I went in skeptical. 38 sources later, I'm more skeptical. Here's why the most popular thought experiment in tech doesn't survive contact with physics, logic, or Occam's Razor.
I'll be honest about my starting position: the simulation hypothesis always seemed freaking stupid and impossible to me.
Not in the sense that it's not interesting to think about over a beer. In the sense that people — smart people, people running billion-dollar companies — were treating it as a serious probability claim about reality. Elon Musk says there's a "one in billions" chance we're in base reality. Neil deGrasse Tyson initially gave it better than 50-50 odds. Philosopher David Chalmers takes it seriously as metaphysics.
So I did the work. I read the original papers. I watched the explainer videos. I tracked down what the actual physicists — the Nobel laureates, the cosmologists, the people who build particle accelerators — think about it. I read the philosophy. I read the Reddit threads. I read the critiques of the critiques.
38 sources. And I came out the other side more skeptical than when I went in.
But this isn't a dismissive rant. If you believe in the simulation hypothesis, I want you to see your best arguments presented fairly, and then watch them fail. Not because I'm smarter than Bostrom or Chalmers — I'm not — but because the weight of evidence, logic, and physics points overwhelmingly in one direction.
Let's start with the strongest version of the case.
Giving the hypothesis its best shot
Nick Bostrom's 2003 paper "Are You Living in a Computer Simulation?" is the foundation. And it's important to understand what it actually says, because most people get it wrong.
Bostrom does not argue that we're in a simulation. He argues that
(1) Civilizations almost always go extinct before reaching the capability to run detailed ancestor simulations.
(2) Advanced civilizations that reach that capability almost never choose to run them.
(3) We are almost certainly living in a simulation.
The logic goes like this: if even a tiny fraction of advanced civilizations run ancestor simulations, the number of simulated beings would vastly outnumber real ones. If you're a randomly selected conscious being, the odds are overwhelming that you're simulated.
Bostrom himself distributes credence roughly equally across all three propositions. He's not the zealot his followers make him out to be.
Here's the thing that makes this annoying to dismiss: the formal logic of the trilemma is actually sound. If you accept the premise that consciousness can arise from computation (substrate independence), the math follows. A single posthuman civilization running thousands of ancestor simulations would generate billions of simulated minds. The ratio of simulated to real beings would be absurd.
This is why smart people take it seriously. It's not because they think computers are magic. It's because the logical structure of the argument is tight — if you accept the load-bearing assumption.
Elon Musk's version is simpler and more visceral: "40 years ago we had Pong. Now we have photorealistic VR. If you assume any rate of improvement at all, games will eventually be indistinguishable from reality."
This is compelling because it feels empirically obvious. Technology does improve. Games do look better every year. The intuition is powerful even if, as we'll see, it commits a category error.
David Chalmers, arguably the most important living philosopher of mind, takes the hypothesis seriously. Neil deGrasse Tyson initially couldn't find a strong argument against it. These aren't cranks. They're people who understand that our intuitions about what's "obvious" have been wrong before — about the shape of the earth, the age of the universe, the nature of matter.
Let me be clear: these aren't stupid arguments. The trilemma is formally valid. The intuition about improving technology isn't baseless. The philosophical humility about the limits of perception is well-founded.
They just don't survive scrutiny.
What Nobel laureates and working physicists actually say
Frank Wilczek is a Nobel laureate in physics. On Sean Carroll's Mindscape podcast, he raised what I think is the single most devastating empirical argument against the simulation hypothesis.
His point: the laws of physics have hidden complexity that isn't used for anything.
The laws of physics are local. They don't change over time. They don't change from place to place. They're constrained by symmetries that have no functional purpose in a simulation. A competent programmer wouldn't build unnecessary hidden structure.
If you were programming a universe, you'd make the rules as simple as possible while producing the desired output. You wouldn't embed deep mathematical symmetries that nobody inside the simulation would ever need to discover. You wouldn't make the laws time-invariant and location-invariant — that's an unnecessary constraint that makes the code harder to write.
The hidden complexity in physics is the fingerprint of something that wasn't designed.
Wilczek also raises the "embarrassing question" that simulation proponents can never answer: what is the thing in which the simulation is running made of? What are its laws? You need a reality to simulate a reality. It's turtles all the way down.
Sabine Hossenfelder is a theoretical physicist who originally called the simulation hypothesis "pseudoscience and religion." Her core objection is mathematical: quantum mechanics cannot be efficiently simulated on classical computers.
The numbers aren't close. Simulating the quantum state of just 300 particles requires more classical bits than there are atoms in the observable universe. That's 2300 versus roughly 1080. The entire observable universe contains about 1080 atoms, and you'd need more computational elements than that to simulate a few hundred particles at quantum fidelity.
This isn't a technology problem that better engineering will solve. It's a mathematical impossibility at equivalent-scale physics.
Now, in a 2026 update, Hossenfelder notably softened her position to "5 out of 10 on the bullshit meter" — acknowledging that the parent reality could have fundamentally different computational physics. But even she frames this as "I have no idea," not "it's plausible." And she still calls it unfalsifiable and unscientific.
Cosmologist Sean Carroll attacks the argument's own premises. The simulation hypothesis relies on a typicality assumption — we should assume we're typical observers. But if we're typical, and we can't perform full universe simulations, then we can't conclude that other civilizations probably can.
The typicality assumption contradicts itself. If humans are the baseline "typical" observer, and typical observers can't simulate universes, then the argument that most observers are simulated falls apart at the foundation.
"The simulation hypothesis is totally impracticable from a technical viewpoint. It's a late-night pub discussion, not a viable theory."
George F. R. Ellis, cosmologist
Ellis doesn't even bother with the philosophy. The computational resources required to simulate even a small portion of quantum reality are beyond anything physically achievable. The gap between "philosophically interesting" and "technically feasible" is enormous.
When the argument's own math turns against it
This might be the most elegant objection in the entire discourse, and it comes from the most unlikely place: Reddit.
If Bostrom's statistical reasoning works for our reality, it works for every reality above us too. Each simulated universe is probably itself simulated. This creates an infinite chain of nested simulations, each dependent on the one above it.
Now here's the kill shot: what's the probability that at least one link in an infinite chain fails?
It's essentially 1.
A supercomputer gets turned off. A civilization goes extinct. A power supply fails. In an infinite chain, something breaks. And when any single link breaks, every simulation below it vanishes.
The more probable you think simulation is, the more probable an infinite chain becomes, and the more probable our imminent nonexistence becomes. Our continued existence is itself evidence against the hypothesis.
Brandon Carter's 1983 Doomsday Argument uses identical anthropic reasoning to the simulation hypothesis, but applies it to human extinction. If humanity survives for billions of years, then being born in the first few hundred thousand years is vanishingly unlikely. Therefore, humanity probably ends soon.
Same logic. Same structure. Absurd conclusion.
Think of it as two bags of numbered balls — one with 10 balls, one with a million. You draw ball #7. Which bag did you probably draw from? The small one. Apply that to human history, and being alive now means humanity is a "small bag" species.
If the same reasoning style proves both "we're simulated" and "humanity ends soon," the reasoning style itself is suspect. A method that generates absurd conclusions when applied elsewhere probably isn't generating reliable conclusions when applied to simulation.
The simulation hypothesis requires at least five major assumptions:
(1) Simulated universes must be possible.
(2) Other intelligent life must exist.
(3) They must develop computers powerful enough to simulate consciousness.
(4) Those computers must actually produce conscious experience (substrate independence).
(5) They must choose to run ancestor simulations.
Standard cosmology requires fewer assumptions. And simulation theory doesn't even solve the hard problems of creation — it just moves them one level up. "Who created the simulators?" is functionally identical to "Who created God?"
A child's question defeats it. And the theory has no answer.
If the simulation hypothesis is correct, then simulated civilizations should themselves create simulations — the theory predicts this. But nested simulations create exponentially growing computational demands.
We've already built AI and VR. We're already pushing computational boundaries. If we were simulated, our increasing computational complexity should have crashed the parent system by now.
Our continued existence, with growing technological capability, is harder to explain under the simulation hypothesis than under standard physics.
Quantum misinterpretations, circular reasoning, and the Mandela Effect
Melvin Vopson (University of Portsmouth) claims to have empirical evidence for the simulation hypothesis. His argument: information entropy in biological systems decreases over time rather than increasing, which he says implies built-in "data optimization and compression" — exactly what a simulation would need.
The problems are substantial.
He published in AIP Advances — a mega journal that checks basic rigor but not theoretical validity. He self-published his book through his own physics institute. Popular Mechanics published a devastating critique calling his methodology circular reasoning: he defined his own terms, measured by those terms, then concluded his theory was supported.
There's also the self-defeating meta-problem: a civilization capable of simulating an entire universe would not be so sloppy as to leave a detectable fingerprint in data handling. If the simulation is competent enough to fool us about everything else, it wouldn't leave footprints in information entropy.
Vopson has also appeared on Bible-simulation podcasts, blending physics with religious apologetics. The company his ideas keep should tell you something.
This is the single most-cited "evidence" in pro-simulation videos. The interpretation: particles behave as waves when unobserved but "collapse" to definite positions when measured. The simulation spin: reality only renders details when someone is looking, like a video game.
This is a fundamental misunderstanding of quantum mechanics.
The "observer" in quantum mechanics is any interaction with another system — not conscious observation. A detector counts. A stray photon counts. A wall counts. The double-slit experiment demonstrates wave-particle duality and decoherence. It has perfectly good explanations within standard physics.
Confusing "resembles a computer" with "is a computer" is a category error. Water flowing through pipes resembles electricity flowing through circuits. That doesn't mean your plumbing runs on electrons.
Sylvester James Gates, a respected theoretical physicist, discovered what appear to be error-correcting codes embedded in the equations of supersymmetric string theory — specifically, the same type of codes used in web browsers.
This is genuinely interesting physics. And Gates himself is careful to say it could be mathematical coincidence or reflect deep mathematical structure. He does not claim it proves simulation.
Mathematical structures recur across unrelated domains constantly. Finding computer-science-like patterns in physics equations may reflect universal mathematics, not a simulation. The Fibonacci sequence appears in sunflowers, galaxies, and financial markets — that doesn't mean sunflowers are running stock trading algorithms.
Yes, people actually argue this. The Mandela Effect — large groups misremembering the same thing (Berenstain vs. Berenstein Bears, Nelson Mandela dying in prison) — is cited as evidence of "simulation patches" or "timeline edits."
This requires believing that a civilization sophisticated enough to simulate an entire universe is simultaneously sloppy enough to leave memory inconsistencies when editing its code.
This is the weakest argument in the entire discourse. Human memory is demonstrably unreliable. Confabulation, social reinforcement, and schema-based memory explain every Mandela Effect without invoking simulation theory. We don't need a glitch in the Matrix to explain why people misremember the spelling of a children's book series.
The universe's fundamental constants do appear "fine-tuned" for life. Change the strong nuclear force by 1% and atoms don't form. This is genuinely mysterious.
But fine-tuning has at least four explanations: coincidence, multiverse, God, or simulation. The simulation hypothesis is just one option, and it's structurally identical to the teleological argument for God's existence that philosophers have been debating for centuries.
The anthropic principle offers a simpler answer: we can only observe a universe compatible with our existence. The "tuning" may not be remarkable — it's selection bias.
Same psychological needs, different vocabulary
Philosopher Duncan Clarke makes the argument that should end the simulation hypothesis's pretensions to scientific status: it is functionally identical to religious creation myths.
The mapping is exact:
God = simulator. Creation = simulation. Heaven = base reality. Afterlife = logging out. Prayer = hoping the simulators are watching. Unfalsifiability = unfalsifiability.
Both posit an intelligent creator. Both describe a designed reality. Both invoke a higher plane of existence. Both offer the possibility of transcendence beyond death. Both claim moral purpose (the creator watches your behavior). And both are completely unfalsifiable.
"Isn't this just the creation philosophy used in most religions, just on Windows?"
This single Reddit comment might be the most efficient demolition of the simulation hypothesis ever written. It captures in one sentence what takes Duncan Clarke an entire essay to argue: the simulation hypothesis maps 1:1 onto religious creation, substituting "programmer" for "God" and "parent reality" for "divine realm."
The cultural critique matters too. As Popular Mechanics noted, the simulation hypothesis is "being amplified by extremely wealthy people who have financial incentive to make people feel like they aren't real" — devaluing human experiences, institutions, and rights. If nothing is real, why worry about labor conditions, or inequality, or the environment?
The hypothesis appeals to the same psychological needs as religion: fear of death (maybe death is just logging out), desire for meaning (someone designed all this), and hope for transcendence (there's a higher reality beyond this one). It gives tech-literate atheists every comfort of faith while letting them feel superior to people in churches.
Clarke calls it "a faith-based belief system with parallels to religion." That's generous. I'd call it creationism in a lab coat.
Same question, different vocabulary, still no answer
The simulation hypothesis feels cutting-edge. It isn't. It's the latest version of a question humans have been asking since at least the 4th century BC.
Zhuangzi's Butterfly Dream (4th century BC): Am I a man dreaming of being a butterfly, or a butterfly dreaming of being a man?
Plato's Cave (4th century BC): Chained prisoners see shadows on a wall and mistake them for reality.
Descartes' Evil Demon (1641): What if a malicious entity feeds you entirely false perceptions?
Aztec philosophy: The world is a painting by the gods, not the ultimate reality.
Gnosticism: The material world was created by a flawed deity (the Demiurge) — we live in an imperfect creation, not the true divine realm.
The simulation hypothesis replaces "evil demon" with "computer." It replaces "Demiurge" with "programmer." It replaces "cave shadows" with "rendered pixels."
Every version boils down to the same question: how do you know your perceptions match reality?
And every version goes nowhere. 2,400 years of asking the same unanswerable question in different clothing. Bostrom formalized it elegantly. But formalization isn't progress. Putting ancient confusion into a mathematical framework doesn't resolve the confusion — it just makes it more precise.
The simulation hypothesis is the computational-era restatement of Cartesian skepticism. And Cartesian skepticism was the 17th-century restatement of Plato's Cave. And Plato's Cave was the Western restatement of Zhuangzi's Butterfly Dream. None of these versions were ever resolved. They are inherently unfalsifiable.
Intellectual honesty requires admitting what we don't know
If this analysis stopped at "the simulation hypothesis is wrong," it would be dishonest. There are things the skeptical side cannot prove, and pretending otherwise would make me as sloppy as the proponents.
Bostrom's formal logic IS valid if substrate independence is true. The trilemma's structure holds. If consciousness can arise from computation, and if any civilization ever runs ancestor simulations, the statistical argument has force. You can't just wave it away.
Nobody has proven consciousness is substrate-dependent. We don't know whether consciousness requires biological brains. This is the hard problem of consciousness — possibly the deepest unsolved problem in philosophy — and nobody is close to solving it. The entire simulation hypothesis hangs on this question, and the question is genuinely open.
Quantum mechanics IS genuinely weird. The simulation interpretation of quantum phenomena is wrong — but the mystery is real. Wave-particle duality, entanglement, the measurement problem — these are strange features of reality that we don't fully understand. The simulation explanation fails, but dismissing the weirdness itself would be intellectually dishonest.
AI IS making the concept more imaginable. LLMs demonstrate that computation can produce convincing "understanding." AI-generated images and video show that synthetic realities are possible at the visual level. This doesn't make simulation more probable — generating a convincing video is infinitely easier than simulating quantum physics — but it does make the premise feel less absurd than it did a decade ago.
Even Hossenfelder, one of the fiercest critics, moved from "pseudoscience" to "5 out of 10" after reconsidering that the parent reality might have different physics. When a prominent skeptic softens, intellectual honesty requires noting it.
Where 38 sources leave us
After 38 sources, the simulation hypothesis fails on physics (hidden complexity no programmer would build, quantum scaling that requires more compute than the universe contains), logic (infinite regress, cascading failure that makes our existence evidence against it, self-defeating typicality assumption), parsimony (five-plus assumptions where standard cosmology needs fewer, and "who created the simulators?" loops forever), evidence (circular reasoning from Vopson, misinterpreted quantum mechanics, error-correcting codes that reflect math not simulation), and cultural analysis (it maps 1:1 onto religious creation myths while claiming scientific status).
The one thing proponents have going for them is an unsolved problem — consciousness — and building an entire worldview on an unsolved problem is called faith, not science.
Tyson got it right when he reversed his position after hearing Gott's objection: "That changes my life." The common trait of all hypothetical simulated universes is the ability to produce simulated universes — and since ours can't do that, we're either in base reality or at the very end of a chain.
Even the behavioral implications are a dead end. Robin Hanson argues that if we're simulated, we should be "entertaining to avoid being turned off." Preston Greene says we shouldn't try to find out because discovering the truth might end the simulation. Both positions are unfalsifiable, unprovable, and indistinguishable from superstition.
The simulation hypothesis is a genuine intellectual achievement by Bostrom. The trilemma is elegant. The formal logic is tight. But elegance isn't truth, and tight logic with an unproven premise is just a beautiful argument for something that's almost certainly not the case.
The universe is strange enough without needing to be a computer program. Quantum mechanics is genuinely weird. Consciousness is genuinely mysterious. Fine-tuning is genuinely unexplained. These are real mysteries.
But mysteries aren't evidence. And the simulation hypothesis, for all its Silicon Valley polish, is just the latest way of saying: "I don't understand reality, so someone else must have made it."
People have been saying that for 2,400 years. They've never been right yet.
AI, predictions, science, the stuff everyone gets wrong. Sign up and I'll send you the interesting ones.
No spam. Unsubscribe anytime.