Simulation Hypothesis
English

Tests · could we tell from inside?

You could only detect a simulation if it cut a corner that physics doesn’t. That could be a grid where space should be smooth, detail that only appears when measured, or a computer too small for what it runs. No experiment has reported such a sign. Every measurement so far has set a limit instead. Each entry below says what was tested or argued, what came of it, which dial of the Sandbox it bears on, and what it doesn’t show.

Run, and found nothing

Measurements that could have shown a sign of a computed world. None did. Each one sets a limit instead. A careful simulator would also give a null result. So these limits rule out particular designs of simulation, not the hypothesis as a whole.

Fermilab Holometer search for Planck-scale correlated noise

Two 39-40 m power-recycled Michelson interferometers sat side by side, cross-correlated up to 25 MHz. They searched for correlated spacelike position noise, which Hogan's model of holographic/quantum-geometric spacetime predicts.

Result. The interferometers reached a sensitivity below one Planck time in strain power spectral density. They saw no correlated noise. The collaboration reports that Hogan’s original model is ruled out at high statistical significance. The final analysis also limits shear-noise variants.

What it does not show. It rules out specific noise-correlation models of quantum geometry. It doesn't rule out discreteness or simulation in general. Hogan noted such noise could be too small to detect, or could show up in setups nobody has probed.

Bears on: Matter floor · Holometer Collaboration: Chou, A. S. et al. (2016). First measurements of high frequency cross-spectra from a pair of large Michelson interferometers. Phys. Rev. Lett. 117, 111102. arXiv:1512.01216.; Holometer Collaboration: Chou, A. et al. (2017). Interferometric constraints on quantum geometrical shear noise correlations. Class. Quantum Grav. 34, 165005. arXiv:1703.08503.; Fermilab News (3 Dec 2015). Holometer rules out first theory of space-time correlations.

Photon time-of-flight limits on Lorentz violation: Fermi GRB 090510

Compares arrival times of photons up to about 31 GeV from a short, distant gamma-ray burst. The test is for a speed of light that depends on energy, which some discrete or quantum-gravity spacetimes predict.

Result. They found no delay that depends on energy. For a linear dependence, the quantum-gravity mass is limited to M_QG > 1.2 E_Planck (Abdo et al. 2009, Nature 462, 331). A later analysis of the same burst limited a random 'spacetime fuzziness' model to a scale above 2.8 E_Planck (95% confidence) and 1.6 E_Planck (99%).

What it does not show. It only applies to models that predict a photon speed that depends on energy. A discrete spacetime that keeps Lorentz symmetry on average, or whose effects are quadratic or smaller, isn't ruled out by the linear bound.

Bears on: Matter floor · Abdo, A. A. et al. (Fermi LAT and GBM Collaborations) (2009). A limit on the variation of the speed of light arising from quantum gravity effects. Nature 462, 331-334.; Vasileiou, V., Granot, J., Piran, T. & Amelino-Camelia, G. (2015). A Planck-scale limit on spacetime fuzziness and stochastic Lorentz invariance violation. Nature Physics 11, 344-346.

Lorentz-violation limits from GRB 221009A (LHAASO data)

Time-of-flight analysis of TeV photons from the brightest gamma-ray burst ever observed. It uses public LHAASO data in the 0.2-7 TeV range.

Result. Piran & Ofengeim: for a linear (n=1) energy dependence, the lower limits on the LIV scale are 5.9 (subluminal) and 6.2 (superluminal) Planck energies. For quadratic (n=2), they are 5.8 and 4.6 × 10-8 E_Planck, the best time-of-flight limits of that order. Yang, Bi & Yin report different values, because their method differs. For example, n=1 subluminal gives E_QG > 14.7 × 1019 GeV, using the LHAASO afterglow.

What it does not show. The limits depend on the analysis method and on assumptions about how the source emits. They limit how photons travel at different energies. They don't tell us whether spacetime is discrete.

Bears on: Matter floor · Piran, T. & Ofengeim, D. D. (2024). Lorentz invariance violation limits from GRB 221009A. Phys. Rev. D 109, L081501. arXiv:2308.03031.; Yang, Y.-M., Bi, X.-J. & Yin, P.-F. (2024). Constraints on Lorentz invariance violation from the LHAASO observation of GRB 221009A. JCAP 04 (2024) 060. arXiv:2312.09079.

Bounds from physics and computation

Arguments from known physics or the theory of computation about what it would take to simulate this universe.

Beane, Davoudi & Savage: lattice spacing from cosmic rays

Assumes the universe is a numerical simulation on a cubic space-time lattice (unimproved Wilson fermions, like early lattice QCD). Then works out what we could observe, including broken rotational symmetry in the highest-energy cosmic rays.

Result. Their strongest bound is an inverse lattice spacing above about 1011 GeV. That's a spacing below about 2 × 10-27 m. It comes from the high-energy cutoff of the cosmic-ray spectrum. A lattice would also make the highest-energy cosmic rays arrive along its axes more often. Nobody has reported that.

What it does not show. It bounds one specific model, a cubic lattice with Wilson-type discretisation. It doesn't rule out other discretisations, finer lattices or non-lattice methods. And it reports no detection.

Bears on: Matter floor · Beane, S. R., Davoudi, Z. & Savage, M. J. (2014). Constraints on the Universe as a numerical simulation. Eur. Phys. J. A 50, 148. arXiv:1210.1847.

Ringel & Kovrizhin: gravitational responses and the sign problem

The authors show that quantised gravitational responses (such as thermal Hall conductance in fractional quantum Hall phases) block local sign-free quantum Monte Carlo for certain bosonic systems. That suggests some quantum systems resist efficient classical Monte Carlo simulation.

Result. The result is about sign-free QMC representations of specific condensed-matter phases. The paper isn't about the simulation hypothesis. Popular coverage claimed it showed we're not in a simulation. Commentators rejected that.

What it does not show. It doesn't show the universe can't be simulated. It covers one class of classical algorithm. It isn't a formal complexity-theory hardness proof. And a simulator built on a quantum computer is unaffected (Aaronson, 2017).

Bears on: Matter floor, quantum rungs · Ringel, Z. & Kovrizhin, D. L. (2017). Quantized gravitational responses, the sign problem, and quantum complexity. Science Advances 3(9), e1701758. (arXiv:1704.03880, earlier title 'Quantized gravitational responses and the sign problem'); Aaronson, S. (3 Oct 2017). Because you asked: the Simulation Hypothesis has not been falsified; remains unfalsifiable. Shtetl-Optimized.

Vazza: astrophysical energy and information budget

Uses the holographic bound and Landauer's principle to estimate the energy needed to encode the state of the visible Universe, of Earth at Planck resolution, and of Earth at a coarser resolution set by high-energy neutrino observations.

Result. Encoding the visible Universe would need about 3.5 × 10124 bits and 8.9 × 10101 J. The observable Universe holds about 2.7 × 1071 J. Earth at Planck resolution needs about 3.0 × 1052 J. Earth at the coarsest resolution that high-energy neutrino observations allow (about 10-23 m) needs about 4.3 × 1028 J. That energy exists in principle, but not at the rate a running simulation needs. Conclusion: a universe with our physics can't simulate this one. Figures are from the preprint (arXiv v1).

What it does not show. It doesn't cover simulators with different physics, or simulations that only render what observers need. A 2026 Frontiers commentary (Edge & Brown) argues the paper targets a stronger, shared-world version than Bostrom's claim.

Bears on: What a computer can run · Vazza, F. (2025). Astrophysical constraints on the simulation hypothesis for this Universe: why it is (nearly) impossible that we live in a simulation. Frontiers in Physics 13, 1561873. arXiv:2504.08461.; Edge, E. & Brown, C. A. (2026). Commentary: Astrophysical constraints on the simulation hypothesis for this Universe. Frontiers in Physics (article 1808725), 21 April 2026.

Wolpert: computer science theory and the simulation hypothesis

Links computer science theory to physics through the physical Church-Turing thesis. Defines simulation formally. Then uses Kleene's second recursion theorem and Rice's theorem to get results on simulation and self-simulation.

Result. Proves that, under his assumptions, we could be in a simulation that we ourselves run on a computer. That's self-simulation: two identical, equally real instances. Also derives impossibility results from Rice's theorem, and looks at homomorphic encryption and nested-universe graphs.

What it does not show. It's a formal analysis that depends on the physical Church-Turing thesis. It doesn't say whether our universe is a simulation. Its results hold for any such pair of universes, not only this one.

Bears on: Infinite regress · Wolpert, D. H. (2025). What computer science has to say about the simulation hypothesis. J. Phys. Complexity 6(4), 045010. (arXiv:2404.16050, 'Implications of computer science theory for the simulation hypothesis')

Proposed, not yet run

Experiments that have been published but not carried out.

Campbell et al.: proposed rendering-on-observation tests

Assumes a simulator with limited resources only renders something when the information reaches an observer, not when a machine detects it. Proposes double-slit and delayed-choice experiments, in concept only, where the result would differ if the system were saving effort this way.

Result. The paper describes the experiments in concept and reports no data. We found no published run of these designs.

What it does not show. Standard quantum mechanics already predicts the outcomes of such experiments without a simulator. So a null result wouldn't tell the two apart cleanly. And the proposal rests on a rendering rule it assumes, not one it derives.

Bears on: Rendering · Campbell, T., Owhadi, H., Sauvageau, J. & Watkinson, D. (2017). On testing the simulation theory. Int. J. Quantum Foundations 3(3), 78-99. arXiv:1703.00058.; IJQF article page

Claimed, and contested

Published claims for or against the hypothesis that the field hasn’t accepted.

Vopson: second law of infodynamics

Proposes that the information entropy of systems that hold information states stays constant or falls over time. A 2023 follow-up argues this fits a simulated universe. It points to digital data, genetics, atomic physics and cosmology.

Result. Original paper: Vopson & Lepadatu (2022). The simulation argument: Vopson (2023). A 2025 paper derives gravity from the same idea. It drew public criticism of its mathematics and its use of entropy, and the author answered. Nobody has published an independent experimental confirmation.

What it does not show. Even if the proposed law held, it wouldn't imply a simulator. Other explanations for falling information entropy remain open. And the law itself is disputed.

Bears on no dial directly · Vopson, M. M. & Lepadatu, S. (2022). Second law of information dynamics. AIP Advances 12, 075310.; Vopson, M. M. (2023). The second law of infodynamics and its implications for the simulated universe hypothesis. AIP Advances 13, 105308.; Vopson, M. M. (2025). Is gravity evidence of a computational universe? AIP Advances 15, 045035.

Faizal et al.: undecidability and 'cannot be simulated'

Argues from Gödel, Tarski and Chaitin that a fully algorithmic theory of everything is impossible. So some parts of reality need non-algorithmic understanding. A simulation would be algorithmic, so the universe can't be one.

Result. Published in J. Holography Applications in Physics 5(2), 10-21 (2025). A comment by E. Redden (arXiv:2512.11807) uses Conway's Game of Life to argue that undecidability limits what can be proved, not what can be run. It also says the conclusion isn't justified without evidence of hypercomputation in nature. A second JHAP letter (Khan, 2025) supports the argument and reinterprets it.

What it does not show. It doesn't establish that physical processes need a non-algorithmic way to settle undecidable statements. A no-simulation conclusion needs that premise.

Bears on: Matter floor, quantum rungs · Faizal, M., Krauss, L. M., Shabir, A. & Marino, F. (2025). Consequences of undecidability in physics on the theory of everything. J. Holography Applications in Physics 5(2), 10-21. arXiv:2507.22950.; Redden, E. (2025). Provability vs. execution: a comment on 'Consequences of undecidability in physics on the theory of everything'. arXiv:2512.11807.; Khan, A. K. (2025). Discussion on the Faizal-Krauss-Shabir-Marino argument about the theory of everything. J. Holography Applications in Physics 6(1).

Compiled 4 October 2026. We checked the citations against the published papers.