulary disappears at lower levels.

It also reminds us that observation has a readout. We do not encounter a quark as a tiny coloured dot. We encounter detector signals and statistical patterns interpreted through theories with precise consequences. A particle track is not the particle itself, just as a folder icon is not the directory itself. Both are causally connected representations.

“Representation” does not mean fiction. The icon responds reliably to changes in the file system. Particle models reliably predict scattering rates and decay patterns. An interface can hide implementation while still conveying genuine structure.

Where the analogy breaks

A computer is engineered. Its layers were designed, and in principle we can trace the chain from a click on an icon through software instructions, logic gates, device physics, and the materials of the machine. The desktop is deliberately an interface to a known implementation.

Nature may not be organised that way. There is no evidence that atoms are icons drawn by a cosmic operating system, or that an unseen substrate stores reality in a lower-level code. Layered descriptions do not prove that there must be a final layer resembling computer hardware. The search may end in entities that are not made of anything smaller, or a deeper theory may replace objects with relations and structures.

Physics also does more than enlarge the same interface. It changes the interaction, constructs new instruments, increases probe energy, and tests competing models. Rutherford did not merely magnify an atom; scattering experiments exposed the concentration of charge in a nucleus. Modern colliders extend this strategy to smaller scales.

We may never obtain a view from outside reality. But that does not mean we are confined to its surface.

The thing beneath the thing

The deeper insight is therefore not that close inspection can never tell us what matter is made of. It repeatedly has. The insight is that each successful descent can change what “made of” means.

At the desktop level, folders and files are objects with names and behaviours. At the file-system level, they are records, identifiers, mappings, and data. At the hardware level, they are no longer folder-like objects at all, but patterns maintained across physical states. None of these descriptions is merely the same picture with more detail.

Physics may confront us with an analogous transition. Everyday objects give way to atoms; atoms give way to quantum systems; composite particles give way to fields and interactions. The deeper account may not supply smaller pieces carrying the same familiar qualities. It may replace the ontology—the kinds of entities the theory asks us to use.

In that restrained sense, the foundational may indeed be “other.” Not necessarily hidden behind an impenetrable screen, and not necessarily unknowable, but unlike the interface through which it first became visible. The limit may not be how far we can zoom. It may be how long our inherited concepts remain appropriate as we do.

Boundary of the analogy: a bit is an abstract binary value realised through distinguishable physical states; voltage levels are one common implementation, not the universal substance of computation. Current physics treats protons and neutrons as composite and quarks and electrons as elementary within the Standard Model. The claim that reality must contain an inaccessible final layer is philosophical, not an established scientific result.