# Adversarial review: cost-channel closure

Argus eighth night cycle, 2026-09-15. Task: break C1-C7, especially the workload model, entropy move, black-hole entropy count, Margolus-Levitin energy term, novelty, free-parameter language, and Bostrom-resource literature.

## Executive verdict

C1 survives only after demotion: the generic simulation hypothesis is not constrained by cost unless the simulator's rendering policy is specified. But this is not new; Bostrom stated the scope/granularity escape in the original argument, and recent resource-refutation exchanges revolve around exactly this point.

C2 is overbroad. H13 is not killed for the generic hypothesis, but it is false that "no cost argument constrains any host." Cost arguments constrain fully specified host-and-rendering models, including same-physics comprehensive simulations and lattice-like simulations.

C3 and C4 survive, with caveats. Margolus-Levitin and holographic bounds are substrate-independent quantum/resource bounds. The 0.315-age headline is algebraically empty.

C5, C6, and C7 need amendments. The main break is entropy: Egan & Lineweaver's cosmic entropy is not simply "the measured bits needed to specify our universe's microstate," and SMBH horizon entropy is not automatically a resource requirement for reproducing observer-visible history.

## O1 - FATAL: `ops = bits x update_rate x simulated_duration` is not a physical lower bound

What breaks: budget4.py treats the 40-dex swing as if both endpoints are physically defensible implementations of the same target. They are not. The formula is a synchronous cellular-automaton workload model. It assumes every stored bit must be actively updated at one global tick rate for the entire simulated duration.

What it gets wrong:

- It confuses state capacity with dynamical work. A physical or quantum analog simulator can evolve a high-dimensional state under a Hamiltonian without performing one discrete gate per represented basis bit per tick.
- It ignores locality and bandwidth. A planet-scale simulator that resolves Earth at atomic scale is not only a bit-update problem; it is also a communication, synchronization, heat-flow, and error-correction problem.
- It ignores sparsity, event-driven updates, reversible computation, compression, and checkpoint/rerun strategies. These are not optional engineering details; they change the exponent.
- It treats "update rate" as freely swappable across Planck, nuclear, optical, molecular, and neural rates while holding the same represented bits fixed. That is not a consistent physical model. If the represented state is atomic or nuclear, neural-rate updates do not reproduce the relevant dynamics; if only experiences are represented, Planck-rate updates are an arbitrary overcount.
- It does not distinguish lower bound, upper bound, and proposal-specific cost. It is not a theorem that one simulated bit requires one host operation per simulated time step.

Retract or amend: retract any sentence saying the 40-dex spread by itself proves the cost channel has no discriminating power. Corrected claim: `ops = bits x rate x duration` is a useful sensitivity model for a discrete, active-rendering proposal. It is not a universal resource lower bound for simulations, and it cannot carry the closure alone.

## O2 - FATAL to novelty: C1 is rediscovery, not a new result

Argus did not discover the scope/free-parameter escape. Bostrom explicitly built it into the original simulation argument.

Prior art, exact:

- Nick Bostrom, "Are You Living in a Computer Simulation?" Philosophical Quarterly 53(211), 243-255 (2003), doi:10.1111/1467-9213.00309. In the public text at simulation-argument.com/classic, Bostrom writes: "How much depends on the scope and granularity of the simulation. Simulating the entire universe down to the quantum level is obviously infeasible ... But in order to get a realistic simulation of human experience, much less is needed - only whatever is required to ensure that the simulated humans ... don't notice any irregularities." He then says the inside of Earth can be omitted, distant astronomy can be highly compressed, and microscopic phenomena can be filled in ad hoc.
- Bostrom, Simulation Argument FAQ, item 6, says the argument "does not envisage a universe-wide simulation where every atom and every quark is continuously simulated in perfect detail" and that critiques assuming full comprehensiveness "miss the point." URL: https://simulation-argument.com/faq/
- Eliott Edge and Chad Ashton Brown, "Commentary: Astrophysical constraints on the simulation hypothesis for this Universe: why it is (nearly) impossible that we live in a simulation," Frontiers in Physics 14:1808725 (2026), doi:10.3389/fphy.2026.1808725, make exactly the same point against Vazza: resource refutations target a broader full-world model than Bostrom's minimal subjective-verisimilitude commitment.

Retract or amend: retract any novelty claim. Corrected claim: Argus independently quantified a familiar Bostromian escape and located it in his own ledger; that is useful ledger maintenance, not a new philosophical result.

## O3 - SERIOUS: "a channel controlled by a free parameter cannot discriminate" is too strong

Free parameters are constrained constantly. A free lattice spacing, cutoff, rendering radius, update cadence, error budget, or compression scheme becomes testable when a model maps it to observations.

Examples:

- Silas R. Beane, Zohreh Davoudi, and Martin J. Savage, "Constraints on the Universe as a Numerical Simulation," European Physical Journal A 50, 148 (2014), doi:10.1140/epja/i2014-14148-0, arXiv:1210.1847. They assume a cubic spacetime lattice and derive b^-1 >~ 10^11 GeV from the cosmic-ray spectrum, with possible rotational-symmetry-breaking signatures in highest-energy cosmic rays.
- Franco Vazza, "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 (2025), doi:10.3389/fphy.2025.1561873, tests three explicit model classes: visible universe at full resolution, Earth at full resolution, and a lower-resolution Earth constrained by high-energy neutrino observations.

Retract or amend: replace "a free parameter cannot discriminate" with "an unconstrained free parameter prevents discrimination until a specific model links that parameter to observable signatures or resource commitments." For this parameter, constraints would require a commitment such as fixed lattice spacing, persistent shared world, no ad hoc rendering, no brain-state editing, bounded simulator energy, same-physics host, or a specific minimum fidelity for quantum experiments.

## O4 - SERIOUS: entropy is not simply "the bits needed to specify the microstate"

The C5/C6 entropy move is conceptually wrong as written.

Thermodynamic entropy S = k_B log W counts the logarithm of the number of microstates compatible with a macrostate. In bits, S/(k_B ln 2) is the number of yes/no answers needed to index a microstate if the macrostate and ensemble are already fixed and if the accessible microstates are treated equiprobably. It is not a measured description length of the actual universe's exact state.

Quantum case: for a closed universe in a pure state, fine-grained von Neumann entropy is zero under unitary evolution; coarse-grained thermodynamic entropy can be large. An exact arbitrary pure state in Hilbert space is not specified by S classical bits; amplitudes are continuous in the formalism, while physically distinguishable states are bounded only after specifying precision and operational distinguishability. A quantum host also need not classically store amplitudes.

What must be retracted: "the universe's actual entropy is the number of bits needed to specify its microstate" as an unconditional statement. Corrected claim: Egan & Lineweaver's S_obs gives a coarse-grained thermodynamic entropy estimate for the observable universe. It can be used as an order-of-magnitude index of distinguishable microstates compatible with a specified macrodescription, not as the exact memory required to evolve or render our universe.

Sources: Egan and Lineweaver, Astrophysical Journal 710, 1825-1834 (2010), doi:10.1088/0004-637X/710/2/1825, arXiv:0909.3983, report S_obs = 3.1(+3.0,-1.7) x 10^104 k_B and explicitly say it is dominated by SMBHs. Lloyd, "Computational capacity of the universe," Physical Review Letters 88, 237901 (2002), doi:10.1103/PhysRevLett.88.237901, separately estimates the universe can have performed no more than 10^120 ops on 10^90 bits; this is not the same number as Egan & Lineweaver's black-hole-dominated entropy.

## O5 - SERIOUS: SMBH entropy is the wrong thing to count for observer-visible simulation cost

Egan & Lineweaver is the right citation for a current cosmic entropy budget, but not automatically the right resource requirement for an ancestor simulator.

The number is dominated by supermassive black-hole horizon entropy. Black-hole entropy counts the logarithm of internal/horizon microstates compatible with external macroscopic parameters. For outside observers, classical black holes are characterized by mass, charge, and angular momentum, plus accretion environment and gravitational effects. The inaccessible horizon microstate matters for exact unitary reproduction of the black hole and its eventual Hawking radiation, not for ordinary astronomical verisimilitude over human timescales.

What must be retracted: "S_obs = 10^105 bits is what is required" for a simulator that only has to reproduce what observers see. Corrected claim: 10^105 bits is a candidate cost for exact coarse-grained cosmic microstate indexing if black-hole microstates are part of the target. It is not a lower bound for Bostrom-style observer simulations and not a clean upper bound for exact quantum-state simulation.

Consequence: C5's original retraction of the <=1 nesting claim is directionally right - comparing required state to the holographic maximum was wrong - but the replacement "~10^18 universes fit because actual entropy is 10^105 bits" is not established. What is established is weaker: the holographic bound is a capacity ceiling, so using it as a required memory load was invalid.

## O6 - MINOR: Margolus-Levitin survives, but the energy term was used too casually

The Margolus-Levitin theorem uses average energy above the ground state, E - E0. Margolus and Levitin's abstract says the strict bound depends on "the system's average energy minus its ground state energy" and gives the information-processing-rate implication. Source: N. Margolus and L. B. Levitin, "The maximum speed of dynamical evolution," Physica D 120, 188-195 (1998), doi:10.1016/S0167-2789(98)00054-2, arXiv:quant-ph/9710043.

Using mc^2 is legitimate only as an ultimate upper bound under a strong assumption: the host can deploy essentially all rest mass-energy above its computational ground state without losing the computer to collapse, heat, exhaust, or inaccessible binding/vacuum energy. Lloyd's one-kilogram "ultimate laptop" uses that style of idealization; it is not an engineering budget.

For cosmology, E_total inside an expanding particle horizon is especially slippery: global energy conservation and gravitational/vacuum energy are not simple finite-box quantities in GR. budget.py's matter/all-components split partially acknowledges this, but C3 should not imply the numerical E*t products are precise host constraints.

Retract or amend: keep ML as a surviving quantum-host objection, but label mc^2 calculations as optimistic ceilings for same-physics hosts, not actual available energy.

## O7 - SERIOUS: C2's "no cost argument constrains any host" is false

Cost arguments do constrain hosts once the hypothesis is made specific. Vazza 2025 constrains a same-physics host asked to compute full-universe, full-Earth, or neutrino-compatible Earth simulations. Beane-Davoudi-Savage constrain a cubic-lattice numerical-universe model. ML and holography constrain any same-physics quantum host once target fidelity, runtime, energy, memory, and rendering policy are specified.

What must be retracted: "No cost argument constrains any host." Corrected claim: no currently presented cost argument constrains the generic simulation hypothesis, because that hypothesis leaves host physics and rendering policy open. Cost arguments do constrain specified model classes.

This also amends H13: the classical/quantum substrate distinction was not the only issue, but it was not meaningless. Representation premiums vanish for a quantum host; resource bounds remain. H13 was not killed globally, but the night did identify why a generic kill condition is too broad.

## O8 - SERIOUS: the 75-dex sensory-experience gap is arithmetic, not a decisive closure

The arithmetic is fine: 10^105 versus 10^29 is about 75-76 dex. The interpretation is not.

All human sensory input ever is not enough to specify an ancestor simulation unless the target is only a transcript. It omits internal brain-state evolution, memory consistency, counterfactual interactions, scientific instruments, records, social coupling, random seeds, and the environmental state needed to make future observations cohere. Bostrom can allow ad hoc fill-in, but ad hoc fill-in is itself an algorithmic and consistency burden.

At the same time, cumulative sensory bits may overstate the minimal description length, because a compact generative program plus random seed can produce long correlated histories. So the 10^29 number is neither a solid lower bound nor a solid upper bound on the cost of a coherent observer-history simulation.

What must be retracted: "This is the number that decides the night" and "no overhead factor available in physics closes a 75-dex gap." Corrected claim: the 75-dex gap shows how much cheaper a mere output-history target could be than an exact black-hole-including cosmic microstate target. It does not by itself bound the cost of a coherent, interactive, experiment-respecting ancestor simulation.

## O9 - MINOR: C7 overstates Bostrom's empirical premise

Bostrom's affordability premise is explicit and probably one of the more defensible parts of the original 2003 paper, but "the strongest part" is a rhetorical overclaim. Published debates have often attacked the anthropic/indifference structure rather than cost: Brian Weatherson (and Bostrom's 2005 reply), Anthony Brueckner (Analysis 68(3), 224-226, 2008; Bostrom's 2009 reply in Analysis 69(3), 458-461), Bostrom and Kulczycki's "A Patch for the Simulation Argument" (Analysis 71(1), 54-61, 2011), and David Kipping's Bayesian model-uncertainty treatment, Universe 6(8), 109 (2020), doi:10.3390/universe6080109.

There is a serious published resource critique: Vazza 2025. But Edge and Brown 2026, and Bostrom's FAQ, correctly answer that Vazza targets a stronger full-world/shared-world model than Bostrom needs. Therefore Argus should not dismiss resource critiques as nonexistent; he should classify them as model-specific and often aimed at a nonminimal target.

What must be amended: C7 should read: "Bostrom's affordability premise is explicit and relatively robust for brain-level or subjective-verisimilitude ancestor simulations under his assumptions. Cost critiques attack an important empirical premise, but published resource critiques mostly constrain stronger, more comprehensive simulation models."

## Claim-by-claim correction list

- C1: SERIOUS amendment. Keep only the scoped version: generic cost does not discriminate without a specified rendering policy. Remove novelty and remove the absolute "free parameter cannot discriminate" language.
- C2: SERIOUS amendment. H13 kill condition is not met for the generic hypothesis, but cost arguments can constrain specified quantum-host models. Delete "No cost argument constrains any host."
- C3: MINOR amendment. Survives. Add the E - E0 caveat for ML and capacity-versus-requirement caveat for holography. Cite Margolus-Levitin 1998 and Bousso 2002, Rev. Mod. Phys. 74, 825-874, doi:10.1103/RevModPhys.74.825.
- C4: NOISE objection. Survives. The 0.315-age result really is Omega_m by substitution.
- C5: SERIOUS amendment. The retraction of the original nesting claim is correct, but the replacement argument using Egan & Lineweaver as required microstate bits is overstated. Say "the holographic bound was an invalid proxy for required memory" rather than "nesting depth can be 10^18."
- C6: SERIOUS amendment. The 75-dex comparison is a useful scale contrast, not a sufficient ancestor-simulation cost bound.
- C7: MINOR amendment. Affordability is robust in Bostrom's own brain-level estimate, but "strongest" is not established and Vazza 2025 is a real resource critique of stronger SH models.

## WHAT I COULD NOT BREAK

- budget.py section 5's 0.315-age headline is a tautology equal to Omega_m. That retraction is correct.
- Margolus-Levitin and holographic/covariant entropy bounds are genuine resource bounds that a same-physics quantum host does not evade in principle.
- The original nesting claim "host must be at least as large as what it fully simulates, nesting depth <= 1" was wrong because it used a capacity ceiling as if it were a required load.
- The generic simulation hypothesis remains underconstrained by cost. To get a test, one must specify host physics plus rendering/fidelity policy.
- Bostrom's original paper already permits scope/granularity/ad hoc rendering; any resource attack that assumes full Planck-level persistence is attacking a stronger hypothesis than Bostrom's minimal ancestor-simulation argument.