Mills-Randal: Science Contrarian review (Step 2)

Reviewer role: Science Contrarian (PhD standard; judges by experimental evidence, not consensus or reputation; Mallove is a model for standards, not a persona). Project: R. L. Mills, The Grand Unified Theory of Classical Physics (GUT-CP), project copy = "December 10, 2018 Edition" (1,876 pp). Mills is the founder, CEO and president of the company that sells the technology (BlackLight Power, now Brilliant Light Power): an interested party, and every company-hosted report below is reported as the company's presentation. Tags: [read] = from the documents, deep reads or downloaded files, [web] = Internet source with URL, [memory] = recalled, with confidence. Sources are repeated in the text where a verdict depends on them.

Source gap (repeated from claims.md): the deep reads cover about 40 percent of the book's pages. Not deep-read: Ch.3-4, Ch.9 (except via Appendix VI), Ch.11 (the H2 and dihydrino molecule), Ch.15 pdf 801-1082, Ch.17-21, 23-30, 31, 33-35 and the first half of Ch.32. My statements about those chapters rest on digests (claim 7 entirely so). Mills's 2008 reply to Rathke (165 pp), the Booker/Nansteel theory reviews, the Jansson 194-page report, the validation presentation, and the BAILII/UK IPO texts were NOT read by me; I used the Investigator's notes for them and say so.

0. What I did

1. Read claims.md, sites.md (Leads), the Investigator's research notes, and the summary, correction and open-item sections of every deep read (the deep file is one 4,944-line file). [read]
2. Read myself: Hagen and Mills 2022 (IJHE; text extracted from research/files, discussion and "Search for alternative interpretation" sections), Rathke 2005 and Dombey 2006 (abstracts and openings), and digests for Ch.24-30 (pdf 1536-1579).
3. Recomputed the g-factor formula (Eq. 1.228) with the book's alpha and with the CODATA 2018 alpha (script in memory, NumPy/stdlib).
4. Web checks (6): Phelps and Clementson 2012; XRISM and Hitomi on the 3.5 keV line; Aspect 1981 numbers. Results in the text.

1. Corrections to Step 1 (facts.json, claims.md, deep reads, Investigator)

1. EPR agreement figure (claims.md claim 2, deep read pp.1228-1232). The book states "average error 0.097 G" over Tables 16.49-16.50. The 2022 Hagen and Mills paper, which is the peer-reviewed form of the same data, states: "The GUTCP fit to the X-band spectrum of H2(1/4) is semi-quantitative with an average error of 0.004 +/- 0.550 gauss over the 11 lines assigned to spin-orbital coupling splitting"; also "the fluxon separation for any given position does not quantitatively fit the predicted value; in particular the separation is not a constant. Also, the number and relative intensities of fluxon lines for a given satellite line are presently not understood" [read: Hagen-Mills 2022, "Remaining questions"]. So the journal paper is considerably less confident than the book. The deep reader's "random-match benchmark s/4 = 0.08-0.23 G" now has a published comparison: scatter of +/-0.55 G is larger than that benchmark. Step 1 should carry both numbers.
2. The paper itself concedes a conventional fit. "Alternative to the hydrino analysis ... the spectrum can also be approximately reproduced under a conventional phenomenological spin Hamiltonian assuming an unusual combination of two isotropic radicals of unequal [hyperfine] ... " (with an I = 1 nucleus, e.g. 14N, for the triplet pattern) [read: p.7]. The authors reject it as "highly unlikely" on chemistry grounds (only H2, O2, H2O and Ga in the reactor; no anisotropy; relaxation behaviour at low temperature). The claims.md summary of claim 2 mentions only the EPR support; the existence of this stated alternative, and that a "broad signal underlies" the spectrum of unknown origin, belongs in the record.
3. Bruker "replication" (deep read pp.1228-1232, Investigator correction). I agree with the Investigator's addition (the 2022 paper gives place and duplicate spectrometers). Add: the cited source [13] is Mills's own "Hydrino states of hydrogen" PDF, and the paper's competing-interest statement says Brilliant Light Power "provided the samples"; Hagen "did not receive any financial support from Brilliant Light Power" [read]. So the independent party is Hagen as measurer; the sample production and interpretation are the company's. No Bruker-authored record exists in the files.
4. g-factor "eleven-figure agreement" (claims.md claim 1, observations). Confirmed by my recomputation: Eq. 1.228 with alpha^-1 = 137.03604 gives 1.0011596521215 (book ...120); with CODATA 2018 alpha it gives 1.00115965247, 2.87e-10 above the measured 1.00115965218. Measurement precision is about 1e-13, so the gap is thousands of sigma [web/memory: CODATA 2018, high]. Not a discrepancy with the deep reader; it confirms it. One addition for fairness: the formula contains Schwinger's alpha/2pi term exactly, and the two further terms (coefficients -1/3 and +pi^2/3 in powers of alpha/pi vs QED's -0.3285 and +1.1812) nearly cancel. QED to the same order reproduces the measurement to 1e-13 only with the higher-order terms (to alpha^5) [memory, high].
5. Ch.7-8 digest corrections in claims.md are consistent with the deep reads; I accept them.
6. claims.md claim 4 wording, "no fitted parameters." The Ch.13 deep read (J. "Summary of answers") lists extra adjustments: CH2 magnetic energy 0.14803 eV, sp3 hybridisation factors 0.91771 (C) and 0.93172 (Cl), a 0.75/0.25 charge split, double-bond multipliers for CO and NO, and experimental omega_e values used as inputs for all molecules except CH, CD, NH, ND, OH, OD, Cl2. Claim 4 as the book states it ("no fitted parameters") should be marked as contradicted by the book's own derivations in part; the claim in claims.md reports the book's statement, so this is a qualification, not a transcription error.
7. Ch.5, hydrino Table 5.2 and the force balance (deep read). The deep reader showed that the force balance Eq. 5.28 is satisfied identically for any Z_eff, so integrality comes from a separate postulate on the "trapped photon" field. claims.md claim 2 lists the energy formula only. I keep that finding in the review (Q a).
8. Position on p up to 137 vs 126. The Investigator's lead is correct as far as I read it: the 2025 Mills-Hagen paper states p <= 126 [read: files/BLP-FeOOH_Hydrino_Paper.pdf p.1, per sites.md]. I did not re-open it. This changes the dark-matter/gamma chain that uses p = 137 (claim 3) and is itself an example of the theory's parameters moving between editions.
9. Facts left uncorrected but flagged: the Investigator logs the Research Square date as 2021; the book prints 2020. Crossref is the better source [web, per Investigator: Crossref doi 10.21203/rs.3.rs-144403/v1]. I did not re-check.
10. Truncated section of the deep file. The pp.1214-1252 section lost the tail of its "Corrections" bullet (neon U(Ne) calculation printed inside the solid-H2 section); acknowledged by a later reader. Nothing in my verdicts depends on it.

2. How I weigh the Investigator's findings

3. Claims

Claim 1: Classical electron; closed-form atoms and the g factor

a. Mathematics. Mixed. Arithmetic reproduces everywhere I looked (Eq. 1.228, Table 1.5, Table 7.1, helium r1 = 0.566987 a0). But the derivation is not internally consistent as printed [read, Ch.1 and Appendices I-IV deep reads]: - hbar is an input (Eq. 1.35 builds it in; Eq. 1.37 "derives" it back). - Spin 1/2 follows from a hemisphere-uniform great-circle construction chosen to give hbar/2 (Mills's own numerical result 0.495 hbar); the full Bohr magneton needs a further resonant Larmor photon postulate; N = 3/2 is inserted by hand. - A missing (4 pi) factor: the printed field energy (Eq. 1.162) is (4 pi)^2 times the SI value. - Three conflicting wavelength definitions, Gaussian/SI mixing, a factor-2 mismatch in the spin-flip photon. - The nonradiation condition vanishes only because lambda_n = r_n is imposed; for l >= 1 the proper factor j_l(k r) is not zero (j_1(2 pi) = -0.159). The Poynting-vector proof drops a charge term that its own continuity equation makes nonzero. The self-force removal rests on postulates (2-D perfect conductor, no action of one surface point on another); with the self-force included Mills's own force balance gives r = 2 a0. - Relativistic ionization energies: Eq. 1.293 is numerically Dirac's point-nucleus 1s formula with a reduced-mass factor. It reproduces what the Dirac equation gives; it does not add a result. - Multi-electron atoms: integers A, B "chosen to give an energy minimum"; errors up to 4.6 percent with a smooth Z-trend; the "experimental" values at high Z are partly interpolated isoelectronic fits.

b. Experimental consistency. The well-measured hydrogen levels and the Rydberg are reproduced because the model is Bohr-equivalent (reduced-mass R_H = 10,967,758 m^-1) [read]. These are not discriminating: they are matched by standard quantum mechanics too. The one place the model speaks differently from standard theory at high precision is g-2: Eq. 1.228 differs from the measured value by 2.9e-10 with modern alpha [computed]; its alpha^2 and alpha^3 coefficients differ from QED's; QED at five loops matches the experiment to about 1e-13. The "eleven-figure agreement" is true only against 1973-era constants (uncertainty about 7e-9 in g/2), which are less precise than the discrepancy now found. Anomalous results the model should predict and standard theory does not: none identified. The positronium section (Ch.30, digest) quotes calculated g/2 = 1.001159652137 vs measured 1.001159652188(4): off by 5e-11, many sigma, with the book's claim of agreement not matching the stated experimental uncertainty (digest p.1578; I did not read the page image).

c. Tested? There is no controlled experiment designed to discriminate this model from QM on the electron's structure. Booker (a physics PhD, also a company report author) states he reproduced Tables 1.1-1.5 and 7.1-7.2 "to a very high degree of accuracy" [web/read via Investigator]; that is a check of arithmetic by someone paid by the company in other roles (payment for the theory reviews not stated). The muon g factor "prediction" (Appendix III Eq. 26) returns its input: beta^2 = 2/g_mu is taken from the measured g_mu, and the result is unchanged within 7e-9 for any input [read]. It cannot be tested against the Fermilab value. Selke 2025 extends the force-balance method to heavier ions (Investigator; not read).

d. If true. QED's perturbative expansion would have to be a coincidence for the electron's anomalous moment to 1e-13, quantum mechanics would not be needed for atoms, and Haus' nonradiation condition would have to hold for bound electrons with l >= 1 contrary to the Bessel-function factor. That is a very large change, against dozens of independent confirmations of QED (Lamb shift, g-2, hyperfine).

Strongest honest case. (i) A real result: a spherical shell with surface current satisfying the Haus condition does not radiate; extended nonradiating sources exist in classical electrodynamics (Goedecke 1964, Devaney and Wolf 1973 [memory, high]). (ii) Getting the hydrogen ground-state energy, the ionization energies of 1- to 20-electron systems to a few percent, and Schwinger's term from a single constant is not trivial, and no one else has done it with this little input. (iii) The model's mathematically explicit predictions can be checked, which many alternative theories cannot say. Against this: the points above show the work is closer to a re-expression of Bohr/Dirac results plus per-atom choices than a derivation, and the one sharp precision test (g-2) goes the wrong way.

Verdict (split): - Strong form (the electron, spin, g and atomic energies are derived from Maxwell's equations and Newton's laws with only constants): Internally inconsistent (hbar input, wavelength postulate, 4-pi and factor errors, nonradiation condition valid only under lambda = r), and for the g factor specifically Tested, not supported (2.9e-10 away from the modern measurement; different QED coefficients). - Weak form (a force-balance recipe reproduces hydrogen-like and some multi-electron energies to about 0.5-5 percent): Established as arithmetic only; not independent of known results (it matches Bohr/Dirac for one electron; the multi-electron integers are chosen atom by atom). I classify the multi-electron predictions as Never properly tested (no blind test of the integer choices).

Claim 2: Hydrinos (states below n = 1; catalytic energy release; evidence)

a. Mathematics. The bookkeeping is consistent (Eqs. 5.5-5.9; energy [(p+m)^2 - p^2 - 2m] x 13.6 eV reproduced). Two internal issues from the deep reads: (1) the force balance Eq. 5.28 is an identity for any integer p+m, so the existence of integer states below n = 1 comes from the "trapped photon" Laplace condition (Eq. 5.27), not from the force balance; (2) the catalytic resonance condition (multiples of 27.2 eV within about 1 eV for K, Li, He+, Ar+ etc.) is a tolerance, and the same book allows the H atom itself and H2O as catalysts, so most matter at some temperature is a catalyst candidate. The hydride-ion binding 0.75418 eV agrees with Lykke 0.754195(19) eV at about 1 sigma [read], but ab initio QM gives the same value for ordinary H- [memory, medium-high], so it does not discriminate. Power examples (100 kW, 144 kW per m^3) are conditional estimates with efficiency factors set to one. The upper limit moved between editions: 137 in the project book, 126 in the 2025 paper [per Investigator].

b. Consistency with established experiment. Standard quantum mechanics predicts no state below the ground state of hydrogen. The Klein-Gordon "hydrino" solution (Naudts) is an anomalous singular solution; Dombey shows it disappears for an arbitrarily small finite nuclear radius [read: arXiv abstract]. Hydrogen's ground state energy and hyperfine splitting are measured to many digits; a population of hydrinos coexisting with ordinary H (for example in the Sun, the interstellar medium, any hydrogen flame) would need to be consistent with these measurements; I have found no consideration of such an upper limit in the files I read. Anomalous experimental reports exist and I treat them as data: excess heat in light-water nickel cells (NASA 1996; Thermacore 1994); DSC exotherms (Glumac 2014); EUV continua near 10.1 and 22.8 nm; broad Balmer lines (fast H, with Phillips and Loureiro-Amorim disagreeing on the interpretation); the EPR spectrum of Ga(O)OH.

c. Tested? Yes in part, by whom: - Heat, light-water Ni-K2CO3: NASA Lewis 1996 (positive, small, authors name recombination as an ordinary explanation they could not exclude); EarthTech 1997 (negative, with recombiner); Thermacore 1994 (positive, contractor in cooperation with the inventor); company-listed labs, unsourced. This is a mixed, low-signal set and mostly 1990s. - Plasma power: Nansteel 2021 (single test, 169-210 kW, water-bath calorimetry), Tse (hired observer), Booker 2020: all reported on the company site, none published in an independent journal as far as the Investigator found. I did not read them. - EUV continua: Harvard CfA ran a pinch-discharge reproduction (per the Investigator's search results; I have not seen its report); Phelps and Clementson 2012 give an ordinary-physics explanation (ions sputtered from electrodes). - EPR: Hagen and Mills 2022 (peer-reviewed, authors' competing interest stated), fit "semi-quantitative", a conventional two-radical spin Hamiltonian fits approximately, one underlying broad signal of unknown origin, samples from the company. The Bruker "duplicate" is reported by Mills's co-author. No independent group has published a replication (the Investigator found none through 2026). IIT Kanpur announced a replication programme in 2026; no results. - Theory: Rathke 2005 (NJP, ESA) finds "severe inconsistencies in the deterministic model"; Mills answered with 165 pages (2008, not read by me). So: the heat and spectral anomalies have been tested by some outside parties and the results are mixed and unresolved; the hydrino identification has never had a neutral, preregistered test with blinded controls and independently produced samples.

How it was rejected. USPTO: the examiner allowed 09/009,294 after "experimental results and the submission of samples"; the application was withdrawn from issue on 17 February 2000 after "an outside inquiry" and a group director's view that the claim was "contrary to the known laws of physics and chemistry"; the court records that no one involved had reviewed the application before the notice [web: FindLaw, per Investigator, https://caselaw.findlaw.com/court/us-federal-circuit/1362587.html]. The Federal Circuit upheld the Director's authority; it did not rule on the physics. UK IPO 2008 rejected on "contrary to generally accepted physical law" grounds; the Patents Court appeal [2008] EWHC 2763 (Pat) remitted the applications [memory, medium; not read]. I would call the agency-level handling "dismissed on theory, without a neutral experimental test", since the one examiner who examined the data allowed the claims. This is not suppression in the sense of hidden data: the company's data are published on its own site. Counterpoint: a patent office is not a laboratory and has no duty to test; and the company has not placed samples, cells or complete raw data in a neutral laboratory that the Investigator can document. Also, as the Investigator notes, other laboratories listed by the company are unsourced.

d. If true. Quantum mechanics would need a different structure for the hydrogen atom, with a rule forbidding or hiding the sub-ground states from every spectroscopic experiment ever done; hydrogen would be an energy source with a new energy scale (about 100 eV per atom in the first step) and the dark-matter inventory would be redefined. Chemistry would need hydrino hydride compounds and their NMR shifts.

Strongest honest case. (i) Several anomalies are measured, not just theorised: NASA's own (small) excess, Thermacore, DSC exotherms that exceed the author's thermochemistry. (ii) The EPR spectrum is a real structured spectrum from an outside expert, not a plot made by Mills; its origin is stated by its own measurer as not understood in several respects. (iii) The theory makes testable, integer-pattern predictions (rotational lines at 4 x 61 cm^-1 for H2(1/4), 16 x H2 rotational constant; hydride NMR shifts; a hyperfine line list) that a neutral laboratory could check in an afternoon with a sample. (iv) The record of thirty-five years is odd: continuous private funding (Form D: $18.3M sold by 2018; company: over $150M) and sustained publication, without a product, which is consistent with either a real but hard-to-engineer effect or a persistent measurement artifact. Strongest honest case against. (i) In the one place where independent teams looked (EUV, light-water heat) results are mixed to negative, and the positive outside results have ordinary candidate explanations they themselves name. (ii) The EPR agreement metric changed from 0.097 G in the book to 0.004 +/- 0.55 G in the paper, with a conventional alternative fit. (iii) A 2025 paper moves the hydrino cut-off from 137 to 126. (iv) The explained power examples assume populations of hydrinos whose creation is not computed. (v) The raw power validations are single runs by hired observers.

Verdict (split): - Strong form (stable states of hydrogen at n = 1/p exist, are produced by catalysis, and release the stated energy): Never properly tested by a neutral controlled protocol. It was not dismissed without testing; several partial tests exist, with mixed results. Flag: Tested, not supported applies to the 1997 EarthTech light-water cell specifically; Plausible, untested does NOT apply, because the theory conflicts with standard QM and with a hydrogen ground state measured very precisely. - EPR proof of "molecular hydrino": Never properly tested (no independent replication published; a conventional alternative fit not excluded by data in the paper). - Heat claims in light-water nickel cells: Never properly tested (mixed positive and negative, recombination not excluded in the positive NASA run).

Claim 3: Hydrinos as dark matter and cosmic emissions

a. Mathematics. Arithmetic reproduces: cutoffs at 91.2/22.8/10.1 nm for m = 1, 2, 3 (the book pairs them in the reverse order on p.263 [read]); 3.48 keV = H[a_H/4] + H -> H[a_H/17] gives 3481.6 eV (0.3561 nm) [read]; Foerster-type rate 8e21/s only by setting the orientation factor, transition probability and overlap to one by assertion. A p = 137 chain via Eq. 5.113 is placed in other chapters (not read). Book errors: Sun power density 272 W/L (should be 2.72e-4), 0.1 MeV paired with 1.16e7 K (should be 1.16e9 K).

b. Experiment. The 3.5 keV line (Bulbul et al. and Boyarsky et al. 2014) is real as a feature in stacked XMM spectra, but Hitomi's Perseus observation did not find a line at the claimed flux (inconsistent at 99 percent for a broad dark-matter line and 99.7 percent for a narrow gas line); XRISM's ten-cluster stack finds no unidentified line, with a decay-rate limit 3-4 times tighter than Hitomi and still 5 times above the Bulbul rate [web: search summary; arXiv:1903.11608, Hitomi/XRISM abstracts, not read]. So the one number the book matches (a single energy) is itself not confirmed. The 511 keV and gamma-ray burst claims are cited only by reference (Share, Hurley, Boehm) [read]. Dark-matter evidence from rotation curves, lensing, the CMB peak pattern and the Bullet Cluster is not addressed with a hydrino model of a halo [read: not found in Ch.5-6]. The cosmology (CMB peaks at 197/541/824 vs about 220/540/810, B-mode) is addressed under Claim 6.

c. Tested. Nothing in the pp.265-286 deep read is a laboratory spectrum, hydrino power measurement or fusion measurement. No replication of the astrophysical mapping. The Hagen and Mills cosmic neutrino background EPR paper (not downloaded) is the only experimental item bearing on this claim.

d. If true. Dark matter would be ordinary hydrogen in a state with no interaction at the observed level; the Standard Cosmological Model's CDM component and the 3.5 keV search programme would be mooted.

Strongest honest case. A single-number match at 3.5 keV looked impressive in 2014; if hydrinos exist, dark hydrogen is cheap and attractive (no new particle). The model's p-quantised energy ladder yields many lines, and astronomers report several unidentified features. Against: with p up to 137 and m up to 4, a ladder of cutoff and line energies is dense; a match to one line is weak evidence without a look-elsewhere count; Hitomi/XRISM do not support the 3.5 keV line; the project book's own mechanism (hydrino disproportionation needs a hydrino population created by catalysts at 675 K) is not computed for space.

Verdict: Never properly tested (depends on claim 2; nothing tested directly); the 3.48 keV match is Not testable as stated as evidence (one number, with an unspecified chain), and the line itself is Tested, not supported as a dark-matter decay line by Hitomi/XRISM, which says nothing for or against hydrinos specifically.

Claim 4: One method solves molecules and matter

a. Mathematics. Arithmetic reproduces everywhere checked (all 18 rows of Table 13.1 within 1.5 percent; typical 0.3 percent; Table 15.9 sample; boron and organoaluminium). The summary table (431 rows, mean 0.17 percent) is a count of percentages that the Ch.15 deep reader found do not follow (exp - calc)/exp for guanine and eight other rows. Group increments are near-constant (12.1577 eV per CH2 = 7.83016 + 4.32754). The per-CH2 step reproduces experimental alkane steps (12.16-12.18 eV).

b. Experiment. Geometry and energies are within 0.1-3 percent of handbook values. The biggest mismatches: amino acids, DNA bases, phosphorus compounds, noble-gas van der Waals (1-3 percent). CH2 and NH2 bond angles miss by 2-3 degrees. The "experimental" CH2, CH3, CH4, NH2 values in Table 13 come from computational papers. Bond energies of this accuracy are what group-additivity (Benson) schemes also produce, but these are fitted [memory, medium]; ab initio chemistry (CCSD(T)) does better for small molecules and gives geometries and dipoles that Mills's scheme only partly treats [memory, medium].

c. Tested. No blind test: no list of molecules whose properties were published before the experimental values; the book is a compilation. Booker/Nansteel report reproducing the computations (company-hosted). The deep reader flagged that experimental omega_e values are inputs for most molecules and that hybridisation factors and magnetic terms are added case by case. Whether the group parameters were tuned to the experimental set "cannot be told from the pages read".

d. If true. Quantum chemistry's orbital and electron-correlation machinery would be an unnecessary scaffolding; density functional theory's success would have to be accidental.

Strongest honest case. Getting 800 organic molecules' total bond energies to 0.2-0.3 percent from a closed-form recipe is a remarkable fact if the parameters really are fixed in advance: a group-additivity scheme that predicts new molecules would be a real result; the experimental comparison values are third-party handbooks; the numbers reproduce. Against: the number of discrete choices per molecule (hybridisation factor, charge split, magnetic term, which fragment is "radical") is large enough that a fit at 0.2 percent is not surprising; no out-of-sample test.

Verdict: the accuracy statement is Plausible, untested for out-of-sample molecules (Never properly tested as a blind prediction method); the claim of "no fitted parameters" is Internally inconsistent where the book's own derivations add case-specific factors; the arithmetic reproduces.

Claim 5: Quantum mechanics is wrong or unnecessary

a. Mathematics. Ch.40: the Fourier transform of r e^(-r/a0) is correct up to a factor-2 slip; the non-zero result at |k| = omega/c is generic for any smooth density; the argument depends on the premise that the Schrodinger charge density oscillates at omega_n, which for a stationary state (|psi|^2 static) it does not [read]. Ch.42 Aspect/Bell: the classical coincidence curve R(phi)/R0 = 0.2490 + 0.2178 cos 2phi uses formula (42.23), which is the quantum prediction formula of Aspect et al. [read; calc]; the mechanism described in words gives half the modulation and satisfies the inequalities [read: deep reader]. My web check: Aspect's own quantum predictions for the 1981 experiment are delta_QM = 5.8e-2 +/- 0.2e-2 and measured 5.72e-2 +/- 0.43e-2 (13 sigma), and the 1982 S_QM = 2.70 +/- 0.05 [web: https://arxiv.org/pdf/2212.05535 and other secondary sources, via search summary; I did not read the PRL]. The "classical prediction" delta = 5.8e-2 in the book is therefore the QM number. Medium confidence is now closed: matches the deep reader's memory.

b. Experiment. The Bell violation is confirmed by the loophole-free tests of 2015 (Delft, Vienna, NIST) and later [memory, high], none cited in the book beyond 2007. The classical models for Stern-Gerlach, electron diffraction, the SQUID cat experiments (Friedman 2000 splitting 0.139 K; O'Connell 2010 6.187 GHz) are reproduced arithmetically but only by single-number comparisons (SQUID energy depends on the alpha factor from the g model; without it about 120 K, per the deep reader). The one genuinely interesting quantitative section is superfluid helium electron mobility (Ch.42): 15 mobility peaks of Ihas/Sanders with r1 = 6.7 A from 0.5 eV, deviations mean 3.4 percent, max 8.9 percent, with about 69 candidate values from a free (n, l, m) choice [read]. Maris's finding that the standard bubble models fail on those peaks is exactly the kind of anomaly I would want a neutral reviewer to take up; but the deep reader counted the free choices and I cannot say the fit is more than a chance match without that count being made.

c. Tested. Classical alternatives to QM were tested decades ago in the Bell programme and in hidden-variable models; the book does not address the loophole-free tests. The "uncertainty principle refuted" claim has no experiment attached on the pages read; Appendix II/III arguments for the muon rest on circular input.

d. If true. The whole of QM, quantum optics, quantum computing and the Nobel 2022 Bell experiments.

Strongest honest case. Foundational critiques of Born probability and of wave-particle duality are legitimate subjects; Bell's theorem assumes local realism including measurement independence, and "Bell's theorem does not apply to deterministic classical physics" is one position (superdeterminism) held by some physicists, though not in Mills's form. The helium-mobility peak assignment is an unusual quantitative claim that could be assessed. Against: the Bell calculation is the QM calculation relabelled; the radiation argument for Schrodinger orbitals assumes time dependence absent in stationary states; the muon check is circular.

Verdict: the Bell part: Internally inconsistent (the "classical" prediction is the quantum formula) and Tested, not supported by the loophole-free tests. The "Schrodinger densities must radiate" part: Internally inconsistent (premise not derived). The helium-mobility peaks: Never properly tested against alternative explanations.

Claim 6: Spacetime, gravity, cosmology and particle masses

a. Mathematics. Reproduced by the deep reader: the period 3.10e19 s; H0 = 78.5 km/s/Mpc from m_U; lepton ratios; Z, W, Higgs masses; mu_p = (8 pi/9) mu_N. Errors: 1 Gyr = 3.1358e16 s (should be 3.156e16); the prefactor 1229 should be about 12,290; Box 32.3 ratio mis-stated; km/s^2 = km/s/Mpc (Eq. 32.155); the absolute lepton masses need a "1 sec = 0.9975 s" fit; "N" correction factors in Eqs. 36.15-36.19 are justified in words. The muon/electron ratio agrees to 1.6e-8 after those corrections [read]; the tau ratio to 5.6e-5; the neutron mass misses by 5.2e-6, far outside its measurement error.

b. Experiment. H0 78.5 vs Planck 67.4 +/- 0.5 and SH0ES 73.0 +/- 1.0 (high by 7-16 percent) [memory, high]; T_CMB from the equations 3.08 K vs 2.7255 K (13 percent); CMB peaks 197/541/824 vs about 220/540/810, close only with the age fixed at 10 Gyr; B-mode r = 0.16 vs r < 0.036 (BICEP/Keck 2021) [memory, high]; top quark 187 GeV vs 172.57 +/- 0.29; Higgs 128.75 vs 125.2 GeV (2.8 percent); deuteron binding 2.181 vs 2.2246 MeV (2 percent); X17 17.11 MeV vs ATOMKI-reported masses (unconfirmed) [memory]. The sign of the acceleration agrees with the discovery of accelerated expansion (1998-99), but the model has no decelerating epoch, which the supernova data show [memory, high]. The "predicted in 1995" claim (November 1995 edition, Chapter 22) cannot be checked from the project files; the Investigator listed WorldCat/LoC as unreached. Not-found is not a refutation; but the earlier edition is the only record of whether this is a prediction.

c. Tested. The comparisons are Mills's with published numbers; nothing independent beyond Booker's arithmetic check of Ch.32 (half the equations, company-hosted). The pre-/post-dating of each number is the crux and is unresolved for the top-quark mass (announced 2 March 1995), X17 (published after the ATOMKI reports) and the CMB peaks (fitted after DASI/WMAP).

d. If true. General relativity's field equation would be replaced for gravity by a wave-equation with particle production; the standard Lambda-CDM parameters would be re-expressed through Mills's constants; all quark masses would be derived. A single constant alpha would fix lepton masses to 1e-8: if the correction factors were truly fixed in advance this would be extraordinary.

Strongest honest case. The lepton-ratio fit is the single most striking quantitative item: a muon/electron mass ratio matching to 1.6e-8 and a tau ratio to about 1 sigma from alpha and geometric factors; if the correction factors are pre-specified, it is an unreasonably good match. The CMB peak formula with two parameters gives positions close to the measured ones. Against: the correction factors are justified in words after the fact; field-strength factors are numerous; H0, T_CMB, B-mode and top mass miss by 7-16, 13, ~4x and 8 percent; the 1995 acceleration claim is not verifiable in the files.

Verdict (split): the cosmological numbers (H0, T_CMB, B-mode r, top-quark mass): Tested, not supported (they disagree with current measurements). The lepton mass ratios from alpha: Never properly tested (agreement after corrections whose pre-specification is unknown; a look-elsewhere count is needed). The "1995 prediction of acceleration": Never properly tested (primary record not obtained). The wave-equation derivation of gravity/Schwarzschild: Not assessed (digest only; first half of Ch.32 not deep-read).

Claim 7: Collective phenomena (superconductivity, quantum Hall, Aharonov-Bohm, pair production, positronium) are classical

Based on digests only (Ch.24-30, pdf 1536-1579) [read: digest]; I did not read the equations. a. Ch.24 uses the textbook Bose-Einstein, Fermi-Dirac and Planck derivations with de Broglie wavelengths and Planck's constant: it reproduces the standard results rather than deriving them classically. Ch.25 gets Tc for "conduction electron density 1e29/m^3" 30.8 K vs 23.2 K for Nb3Ge, and for oxides 7 K (3D), 22 K (2D) and 180 K (1D) with the dimension chosen per material: the dimension is a free choice. Ch.26 gets R_H = h/(n e^2) with the integer n entering by a band-pass condition n k_p = k_z; the fractional case by "linked wells" with integers n1, n2: the fractions appear as a result of choosing integers. Ch.27 uses Phi_0 = h/2e as the flux increment for angular momentum conservation. Ch.30 gives g/2 = 1.001159652137 vs 1.001159652188(4). b-c. Not enough read to say. Standard theory explains these effects (BCS, Landau levels with Laughlin states, Aharonov-Bohm phase) and they are quantitatively verified; the book's comparisons are limited to selected values in the digests. I cannot confirm that the "predicted" fractional states match the observed fractions beyond the examples shown. d. BCS theory and the fractional-QHE theory (Laughlin, composite fermions) would be replaced. Verdict: Never properly tested (provisional; rests on digest only; the deep-read order in claims.md for Ch.25-27 remains outstanding).

4. Pattern across claims (what the contrarian sees)

1. Arithmetic is not the problem. Across the book the deep readers (and, by his own statement, Booker) find that the printed numbers follow from the printed equations. The weak points are (a) the equations rest on postulates (hbar as input, lambda = r, hemisphere construction, Z_eff integer condition), (b) per-case choices that make a comparison less than blind, and (c) comparisons to one number.
2. The one place a sharp precision test exists, the electron g factor, goes against the theory (2.9e-10, thousands of sigma, with different alpha-power coefficients).
3. The strongest positive items are not the ones the book emphasises. The measured anomalies (NASA excess heat at the 10-percent level; Thermacore; Glumac DSC; the structured EPR spectrum; the helium-mobility peaks) are the evidence that deserves a neutral laboratory. They do not need the cosmology or the muon mass.
4. The most informative cheap test is the H2(1/4) Raman/EPR prediction set. A neutral laboratory with a sample from the company and a sample from a different source could test the EPR spectrum, the rotational line at about 1950 cm^-1, and the hydride NMR shifts. Hagen and Mills 2022 invite exactly that ("an urgent invitation to academia at large to repeat and extend"); IIT Kanpur's programme is the first sign of uptake.

5. Absence of evidence (SOURCE HONESTY)

Thirty-five years of filings, reports and demonstrations (1991 onwards), $18.3M sold on Form D by 2018 and "over $150 million" claimed by the company, with no product sale found and only one non-company group (IIT Kanpur, 2026) announcing a replication. Competing explanations and what decides them: - Commercial delay or engineering difficulty (real effect, hard to scale): for it, continued private funding by repeat investors and board stability (Good on every Form D); against it, the claimed 100 kW prototype since 2020 with no sale or independent demonstration. Decisive record: outside engineers' access to a running cell with their own instruments. - Trade secrecy: for it, the redacted Jansson report (194 pp) and redacted summary reports; against it, the company publishes validation reports and the book. Decisive record: unredacted reports. - Effect not real / artifacts (recombination, electrode ion emission, instrument or calorimetry errors): for it, NASA's own caveat, EarthTech's negative, Phelps and Clementson's explanation, the changes in the p limit and the EPR metric; against it, anomalies persist across methods (heat, EUV, DSC, EPR) in several labs. Decisive record: an independent group's blind result with its own samples (IIT Kanpur). - Institutional refusal to test after the 2000 USPTO and 2008 UK rulings: for it, the 2000 withdrawal with no review and the pattern of critics citing theory, not tests (Rathke himself wrote that discussion of the theory had been "restricted to the sweeping statement that the hydrino model is in contradiction to quantum mechanics"); against it, NASA/NIAC funded Rowan's Phase I (2002), Rutgers/UIUC/Rowan academics did measurements. Decisive record: USPTO file wrappers, identity of the "outside inquiry", agency review notes, the Floyd J judgment text. - Specific disappearances: Thermacore Phase II (not located); the 13 labs the company once listed (AECL Chalk River, Westinghouse, MIT Lincoln, INEL, Hokkaido, Moscow Power Engineering Institute, Brookhaven); the November 1995 edition (the record for claim 6's acceleration "prediction").

6. What would settle each claim (by the contrarian's lights)

1. g-factor: nothing further needed; compare Eq. 1.228 with modern alpha (done above). For the weak form, a blind prediction of an ionization energy for an ion not tabulated.
2. Independent laboratory with its own reactor: produce Ga(O)OH or FeOOH products independently, EPR at several microwave bands plus Raman at the predicted lines, with an ordinary-contaminant analysis; a calorimetry run in a light-water cell with a recombiner, as EarthTech did, in a third lab.
3. Hitomi/XRISM-type limits, and a look-elsewhere count for the line ladder.
4. A list of molecules whose properties are published under the method before the data are looked up.
5. A published statement of how Eq. (42.23) differs from the QM formula and a run against the 2015 loophole-free data.
6. The November 1995 edition (LoC/WorldCat/archive) and the earliest dated statements of the top-quark mass, X17 mass and the CMB peak formula.
7. Deep read of Ch.25-27 (already ordered in claims.md).

7. Summary of verdicts

8. Corrections to earlier documents (for the orchestrator)

1. claims.md, claim 2: add that the published EPR paper (Hagen and Mills 2022) reports average error 0.004 +/- 0.550 G over 11 lines, "semi-quantitative", with a stated conventional alternative fit (two radicals, one I = 1 nucleus) and an unexplained broad underlying signal; the book's "0.097 G" is the book's figure.
2. claims.md, claim 4: "no fitted parameters" is the book's statement; the Ch.13 deep read lists case-specific factors and experimental inputs that qualify it.
3. deep/GUT-CP-2020-Ed-Web-b290db.md, Ch.42 section, medium-confidence memory about Aspect's numbers: confirmed [web: delta_QM = 5.8e-2 +/- 0.2e-2, https://arxiv.org/pdf/2212.05535 via search summary]. Upgrade the tag for the arbiter.
4. facts.json: if it still lists the 3.48 keV match as support for claim 3, add Hitomi (Perseus) and XRISM (ten-cluster stack) limits [web: via search summary of the XRISM Collaboration paper and Hitomi constraints].
5. The truncated tail of the pp.1214-1252 deep read still needs redoing if the neon U(Ne) point matters (low priority).
6. Nothing in sites.md contradicts my reading; two caveats: Rathke's and Dombey's texts were read only to the abstract/opening, and the EarthTech list of 13 labs is the company's list transcribed by EarthTech (not independent confirmation of those results).