arxiv: 2604.17516 · v3 · submitted 2026-04-19 · 🌌 astro-ph.IM · astro-ph.GA

Recognition: unknown

Kardashev's Conundrum: Statistical Falsification of the Standard Kardashev Model and the Kardashev--Sagan--Nakamoto Resolution

Sebastian Gurovich

Authors on Pith no claims yet

Pith reviewed 2026-05-10 05:26 UTC · model grok-4.3

classification 🌌 astro-ph.IM astro-ph.GA

keywords Kardashev scaleglobal primary energyexponential growthlinear regressionBitcoin hashraterenormalizationcivilizational developmentstatistical falsification

0 comments

The pith

Global energy data falsifies the Kardashev 1% exponential model and requires renormalization by hashrate to restore physical coherence.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tests the long-standing assumption that global power consumption grows exponentially at 1% per year as a marker of civilizational advance. Markov Chain Monte Carlo analysis of six decades of primary energy data yields a growth rate near 2% per year, yet a simple linear model fits the observations far better and is favored by information criteria. Extrapolating that linear trend to the Sun's total luminosity produces a timescale of 1.6 quadrillion years, roughly 100,000 times the age of the universe. This reductio demonstrates that any function of power consumption alone is dimensionally incomplete. The authors therefore introduce a parameter-free renormalization that divides annual energy production by annual Bitcoin hashrate, yielding a quantity whose 15-year record spans 14 orders of magnitude.

Core claim

The standard Kardashev model assumes power consumption P(t) grows as an independent-increment geometric series, yet six decades of data reject both the 1% rate and the required statistical structure. A linear OLS fit is statistically preferred and extrapolates to solar luminosity in 1.6E15 years, a physical reductio termed Kardashev's Conundrum. No functional form fitted to P(t) alone can satisfy both statistical adequacy and physical coherence because the Kardashev variable is dimensionally incomplete. The Kardashev-Sagan-Nakamoto resolution defines the renormalized quantity B(t) = P(t)/H(t) in J/Hash, motivated by the Landauer limit and Sagan's information-richness criterion; this quantity

What carries the argument

The Kardashev-Sagan-Nakamoto (KSN) renormalisation B(t) = P(t)/H(t) in J/Hash, which divides annual primary energy production by annual Bitcoin hashrate to supply the missing dimensional completeness without introducing free parameters.

If this is right

The 1% exponential growth rate lies well outside the 95% credible interval obtained from MCMC inference on 1965-2024 energy data.
A linear OLS model is preferred over free-rate exponential models by the Widely Applicable Information Criterion.
Linear extrapolation places the Type II civilizational timescale at approximately 1.6E15 years.
The renormalized B(t) spans 14 orders of magnitude over 2009-2024 and fulfills Sagan's information-richness requirement.
No functional form of P(t) alone can meet both statistical adequacy and physical coherence.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Civilizational progress metrics may need to track energy per unit of computation rather than raw power consumption alone.
The 14-order variation in B(t) offers a quantitative basis for comparing technological epochs once longer hashrate and energy records become available.
Substituting other global compute proxies for Bitcoin hashrate would test whether the renormalization remains robust across different information-processing measures.
The approach connects energy accounting directly to thermodynamic limits on computation, suggesting extensions to other planetary-scale energy budgets.

Load-bearing premise

The linear trend fitted to recent global primary energy data will continue without bound or saturation long enough for the extrapolation to solar luminosity to serve as a meaningful physical reductio.

What would settle it

A sustained change in the growth rate of global primary energy production, such as saturation or acceleration, that would allow solar-luminosity levels to be reached within roughly the main-sequence lifetime of the Sun.

Figures

Figures reproduced from arXiv: 2604.17516 by Sebastian Gurovich.

**Figure 1.** Figure 1: Global energy production 1965–2024 (OWID dataset) in watts, with the standard Kardashev one-percent exponential model [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗

**Figure 2.** Figure 2: Zoomed view of the OWID global primary-energy production data (1965–2024) with the linear and exponential best-fit models [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

**Figure 3.** Figure 3: Year-over-year differences ∆Pi = Pi − Pi−1 in global primary-energy production (W) as a function of calendar year 1965– 2024, derived from the OWID direct-method energy dataset. Two stray negative outliers — clearly separated from the general trend of the data — occur at 2008 and 2020, corresponding respectively to the subprime great financial crisis and the COVID-19 pandemic. These are identifiable histor… view at source ↗

read the original abstract

We test the standard Kardashev one-percent exponential conjecture against six decades of global primary-energy production data (1965-2024; Our World in Data). Markov Chain Monte Carlo inference yields a posterior growth rate of r = 2.01 +/- 0.03% per year (95% credible interval [1.94%, 2.08%]), placing the Kardashev 1% value well outside the credible interval. A linear OLS model fits the data with remarkably low dispersion (R^2 = 0.987) and is preferred over the free-rate exponential by the Widely Applicable Information Criterion ({\Delta}WAIC = 5.5). Year-over-year increments are non-Gaussian (Shapiro-Wilk W = 0.925, p = 0.0014; skewness = -0.664) with identifiable crisis outliers (2008, 2020), rejecting the independent-increment multiplicative structure with positive drift required by Kardashev's (1+x)^t geometric series. Extrapolation of the linear model to the solar luminosity yields a Type II civilisational timescale of approximately 1.6E15 years -- approximately 1E5 times both the age of the Universe and the main-sequence lifetime of the Sun -- a physical reductio we term Kardashev's Conundrum. No functional form fitted to P(t) alone can simultaneously satisfy statistical adequacy and physical coherence: the Kardashev variable is dimensionally incomplete. We propose the Kardashev-Sagan-Nakamoto (KSN) renormalisation B(t) = P(t)/H(t) [J/Hash, the KarNak unit], where H(t) is the annual Bitcoin hashrate. The renormalisation adds no free parameters, is motivated by the Landauer limit, and fulfils Sagan's information-richness requirement. Over 2009-2024, B(t) spans 14 orders of magnitude.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper rejects the 1% exponential Kardashev model via MCMC and WAIC on energy data and proposes a hashrate renormalization, but the linear extrapolation reductio assumes an unrealistically stable trend over 10^15 years.

read the letter

The punchline is that this paper uses statistical model comparison on six decades of energy data to reject the exponential growth at the heart of the standard Kardashev scale, then argues for renormalizing it with an information proxy based on Bitcoin hashrate. They fit an exponential via MCMC and get a growth rate of about 2% per year with a narrow credible interval that puts the 1% value outside it. The linear OLS model has an R-squared of 0.987 and beats the exponential on WAIC by 5.5 points. They also show that year-over-year changes are not normally distributed and flag specific outliers from economic crises. Extrapolating the linear fit to solar luminosity gives a timescale of 1.6 times 10 to the 15 years, which they call Kardashev's Conundrum to show that power alone can't be both statistically good and physically sensible. The proposed fix is B(t) equals power over hashrate in a new unit they name after Kardashev, Sagan, and Nakamoto. The new elements are the specific credible interval rejection of 1%, the linear conundrum as a reductio, and the hashrate renormalization over 2009-2024 data. The analysis is transparent with public data and common tools like MCMC and information criteria. The main limitation is in the reductio step. The linear model is fitted to recent data that includes non-Gaussian increments and known shocks, so there's little basis for assuming it persists unchanged over astronomical timescales. Energy growth will hit physical or economic ceilings well before solar levels, which undercuts the claim that no P(t)-only form can work. The choice of Bitcoin hashrate as the information measure is reasonable given the Landauer connection but would benefit from checks against other proxies like total FLOPS or data storage. This work is for researchers in SETI, astrobiology, and long-term energy or tech modeling. Anyone looking at how to update civilizational scales with modern data will find the statistical section useful. The paper shows clear thinking on the data side and engages the literature directly, so it deserves a serious referee even if the interpretation needs tightening. I would send this to peer review.

Referee Report

2 major / 2 minor

Summary. The manuscript tests the standard Kardashev 1% exponential growth model against global primary-energy production data (1965-2024). MCMC inference yields a posterior growth rate of 2.01% yr^{-1} (95% CI [1.94%, 2.08%]), inconsistent with the conjecture. An OLS linear model is statistically preferred (R²=0.987, ΔWAIC=5.5 over exponential) despite non-Gaussian increments (Shapiro-Wilk W=0.925, p=0.0014) and crisis outliers. Extrapolation of the linear trend to solar luminosity produces an unphysical timescale of ~1.6×10^{15} years, termed Kardashev's Conundrum, implying no P(t)-only functional form can be both statistically adequate and physically coherent. The authors propose the parameter-free Kardashev-Sagan-Nakamoto renormalization B(t)=P(t)/H(t) (J/Hash) using annual Bitcoin hashrate as an information proxy motivated by the Landauer limit.

Significance. If the statistical preference for the linear model and the proposed renormalization hold after addressing extrapolation concerns, the work supplies a reproducible, data-driven critique of the Kardashev scale and introduces a dimensionally consistent information-theoretic extension. Strengths include use of public data, MCMC with WAIC model comparison, and explicit falsification of the independent-increment multiplicative structure; these elements support falsifiable predictions for future energy-information ratios.

major comments (2)

[Abstract] Abstract and extrapolation claim: The central reductio that the linear model yields a 1.6×10^{15}-year timescale to solar luminosity (3.8×10^{26} W) assumes indefinite continuation of the fitted trend without saturation or shocks. However, the manuscript itself documents non-Gaussian year-over-year increments (Shapiro-Wilk W=0.925, p=0.0014; skewness=-0.664) and explicit crisis outliers (2008, 2020), which contradict the stable additive process required for such an extrapolation to serve as a physical reductio. This assumption is load-bearing for the claim that no P(t)-only form satisfies both statistical adequacy and physical coherence.
[Abstract] Renormalization proposal: The choice of Bitcoin hashrate H(t) as the information-processing proxy is motivated by the Landauer limit and fulfills Sagan's information-richness criterion, yet the manuscript provides no independent cross-validation against other computational metrics (e.g., total CPU cycles, data-center energy efficiency, or global transistor counts). The reported 14-order-of-magnitude span in B(t) over 2009-2024 may therefore reflect Bitcoin-specific dynamics rather than a general information capacity measure.

minor comments (2)

[Abstract] The abstract states the linear model has 'remarkably low dispersion' but does not report the residual standard error or perform a formal test for heteroscedasticity, which would strengthen the model-comparison section.
[Abstract] Notation for the proposed unit (KarNak) is introduced without a dedicated definition equation; adding B(t) ≡ P(t)/H(t) [J hash^{-1}] as an explicit equation would improve clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments on the manuscript. We address each major comment below, providing our strongest substantive defense while indicating where revisions will be made.

read point-by-point responses

Referee: Abstract and extrapolation claim: The central reductio that the linear model yields a 1.6×10^{15}-year timescale to solar luminosity (3.8×10^{26} W) assumes indefinite continuation of the fitted trend without saturation or shocks. However, the manuscript itself documents non-Gaussian year-over-year increments (Shapiro-Wilk W=0.925, p=0.0014; skewness=-0.664) and explicit crisis outliers (2008, 2020), which contradict the stable additive process required for such an extrapolation to serve as a physical reductio. This assumption is load-bearing for the claim that no P(t)-only form satisfies both statistical adequacy and physical coherence.

Authors: The linear extrapolation is presented explicitly as a reductio ad absurdum to demonstrate that even the statistically preferred model (R²=0.987, ΔWAIC=5.5) produces an unphysical timescale when extended to solar luminosity. This supports the manuscript's central claim that no P(t)-only functional form can be both statistically adequate and physically coherent. The documented non-Gaussian increments and outliers are invoked to reject the Kardashev exponential structure, not to validate the linear model for forecasting. We will revise the abstract and discussion sections to state more explicitly that the 1.6×10^{15}-year figure is a reductio highlighting dimensional incompleteness rather than a literal projection, and to address the effects of non-stationarity on long-term extrapolation. revision: partial
Referee: Renormalization proposal: The choice of Bitcoin hashrate H(t) as the information-processing proxy is motivated by the Landauer limit and fulfills Sagan's information-richness criterion, yet the manuscript provides no independent cross-validation against other computational metrics (e.g., total CPU cycles, data-center energy efficiency, or global transistor counts). The reported 14-order-of-magnitude span in B(t) over 2009-2024 may therefore reflect Bitcoin-specific dynamics rather than a general information capacity measure.

Authors: Bitcoin hashrate is adopted as the proxy because it supplies the longest continuous, publicly verifiable record of large-scale computational work directly coupled to energy consumption, satisfying both the Landauer limit and Sagan's information-richness criterion without introducing free parameters. While we acknowledge that cross-validation against metrics such as aggregate CPU cycles or transistor counts is absent from the current analysis, such datasets lack comparable temporal coverage and precision. The observed 14-order span in B(t) nonetheless illustrates the renormalization's utility in exposing the divergence between energy and information scales. We will add a limitations paragraph justifying the proxy choice on grounds of data availability and suggesting avenues for future cross-validation. revision: partial

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper's reductio—that no P(t)-only functional form satisfies both statistical adequacy and physical coherence, rendering the Kardashev variable dimensionally incomplete—follows from external data fits (OLS linear R²=0.987 preferred over exponential by ΔWAIC=5.5; MCMC r=2.01% outside 1% conjecture) and extrapolation to solar luminosity (1.6E15 years). This is not equivalent to inputs by construction, as the linear model's preference and unphysical timescale are independent data consequences, not self-referential. The KSN renormalization B(t)=P(t)/H(t) incorporates external hashrate data, adds no fitted parameters, and draws motivation from the external Landauer limit without self-definition or renaming. No self-citations, uniqueness theorems, or ansatzes are load-bearing. The chain is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 1 invented entities

The central claims rest on fitted growth parameters from data, an assumption of unbounded linear continuation, and an ad-hoc proxy for information processing; the new unit is introduced without additional free parameters but depends on the proxy choice.

free parameters (2)

exponential growth rate r = 2.01%
Posterior mean 2.01% per year obtained via MCMC from 1965-2024 energy data
linear model slope
Ordinary least squares fit to the same energy time series used for long-term extrapolation

axioms (2)

domain assumption Global primary energy production can be adequately described by either a constant-rate exponential or a linear functional form over multi-decade intervals
Invoked to perform MCMC inference, OLS fitting, and extrapolation to solar luminosity
ad hoc to paper Annual Bitcoin hashrate constitutes a suitable proxy for information-processing capacity governed by the Landauer limit
Basis for defining the renormalized quantity B(t) without further justification or alternative proxies

invented entities (1)

KarNak unit (J/Hash) no independent evidence
purpose: Renormalized measure of civilizational progress that incorporates both energy and computation
Defined directly as P(t)/H(t) and reported to span 14 orders of magnitude from 2009-2024

pith-pipeline@v0.9.0 · 5674 in / 1954 out tokens · 95678 ms · 2026-05-10T05:26:02.185806+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

13 extracted references · 13 canonical work pages

[1]

International Journal of Theo- retical Physics 21, 905–940

The thermodynamics of computation—a review. International Journal of Theo- retical Physics 21, 905–940. doi:10.1007/BF02084158. Blumer, A., Ehrenfeucht, A., Haussler, D., Warmuth, M.K.,

work page doi:10.1007/bf02084158
[2]

Warmuth , Akeywords =

Occam’s razor. Information Processing Let- ters 24, 377–380. doi:10.1016/0020-0190(87)90114-1. Box, G.E.P., Cox, D.R.,

work page doi:10.1016/0020-0190(87)90114-1
[3]

Journal of the Royal Statistical Society Series B: Statistical Methodology , author =

An analysis of transfor- mations. Journal of the Royal Statistical Society: Se- ries B 26, 211–252. doi:10.1111/j.2517-6161.1964. tb00553.x. Dallaire-Demers, P.L., BTQ Technologies Team,

work page doi:10.1111/j.2517-6161.1964 1964
[4]

arXiv preprintarXiv:2603.25519

Kardashev scale quantum computing for bitcoin min- ing. arXiv preprintarXiv:2603.25519. Freeth, T., Higgon, D., Dacanalis, A., MacDonald, L., Georgakopoulou, M., Wojcik, A.,

work page arXiv
[5]

Gray, R.H.,

URL: https://doi.org/10.1038/s41598-021-84310-w, doi:10.1038/s41598-021-84310-w. Gray, R.H.,

work page doi:10.1038/s41598-021-84310-w
[6]

Hastings, W.K.,

doi:10.3847/1538-3881/ ab792b. Hastings, W.K.,

work page doi:10.3847/1538-3881/
[7]

Monte Carlo sampling methods using Markov chains and their applications,

Monte carlo sampling methods us- ing markov chains and their applications. Biometrika 57, 97–109. doi:10.1093/biomet/57.1.97. Kardashev, N.S.,

work page doi:10.1093/biomet/57.1.97
[8]

(1961) Irreversibility and heat generation in the computing process.IBM Journal of Research and Development5(3) 183–191https://doi

Irreversibility and heat generation in the computing process. IBM Journal of Research and Development 5, 183–191. doi:10.1147/rd.53.0183. Lares, M., Funes, J.G., Gramajo, L.,

work page doi:10.1147/rd.53.0183
[9]

arXiv e-prints , arXiv:1510.07674arXiv:1510.07674

IAU 2015 Resolution B3 on Recom- mended Nominal Conversion Constants for Selected Solar and Planetary Properties. arXiv e-prints , arXiv:1510.07674arXiv:1510.07674. Metcalfe, B.,

work page arXiv 2015
[10]

Computer 46, 26–31

Metcalfe’s law after 40 years of ether- net. Computer 46, 26–31. doi:10.1109/MC.2013.374. 10 Metropolis, N., Rosenbluth, A.W., Rosenbluth, M.N., Teller, A.H., Teller, E.,

work page doi:10.1109/mc.2013.374 2013
[11]

The Journal of Chemical Physics21(6), 1087–1092 (1953) https://doi.org/10.1063/1.1699114

Equation of state calcula- tions by fast computing machines. Journal of Chemical Physics 21, 1087–1092. doi:10.1063/1.1699114. Nakamoto, S.,

work page doi:10.1063/1.1699114
[12]

Biometrika 52, 591–611

An analysis of variance test for normality (complete samples). Biometrika 52, 591–611. doi:10.1093/biomet/52.3-4.591. Shklovskii, I.S., Sagan, C.,

work page doi:10.1093/biomet/52.3-4.591
[13]

In: Dancer, A., Garc´ıa-Prada, O., Kirwan, F

Turing, Lovelace, and Babbage, in: The Turing Guide. Oxford University Press. URL:https: //doi.org/10.1093/oso/9780198747826.003.0033, doi:10.1093/oso/9780198747826.003.0033. Watanabe, S.,

work page doi:10.1093/oso/9780198747826.003.0033