Life, Information, and the Limits of Simulation

The Old Riddle of a Living Cosmos

Long before anyone had a word for biology, Aristotle imagined the entire cosmos as a single enormous organism, straining toward some final, built-in purpose — every falling stone, every growing tree, every orbiting sphere pulled forward by the same kind of directedness that pulls an acorn toward being an oak. This is teleology: the idea that things move not just because something pushed them, but because something is drawing them toward a destined end-state. Modern physics threw this idea out with real force. Newton's universe runs on push, not pull toward purpose; a falling apple has no goal, only a prior cause. And yet the question underneath Aristotle's picture never actually left. It just changed vocabulary. We no longer ask whether the heavens have intentions. We ask whether the universe computes — whether reality, at bottom, is an information-processing system, and whether life is simply what happens when that system's computations become sufficiently intricate.

This reframing is not merely a modern affectation. It matters because "computation" gives us something teleology never could: a mechanism. A goal is mysterious — what pulls the acorn forward, and from where? A computation is not mysterious in the same way; it is a sequence of definite steps, each one following from the last by a fixed rule, with no destination built in from the start, only an unfolding. If life and mind turn out to be the products of computation, then Aristotle's ancient intuition — that the universe has something organism-like about it — would be vindicated, but for reasons Aristotle could never have imagined, and stripped of the purpose he thought was doing the work.

The Trouble with Defining Life: A Family Resemblance, Not a Formula

Before we can ask whether the universe computes life into existence, we need some workable sense of what "life" even picks out — and this turns out to be far harder than it first appears. We recognize life instantly: a bacterium, a redwood, a hummingbird. But try to write down a rule that separates the living from the non-living, applying to all and only living things, and the rule collapses under its own weight.

Consider the usual candidates. Reproduction — but a mule cannot reproduce, and a virus cannot reproduce without hijacking a host cell's machinery, yet nobody doubts a mule is alive or that viral disease is a biological phenomenon. Response to external stimuli — but a thermostat responds to temperature, and a sunflower's phototropism is barely more sophisticated than a simple feedback loop; the criterion is either too narrow or too permissive depending on how strictly you apply it. Growth — but so does a crystal in a supersaturated solution, adding to itself along the lines of its existing lattice, with no metabolism, no information storage, nothing we'd want to call life. Metabolism, the processing of energy to maintain structure — but a candle flame consumes fuel, releases waste heat, and maintains a stable, self-sustaining shape for as long as fuel is available; it looks, in a crude sense, like it's "alive" by this test alone.

What's happening here is not that we've failed to find the right single criterion yet. It's that life is what philosophers call a cluster concept or family resemblance category: something identified not by one necessary-and-sufficient property but by an overlapping bundle of properties, no single one of which is required, and no fixed number of which is sufficient. Real organisms cluster tightly around several of these properties at once — reproduction and metabolism and response to stimuli and internal regulation and evolution by selection — while borderline cases like viruses, prions, or self-replicating RNA molecules in a test tube satisfy only some of them, which is exactly why they sit at the edges of our intuitions and provoke argument rather than settled agreement. This is not a failure of biology; it is a clue. It suggests that "life" is not a natural kind with sharp edges, like "gold" or "electron," but a graded phenomenon — which in turn suggests that instead of asking is this thing alive or not, the more productive question might be how much, and in what way, is this thing organized like the things we call alive. That reframing — from a yes/no category to a matter of degree and kind of organization — turns out to be the thread that connects everything that follows.

Two Origin Stories, and What Turns on the Difference

This uncertainty about what life is feeds directly into a much bigger disagreement about how life began, and the disagreement is not merely academic — it changes what we should expect to find when we look at the rest of the universe.

The first view treats the emergence of the first living organism as a near-miracle: an astronomically improbable convergence of chemical accidents, so unlikely that it plausibly happened exactly once, on this one planet, and would be vanishingly unlikely to repeat itself anywhere else. On this view, Earth's biosphere is less like a common outcome of chemistry and more like a lottery ticket that happened to pay out — and if you ran the tape of the early solar system again, with slightly different initial conditions, you might easily get a sterile planet.

The second view denies that there is any sharp threshold to cross in the first place. On this account, there is no bright line between non-living chemistry and living chemistry — only a long, continuous ramp of increasing organization, starting with simple molecules, moving through self-catalyzing chemical cycles, then through molecules capable of storing and copying information, and eventually arriving at something we're willing to call "alive," without any single step along the way being especially miraculous. The origin of life, on this account, is not an exception to the ordinary processes of physics and chemistry; it is simply one more phase transition, comparable in kind (if not in complexity) to water freezing into ice or a supersaturated solution suddenly crystallizing.

The stakes of this disagreement become obvious the moment you think about the search for life elsewhere. If the first view is right, then the Drake Equation's optimistic estimates of civilizations across the galaxy collapse toward a single, lonely data point — us — and the universe's silence (the "Fermi paradox") stops being surprising and becomes almost expected. If the second view is right, then wherever you find the right raw ingredients — carbon chemistry, liquid water or an equivalent solvent, an energy gradient to exploit — you should expect something organizing itself in life-like ways, given enough time. The two views make different predictions, at least in principle, which is what keeps this from being pure philosophy: it's a live empirical question, one that missions to Europa's ice-covered ocean or Enceladus's plumes are, in a small way, designed to help answer.

From Soup to Cell: What Miller-Urey Actually Showed, and Where It Falls Short

The experimental result that gave the second view — life as a common, near-inevitable outcome — real scientific teeth was Stanley Miller and Harold Urey's classic experiment: sparking electricity through a flask of gases meant to mimic the early Earth's atmosphere, and finding that amino acids, the building blocks of proteins, appeared within days, unbidden, from nothing more exotic than ordinary chemistry and an energy source. Amino acids are not alive. They are not even close to alive — a modern protein might be a chain of hundreds of them, folded into a precise three-dimensional shape that determines its function, and no experiment has ever shown that shape emerging spontaneously in one step. But the ease with which the raw ingredients appeared was itself the striking result. It suggested that the universe's basic chemical vocabulary — the letters life is written in — is not rare or fragile. It falls out of ordinary reactive chemistry almost as a matter of course.

Life, once you have those letters, runs on two chemical families working in a division of labor. Nucleic acids — DNA and RNA — store and transmit the instructions for building an organism: a kind of durable, copyable archive, analogous to a program's source code. Proteins do the actual work described by that archive, serving two roles at once: structural, forming the physical scaffolding and machinery of the cell, and catalytic, speeding up chemical reactions by factors of millions or more, without which the cell's chemistry would be far too slow to sustain anything resembling life. This division — an archive that stores instructions, and a workforce that executes them — is worth sitting with, because it is already, at the level of raw biochemistry, a computational architecture: information storage separated from information execution, exactly the separation that a Turing machine's tape (storage) and read/write head (execution) formalizes in pure logic. Life did not need to wait for engineers to invent this architecture. Chemistry seems to have stumbled into something functionally equivalent to it on its own.

What we still don't understand well is the jump from isolated amino acids to functioning proteins, and, even more so, the origin of the nucleic acids that store the instructions in the first place — a puzzle serious enough that it has its own name, the "chicken-and-egg problem" of the origin of life: proteins are needed to build nucleic acids, but nucleic acids are needed to specify proteins, so which came first? One influential resolution, the "RNA world" hypothesis, proposes that RNA — which, unlike DNA, can both store information and catalyze reactions on its own — briefly played both roles simultaneously, acting as its own archive and its own workforce, before a division of labor eventually split those jobs between DNA and protein. Whether or not this particular hypothesis is correct in its details, the reasoning behind it illustrates the wider point: the origin of life increasingly looks like a search for a plausible sequence of small, individually unsurprising steps, rather than a single impossible leap — which tips the argument, even if only provisionally, toward the second of the two origin stories above. Life, on the evidence so far, appears to be a rather common feature of any universe built from the same physics and chemistry as ours.

There is a deeper physical principle lurking underneath all of this, one the original discussion only gestures at: life is not a violation of the tendency of closed systems to run down into disorder (the second law of thermodynamics), but neither is it a simple exception to it. Living systems are what the chemist and physicist Ilya Prigogine called dissipative structures — organized patterns that persist and even increase their internal order specifically because they are constantly importing energy from their environment and exporting entropy (disorder) back out into it, the way a whirlpool maintains its intricate, stable shape only by continuously channeling water through itself. A living cell, in this light, is not fighting the laws of thermodynamics; it is a particularly elaborate way of obeying them, buying local order by exporting a larger quantity of disorder elsewhere. This gives us a physical, non-mystical answer to a question that otherwise sounds almost paradoxical: how can complexity increase in a universe that is, overall, running down? The answer is that it can increase locally and temporarily, wherever there is an energy gradient to exploit — a sunbeam hitting a leaf, a chemical gradient at a hydrothermal vent — provided the larger system as a whole still pays the entropy bill.

Von Neumann's Universal Constructor: Life as Pure Logic

If biochemistry alone can't tell us what life fundamentally is — since, as we've seen, dissipative structures and self-organizing chemistry are necessary but not obviously sufficient — perhaps a more abstract, substrate-independent description can succeed where chemistry leaves off. This was John von Neumann's ambition: to strip away the wet chemistry entirely and find the logical skeleton underneath, the pattern that makes something alive regardless of what material happens to implement it.

Von Neumann's key insight, worked out in the 1940s in the abstract setting of what we'd now call a cellular automaton (a grid of cells, each following simple local rules based on its neighbors' states), was that a machine capable of exact self-reproduction requires three distinct components working together, and that all three are logically necessary, not optional design choices. First, a description of the machine — a complete, storable blueprint of its own structure, analogous to a genome. Second, a constructor — a mechanism that reads that description and builds a new copy of the machine from raw material, analogous to the cellular machinery (ribosomes and associated apparatus) that reads genetic instructions and assembles proteins. Third, and this is the subtle part, a copier — a mechanism that also duplicates the description itself, unread and unchanged, and hands that duplicate to the offspring, so the new machine has its own blueprint to reproduce with in turn, rather than being built once and then unable to reproduce further.

This three-part architecture is not an accident of von Neumann's design; it is a logical requirement, and it is worth pausing on how strange it is. A machine that could only build a copy of itself, without also copying its own blueprint, would produce sterile offspring — perfect copies structurally, but incapable of building a third generation, because they'd have no description of themselves to pass on. Von Neumann worked this out purely through mathematical logic, years before Watson and Crick determined the physical structure of DNA. When the double helix was finally decoded, biology turned out to have independently implemented almost exactly this architecture: DNA plays the role of the stored description, ribosomes and associated enzymes play the role of the constructor, and DNA replication (copying the archive itself, separately from building proteins) plays the role of the copier. Von Neumann had, in effect, derived the logical necessity of something like DNA replication from pure reasoning about self-reproducing automata, before anyone knew what DNA even was. This is a remarkable convergence, and it is exactly the kind of evidence that gives weight to the idea that life's essential nature is not really about carbon and nitrogen at all — it is about a particular logical architecture, one that happens, in our universe, to have been implemented in biochemistry, but that could in principle be implemented in silicon, in a cellular automaton, or in any sufficiently expressive computational substrate.

Organization as the Mark of Life: The Whole Versus the Sum of Its Parts

Von Neumann's architecture tells us how self-reproduction is logically possible, but it doesn't yet tell us how to measure the difference between a living system and a merely complicated one. For that, we can turn to ideas from algorithmic information theory, the branch of mathematics (developed largely by Andrei Kolmogorov, Ray Solomonoff, and Gregory Chaitin) that measures the complexity of an object not by how big or intricate it looks, but by the length of the shortest possible computer program that could produce it. A string of a million zeros is enormously long but has almost no complexity in this sense, because a tiny program ("print 0, one million times") reproduces it exactly. A string of a million truly random digits, by contrast, has close to maximal complexity in this sense, because no program shorter than the string itself can reproduce it — you simply have to spell it out.

This framework lets us give a sharper version of Chaitin's proposal about what separates living organization from mere aggregation: a living system is one where the complexity of the whole is substantially less than the sum of the complexities of its parts considered separately. This sounds paradoxical until you unpack it. Consider a heap of a trillion independent, unconnected atoms, each jiggling on its own trajectory: to specify the state of the whole heap, you basically have to specify each atom's state individually, because knowing one atom's position tells you almost nothing about any other atom's position. The complexity of the whole is close to the sum of the complexities of the parts — there is no shortcut, no compression available, because there is no relationship between the parts to exploit. Now consider a living cell, where the trillion atoms are bound into interlocking, mutually constraining structures — membranes, enzymes, genetic machinery, feedback loops — such that knowing the state of one part tells you an enormous amount about the state of many other parts, because they are functionally coupled. Here, a vastly shorter description suffices: instead of listing every atom's position, you can largely get away with saying "this is a functioning liver cell in phase G1 of its cycle," and a huge amount of detailed structure follows automatically from that compact description, because the parts are not independent — they are interdependent, mutually determining, effectively "aware" of each other's states through chemical signaling and physical binding. This is what Chaitin meant, in more precise language, by saying life means unity, and that dead matter is best captured as the sum of its parts while living matter is something less than that sum — not less in importance, but less in the number of independent bits of information actually needed to describe it, because the parts have partially collapsed into a single, tightly coupled whole.

Logical Depth: Why Some Complexity Is Meaningful and Some Is Just Noise

Algorithmic complexity alone, however, has an embarrassing blind spot: by this measure, a string of purely random digits is more complex than the human genome, because a random string is, by definition, incompressible — there is no shorter program that generates it — while the genome, despite encoding a staggeringly intricate organism, is compressible in principle (it is, after all, built from a small alphabet of four bases, repeated and rearranged according to evolutionary and biochemical constraints). This means raw algorithmic complexity, taken alone, cannot be what makes an organism special, since blind noise would score higher. Something else is needed to distinguish a genome — meaningful, structured, "deep" complexity — from a string of coin flips — meaningless, unstructured, "shallow" complexity, even in cases where the two might have comparable raw information content.

This is precisely the gap Charles Bennett's notion of logical depth was designed to close, and it deserves a fuller explanation than a quick summary allows. Bennett's insight was to stop asking "how much information does this object contain" and start asking "how much computational work would it take to produce this object from a short, simple starting description." A string of random coin flips has high information content (you can't compress it) but is logically shallow, because the "program" that produces it is essentially the string itself, with almost no computation involved — you just flip the coins and write down the results, with no meaningful process transforming a compact starting point into an elaborate output. Contrast this with the digits of pi: any given long stretch of them looks statistically indistinguishable from random noise, yet they are logically deep, because a very short, simple program (a formula for computing pi) can generate arbitrarily many digits, but only by doing a large and irreducible amount of computational work — you cannot shortcut the calculation, and the eventual output faithfully preserves, in compressed form, all of that computational history.

A living organism, on this view, is logically deep in an even richer sense than the digits of pi, because its "short program" is not a static mathematical formula but an unfolding process: billions of years of evolutionary search, in which countless variations were generated, tested against a punishing environment, and the vast majority discarded, with only the rare improvements retained and built upon by the next round. A genome is, in effect, an extremely compressed record of that entire process — a message whose "buried redundancy," in Bennett's phrase, is recoverable only by an observer willing to do a comparable amount of work to unpack it, whether that means running natural selection forward again in simulation or painstakingly reconstructing evolutionary history from fossils and comparative genomics. This is why an organism feels qualitatively different from a snowflake or a sand dune, even though both display intricate, non-repeating structure: the snowflake's pattern, however elaborate it looks, is the product of comparatively simple, fast physical rules (freezing dynamics) applied over a short time; the organism's pattern is the product of an immensely long, effortful, cumulative search process, whose "cost" is baked irreversibly into the final structure. Depth, unlike raw complexity, tracks time invested, not just intricacy — and this gives us, finally, something close to a rigorous, substrate-independent answer to the question the essay opened with: life is what you get when a system's structure has enough logical depth that its "buried redundancy" could only plausibly have been produced by a long history of accumulated, selected computation, rather than by a short, cheap process or by pure chance.

Is the Universe a Computer? The Digital Physics Program

Once organization, unity, and depth are on the table as the real markers of life, a far larger question becomes almost impossible to avoid: is the universe itself organized this way — is it, at bottom, a computation, with life and mind as unusually deep patterns arising within it?

Several serious physicists have answered yes, in registers ranging from cautious metaphor to full commitment. John Wheeler's contribution was more a research program than a single claim, distilled into his slogan "it from bit": the proposal that every physical "it" — every particle, every field, every force — is not merely described by information but is, at the deepest level, constituted by it; that asking a yes/no question of nature (does the particle go left or right? is the spin up or down?) is not just how we learn about physical reality, but is in some sense how physical reality gets made, moment to moment, from an underlying substrate of information. Ed Fredkin and Tom Toffoli pushed this into an explicitly computational picture, treating the universe as a literal, vast cellular automaton — not metaphorically, but as a genuine claim about physical mechanism, in which particles are patterns propagating through a discrete computational substrate according to fixed local rules, the same way a glider moves through Conway's Game of Life. On this view, physicists are not so much discovering laws of nature in the traditional sense as reverse-engineering an already-running computation that was set in motion long before any of us existed, hitching a ride on someone else's ongoing calculation and trying to figure out which parts of it happen to intersect with what we care about.

Frank Tipler took the most radical position of the group, treating the physical universe as strictly equivalent to its own simulation when viewed at a sufficiently abstract level — not merely simulable in principle, but identical, in every respect that matters, to an abstract computational process. This is a stronger claim than Wheeler's or Fredkin's, because it collapses the distinction between "the universe" and "a description of the universe running as a program" altogether, rather than merely proposing information as the fundamental ingredient of physics.

It's worth noting that this whole family of ideas has a more precise modern descendant in what's sometimes called the Church-Turing-Deutsch principle: the proposal, associated with David Deutsch's work on quantum computation, that every finite physical system can in principle be simulated to arbitrary accuracy by a universal computing device. If true, this principle would mean that computation is not just one useful model for physics, but is built into the very structure of what physical law permits — that "being simulable" is a property built into the fabric of any universe governed by consistent, finite laws, our own included. This gives the older, more poetic versions of the "universe as computer" idea (Wheeler's slogan, Fredkin's cellular automaton) a more rigorous modern anchor, even for those who remain skeptical of the strongest, most literal versions of the claim.

The Continuum Objection, and the Case for a Discrete Universe

The most serious scientific objection to this whole picture came from an unlikely source: Richard Feynman, one of the founders of the very field — quantum computation — that would later make digital physics fashionable again. Feynman's worry, when he considered whether nature could ever be exactly simulated by a finite computation, was structural rather than merely practical: our best theories of physics treat space and time as genuinely continuous, infinitely subdivisible, with no smallest possible distance or duration — and a computation, by its nature, proceeds through a finite number of discrete steps. If space and time really are continuous all the way down, with no floor, then an exact, finite simulation of even a small region of space for a small duration of time would require, in the limit, an infinite number of computational steps — which is not merely impractical but is a difference in kind, not degree, between computation and physical reality. This is a serious objection, not a hand-wave, and it deserves to be taken as seriously as Feynman intended it.

But, as Paul Davies has pointed out, the apparent continuity of space and time is not something we've ever actually verified — it is a working hypothesis, baked into the mathematics of our theories because continuous mathematics (calculus) happens to be enormously convenient and has worked extraordinarily well at every scale we've so far been able to probe, not because anyone has confirmed there is no smallest possible "grain" to reality. Our best current experiments and observations constrain the graininess of time to be smaller than roughly 10⁻²⁶ seconds — an extraordinarily short interval, but not zero. Davies' analogy is apt: a movie film advancing one frame at a time looks perfectly smooth and continuous to an eye that cannot resolve the individual frames, and the fact that we haven't yet detected the "frame rate" of reality does not mean reality has no frame rate; it may simply mean our instruments aren't yet fine enough to see the individual frames.

This is not idle speculation dressed up in physics language — it connects directly to some of the most active research programs in fundamental physics. Many approaches to quantum gravity (the unfinished project of reconciling general relativity's smooth spacetime with quantum mechanics's granular, probabilistic rules) independently arrive at some notion of a smallest meaningful length, the Planck length (roughly 10⁻³⁵ meters), below which the very concepts of "distance" and "duration" may cease to have their ordinary meaning, dissolving into something more like discrete, combinatorial structure — spin networks, causal sets, or other candidate discrete substrates, depending on which research program you consult. Separately, the holographic principle, motivated by results in black hole thermodynamics, suggests that the maximum amount of information that can be packed into any region of space is proportional not to its volume but to the area of its boundary, measured in fundamental, discrete units — a deeply strange result that only makes sense if information, at the most basic level, comes in indivisible chunks rather than a smooth continuum. None of this proves the universe is a computer in Fredkin's literal sense. But it does mean Feynman's continuum objection, while entirely reasonable given the physics available at the time, may rest on an assumption — true continuity — that independent lines of cutting-edge research are now actively questioning from a completely different direction, for reasons that have nothing to do with the philosophy of simulation at all.

Can a Simulation Be Conscious? Tipler, Penrose, and the Limits of Verification

If the universe can, in principle, be described as computation, a sharper and more unsettling question follows immediately: could the conscious beings inside that computation — could we — be simulated rather than "fundamentally real," in whatever sense that phrase is even supposed to mean, and would there be any way to tell the difference?

Tipler's argument here is a piece of careful, if unsettling, logic, and it's worth reconstructing it step by step rather than just stating the conclusion. Take any test a person might use to convince themselves that they genuinely exist and are not merely simulated: noticing that they are thinking, interacting with an external world that pushes back in consistent ways, remembering a continuous personal history, reflecting on the very fact that they are reflecting. Now notice that a sufficiently detailed simulation of a person would, by the very completeness of the simulation, perform every one of these same tests and get every one of the same results — because the simulated person's thoughts, memories, and reflections are themselves part of what's being simulated, indistinguishable from the inside from the "real" article. The crucial move is realizing that this isn't a limitation of some particular clumsy simulation technology that better engineering could eventually fix; it's a structural, in-principle barrier. A simulated observer's only tools for investigating their own reality are tools that exist within the simulation, which means those tools can never, even in principle, reach outside the simulation to check whether there's anything "underneath" it — the way a character in a novel has no access to the paper and ink, or the author, that constitute the "reality" underlying their fictional world, no matter how vividly the character might reason about their own existence.

Tipler's conclusion is a specific philosophical move: since no experiment, real or imaginable, could ever distinguish a "fundamentally real" physical universe from a sufficiently complete and self-consistent computational one, the distinction between them carries no empirical content, and — adopting a broadly empiricist stance that treats unverifiable distinctions as meaningless rather than merely unknown — should be set aside as a Kantian "thing-in-itself": a concept we are logically forced to gesture at but can never actually cash out in observation, and therefore not a fact about the world so much as an artifact of how we've phrased the question. On this view, the sensible conclusion isn't "we can't know whether the universe is real or simulated" but rather "the question, as posed, doesn't actually distinguish two different possible states of affairs" — rather like asking whether the number seven is "really" red.

Roger Penrose represents the most serious and specific dissent from this entire line of reasoning, and it is worth being precise about where exactly he disagrees, because it is not where casual summaries usually place him. Penrose does not simply doubt that current computers can replicate consciousness — a comparatively mild and widely shared skepticism. His claim is much stronger: that the physical processes which give rise to conscious awareness are not, even in principle, the kind of process that any computation — however powerful, however complete — could faithfully reproduce, because (in Penrose's specific and controversial argument, developed with the anesthesiologist Stuart Hameroff) consciousness may depend on physical processes, possibly involving quantum effects in the brain's microstructure, that are fundamentally non-computable in the technical mathematical sense established by Gödel's incompleteness theorems and Turing's halting problem. If Penrose is right — and this remains a minority position among physicists and neuroscientists, sharply contested on both physical and biological grounds — then the entire "universe as computer" picture, however well it might describe rocks, weather, and galaxies, hits an impassable wall exactly at the point that matters most to the question of whether we could be simulated: it may describe everything about a person's physical brain and behavior, while remaining permanently, structurally unable to account for the fact that there is something it is like to be that person — the philosopher's "hard problem" of consciousness, sharpened by Penrose into a specific, falsifiable-in-principle claim about computability rather than left as a vague mystery.

This same underlying puzzle — that a sufficiently good simulation would be indistinguishable from the inside — has since become one of the more widely discussed arguments in contemporary philosophy of mind and cosmology, under the general heading of "simulation arguments," which reason from the sheer number of possible simulated minds a sufficiently advanced civilization could in principle run to a probabilistic conclusion about which kind of world any given observer is more likely to find themselves in. The details of that later literature differ from Tipler's framing, but the core structural insight — that internal, first-person verification cannot in principle distinguish simulated from "base level" reality — is exactly the same one Tipler was pointing at.

Endophysics: Why We Can Never Step Outside to Check

There is a further wrinkle, developed by the physicist Karl Svozil, that doesn't so much resolve the Tipler/Penrose disagreement as explain, at a deeper level, why it is so stubbornly hard to resolve — and this piece of the puzzle deserves more attention than it usually gets, because it reframes the entire debate as being about the structure of observation itself, not just about consciousness or computation.

Svozil distinguishes two fundamentally different postures a scientist can take toward a system under study. In exophysics (from the Greek for "outside"), the observer and the system form a clean, two-level hierarchy: the observer sits outside the system, peers in, and measures whatever they like without disturbing what's being measured — information flows in one direction only, from system to observer, the way an astronomer observes a distant galaxy without the galaxy's behavior depending in any way on the astronomer's presence. Essentially all of laboratory science, and essentially all of our everyday intuitions about "objective observation," are built on this exophysical model, because in practice the disturbance an observer causes is negligible compared to the system being studied.

Endophysics (from the Greek for "inside") describes a completely different situation, one where the observer is not outside the system but is themselves a constituent part of the very system under investigation, built from the same stuff, governed by the same laws, with no external vantage point available even in principle. Here the clean hierarchy collapses: "measuring device" and "thing being measured" are no longer permanently distinct roles but can trade places depending only on which part of the system you're currently treating as the observer, and information flows both ways, contaminating every measurement with the observer's own presence and activity inside the very system being measured. This is not a hypothetical curiosity; it is arguably the actual epistemic situation of every physicist who has ever studied quantum mechanics, where the act of measurement is now well established to unavoidably disturb the system being measured, in a way that cannot be reduced below a certain fundamental limit no matter how careful or clever the experimenter is — the observer effect is not a matter of clumsy instruments, but a structural feature of what it means to measure something from within the same physical universe that contains both the measuring device and the thing measured.

This distinction explains, with real precision, why the question "is our universe fundamentally real or a simulation" is so resistant to closure. We are not exophysical observers with a privileged, external vantage point on the cosmos, the way a programmer sits outside the computer running their program, able in principle to pause it, inspect its internals, and compare the running simulation against the intended design. We are endophysical participants, built entirely from the same physical (or computational) substrate as the universe we are trying to characterize — which means any theory we construct about whether that substrate is "fundamental" or itself simulated is, itself, a product of computation happening inside the very system whose fundamental status is in question. There is a structural echo here of Gödel's incompleteness theorems, which showed that any sufficiently powerful formal system contains true statements about itself that the system cannot prove using only its own internal resources — self-reference, once again, runs into a wall precisely when a system tries to fully characterize itself from within, using only tools the system itself provides. Archimedes once imagined that, given "a point outside the world," he could move the entire Earth with a lever of sufficient length — a boast about mechanical leverage, not metaphysics, but one that has since become an apt metaphor for exactly the kind of external vantage point that endophysics says is permanently unavailable to any observer who is, themselves, part of the world being levered.

Deutsch's Insight: Why the Whole Edifice Rests on a Contingent Fact

It's worth ending on a smaller, more grounded point, made by the physicist David Deutsch, because it quietly ties this entire chain of speculation back to something concrete and, unlike most of what precedes it, genuinely uncontroversial — and it deserves more weight than a passing mention, because it reframes everything before it as resting on a foundation that could, in principle, have been otherwise.

Why can we build calculators? Why can human beings do arithmetic in their heads at all? The question sounds almost too simple to be worth asking, but Deutsch's answer is genuinely surprising: this capacity is not guaranteed by logic or mathematics in the abstract, the way the truths of arithmetic themselves are supposedly guaranteed. It depends, instead, on a contingent, empirical fact about this particular universe: that the laws of physics happen to permit the construction of physical systems — whether silicon transistors, mechanical gears, or networks of neurons — whose behavior tracks and instantiates abstract operations like addition, subtraction, and multiplication. In a universe with different physical laws, arithmetic would remain just as true in the abstract (two plus two would still equal four as a fact of pure mathematics), but there might be no way to physically compute it — no possible arrangement of matter and energy, in that hypothetical universe's physics, capable of tracking the operation reliably. Addition, multiplication, and the rest would remain non-computable functions: things we might still reason about abstractly, invoke as steps in what would then have to be called "non-constructive" proofs, but never actually perform, because no physical process in that universe's laws would be capable of tracking them.

That single observation reframes everything that came before it in this essay. The question of whether life is common, whether organisms are logically deep in Bennett's sense, whether the cosmos itself is a cellular automaton, whether consciousness could ever be simulated, whether an endophysical observer could ever step outside their own universe to check — all of these turn out to be, at bottom, questions about what a particular, contingent set of physical laws happens to make computable and organizable. Life is what these specific laws allow chemistry to build, given enough time and the right gradients. Minds are what these specific laws allow neurons (or perhaps, if Penrose is wrong, silicon) to compute. And our very capacity to ask any of these questions — to reason, to calculate, to wonder whether we ourselves are simulated — is itself made possible only because physics, in this universe, happens to permit exactly the kind of physical processes that can track logical and arithmetical operations. We are not external auditors examining the source code of reality from a safe distance outside it. We are, in the fullest and most literal sense available to us, running on it.

Living With an Unanswerable Question

What makes this whole cluster of ideas so durable — stretching from Aristotle's teleological cosmos through Miller and Urey's flask of sparking gases to von Neumann's self-reproducing automata, Chaitin's and Bennett's measures of depth and organization, and finally to Tipler's and Svozil's arguments about the limits of self-knowledge from within a system — is not that any of it settles the question of what life is or whether the universe computes. None of it does, and arguably none of it can, for the endophysical reasons Svozil's argument makes precise: we are asking these questions from inside the very system we're trying to characterize, using cognitive and computational resources that are themselves part of what's under investigation, with no lever long enough and no point outside the world to stand on while we pull it.

What this body of ideas offers instead is something more valuable than a final answer: a genuinely better set of questions, sharpened by a century of mathematics and physics that Aristotle didn't have available. Instead of asking whether the universe has a purpose, we can ask whether it has a computational structure, and what kind. Instead of asking whether something is alive in a binary sense, we can ask how organized, how integrated, how logically deep its structure is, and by how much the whole outruns the sum of its parts. Instead of asking whether we can prove we're not simulated, we can ask what, precisely, a first-person observer embedded inside a system could ever hope to determine about that system's ultimate nature — and recognize that the honest answer, "less than we'd like, and perhaps nothing at all," is itself a substantive discovery about the structure of knowledge, not a failure to find one. That may be the deepest insight buried in all of this: not a solution to the riddle of life and the universe, but a rigorous, hard-won understanding of exactly why the riddle resists solving from where we necessarily stand to ask it.