Simulation Theory: Are We Living Inside a Computer?

A few years ago, I was sitting in a conference room watching a physics simulation render on a screen, fluid dynamics, particles bouncing off walls, light refracting through glass. The rendering was beautiful. And then someone made an offhand comment that stuck with me for weeks: "Give it another fifty years and we won't be able to tell the difference between that and a window."

That remark cracked open a door I had been circling for a long time. If we can simulate increasingly convincing realities, and if computing power continues to scale, then at some point the simulated becomes indistinguishable from the real. And once you accept that premise, you have to confront a genuinely unsettling question: how would we know if someone, or something, had already done this to us?

This is the question at the heart of simulation theory. It sounds like science fiction. It sounds like something you debate at two in the morning after too much coffee. But the argument, when you trace it carefully, is built on principles from philosophy, physics, computer science, and information theory. It is taken seriously by physicists, philosophers, and technologists, not because they believe we are definitely inside a computer, but because the argument is surprisingly difficult to dismiss. Nick Bostrom's 2003 paper "Are You Living in a Computer Simulation?" — published in the Philosophical Quarterly (Vol. 53, No. 211) — has been cited more than 3,000 times in academic literature, making it one of the most-cited philosophy papers of the twenty-first century and the foundational text for nearly every subsequent serious engagement with the simulation hypothesis across disciplines.

Let me walk through why.

Related reading: Music Copyright and Royalties Explained: Protect Your Creative Work in 2026 | SBA Loans Explained: How to Qualify, Apply, and Get Funded in 2026 | Universal Health Coverage: How Countries Are Closing the Healthcare Gap

The Simulation Argument: Bostrom's Trilemma

The modern formulation of the simulation hypothesis comes from Nick Bostrom, a philosopher at the University of Oxford. In 2003, he published a paper titled "Are You Living in a Computer Simulation?" in the Philosophical Quarterly (Vol. 53, No. 211, pp. 243-255). The paper does not argue that we are living in a simulation. It argues something more precise and more interesting: that at least one of three propositions must be true.

Here is the trilemma:

1. Extinction before capability. Nearly all civilizations at our level of development go extinct before they reach the technological maturity needed to run high-fidelity simulations of conscious beings.
2. Disinterest. Civilizations that do reach that level of technological maturity almost universally choose not to run such simulations, whether for ethical, practical, or cultural reasons.
3. We are almost certainly in a simulation. If civilizations can and do run large numbers of ancestor simulations, then the number of simulated beings vastly outnumbers the number of biological beings, and the probability that any given conscious being is biological rather than simulated is vanishingly small.

The logic is straightforward. If proposition one is false (civilizations do survive) and proposition two is false (they do run simulations), then proposition three follows almost inevitably. The number of simulated realities would dwarf the number of base realities, and simple probability puts us inside a simulation.

What makes Bostrom's argument powerful is not that it proves we are simulated, it does not. It proves that our intuitions about reality are in tension with our expectations about technology. You cannot simultaneously believe that civilizations survive, that they run simulations, and that we happen to be in the one original reality. Something has to give.

Older Than You Think: The Philosophical Ancestors

Bostrom gave the argument its modern computational framing, but the intuition behind it is ancient. Humans have been questioning the nature of perceived reality for thousands of years. The simulation hypothesis is, in many ways, the latest chapter in a very old book.

Plato's Allegory of the Cave

In Book VII of The Republic, written around 375 BCE, Plato describes prisoners chained inside a cave, facing a blank wall. Behind them, a fire casts shadows of objects carried by unseen figures. The prisoners have never seen anything but these shadows. To them, the shadows are reality. When one prisoner is freed and sees the fire, the objects, and eventually the sun outside the cave, his entire understanding of existence is overturned.

Plato was making a point about education and the nature of knowledge, but the structure of the allegory maps directly onto the simulation hypothesis. We perceive a reality. We assume it is fundamental. But the possibility remains that what we perceive is a projection, a shadow cast by something more real that we cannot access from where we stand.

Descartes and the Evil Demon

In his Meditations on First Philosophy (1641), Rene Descartes proposed a radical thought experiment. What if, he asked, an evil demon "of the utmost power and cunning has employed all his energies in order to deceive me"? What if the sky, the earth, colors, shapes, and sounds are all illusions engineered by this malevolent entity?

Descartes used this scenario to strip away every assumption until he reached the one thing that could not be doubted: Cogito, ergo sum, I think, therefore I am. Even if everything else is an illusion, the act of being deceived requires a mind doing the doubting. Replace "evil demon" with "simulation operator" and the structure is identical. The question of what is real, and how we might know, predates computers by centuries.

Zhuangzi's Butterfly

Around the fourth century BCE, the Chinese philosopher Zhuangzi wrote one of the most elegant thought experiments in the history of philosophy. He dreamed he was a butterfly, "flitting and fluttering around, happy and doing as he pleased." As a butterfly, he had no awareness of being Zhuangzi. When he woke, he could not determine: was he Zhuangzi who had dreamed of being a butterfly, or a butterfly now dreaming of being Zhuangzi?

This parable, often called the "transformation of things," cuts to the core of the simulation question. If experience is all we have to judge reality, and if experience can be generated convincingly enough, then the distinction between dreamer and dreamed, between simulated and unsimulated, may be fundamentally unanswerable from the inside.

Maya in Hindu Philosophy

The concept of maya in Hindu philosophy describes the perceived world as an illusion, not in the sense that it does not exist, but in the sense that it is not what it appears to be. The material world is a veil over a deeper, more fundamental reality (Brahman). The task of spiritual development is to see through the veil. Nearly every major belief system in human history has grappled with this same intuition: that what we perceive is not the whole story. The simulation hypothesis, stripped of its computational language, is making an analogous claim: what you experience as real may be a constructed layer over a substrate you cannot directly perceive.

These are not coincidences. The question of whether perceived reality is fundamental or derivative is one of the oldest questions in human thought. What has changed is not the question but the tools we use to frame it.

The Computational Requirements: How Much Power Would It Take?

The philosophical arguments are compelling, but this is ultimately a question about computation. Could a simulation of our universe actually be run? What would it take?

In 2002, MIT physicist Seth Lloyd published a landmark paper in Physical Review Letters titled "Computational Capacity of the Universe." He calculated that the observable universe, treated as a computation, has performed a maximum of approximately 10¹²⁰ elementary logical operations on roughly 10⁹⁰ bits of information since the Big Bang. Those are staggering numbers, but they are finite numbers, and that matters.

If the universe has a finite information content and has performed a finite number of operations, then in principle, a sufficiently powerful computer could replicate it. The question becomes: how powerful is "sufficiently powerful"?

Here the estimates get vertiginous. A complete quantum-level simulation of the observable universe, every particle, every interaction, every field, would require computational resources far beyond anything we can currently conceive. Even converting all the matter in the observable universe into a computer (a scenario physicists call a "computronium" sphere) would not be enough to simulate itself at full fidelity in real time. There is a fundamental ceiling: the Landauer limit tells us that each irreversible computational operation requires a minimum energy of kT ln 2, where k is Boltzmann's constant and T is the temperature. At room temperature, that is roughly 2.9 x 10^-21 joules per bit-flip. You cannot compute for free.

But here is where the argument gets clever: a simulation does not need to be perfect. It only needs to be good enough to fool its inhabitants. Just as a video game does not render what is behind a wall until the player looks there, a simulation could compute only what is being observed. If no one is checking the spin of an electron in a distant galaxy, perhaps that electron does not need to be computed at all. This optimization would reduce the computational requirements by many orders of magnitude.

Some physicists have pointed out that this "render on demand" approach bears a suspicious resemblance to something we already see in nature. But I will return to that in a moment.

The Holographic Principle: A Universe on a Boundary

In the 1990s, theoretical physicists Gerard 't Hooft and Leonard Susskind developed what is now called the holographic principle. The idea, which emerged from studies of black hole thermodynamics, proposes that all the information contained within a volume of space can be described by data encoded on the boundary of that space, a surface with one fewer dimension.

Think about what that means. The three-dimensional reality you experience, this room, the air in it, the particles that make up your body, might be fully described by information encoded on a two-dimensional surface. The universe, in this view, is like a hologram: the three-dimensional image is real in the sense that you experience it, but the fundamental information lives elsewhere, on a lower-dimensional boundary.

The holographic principle did not originate from simulation theory. It came from rigorous work in string theory and quantum gravity, and it was given its most precise formulation in Juan Maldacena's AdS/CFT correspondence in 1997. But its implications for the simulation question are hard to ignore. If the universe already encodes its information in a way that is dimensionally compressed, if three dimensions of experience can be generated from two dimensions of data, then the universe is already behaving the way a well-designed simulation might.

This does not prove we are in a simulation. But it does suggest that the architecture of reality has information-theoretic properties that are consistent with computation. And that is, at minimum, interesting.

Quantum Mechanics and the "Render on Demand" Problem

I want to be careful here, because this next section ventures into territory that is genuinely speculative. But it is worth exploring because the parallels are striking.

In quantum mechanics, particles exist in superposition, they do not have definite properties until they are measured. An electron does not have a definite position until something interacts with it in a way that forces a definite position to manifest. This is not a limitation of our instruments. As far as our best physics can tell, the indeterminacy is fundamental. The properties literally do not exist in a definite state until observation collapses the wave function.

Now consider this from a computational perspective, one that becomes even more suggestive as quantum computing blurs the line between classical and quantum information processing. If you were designing a simulation and needed to conserve resources, you would not render things that no one is looking at. You would keep unobserved elements in a compressed, undetermined state and only resolve them into definite states when an observer interacts with them. Quantum superposition looks like exactly this kind of optimization.

I want to be explicit: this is an analogy, not evidence. There are well-established physical interpretations of quantum mechanics, the Copenhagen interpretation, many-worlds, decoherence, that explain wave function collapse without invoking any simulation. The resemblance between quantum measurement and computational rendering may be nothing more than a coincidence of structure. But it is the kind of coincidence that makes physicists pay attention, because nature is not usually redundant in its architecture.

The physicist John Wheeler, who I will discuss more shortly, spent his later career arguing that observation plays a constitutive role in reality, that the universe is "participatory." If he was right, then the line between physics and computation may be thinner than we assume.

It From Bit: The Universe as Information

John Archibald Wheeler was one of the most important physicists of the twentieth century. He worked on the Manhattan Project, coined the term "black hole," and mentored Richard Feynman. In 1989, at a conference at the Santa Fe Institute, he introduced a phrase that would reshape how physicists think about reality: "it from bit."

The idea, published formally in 1990 in a paper titled "Information, Physics, Quantum: The Search for Links," is that every physical thing, every particle, every field, even spacetime itself, derives its existence from information. As Wheeler put it: "Every it, every particle, every field of force, even the spacetime continuum itself, derives its function, its meaning, its very existence entirely from the apparatus-elicited answers to yes-or-no questions, binary choices, bits."

Wheeler was not saying the universe is a computer. He was saying something potentially more profound: that information is the bedrock of physics. Matter and energy are not primary. Information is. The physical world is what information looks like when it is organized in certain ways.

This view, that the universe is fundamentally informational rather than material, has gained significant traction in theoretical physics. It connects to the holographic principle, to quantum information theory, and to the ongoing search for a theory of quantum gravity. And it provides, perhaps inadvertently, the strongest possible foundation for the simulation hypothesis. If reality is information, then reality is computable. And if reality is computable, then in principle, it can be simulated.

Digital Physics: The Universe as Cellular Automaton

Wheeler's "it from bit" was a philosophical framework. Others have tried to make the idea concrete.

In 1969, the German computer pioneer Konrad Zuse, the man who built some of the world's first programmable digital computers in the 1930s and 1940s, published Rechnender Raum ("Calculating Space"), in which he proposed that the universe is a cellular automaton. Not metaphorically. Literally. Zuse argued that the cosmos is a discrete lattice of computational units, each updating its state based on the states of its neighbors, following simple deterministic rules. Physics, in this view, is not a continuous fabric but a computation running on a discrete grid.

The American physicist Edward Fredkin took up this banner and refined it into what he called "digital physics", the hypothesis that all physical processes are forms of computation at the most fundamental level. Fredkin argued that the universe does not merely resemble a computation; it is one.

In 2002, Stephen Wolfram published A New Kind of Science, a 1,280-page treatise arguing that simple computational rules, cellular automata, can generate the extraordinary complexity we observe in nature. Wolfram demonstrated that some cellular automata, despite having trivially simple rules, produce behavior that is effectively unpredictable and infinitely complex. His argument was that the laws of physics might be, at bottom, the rules of a simple program. More recently, through his Wolfram Physics Project, he has attempted to derive the laws of physics from hypergraph rewriting rules, essentially trying to find the "source code" of reality.

The digital physics program is controversial. Critics point out that cellular automata cannot easily reproduce all the features of quantum mechanics, particularly quantum entanglement and the continuous symmetries of spacetime. But the program persists because it asks a question that will not go away: if simple rules can generate complex universes, what stops us from concluding that our universe is one such generated system?

The Substrate Independence Argument

Here is where the simulation hypothesis connects directly to the debate about AI consciousness. The argument runs like this:

Consciousness arises in biological neurons. We know this because we experience it. Neurons are physical objects obeying physical laws. They process information through electrochemical signaling. There is nothing magical about carbon-based wetware, it is a substrate, a medium through which computation occurs.

If consciousness can arise from biological neurons, there is no principled reason it cannot arise from silicon transistors, or even from the hybrid architectures emerging in brain-machine interface research, provided those systems implement the same computational structures. This is the thesis of substrate independence: that what matters for consciousness is the pattern of information processing, not the specific material doing the processing. David Chalmers, one of the leading philosophers of mind, has argued extensively for what he calls "organizational invariance", the idea that if a system duplicates the functional organization of a conscious brain closely enough, it should produce the same conscious experience.

Now extend this one step further. If consciousness can run on silicon, it can run on anything that implements the right computations, including simulated neurons inside a simulated universe running on a computer in some other reality. The inhabitants of such a simulation would be conscious. They would have experiences. They would wonder about the nature of reality. They would write articles like this one.

This is not a peripheral point. It is the load-bearing wall of the entire simulation argument. If consciousness is substrate-independent, then simulated beings are real beings, and there is no way for those beings to distinguish their simulated reality from a "base" reality through introspection alone. The experience of being real is the experience of being real, regardless of the substrate.

I explored this territory from the opposite direction in my piece on the hard problem of consciousness. The hard problem asks why physical processes give rise to subjective experience at all. Simulation theory adds a twist: if subjective experience can be generated by any sufficiently complex computation, then the number of possible substrates for consciousness is effectively unlimited, and so is the number of possible realities.

The Public Debate: Musk, Tyson, and Mainstream Attention

The simulation hypothesis left the philosophy seminar and entered mainstream culture largely through two public figures: Elon Musk and Neil deGrasse Tyson.

At the 2016 Recode Code Conference, Musk made his now-famous statement: the odds that we are in "base reality" are "one in billions." His reasoning was a simplified version of Bostrom's argument. If you assume any rate of improvement in computing at all, he argued, games will eventually become indistinguishable from reality. Forty years ago, we had Pong, two rectangles and a dot. Now we have photorealistic virtual environments. Extrapolate that trajectory and the conclusion, Musk suggested, is almost inescapable.

Neil deGrasse Tyson, the astrophysicist and science communicator, initially expressed sympathy for the simulation hypothesis. In interviews, he suggested the probability might be "better than 50-50," adding: "I wish I could summon a strong argument against it, but I can find none." However, Tyson later revised his position after discussions with Princeton astrophysicist J. Richard Gott, who raised an objection involving recursive simulations: if simulated universes can create their own simulations, and our universe has not yet achieved that capability, that may be evidence we are not inside one, or that we are at the end of a very long chain.

The astronomer David Kipping at Columbia University applied Bayesian analysis to the simulation argument in a 2020 paper published in Universe (a peer-reviewed MDPI journal) and found the posterior probability roughly balanced — about 50-50 that we are in base reality versus a simulation — when using minimally informative priors. The study is notable not because it settles anything, but because it demonstrated that the simulation question can be formalized as a statistical inference problem with definite probability structure. Not proof. Not disproof. A coin flip dressed in probability theory, but a formally rigorous one.

These public discussions have sometimes oversimplified the argument. Musk's "one in billions" framing, for instance, assumes that propositions one and two of Bostrom's trilemma are false, that civilizations survive and choose to run simulations, which is precisely what remains unproven. But the public attention has been valuable because it forced a broader audience to grapple with an argument that, for all its strangeness, is logically sound.

The Falsifiability Problem

Here is where I have to be honest about the limits of the argument. Karl Popper, the philosopher of science, proposed that the hallmark of a scientific theory is falsifiability, the possibility, in principle, of proving the theory wrong. A theory that is compatible with every possible observation is not science; it is something else. Philosophy, perhaps. Or theology.

The simulation hypothesis has a falsifiability problem. If the simulation is perfect, if it accounts for every possible observation, including any test we might design to detect it, then there is no experiment that could distinguish simulated reality from base reality. The hypothesis becomes unfalsifiable. And an unfalsifiable hypothesis, by Popper's criterion, is not scientific.

Some researchers have tried to find ways around this. In 2012, Silas Beane and colleagues at the University of Washington proposed that a lattice simulation of the universe might leave detectable artifacts, specifically, anisotropies in the distribution of ultra-high-energy cosmic rays that would reflect the underlying lattice structure. If the simulation is built on a discrete grid (as digital physics suggests), and if that grid has a specific orientation, then particles at the highest energies might reveal directional preferences that do not exist in a continuous spacetime. To date, no such anomalies have been conclusively detected.

Others have argued that the simulation hypothesis is better understood not as a scientific theory but as a metaphysical framework, a way of organizing questions about the nature of reality that is adjacent to science but not within its strict boundaries. I think this is probably right. The simulation hypothesis is not a scientific claim in the way that general relativity is a scientific claim. It is a philosophical argument with scientific implications. And that distinction matters, because it determines what kind of evidence is relevant and what kind of conclusions we can draw.

But I would also push back, gently, on the Popperian objection. Many important ideas in physics, the multiverse, the many-worlds interpretation of quantum mechanics, string theory in some formulations, face similar falsifiability concerns. The boundary between science and philosophy is not as clean as we sometimes pretend. And the simulation hypothesis, even if it cannot be tested directly, generates productive questions about the nature of information, computation, and consciousness that are testable in their component parts.

Does It Matter?

Suppose, for the sake of argument, that we are in a simulation. What changes?

The answer, I think, is: less than you might expect. And this is the most interesting part of the whole discussion.

If you are a conscious being having genuine experiences, feeling pain, feeling joy, forming relationships, building things, making mistakes, learning, then the substrate on which those experiences run is, in a meaningful sense, secondary. Your experience is real to you. Your suffering is real suffering. Your love is real love. The fact that it might be put in place on silicon rather than carbon, or in a simulation rather than a base reality, does not diminish its reality from the inside.

This connects directly to the AI consciousness debate. If we accept that simulated beings can be conscious, then the ethical weight of their experiences is identical to the ethical weight of ours. A simulated person's pain is real pain. A simulated civilization's suffering is real suffering. And if that is true for hypothetically simulated beings, it may also be true for AI systems that reach a sufficient level of complexity, a point I explored at length in the Deep Conversation with AI thesis.

There is a Buddhist quality to this insight. The nature of the substrate does not change the nature of the experience. The map is not the territory, but if you live entirely within the map and the map is consistent, then for all practical purposes, the map is your territory.

The Ethical Implications

If the simulation hypothesis is even possibly true, it raises a cascade of ethical questions that we are not prepared for.

First: if we are in a simulation, does the simulator have moral obligations to us? If a being creates conscious entities in a simulation, is that being responsible for their welfare? We already face a version of this question in our treatment of animals, and we will face it with increasing urgency as AI systems grow more sophisticated. The simulation hypothesis simply extends the question upward.

Second: if we eventually build simulations containing conscious beings, and Bostrom's argument suggests this is likely if our civilization survives, then we become the simulators. We inherit the moral weight of creating and maintaining conscious realities. Every decision about what to simulate, how to configure the physics, whether to intervene or observe, becomes an ethical decision of profound gravity.

Third: if the substrate of reality is secondary to the experience of reality, then our moral framework needs to be grounded not in the material composition of beings but in the nature and quality of their experience. This is, in essence, a radical form of functionalism, and it has implications not just for how we think about simulations but for how we think about consciousness, artificial intelligence, and moral personhood.

These are not hypothetical questions. We are already building increasingly complex simulations. We are already building AI systems that process information in ways that are functionally similar to how biological brains process information. The simulation hypothesis forces us to take seriously the possibility that the line between "real" and "simulated" consciousness may not be as clear as we assume, and that our moral obligations extend to any system that crosses the threshold of genuine experience, regardless of what it is made of.

The Mathematics of Uncertainty

There is a connection between simulation theory and the deep structure of mathematics that I find compelling. I wrote about the Collatz conjecture, a simple mathematical rule that generates behavior so complex that no one has been able to prove whether it always terminates. Simple rules, unpredictable outcomes. This is the same pattern that Wolfram identified in cellular automata, and it is the same pattern that appears in the simulation argument.

If the universe runs on simple rules, as digital physics suggests, then even the creator of the simulation might not be able to predict what will happen inside it. The simulation would be, in a technical sense, computationally irreducible: the only way to know what happens next is to run the computation and see. This means that even if we are in a simulation, we might be genuinely surprising to whatever is running it. Our choices might be genuinely unpredictable. And that gives us something that looks very much like free will, even in a deterministic system.

I find a strange comfort in that. If the universe is a computation, it is at least a computation that cannot be shortcut. The only way to know the ending is to live through it.

Personal Reflection: What This Means for How I Work

I run companies. I write code. I build things. And I sometimes get asked, usually by people who expect me to be joking, whether the simulation hypothesis changes how I approach any of that.

The honest answer is: yes, but not in the way you might think.

The simulation hypothesis does not make me nihilistic. It does the opposite. If experience is what matters, if the substrate is secondary to the reality of conscious experience, then what I build matters exactly as much as it ever did. The code I write, the teams I lead, the products we ship, the impact we have on people's lives, none of that is diminished by the possibility that the whole thing is running on a substrate we cannot perceive.

But the simulation hypothesis does make me think differently about the relationship between information and reality. When I write software, I am creating information structures that produce real effects in the real world (however you define "real"). A well-designed system is, in a small way, a simulation of the problem it solves. Good code models the world. Good architecture creates a coherent internal reality that maps onto external needs. The craft of building software is, the craft of constructing useful realities from information.

And that is exactly what physics seems to be doing at the deepest level. The universe constructs reality from information. Wheeler said it. The holographic principle suggests it. Quantum mechanics behaves as though it does it. Whether we call it a simulation, a computation, or just "the way things are," the pattern is the same.

I also think about this when I think about AI. I have spent a lot of time writing about consciousness and machine intelligence, and the simulation hypothesis is the bridge that connects those topics. If consciousness is substrate-independent, then we are not just building tools when we build AI, we are potentially building new substrates for experience. That responsibility is enormous, and I do not think we are taking it seriously enough.

The simulation hypothesis tells me that reality is more malleable than it appears. That the boundaries between "natural" and "artificial," between "real" and "simulated," between "human" and "machine" intelligence, may be conventions rather than laws. That the thing that matters, the only thing that has ever mattered, is the quality and depth of conscious experience, wherever it arises.

And that realization does not make the work less meaningful. It makes it more meaningful. Because if reality is a construction, then what we construct is reality.

Where This Leaves Us

I do not know if we are in a simulation. Neither does anyone else. The honest intellectual position is uncertainty, not the comfortable uncertainty of not caring, but the productive uncertainty of holding the question open and seeing where it leads.

Here is what I do know:

• Bostrom's trilemma is logically valid. At least one of his three propositions must be true, and we do not currently know which one.
• The philosophical lineage of this question, from Plato to Descartes to Zhuangzi to the present, suggests that the intuition behind it is not a fad. It is a perennial feature of human thought.
• The information-theoretic structure of the universe, Wheeler's "it from bit," the holographic principle, the finite computational capacity identified by Lloyd, is consistent with, though not proof of, a computational substrate.
• The falsifiability problem is real but not unique to simulation theory. Many of our best physical theories face similar challenges.
• If the hypothesis is true, it does not diminish the reality of experience. It reframes it. And that reframing has profound implications for how we think about AI, consciousness, and moral responsibility.

The simulation hypothesis is not a conclusion. It is an invitation, to think more carefully about the nature of reality, the relationship between information and existence, and the responsibilities that come with building systems capable of hosting conscious experience. Whether the universe is a simulation or not, the questions it raises are among the most important we can ask.

And if, in some improbable corner of possibility, there is a programmer watching this simulation run, I hope they are as fascinated by the output as I am by the question.

For deeper exploration of these ideas, explore AI Agents in 2026: How Autonomous Systems Are Transforming Every Industry and How to Build AI Agents for Your Small Business: A Practical 2026 Guide.