Arbiter: Kelly-Patrick, "The Practical Guide to Free-Energy Devices" (V32.9, 2018)

Arbiter (Step 3). Date: 2026-10-08. Inputs read: claims.md, facts.json (by section), physicist.md, mathematician.md, contrarian.md, deep/PJKbook-9cc3a5.md (by section), sites.md and the research/ notes. I went back to the extracted book text and to one page image (p.17) where reviewers differed or where nothing had been read. Tags: [read] = project documents or the extracted book text (with page); [read: page image p.N] = I rendered the page; [calc] = my own arithmetic (Python, .venv); [web] = URL; [memory] = recalled, with confidence.

Source gap note (repeated, as claims.md requires)

This is not a source gap in the strict sense: the text layer of the single document is readable (79 of 2,481 pages were OCR'd). But: - The book is a compilation. Kelly says it is mainly a report of what other people say and "you are welcome to consider this a work of fiction" (p.11) [read]. It is the primary source for what Kelly and the people he reproduces claim. It is almost never evidence that a device works. - About 60 percent of the pages were deep-read. Chapters 1-2, the first part of Chapter 3, Chapter 12 (electronics), Chapter 14 (renewables), Chapter 20 (health) and appendix pp.1901-2238, 2253-2328, 2433-2481 were digest-only. Verdicts on Heins, Flynn (beyond its patent text), Asaoka, Kunel, Tseung's magnet frame, Coler, Plauson, Rayburn TREC and the Chapter 14 devices rest on digest-level text and are weaker than those on Sweet, Benitez, electrolysis, Moray or Newman. - Circuit diagrams, graphs and most patent figures are images that no role rendered. Annis/Eberly's COP table, Ainslie's figures and most circuits were never seen. - Chapters 22 "Dealing With Chemtrails" and 23 "The Mandela Effect" appear only in the contents list [read: claims.md]. - Where a verdict below rests on [memory] alone I say so (listed at the end under "Web checks"). - Update after QE (2026-10-09): the digest-only areas named above have since been deep-read (see "QE revision" at the end of this file). Where the revision gives a different verdict from the one in the body, the revision rules.

How the three reviewers compare

All three reviewers apply the same measuring stick and reach almost the same verdicts. They agree, and I confirm, on these points: 1. The book contains almost no input-and-output measurement by a named independent party. The measured items are bulb brightness, V x I products, nameplate ratings, terminal voltages and patentee examples. 2. The weak form of Claim 1 is fine (open systems can have COP above 1); the strong form needs a named, metered source and none is given. 3. The three places where the inventor's own numbers fail by calculation are Benitez GB 121,561, the Don Smith/Mohamed power formulas and the water-splitting energy figures. 4. The cheapest decisive tests are named by the book itself and have never been reported by a neutral party.

Where they differ, I rule below. The main rulings: - Skinner "1200 percent": the physicist took Kelly's reading (COP 1,200); the mathematician and contrarian favoured COP 12. I side with them and add the primary text (see Claim 4). - Sweet's mechanism: three different labels ("Refuted by calculation", "Never properly tested", "Not testable as stated"). I split it into three separable statements (see Claim 3). - Gray's impedance and flux arguments (mathematician): strong as an inconsistency flag, not a refutation by themselves; the energy bound is the decisive one (see Claim 3). - Newman: "Hastings double-counted" (physicist) is an inference, not a documented finding (see Claim 6). - Where the "10,000 builders" figure comes from (contrarian could not find it): it is on p.17, and the same page carries an image none of the reviewers read (see Claim 7).

Recomputed numbers (where reviewers differed or a verdict turns on them)

Disputed point from the Sweet record: which side of the 10.4 V / 1.84 A reading is the output?

The mathematician found that the AboveUnity notes (via the Investigator) list "E-coil 1 input 10.41 V and 1.84 A (about 19 W)" while Gray's lab notes as printed by Kelly give 10.4 V at 1.84 A as the output. I checked the book: "Input = 7.5 volts at 3.1 microamps = 23 micro watts ... Output = 10.4 volts sine wave at 1.84 amps = 19.15 watts at approx 400 Hz" (input coil 120 turns #20; output coil 12 turns #12) [read: pp.248-249, extracted text line 9129]. So in the book it is unambiguously the output, a ratio of 8.3e5 [calc: 19.15 W / 23 uW]. The AboveUnity notes may describe a different notebook (Roberts's May-June 1986 notes for Sweet's own unit) rather than Gray's; I cannot tell, and the Investigator should not treat the two as the same reading. Correction to the mathematician's flag and the research note: the "input/output swap" is not an inconsistency inside the book; it is a difference between two sources whose relation is unknown.

Claim 1. The "closed system" argument

Agreement: all three. The weak form (open systems can have COP above 1 if the source is real and identified) is Established; Kelly is right that COP is not efficiency [read: pp.12-17, 422]. The strong form (a zero-point or ambient field supplies kilowatts to pulse, resonance or magnet circuits) names no source, no flux and no coupling. All three note that Kelly's own text and Sweet's quotation concede that conservation applies to every system (mathematician, correction 9), so "closed" is a rhetorical frame, and that the Lee and Yang citation (parity violation) does not bear on energy supply.

Disagreement: only in labelling. Physicist: strong form Refuted by calculation (named reservoirs) and Never properly tested (unnamed). Mathematician: Not testable as stated, and Refuted for named sources. Contrarian: Never properly tested at device level, and Not testable as stated while the source has no magnitude.

Which is stronger: they are not rivals. A claim can only be refuted against a source that is named. For the named sources the budgets agree across reviewers [calc, with web support for the atmospheric figures]: fair-weather atmospheric electricity gives 2.6e-10 W/m^2; broadcast radio at 1 V/m gives 2.65 mW/m^2; the CMB energy density is 4.2e-14 J/m^3; Casimir energy gives a one-time 4.3e-10 J/m^2 [web for the atmospheric field and current: https://arxiv.org/pdf/0906.1280 via contrarian; https://dspace.ut.ee/bitstream/handle/10062/50334/Dolezalek_Tammet_Latham_Uman_Atmospheric_Electricity_2006_CRC.pdf via mathematician, exact pair not confirmed; conclusion robust to a factor 10]. For an unnamed source, the claim is untestable until the inventor states a flux in watts per square metre.

What would settle it: Tseung's own framing is the cleanest: a self-looped device run for hours in a calorimeter with every battery counted, or any statement of the claimed ambient flux and the collector area. Tseung's "X units supplied plus Y units led out" has no test data in the forum text [web via sites.md: Energetic Forum thread 7530].

Final verdict (split): - Weak form: Established. - Strong form, for the named reservoirs (atmospheric electricity, radio, thermal at one temperature, vacuum ground state): Refuted by calculation. - Strong form, for an unnamed or unquantified source: Not testable as stated (and not dismissed without testing: the field has not been asked for a number).

Claim 2. Sharp pulses, resonance and a tiny drive tap the field

Agreement: all three on three points. (1) The Benitez worked example is internally inconsistent and contradicted by its own energy figure (above). (2) The Don Smith / Mohamed formulas have wrong units (J/s^2 or W/s presented as watts). (3) The device claims (Smith 160 kW, Kapanadze 5 kW, Ainslie COP 17, FLEET, Bedini, Salty Citrus) have no neutral metered input and output. Mohamed's example (10 nF, 30 kV, 100 kHz) costs 0.5 C V^2 f = 450 kW to charge to deliver "30 kW" [calc: 0.5 x 1e-8 x 9e8 x 1e5]. Kelly's own lux test found the Joule Thief worse than direct LED drive (139-221 lux/W against 470-918 lux/W) [read: p.476].

Disagreement: small. The physicist adds an energy-per-spark and Q check to the Benitez example; the mathematician adds the 72-oscillations impossibility; the contrarian adds the "no instrument" quotations. The contrarian lists Newman here and the physicist and mathematician under Claim 6; I treat Newman under Claim 6. The Ainslie WO 99/38247 examples (1.086, 2.45, 0.976-1.176) include values below 1, and the mathematician shows each is within or near what unmetered phase, shunt inductance and thermal-coefficient errors can produce (example 1: 12.28 V / 0.196 A = 62.7 ohm, less than the 68 ohm resistor alone [calc]). The COP 17 headline has no protocol in the book.

A fair statement for the reader: the true core is that resonance amplifies voltage and circulating current by Q while the source still pays the resistive loss each cycle; a spark gap is a fast switch; a coil's back-EMF is L di/dt. Nothing in the files shows output above metered input. What has been absent on both sides is a wideband true-power measurement (the sampling-wattmeter method NBS used on Newman) on a Tesla-coil or pulse-charging system run for hours into a calorimeter or a timed battery discharge.

What would settle it: (a) Benitez: build or SPICE the 1918 circuit with its R, L, C at 1,000 sparks/s and integrate the battery energy over hours; the stored-energy bound predicts at most 160 W. (b) Smith: log the 12 V battery's integrated energy and measure the output in a calorimeter; 160 kW would show immediately. (c) A timed discharge of the charged batteries through a known load, which Kelly names as the only reliable test (p.544) and never reports for any device.

Final verdict (split): - Benitez GB 121,561 worked example (COP 6): Refuted by calculation and Internally inconsistent. This is a paper calculation by the patentee (an interested party); no build or test is on record. It does not by itself show that a Benitez-type circuit cannot work. - Smith / Mohamed / Kelly power formulas: Internally inconsistent (units; double counting; time constants called frequencies). - Smith 160 kW, Kapanadze 40 W to 5 kW, Ainslie COP 17, Bedini and Callanan charging, FLEET, Salty Citrus, Meyl 0.5 mW to 6 mW, Prentice 3 kW, Ace_Propulsion COP 1.8: Never properly tested. Not dismissed without testing; the proponents never supplied a metered balance and sceptics did not run one. For Prentice's 3 kW from 500 W the bound is energetic (a nameplate count of 50 lamps at 60 W). - Kelly's Joule Thief lux test: Tested, not supported (a small, honest null result). - Resonance as a voltage or current amplifier (the true core): Established.

Claim 3. Switched or "conditioned" permanent magnets give large gains

Agreement: all three. The numbers for Sweet's VTA trace to one filmed 1987 demonstration (10 V x 31.2 uA in; 112.5 V x 4.7 A out; "power factor measured to be 1"; Bearden reading a wattmeter) [web: https://www.aboveunity.com/thread/my-notes-on-floyd-sweet/, a builder's transcription], to Sweet's fund-raising letter of 26 April 1987 ("one milliwatt", "over 200 watts", "a gain of some 200,000", "up to 1,000 watts") written to "approach selected potential financial backers" [read: page image p.1, research/files/rex-sweet-sparky1987.pdf], and to a 1988 paper (10 mW in, 5 kW out) [read: A-817]. They disagree with each other by factors from 3 to 8. Annis/Eberly's "COP 200-400" is output over the switch drive power "without the coupling circuit", not a device COP [read: pp.213, 219]. Flynn's patent claims a factor of 2 in coil power for equal holding force, not COP above 1. Kawai's 3.18 is one run by the patentee (17 V x 1.15 A = 19.55 W in; 5.935 N m at 100 rpm = 62.15 W out [calc]). MEG: Kelly says both "no replication" (p.1145) and "replicated by others, COP>1" (p.1147); the only named test is Dave Lawton's null (laminated-iron version, magnet removal made no difference). Patent status: Annis/Eberly abandoned, MEG lapsed 26 Mar 2006, Flynn lapsed 3 Oct 2007; reasons not stated, so these are facts of the record and not evidence either way [web: Google Patents via sites.md].

The energy bound is the decisive result and all three reviewers reach it. A permanent magnet of the stated size holds at most about 11 J of releasable energy [calc; web: https://www.magnetexpert.com/magnetic-properties-of-ferrite-magnets for 7-30 kJ/m^3]. 500 W would drain it in 0.02 s; Gray's reported 10-12 hours at about 20 W (the book's own figure, p.249; the mathematician's 4 MJ uses 111 W) is 0.8-4 MJ, 10^5 times more. So if any output was real, the magnet is not the source and the quoted input is not the whole input. The contrarian states this fairly: the question is then "was the input fully measured".

Disagreement and my ruling. (a) Gray's impedance and flux arguments (mathematician). The numbers are correct [calc]: 7.2 V at 143 uA is 50 kohm, so L = 20 H at 400 Hz for a 250-turn coil whose air-core inductance is about 0.35 mH; peak flux for 24.2 V rms at 388 Hz in 24 turns is 0.585 mWb, B = 0.51 T over the 1.5 inch coil; the excitation supplies 0.05 ampere-turns. The mathematician reads this as 7e5 times too little magnetomotive force. I qualify this. In a reluctance-switching device the permanent magnets supply the flux and the excitation only modulates which path it takes (the principle of Flynn and of a saturable reactor). The output needs a change of flux, not 0.5 T generated by the excitation, so "7e5 missing ampere-turns" is not by itself a refutation. It is a strong diagnostic: either an unlisted high-permeability core is present, or the 143 uA is not the whole current (Gray himself writes that the signal diode "puts a small microamp current into the power coil as well as the excitation coil" [read: deep A4; contrarian]), or the AC microamp reading at 400 Hz on a distorted waveform is unreliable. The energy bound is what does the refuting. (b) Gray's own statement that "the magnetic field requires no additional power, as it is not loaded by the wires passing through the field" (p.249) contradicts the load-current-reflection argument the physicist and mathematician make: with an output current of 4.6 A, ampere-turn balance puts a reflected current in the primary of any ordinary transformer. Whether this device is exempt is the claim; it is not shown. (c) The Sweet mechanism ("negative energy from the Dirac Sea"). The physicist labels it Refuted by calculation (with a mass-energy comparison of the "weight loss of a pound"); the contrarian calls it a hypothesis in words. I split it. The statement that the magnets or the magnetic circuit supply the energy is refuted by the 11 J bound. The Dirac-sea explanation as stated is Not testable as stated: it gives no equations, no flux and no coupling, and in standard theory the filled negative-energy sea is what forbids falling into it, so it supplies no extractable energy [memory, high]. The "weight reduced by about a pound" side claim (p.507) is untested; the mass-energy comparison (4e16 J for a pound against 3.6e6 J per kWh) shows the two quantities are 10 orders apart [calc], but the claim is an anecdote relayed by "Maurice", with no measurement, and I keep it in the record with that status. "Frost on the wires when shorted" and instruments that "read zero above 1 kW" [read: p.507] likewise stay in the record as unrecorded accounts. (d) Is "Sweet never properly tested" fair? One filmed demonstration with a wattmeter, a named advocate (Bearden) and a funder (Mark Goldes, Aesop Institute, stake) exists; hyiq.org reports having tried to replicate "for a number of years now, without any success" (relayed by Weyrich via Kelly, p.247) with no data; Ashley Gray (1994) reports 111 W from 1.03 mW in a different build (about 400 Hz, unconditioned magnets, a point the contrarian flags) with witnesses unnamed. That is evidence, stated, but it is not an instrumented neutral test with the oscillator's own supply counted. It is a failed replication attempt without a protocol and one successful build without independent measurement.

What would settle it: the 1987 video with the meters in view (frame by frame, with power factor and the oscillator's supply), Roberts's May-June 1986 lab notes, Bearden's 1987 paper, Michael Watson's coil data, then a day on a bench with a true-RMS wattmeter on both sides and a battery log on the oscillator. The same bench settles Kawai, Annis/Eberly and Asaoka.

Final verdict (split): - Sweet VTA, energy supplied by the magnets or magnetic circuit: Refuted by calculation. - Sweet VTA as a measured COP (1.6e6, or 2e5 to 1.7e6 across the sources): Never properly tested. The input measurement, especially the oscillator's own supply and the second current path, is where the claim would break. Not dismissed without testing, but the demonstration record is not itself a test. - Dirac-sea mechanism: Not testable as stated. - Gray's 1994 numbers as a description of the hardware named: Internally inconsistent unless an unlisted core or a second current path exists. - Annis/Eberly COP 200-400: Not testable as stated (the COP definition excludes the coupling circuit; no whole-machine measurement). Kawai 3.18, Heins, Kunel, Tseung frame: Never properly tested (patentee or digest-level). - Asaoka (40-fold drop in drive power for equal flux): Plausible, untested as a whole-system COP above 1; the measured effect is consistent with known magnetics. Flynn (factor 2 in coil power): Plausible, untested; the patent claims no COP above 1, so Kelly's heading goes beyond the patent. - MEG: Tested, not supported, limited to two small tests (Lawton; a Universiti Sains Malaysia prototype known only from a search summary [web: https://erepo.usm.my/handle/123456789/20255]) neither using the nanocrystalline core Bearden specifies. Kelly's book contradicts itself on replication.

Claim 4. Impulse, gravity and buoyancy devices give excess mechanical energy

Agreement: all three. Gravity is conservative (closed-cycle work zero); Tseung's "2mgr" excess per pulse has no derivation (2mgr is the energy to lift the mass by a diameter [calc: 490 J for 50 kg at r = 0.5 m]; a flywheel's centre of mass returns to its height every turn); the buoyancy-generator budget is rho g h V in, rho g h V out at best [calc: 98 kJ for 1 m^3 at 10 m]; Kelly's claim that a vortex "negates the water head" (pp.400-401) conflicts with hydrostatics. Campbell's "750 W in, 2,340 W out" is a ratio of 3.1 on a video with no instrument details and "does not appear to have a heavy flywheel" [read: p.366]. Tseung COP 3.3 (text) against 3.4 (heading) is an unexplained internal conflict. Milkovic's "COP 12" is a force ("12W thrust") ratio, not an energy ratio (contrarian; mathematician agrees) [read: deep 10].

Disagreement: Skinner. The physicist did not examine the "1200 percent" ambiguity; the mathematician and contrarian did. Evidence: (1) Kelly writes that the newsreel commentator says "1200% of the input power", "which is COP=12 but it is highly likely that he should have said '1200 times'", because the commentator says one horsepower would power 3,500 homes; at COP 12 that is 2.6 W per home and at COP 1,200 it is 255 W [read: pp.405-406]. So Kelly himself shows both readings. (2) I fetched the United Press story as transcribed by rexresearch: "Skinner estimated the 'gravity power machine' would increase the horsepower of a motor about 1,200 per cent", for "a one-eighth horsepower motor" [web: https://rexresearch.com/skinner/skinner.htm]. A factor of 12 is the ordinary reading of 1,200 per cent. (3) At COP 12 a 93 W motor yields 1.12 kW, plausible for a small lathe, drill press and hacksaw in light use; at COP 1,200 it would be 112 kW [calc]. A newsreel's promotional "3,500 homes" is loose language. Ruling: the primary text supports COP 12 as Skinner's own estimate; COP 1,200 is Kelly's inference, which Kelly flags as such. A factor of 12 from a 93 W motor is still not available without an energy store or an unmetered input.

Skinner's documented record, kept in full: British Pathe "Gravity Power!", Pathetone, 31 July 1939, film ID 1190.10 [web: https://www.britishpathe.com/asset/68269/ via sites.md]; the UP story (4 April 1939) names A. P. Michaels, a Jacksonville consulting engineer, who wrote "a practical device and will have a definite field where it is necessary to use power to drive equipment" and "It should result in a very large saving of power used"; and George C. Estill, a retired president of Florida Power and Light, who examined it and whose "check was not complete enough to allow him to form a conclusion" [web, fetched today]. That is one favourable letter and one incomplete examination by named engineers, with no measurements reproduced. After 1939 the record I have is Kelly's 2017 video; nothing found on the machine's fate (see absence-of-evidence analysis).

What would settle it: a belt-driven flywheel with a metered motor input and metered generator output, run for hours with the flywheel at the same speed at the start and end, closing the loop on a stated load. The book itself says Campbell's loop closed; a neutral person could show it in an afternoon and none is on record.

Final verdict (split): - Net energy from a conservative gravitational field in a closed cycle (2mgr, "spinning chair", hydrostatic vortex): Refuted by calculation. - Campbell, Tseung 3.3, Hardy, Dmitriev, Kwok "Hidro", Ribero (never built, per Kelly): Never properly tested (videos and seller statements; Kwok has a licensing stake and his own patent text includes a solar collector). - Skinner 1939: Never properly tested. Best-documented historical case: examined by two named engineers, one favourable, one inconclusive; read as COP 12, not COP 1,200. - Milkovic "COP 12": Internally inconsistent as an energy claim (a force ratio labelled COP).

Claim 5. Water splitting and engines on water for far less than the textbook energy

Agreement: all three, and the arithmetic is the cleanest in the file. Kelly's "Faraday" baseline is energy-based: 2.34 Wh per litre of HHO, the thermoneutral figure (1.47-1.48 V per cell); the charge law gives 5,740 C per litre of HHO [calc: 2 x 96,485 / (1.5 x 22.414) = 5,740]. The reversible (Gibbs) floor is 1.229 V = 1.96 Wh/L; between 1.229 V and 1.48 V a cell can absorb heat, so up to about 120 percent of Kelly's baseline and a cold cell are textbook-possible. Implied cell voltages: Boyce DC 0.70 V; Boyce pulsed 0.125 V; Cramton 0.063 V; UK men 0.105 V. Every claim of 211 percent and above is below the floor. Inputs conflict inside the book (Cramton 12 W, 36 W, 24 + 12 W; Boyce 216/211/1,000/1,200/1,170 percent). Zigouras and the 3 L/min generator are separated from their claimed outputs by factors of 10^3-10^4 and 13-15.

Disagreement: only on whether to say "Refuted" or "Never properly tested" for the Boyce and Cramton figures. A calculation can only refute if the stated input is the whole input and the stated gas flow is the real flow. Neither is disclosed: no flow-measurement method, no true-power measurement of the pulsed supply (V x I peak is not power), and Boyce says in part that his gas comes from "a chemical reaction ... between the electrolyte used and the metal plates" [read: p.807], which would be a fuel. So the claims as printed are refuted if the numbers are right; the numbers themselves are untested.

The generator and booster claims are separate. Chemical energy of 3 L/min of HHO is 0.36-0.43 kW, against 4-5.5 kW electrical claimed; cold-water mist cannot add energy (it absorbs latent heat). Kelly's internal-combustion "steam engine" explanation has no energy account. Cramton's diesel saving of 1.4 gal/h is about 53 kW of fuel, 73 times the HHO's chemical energy [calc by mathematician; physicist says 42 to 73]. Booster fuel-economy anecdotes (Ewert, Maynard, Caggiano) were reported without protocol. A web check [web: https://en.wikipedia.org/wiki/Hydrogen_fuel_enhancement and a Jalopnik test via search; secondary] found no independent dynamometer study showing a gain on gasoline cars; Popular Mechanics/Dateline (2009) and MythBusters are reported second-hand and not read. A 2008 Murdoch University thesis on a small diesel generator reported savings of 18 percent in one low-load case but generally under 5 percent, and up to 5 percent more fuel at higher load [web: https://researchportal.murdoch.edu.au/esploro/outputs/graduate/Hydrogen-injection-into-diesel-engines-for/991005543884307891, search summary only]. That is one real measurement that was not uniformly null and I keep it in the record; it is a combustion-efficiency effect, not an energy surplus.

Meyer: the patent figures are the strongest part of the refutation. US 4,936,961 "Example 1" gives geometry and a 26 V drive but no gas volume, power or run time; WO 92/07861 claims "about 5 US gallons of water per hour" on about 1-2 mA at 360-650 V, a factor of about 60,000 below the 83.7 kW thermodynamic floor [calc]. The 1996 Ohio finding is secondary and its experts are weakly sourced: Wikipedia says the court found "gross and egregious fraud" and ordered repayment of $25,000 to two investors, and that three expert witnesses found "nothing revolutionary about the cell ... simply using conventional electrolysis", citing The Sunday Times of 1 December 1996 and The Columbus Dispatch of 8 July 2007; the expert-witness sentence carries a "who?" tag [web: https://en.wikipedia.org/wiki/Water_fuel_cell, fetched today]. The suit was brought by investors; Meyer was the defendant. The court file has not been seen. Meyer's 1998 death: Franklin County coroner and Grove City police found a cerebral aneurysm with high blood pressure, no foul play [web: same page, citing the Dispatch]; his brother's report that Meyer said "They poisoned me" is in the record on the other side [read: research note].

What would settle it: a calorimetric measurement of gas energy plus heat against true electrical input, with a mass-flow gas meter and a true-power meter on the pulsed supply, over hours. Meyer's cell at 5 gal/h needs 84 kW of input to split the water.

Final verdict (split): - Electrolysis below the Gibbs floor (Boyce DC and pulsed, Cramton, Zigouras as printed, UK men): Refuted by calculation if the stated input is the whole input; as measurements they are Never properly tested (no flow method, no true-power input). Cramton's three input figures and Zigouras's printed flow against "COP 5-10" are Internally inconsistent as presented. - Generators and engines running on 3-6 L/min of HHO at kilowatt loads: Refuted by calculation (chemical-energy bound). - Meyer's water capacitor at 5 gal/h on milliwatts: Refuted by calculation on the patent's own figures; the 1996 court examination: Tested, not supported (weakly sourced: secondary, party to a suit, court file not read). Meyer's invitations to examine in an open protocol: Never properly tested. - HHO booster fuel economy gains: Never properly tested in this book; independent tests found by the web check show no gain on gasoline cars (second-hand). - The ordinary electrolyser baseline (60 to 80 percent of ideal; a cell may absorb heat up to about 120 percent of 2.34 Wh/L): Established.

Claim 6. Aerial, passive and fuel-less devices deliver power

Agreement: all three on the budgets (above) and on the Moray record: witnessed lamp and appliance ratings (Hayes's 675 W is 100 W + 575 W by nameplate), Hayes writes "I have no instrument in my laboratory that is able to measure the amperage or the voltage at this frequency" [read: p.604], Eyring (BYU) "found no fault with the demonstration and the worst he could say was that it might be induction" and asked for a test 3-4 miles from power lines [read: p.602; web: rexresearch Moray page via sites.md], Hayes says the device was taken more than 26 miles from a power line and over 100 miles from a radio station [read: p.604]. The device is lost (smashed, shot, the "Swedish stone" dwindled) [read: pp.597, 605]. The examiner's rejection grounds are quoted: "no means was provided for causing the cathode to emit an appreciable number of electrons" and "no natural source of electric wave energy is known to the Examiner" [read: p.605]. Moray's own words, "The valves are not rectifiers ... [they] stop the flow of energy ... returning to the outer circuit" [read: p.610], conflict with the book's "rectify" wording. The germanium diode: the book says the equipment "is said to have contained" one; the earlier digest misread this (digest .24, p.596) and Marshall presents the diode only as Moray's illustration [read: deep Ch.7].

Disagreements. (a) Newman. Physicist: Tested, not supported, with "Hastings's COP 8 contains a double count". Contrarian: Tested, not supported, having read the primary NBS report's summary. I read nothing new; the contrarian's quotation stands: "At all conditions tested, the input power exceeded the output power. That is, the device did not deliver more energy than it used" and "in no case did the device's efficiency approach 100 percent" [read: NBSIR 86-3405, p.i, https://nvlpubs.nist.gov/nistpubs/Legacy/IR/nbsir86-3405.pdf, as read by the contrarian; I did not re-open it]. The court opinion records that Hastings counted mechanical work, heat and "anomalous back spikes" in the MHz range as output and that NBS made 77 measurements and found 27 to 77 percent efficiency [web: https://law.justia.com/cases/federal/district-courts/FSupp/681/16/1800217/ via sites.md]. Whether Hastings "double counted" is the physicist's inference from energy conservation (a coil returns at switch-off the energy the battery put in); it is not a documented finding, and the contrarian did not read how NBS treated the spikes. I record: NBS tested for the Patent Office (an adverse party) at the court's order; Newman's side disputed the method; the report's handling of the MHz spikes is unread. Newman's WO 83/00963 shows 12.47 W in and 12.80 W out (102 percent), V x I products of 126 V x 0.099 A and 640 V x ">20 mA", which are apparent power, and the rotating magnet's kinetic energy (about 440 J on the mathematician's assumed dimensions, flagged as assumption) would cover a 12.8 W output for 35 s [calc]. Newman's coil figure (0.5 L I^2 = 116 J for 25,700 H at 95 mA [calc]) is correct and is returned energy, not a gain. (b) Papp. Physicist: mechanism Refuted by calculation (470 MJ in 35 minutes, 13 L of petrol equivalent, from an engine whose only electrical source is its alternator). Contrarian: Never properly tested. The 300 hp figure is an assertion, so the energy bound applies only if it is taken as stated. Ruling: split (below). The 1968 event: a San Jose Mercury News account by David Ansley (27 August 1989; "A Mattel engineer died", "eight others badly wounded"; "Too many engineers say they couldn't find a hamster inside") and Feynman's own account (three injured; Papp sued him; "settled out of court, and Mr. Papp was paid something") disagree on the injuries [web: https://hoaxes.org/comments/papparticle.html, papparticle2.html; both reprints, the site has a stake as a hoax museum]. A web check [web: search summary of https://en.wikipedia.org/wiki/Josef_Papp, not opened] reports that Wikipedia places the demonstration in 1966, calls it unconfirmed in contemporary sources, and says the victim was never identified; it also says the Ansley article could not be found in Mercury News archives. I cannot verify that second-hand summary and record it as a lead. The "Authorities found no explosives" Medium snippet from the Investigator remains untraced; my search found no source mentioning explosives being found or not found at all. The 1983 "independent certification test" has no tester named [read: p.643]. (c) Atmospheric-electric aerials (Ascanius, Kljajic 96 W, Plauson 100 kW, Tesla): all agree the quasi-static field cannot supply the claimed power: a 0.48 m^2 plate intercepts about 1 pA = 1e-10 W; 96 W would need 4e11 m^2 [calc]; Plauson's 100 aerials of 100 m^2 each give 2.6e-4 W [calc]. Local mains induction, broadcast fields and earth-electrode potentials are the real candidates for any measured output and none was excluded. (d) Gritskevitch (mathematician only): the start-up figures cannot all be true (two 10 F capacitors at 100 kV store 1e11 J, not 20 kJ [calc]); the device was reportedly destroyed in January 1997 and the commission's test report (document 14-451) is not in the record [read]. Hyde's "10 kW" and "COP 10" are Kelly's headings, not in the patent text.

What would settle it: for Moray, a Faraday-cage plus spectrum-analyser plus voltage-and-current log, with a run 3-4 miles from any mains line as Eyring proposed in the 1920s; the USPTO file for serial 550,611 and Moray's notebooks. For Papp, the Roser v. Papp and Papp v. Feynman/Caltech court files and a net-energy measurement. For Newman, NBSIR 86-3405's treatment of the MHz spikes.

Final verdict (split by device): - Moray's kilowatt-level output: Never properly tested (many witnesses, ratings not meters, induction not excluded, device lost; the patent examiner's refusal was an objection to the stated mechanism, not a test). If the source is atmospheric electricity: Refuted by calculation. The "dismissed without testing" label applies to the 1931 patent-office stage specifically. - Papp 300 hp: Never properly tested by a neutral party; if the figures are taken as stated (alternator-only input), Refuted by calculation (470 MJ against nothing). A noble-gas charge has no chemical energy; plasma recombination returns at most the ionisation energy that the electrical drive supplied [memory, high]. (QE revision, Q4: this answered only the electrical-recombination step. The patent's own mechanism is a "minute fusion" of one helium atom per stroke, which is Refuted by calculation on its own numbers; see Q4. The patent contains no measured output.) - Newman: Tested, not supported (NBS 1986, party-commissioned, partly read); patent's 102 percent on V x I products: Never properly tested. - Atmospheric aerial plates delivering watts: Refuted by calculation for the fair-weather source; local-field or earth-potential explanations Never properly tested. - Don Smith (160 kW; 28.8 kVA suitcase, which a 12 V 7 Ah cell sustains for about 10 s [calc]) and Gray/Cal-Tech 10 hp: Never properly tested; Smith's formulas Internally inconsistent. - Gritskevitch COP 100: Internally inconsistent and Never properly tested. - Joe Cell, pyramids, Brock's rabbits, Drbal razor blades: Not testable as stated (operator "persona factor"; the only numbered Joe Cell case is a truck with the fuel line connected in "shandy" mode [read: claims.md]); the razor effect is testable and has a mainstream account (dehydration of the edge) [memory, medium, not web-checked; no verdict depends on it]. - Rossi (US 9,115,913), Coler, Rayburn TREC, Hyde, Eskeli cycle: Never properly tested (patent or inventor text; Eskeli's cycle table is a design calculation and not a violation of the first law as an idealised reversible cycle; Rayburn and Coler digest-level). - Compressed-air "fuel-less" engines: Refuted by calculation if closed (the compressor must be driven); Plausible, untested as a modest heat-pump-assisted engine with an external heat source (contrarian).

Claim 7. Suppression

Agreement: all three. As a physics claim it has no content; as a historical claim the book's counts are uncited. The physicist notes the physical objections already explain why no device is commercial; the contrarian and mathematician add that a thin record does not contradict suppression because the inventors worked outside institutions. Kelly is an interested party (he runs the information site).

Correction and new material (neither the reviewers nor the Investigator had it). - The "more than 10,000 Americans" figure is on p.17 (the "Basics" essay, text line 875, page image checked): "More than 10,000 Americans have produced devices or ideas for devices but none have reached commercial production due to opposition from influential people" [read: page image p.17]. The contrarian could not find it; it is there, uncited. The "40,000 confiscated patents" appears twice: p.17 (as something "people ... state") and the Suppression essay, p.998 ("I am reliably informed that in America alone, more than 40,000 free-energy patents have been confiscated by the (private company called the) Patent Office") [read: text lines 885 and 37642]. Neither is sourced. - Page 17 also reproduces an image none of the reviewers read: an extract headed "B. Subject matter of special interest in TC 2800", which Kelly describes as "an extract from a 2006 reminder to Patent Office staff in America to single out all patents which have to do with free-energy and any related subjects and take those patent applications to their supervisor to be dealt with differently". The list reads: 1 Perpetual motion machines; classes 310 and 290; 2 Anti-gravity devices; 3 Room temperature superconductivity; class 310; 4 Free energy - Tachyons, etc.; 5 Gain-Assisted Superluminal Light Propagation (faster than the speed of light); class 702, 359; 6 Other matters that violate the general laws of physics; classes 73, 290; 7 Applications containing claims to subject matter which, if issued, would generate unfavorable publicity for the USPTO, class 84, 702; 8 Reexamination proceedings involving patents in litigation ... [read: page image p.17]. The page gives no source, date or document number beyond Kelly's "2006". I searched for the heading and found nothing [web: search failed to locate it; it is not confirmed by any USPTO page I could reach]. TC 2800 is the Patent Office's electrical-technology examining group [web: https://www.uspto.gov/about-us/uspto-locations/silicon-valley-ca/tc-2800-circuits-customer-partnership-meeting-webcast, the description of the group's area]. So this is a document of unverified provenance, presented by Kelly as evidence, that is worth the reader's attention. What it shows if genuine: that applications on these subjects were flagged for special review, which fits both readings (a screen for inoperative applications under the utility requirement, which is documented US practice for perpetual motion [web: search summary via blueironip.com, secondary], and the suppression reading that Kelly gives it, in which the flag leads to a gag order). It says nothing about what happens next, and item 7 ("unfavorable publicity for the USPTO") is a flag for applications, not a statement that they are granted or refused. Kelly's "USPTO is a privately owned commercial company" is his characterisation and is not supported on the page. - Invention secrecy orders are real: about 5,445 in force (FY2013), 5,680 (FY2016) and about 5,976 (FY2021), covering all subjects [web: https://sgp.fas.org/news/secrecy/2016/10/103116.html via contrarian]; nothing found links them to energy devices, and the figure is far below 40,000. - Episodes in the record, each with its source and stake: Moray's original application file jackets "are there, the contents ... are gone" (the family's statement) [read: p.605; web: rexresearch]; Papp's investors' and partners' suits and the Caltech settlement (Feynman's account, a party to the suit); Newman v. Quigg, where the Patent Office refused his patent as inoperative and the court ordered an NBS test; Meyer's 1996 finding and 1998 death (two readings); Sweet's death in July 1995 "after two unknown visitors", from "Maurice", an unnamed associate, uncorroborated [read: A-806 to A-808]; the shooting and smashing of Moray's equipment, relayed by Kelly without date or assailant [read: p.597]. Kelly's own text undercuts the universal claim at points: the MEG patent was granted "even though it should not have been".

What would settle it: targeted record searches, not argument. USPTO file wrapper for serial 550,611; National Archives; the Roser v. Papp and Papp v. Feynman/Caltech dockets; the 1996 Ohio judgment and the Franklin County coroner's 1998 report; the original TC 2800 document (provenance and date); a count of any secrecy orders tied to energy classes.

Final verdict (split): - Global claim (inventors of working free-energy devices are suppressed by vested interests): Not testable as stated. - "40,000 confiscated patents" and "10,000 builders": Never properly tested and unsourced (no source named for either). - The individual episodes (Moray's missing file contents, Papp's suits, Meyer's death, the TC 2800 list): Never properly tested historically; the primary records are not obtained. Not dismissed; not corroborated.

Absence-of-evidence analysis (where the record goes dark)

1. Moray's 1931 application file (serial 550,611, "the contents ... are gone"). Competing explanations: (a) loss or misfiling over eighty years of a patent office's records (the family's statement is the only source and the family has a stake); (b) removal for secrecy or national-security reasons (a secrecy order would normally leave a record in the file wrapper, which has not been seen); (c) the application was abandoned and the papers discarded or transferred; (d) the contents never matched what the family says. For: the rejection grounds are quoted in the book (p.605) so the examiner saw something; Moray's only grant found is US 2,460,707 (1949, electrotherapy) [web: Google Patents via sites.md]. Against suppression: no source ties the missing contents to any order. Deciding record: the USPTO file wrapper and National Archives index for 550,611; Moray's notebooks; police or hospital records of the alleged shooting.
2. Sweet's lab notes, the 1987 video's meter views and Goldes's funding records. Competing explanations: death of the inventor (1995) and dispersal of papers; trade secrecy by the funder (Aesop Institute, later Magnetic Power Inc and Chava Energy); commercial failure; the device did not work as claimed. The notes by Darryl Roberts and the hyiq.org replication logs are named but unobtained. Deciding record: those documents, and Michael Watson's coil data.
3. Papp after 1968: a patent (US 3,680,431; later US 4,428,193, granted 31 Jan 1984), investor suits and a Caltech settlement, then silence about any measured result [web: Google Patents via sites.md]. Competing explanations: the engine was a deception that was never repeated; secrecy by investors; the litigation consumed the project. Deciding record: the suit files, TRW and NUWC records of the demonstration, and the coroner's record for the person said to have died (whose identity Wikipedia says no later account gave).
4. Meyer: 1996 finding, 1998 death. Two readings are on the record side by side (aneurysm; "They poisoned me"). Deciding record: the court judgment and coroner's report.
5. Kapanadze: two WO patents (2008/103129 and /103130, with Metin Turk) exist [web: Google Patents via sites.md]; the inventor "refuses to share the details"; no independent test appears in the Investigator's searches or in mine (hobbyist logs and promotional documents only) [web: search, 2026-10-08]. Competing explanations: trade secrecy; commercial failure; a result that does not survive measurement. Deciding record: a metered test or the patent disclosures; the two patents have not been read in this project.
6. Ainslie / Panacea COP 17: the 250-page document Kelly cites is not in the files and no independent test turned up [web: search, failed, 2026-10-08 (twice, by the physicist and the contrarian, and by me)]. Deciding record: that document and its measurement protocol.
7. Skinner after 1939: no later record in the files. Competing explanations: the machine did not scale; the wartime period (1939-45) redirected attention; the inventor died or ran out of backing; suppression. Deciding record: Florida newspapers 1939-1945 (the Breckenridge American scan at the Portal to Texas History is on the Investigator's not-reached list), the Michaels letter, Estill's papers.
8. Gritskevitch dynamo: said to have been destroyed in January 1997 "during the war"; commission document 14-451 is absent. Competing explanations: loss in war; the inventor's stated practice of "misleading information"; loss of archives; a result that was never real. Deciding record: the commission's test report.
9. Kelly's own record. It ends in February 2020 (V34.8) and the main site's last dated item is May 2020; he died on 11 July 2020 [web: FESIG memorial video https://www.youtube.com/watch?v=MnrNmV-vYO4; a members-only OverUnity Research thread, snippet only]. The end of the record is fully explained by his death. This is not an anomaly. The book's real name for the author (Edward Dickson Williams) rests on the memorial video.

Corrected claim list

The claims as they should now stand (claims.md should be updated by the orchestrator).

1. The "closed system" claim. Conservation of energy holds for every system, open or closed (Kelly's own text and Sweet's quotation concede this); an open system can have COP above 1 when an identified inflow is metered (heat pumps, solar, sailing). Kelly's strong form is that sharp-pulse, resonance and magnet circuits draw net useful power from an unspecified ambient or zero-point field. No flux, coupling or collector area is given. Verdict: weak form Established; strong form Refuted by calculation for named reservoirs, Not testable as stated for an unnamed one. [pp.12-17, 422]
2. Pulse, resonance and tiny-drive devices. The book's main examples are Don Smith, Kapanadze, Benitez, Tseung's FLEET, Bedini and Ainslie. QE revision: Chapter 2 adds the Adams motor (a forum builder's "27.6 W in, 33.78 kW out", COP 1,223; the earlier "27 W electrical output" was a misreading), RotoVerter, coil shorting (a 16 V to 440 V voltage ratio presented as power), Tewari (the patent claims only "partial nullification" of back-torque), Hatem and the UFOpolitics motors; Chapter 14 adds Kelly's own LM2587 "COP = 1.5" (800 mA at 3.33 V in, 389 mA at 10.3 V out, no meter or method stated), and Chapter 19 a South African builder's COP 45. Kelly's own true-power warning (pp.1089-1091) is kept beside them. Corrections: (a) Benitez GB 121,561's "400 W in, 2,400 W out" is a paper calculation by the patentee whose own R, L and C give Q = 0.67 (no 72 oscillations), whose spark energy is 0.16 J (160 W at 1,000 sparks/s) and whose 40 A through 15 ohm needs 24 kW; the patent family has four GB documents (17,811/1914, 5,591/1915, 14,311/1915, 121,561/1918), not three. (b) Don Smith's rule is printed as W = 0.5 C V^2 Hz (general) and W = 0.5 C V^2 Hz^2 (at resonance); both are dimensionally wrong as powers. (c) Kapanadze's "40 W start, 5 kW out" is unsourced. (d) Ainslie's measured examples are COP 1.086, 2.45 and 0.976-1.176; COP 17 has no protocol in the book.
3. Switched or conditioned permanent magnets (and magnet-only motors). QE revision: Chapter 1 adds a sub-class of motors driven by permanent magnets alone (Wang "six-month Reliability and Safety" test, Johnson US 4,151,431, Perendev, Hohl/Ascanius V-track): no output was measured by a named neutral party in the book, Kelly himself calls Perendev unconfirmed and PM-only motors "very difficult", and a closed cycle of rigid magnets does no net work (Q1). Chapter 3 adds Heins (patent test 0.29 W in, 11.51 W out, "COP of 39.6", where the printed input matches the primary winding's own I^2R loss), Klingelhoefer (COP 8 from V x I products), Markov (a voltage-ratio claim) and Barbosa/Leal (see 6). Corrections: Annis/Eberly's "COP 200-400" is output over the switch drive power "without the coupling circuit", not a device COP; Flynn's patent claims a factor of 2 in coil power for equal holding force; Kawai's 3.18 is one patentee run; MEG's replication status is contradicted inside the book (pp.1145, 1147). Sweet's VTA numbers (0.31 mW in, 500 W out, COP 1.6 million) are one 1987 filmed demonstration (10 V x 31.2 uA; 112.5 V x 4.7 A, ratio 1.69 million) plus a 1987 fund-raising letter (1 mW in, 200 W to 1 kW out) and a 1988 paper (10 mW in, 5 kW out). The conditioning procedure in Kelly's text is his reading of a YouTube video supplied by Horst Weyrich and a lab photo; Ashley Gray's build used unconditioned magnets at about 400 Hz.
4. Impulse, gravity and buoyancy devices. Skinner's "1200 percent" in the primary text is Skinner's own estimate of a factor of 12 for a 1/8 hp motor; COP 1,200 is Kelly's inference. Milkovic's "COP 12" is a force ratio. Tseung's COP is 3.3 in the text and 3.4 in the heading.
5. Water splitting and engines on water. QE revision: the appendix adds Puharich (a printed 15 percent surplus that, on his own measured 10.80 cc of H2, is 63.5 percent of the Gibbs energy and returns 138 J for 180 J in, with the gas volume 1.4 to 2.1 times what the stated current could make), Hasebe (22 times Faraday, 0.72 J/cc against a 7.05 J/cc floor, no method), Chambers (3.6 W said to run an engine, no input for the run), Spiros (a pressure loop whose "free" energy is the Nernst term the cell must also pay) and the high-mpg patents (Shelton "eight times", Schwartz "over 200%", Ogle "over 100 mpg", one-sentence claims; Hotsabi and Pearson booster reports). Kelly's "Faraday" figure is energy-based (2.34 Wh/L, thermoneutral), not the charge law (5,740 C per litre of HHO); under it Cramton's "2,340 percent" is 36 W for 6 L/min (0.063 V per cell), 7,020 percent on the 12 W of p.827. Claims above about 120 percent of 2.34 Wh/L, and all below 1.229 V per cell, fall under the thermodynamic floor.
6. Aerial, passive and fuel-less devices. QE revision: Papp's patent US 4,428,193 states the mechanism as a "minute fusion" of one helium atom per stroke in a noble-gas mixture with about 4 g of thorium-232, "100,000,000 degrees F" and a 1,000-hour charge, with no measured output (the energy of one helium atom at total mass conversion is about 1.7e11 times too small); Britt US 3,977,191 is a similar gas engine with no figures; Barbosa and Leal's COP appears as 100, 273 (22 W to 6 kW), 104 and 102.4, and Clarence's 3 kW runs from a battery, charger and inverter; Matchett's mains-rod circuits are doubted by Kelly himself. Moray's "635-675 W" are nameplate sums (635 W and "0.878 hp" = 655 W conflict inside Hayes's account); Hayes could not measure voltage or current. The germanium diode is "said to have contained", per Marshall only as an illustration. Moray's valves are "not rectifiers": one-way couplers returning energy to the outer circuit. Newman: NBS (1986, for the Patent Office) found input above output at every condition and 27-77 percent efficiency; Hastings counted MHz back spikes as output.
7. Suppression. QE revision: add Adams (told he would be killed if he shared the information, Kelly's introduction, single statement), Wang (Expo display restricted, unsourced), the compiler's "oil industry does not want functional high mpg carburettors" preface (pp.2041 ff., unsourced) and Maurice's further statements about Sweet (Senator, DOE, GM, FBI visitors, an MIT MSEE; p.2243, single source, interested associate). "More than 10,000 Americans" and "more than 40,000 confiscated patents" are on p.17 (and p.998), uncited. Page 17 also carries an image of a "Subject matter of special interest in TC 2800" list (perpetual motion, anti-gravity, free energy, "unfavorable publicity for the USPTO"), of which Kelly says it is a 2006 reminder to staff; provenance unconfirmed.
8. About the source (add to claims.md). "Patrick J. Kelly" is a pen name for Edward Dickson Williams (FESIG memorial, 2020); he died 11 July 2020. V32.9 is not the final edition (V34.5, June 2019; V34.8, 26 Feb 2020; new sections include "The Buie-Morin Power System" and "Art Porter's Magnetic device"; preface caveats unchanged). Changed figures in V34.x have not been checked.
9. Other content in the source (QE revision). The book also makes medical claims (Beck blood electrification and its HIV protocol, Lakhovsky tumour settings, suramin for autism, an Alzheimer audio, colloidal silver as a cure for cancer, malaria, AIDS and Ebola) and essays on finance, media and climate. They were inventoried, not reviewed (no medical reviewer); the book's own safety warnings (pacemakers within ten metres, lethal capacitors, the p.1408 disclaimer) are recorded in Q6.

What would settle it (single decisive experiments and records)

Open questions

1. Which source does the 10.41 V x 1.84 A "input" in the AboveUnity notes come from, and is it the same run as Gray's "output"? The two accounts cannot be reconciled from the files.
2. Does the 143 uA reading on Gray's input coil include every current path? (Gray's text describes a second branch through the signal diode.)
3. How did NBS treat Hastings's MHz back spikes, and was the heating of water by the spikes ever measured calorimetrically?
4. Is the TC 2800 extract on p.17 an authentic Patent Office document, who issued it and when, and what happens to flagged applications?
5. What does the 1996 Ohio judgment against Meyer say, and who were the three experts?
6. Why did the Annis/Eberly application, the MEG patent and Flynn's patent lapse or go abandoned (cost, loss of belief, or other)?
7. Do V34.5 and V34.8 change any figure or caveat in the book (Kelly's last two years)?
8. What measured result, if any, stands behind Ainslie's COP 17, the Kapanadze 5 kW and Tseung's "independent replication"?
9. Is Wikipedia's account of Papp (1966 date, unidentified casualty) supported by a primary source, and is Ansley's 1989 article in the Mercury News archive?
10. (Updated by QE revision.) Chapter 14 and appendix pp.1901-2238 are now deep-read except Aguero, Horvath, Eccles and Blue (pp.1938-1960 region and elsewhere) and the Meyl source for 0.5 mW in, 6 mW out, which remain unread. Further open items from the QE revision: 11. Where is the Chinese government test record for Wang's generators (2008)? 12. What did the Heins source measure as input (real power or the primary winding loss)? 13. What is the source of Puharich's 10.80 cc against a Faraday limit of 5-8 cc (current, waveform or volume)? 14. What did the Hasebe gas-volume measurement actually measure? 15. Does Spiros's stack, with its cell count and a pressure-versus-voltage run on a regulated supply, show the Nernst rise? 16. Where is the 1902 Daily Mail item on Figuera?

Corrections to earlier documents

(For the orchestrator, under CORRECTIONS ARE WELCOME. I did not edit any other file.) 1. claims.md claim 7 gives "more than 10,000 US builders" and the contrarian wrote that she or he could not find it. It is on p.17 [read: page image p.17]. Also claims.md omits the TC 2800 list image on the same page (new material above); add it to claims.md and facts.json. 2. claims.md claim 4 states Skinner "1200 percent" without the reading. Add that the primary UP text says "about 1,200 per cent" as Skinner's estimate for a 1/8 hp motor (a factor of 12) [web: https://rexresearch.com/skinner/skinner.htm] and that COP 1,200 is Kelly's inference (p.405-406). 3. claims.md claim 2 (Benitez 400 W in, 2,400 W battery power) and the physicist's and mathematician's Step 1 corrections: carry the arbiter's framing (stored energy per spark 0.16 J; maximum 160 W at 1,000 sparks/s). The mathematician's decay figure e^-7.1 per cycle is right when the damped period (47 kHz, 21 us) is used; the patent's 71.9 kHz is the undamped frequency. 4. claims.md claim 3 and facts.json: give Annis/Eberly's COP as "without the coupling circuit"; give Flynn as a factor-2 coil power claim; correct the Sweet COP to the two pairs (1.61e6 from 500 W / 0.31 mW; 1.69e6 from 528.75 W / 0.312 mW). 5. claims.md claim 5: "211-216 percent of Faraday" and "2,340 percent" are percentages of the 2.34 Wh/L thermoneutral energy, not the charge law; the physicist and mathematician agree. 6. claims.md claim 6: "635-675 W" are nameplate sums, 635 W conflicts with "0.878 hp" (655 W). 7. The physicist's statement that Hastings's COP 8 "contains a double count" is an inference from energy conservation, not a documented finding of NBS or the court; the documented fact is that Hastings counted back spikes as output and NBS did not find them to produce a gain (NBS's treatment unread). 8. The mathematician's flag that the same 10.4 V x 1.84 A is "input" in one account and "output" in another: in the book it is the output (p.248). The research note should say the AboveUnity notes may describe a different notebook. 9. The physicist's attribution of "Sweet VTA mechanism: Refuted by calculation" rests partly on a mass-energy comparison of an unmeasured anecdote; I record the mechanism as Not testable as stated and keep the magnet-supply refutation. 10. The mathematician's "7e5 missing ampere-turns" reading of Gray's coil is a diagnostic, not a refutation (flux can be switched by reluctance modulation); see Claim 3. 11. facts.json Chapter 5 Benitez list (three GB patents, 17,811 as 1915): four documents; 17,811 was filed in 1914 and published in 1915 [web: Google Patents, Investigator]. facts.json and claims.md also should note the pen name and the later editions (Investigator correction, adopted). 12. digest/PJKbook-9cc3a5.24.md p.596 (germanium diode "tested many times") misreads the text (already recorded by the deep reader).

Web checks (budget 8; used 8)

1. Skinner United Press text [web: https://rexresearch.com/skinner/skinner.htm]: confirmed "about 1,200 per cent", 1/8 hp, Michaels and Estill quotations (retagged from the Investigator's second-hand note).
2. Meyer 1996 ruling and coroner [web: https://en.wikipedia.org/wiki/Water_fuel_cell]: confirmed as secondary, with the expert-witness sentence flagged "who?" and the court file not reached.
3. Ainslie COP 17 independent test [web: search]: failed; nothing found.
4. HHO booster independent tests [web: search]: no independent dyno study showing a gain found; Popular Mechanics, MythBusters and Jalopnik reported second-hand; one Murdoch thesis mixed result on a diesel generator [web: https://researchportal.murdoch.edu.au/...]. Retags the physicist's [memory, medium].
5. Ferrite relative permeability about 1.05-1.1 and saturation 0.3-0.5 T [web: search]: not confirmed by any source found; remains [memory, medium]. The Gray ampere-turn argument, which depended on it, is already downgraded to a diagnostic, so no final verdict depends on it.
6. Kapanadze independent test [web: search]: none found; hobbyist logs and promotional material only.
7. Papp 1968 event and explosives [web: search summary of https://en.wikipedia.org/wiki/Josef_Papp, https://hoaxes.org/comments/papparticle.html, papparticle2.html]: no source on explosives; Wikipedia's 1966 date and unidentified victim reported second-hand.
8. TC 2800 "Subject matter of special interest" list [web: search]: not found on any USPTO page; the group's electrical area confirmed; perpetual-motion handling as an inoperability/utility matter per a secondary summary. Earlier reviewers' web checks (ferrite energy product, fair-weather field and current density, NBS Newman statement, secrecy orders) I accept as retagged [web] as they report them. Not web-checked and not needed for a verdict: NE555 maximum frequency, the Navy seawater-fuel background, razor-blade dehydration, cosmic-ray flux.

Final verdicts in brief

QE revision (2026-10-09)

Why this section exists. The QE (qe.md, verdict MINOR) found that about 40 percent of the book had been sampled, not read, by the fact-finder, and listed eight remediation deep reads. Those deep reads are now in deep/PJKbook-9cc3a5.md: Chapters 1-2 (pp.42-146), Chapter 3 (pp.147-211), Chapter 14 (pp.1151-1277, Kelly's LM2587 measurement and the heaters), the medical pages (pp.1060-1071, 1203, 1210-1212, 1375-1417), the high-mpg patents and booster reports (pp.2041-2087, 2280-2313), Spiros (pp.2003-2037), Puharich, Hasebe and Chambers (pp.1901-1960) and the Papp and Britt patents (pp.2196-2238). I read all of those sections in full, then re-did the arithmetic myself (.venv Python; results tagged [calc]). The three reviewers did not see this material, so the rulings below are mine alone and rest on the arbiter's own checks and the deep readers' page-image reads. Where a verdict below changes an earlier one, I say so. Where I only add a verdict for material nobody had judged, I say "new".

Rules I applied: the same measuring stick as above (energy per event times rate; the Gibbs floor; closed-cycle work of conservative forces; chemical-energy bounds); a claim is called refuted only when accepted laws, applied to the numbers as printed, rule it out, and the assumption that makes it so is stated; interested parties are named; absence from a source is not contradiction.

Web checks for this revision (4 further calls, beyond the 8 above, because the new material contained dated facts a verdict could lean on). (1) Shelton US 2,982,528: Google Patents shows filed 13 June 1958 and published 2 May 1961, inventor Robert S Shelton [web: https://patents.google.com/patent/US2982528A/en]. The book prints "2nd May 1940"; that is a transcription slip (the deep reader's suspicion confirmed). (2) Chambers (Xogen Power Inc) US 6,126,794: filed 26 June 1998, published 3 October 2000 [web: https://patents.google.com/patent/US6126794A/en]; Kelly's "16th July 2002" is wrong. (3) Johnson US 4,151,431: filed 6 December 1973, published 24 April 1979, applicant "Individual" [web: https://patents.google.com/patent/US4151431A/en]; Kelly's date is right. (4) Papp US 4,428,193, claim 1 (from the copy saved by the Investigator, research/pages/gp-US4428193A.html): claim 1 is an engine with a closed chamber and "means, including a plurality of electrodes ... for ionizing the working fluid, said working fluid consisting essentially of an inert gas mixture of helium, neon, xenon, krypton and argon, said argon constituting approximately 17%"; I read only claim 1 and the start of claim 4, so I do not state what the other claims say [web]. The claim does not say "fusion"; the fusion statement is in the specification (p.2219 of the book). Not web-checked, and no verdict depends on them: ortho/para hydrogen combustion (Chambers), argyria (medical paragraph), Th-232 half-life [memory, high].

Q1. Permanent-magnet-only motors (Chapter 1): new sub-class under Claim 3

What the book says (all [read], deep read pp.42-89). Kelly's premise is that "the power of a magnet flows from the environment and not from the magnet" and that a magnetising pulse "polarises the quantum environment" (p.42). No measurement supports it. The examples: - ShenHe Wang (China), 5 kW colloidal-magnet generator (pp.42-44): "Two of Mr Wang's 5 kilowatt generators successfully completed the Chinese government's mandatory six-month 'Reliability and Safety' testing programme in April 2008", and a consortium is said to be buying coal-fired stations to refit them. The source cited is Kelly's own site (Wang.pdf). No tester, report, input, or measured output is named, and "Reliability and Safety" testing does not by its name measure output. The Expo story (the government allowed only a wristwatch-size version) is likewise unsourced. This is one statement, single-source, from an advocate; it stays in the record as a lead for the Investigator (Chinese test records), not as a test result. - Howard Johnson, US 4,151,431 (granted 24 April 1979 [web], demonstrated for Science and Mechanics, Spring 1980): no output figure anywhere in the chapter; Kelly himself later writes that Johnson had difficulty replicating it (p.79), and Diduck's 2007 patent background says of it "in most such devices no working models have been achieved" (p.83; Diduck is a competing applicant). - Perendev (Mike Brady): "It is said that dozens ... have been made and sold ... with an output of not less than 100 kilowatts. As far as I am aware, this has not been confirmed, nor have there been independent tests ... other than a brief test by Sterling Allan" (p.78). By mid-2010 the backers were "most unhappy" (p.79). Kelly's own caveats are kept: PM-only motors are "very difficult" to make work, he recommends designs with moving stator magnets or shielding instead, and he reports repulsion-mode magnets that "lost their magnetisation" after about three months (p.82). - Hohl / Ascanius V-track (pp.71-72): the "replication" that "did work" in one hour is a magnet stack rolling along a V-track, not a self-running drum; no speed, force or run time. Kelly: "it is VERY hard to change a straight magnetic track into one which forms a complete circle" (p.75). - Others, with no data: Ecklin, Tracy, Teal, Jines, Kundel, Flynn, Steorn, "Mr Sung" 20 kW and a 225 hp machine (pp.45-68); the Diduck application (pp.82-89, "currently operating models" claimed in a patent, units mixed: ft-lb for a force, gauss for a force).

The calculation (new). For rigid permanent magnets with fixed magnetisation, the magnetostatic force on a moving body is the gradient of a potential energy, so it is conservative: round a closed cycle the net work is zero [memory, high; this is the same argument as gravity in Claim 4]. Any friction, eddy loss or demagnetisation makes the cycle lossy, not gainful. A motor whose only energy source is its magnets can therefore run only until the magnets' releasable energy is spent, and that is small: a 10 mm neodymium cube with (BH)max about 300 kJ/m^3 holds at most about 0.3 J [calc: 300e3 x 1e-6; (BH)max value memory, high]. Designs that move or gate magnets with a second (electric or mechanical) drive move the question to that drive's energy, which is what Kelly himself says when he points readers to moving-stator and shielded designs.

Verdicts (new): - Continuous rotation driven by rigid permanent magnets alone, closed cycle: Refuted by calculation. - Wang 5 kW with a "six-month Reliability and Safety" test: Never properly tested (a test is asserted, with no tester, data or measured quantity; not dismissed without testing). Perendev 100 kW: Never properly tested (one brief test by an advocate, per Kelly). Johnson: Never properly tested (patent granted; no output stated). Hohl V-track as a motor: Never properly tested (the replication shows the track, not a rotor). - Kelly's own caveats on Perendev, Johnson and PM-only motors: kept in the record beside the claims.

Q2. Large COP figures from Chapters 2-3 (Claims 2 and 3): additions and one correction

All [read] from the deep reads; the arithmetic is mine unless stated.

Adams motor (Chapter 2). - Correction to facts.json confirmed: p.93 reads "an Adams Motor builder with a forum ID of 'Maimariati' who achieved a Coefficient Of Performance of 1,223 ... His input power is 27.6 watts and the output power is 33.78 kilowatts" [calc: 33,780 / 27.6 = 1,224]. The Introduction (PDF p.28, not p.29) gives "an electrical input of 27 watts and a 32 kilowatt output" for "one replication", almost certainly the same report at a lower precision (the text does not say so). The "at least 800% efficient" headline (p.28, p.90) has no measurement behind it, and the Introduction adds, uncorroborated, that Adams "was told that if he shared the information, he would be killed" (single statement by Kelly; entered under Claim 7 as an episode in the record). - Kelly's own caveat, kept: "If a motor is built like this [the basic diagram], then it will most certainly work but it will never reach 100% efficiency ... It is only with a specific configuration which is hardly ever publicised that high performance figures can be achieved" (p.90). The 95 percent back-EMF return diagram is of unknown origin (p.98). - Bound: 33.78 kW is 45 hp. A rotor of a few tens of 10 mm magnets (the book's worked example uses a 10 mm magnet, p.91; the builder's rotor is not described) stores of order 1-30 J [calc, as in Q1]; 33.78 kJ per second is 1,000 times that every second. So if the output were real and sustained, the magnets are not the source and the quoted 27.6 W is not the whole input (the same logic as the Sweet energy bound). The figure comes from a pseudonymous forum builder with no instrument, method, load, duration or second observer. - Verdict (new): Adams COP 1,223 and the 800 percent headline: Never properly tested. "Magnets (or the 'local environment') supply the surplus" as the explanation of a sustained 33 kW: Refuted by calculation (energy bound). Adams-type design rules (duty cycle, coil ratios): not claims about energy; no verdict.

Heins bi-toroid transformer (Chapter 3; patent figures read from page images p.150). The patent test as quoted by Kelly: primary 2.5 ohm winding "carrying 0.29 watts"; secondary 1 2.9 ohm, 0.18 W; load 1 180 ohm, 11.25 W; secondary 2 2.5 ohm, 0.06 W; load 2 1 ohm, 0.02 W; "the input power was 0.29 watts and the output power 11.51 watts, which is a COP of 39.6" (11.51 / 0.29 = 39.7 [calc]; Kelly truncates). Two things are new. (a) The "output" of 11.51 W includes the secondary windings' own resistive loss (0.18 + 0.06 W). That is a minor point. (b) Each listed power equals the current squared times the resistance of the element it sits on, at a consistent current. Coil 1 and load 1 are in series: sqrt(0.18 / 2.9) = 0.249 A and sqrt(11.25 / 180) = 0.250 A [calc]. The 0.29 W in the 2.5 ohm primary corresponds to 0.34 A [calc]. That pattern is what one would see if the "input" is the heat in the primary winding. If so it is not the input power: a source that sustains 11.5 W in the secondaries must supply at least that through the primary's coupled flux, and the book does not give the source voltage, frequency, phase or instrument. This is my inference from the printed numbers, not a statement in the patent; Kelly says only that the primary "should be driven at its resonant frequency", and at resonance the reactive current is large and a V times I product is not power. A second figure, 106.9 mW in and 403.3 mW out (ratio 3.77, p.151), has no stated source. Kelly's lead-in says "thirty times" (p.149), against 39.6 on p.150; he does not reconcile them. The patent is the patentee's document (interested party). Verdicts (new, replacing the digest-level "weaker" ones): the COP 39.6 as stated: Not testable as stated (the input quantity is not defined, and the printed numbers match the primary's winding loss); the bi-toroid device as a transformer with reduced back-flux: Never properly tested (no neutral measurement of real input power). Nothing in the files shows it cannot work as an ordinary, possibly well-designed, transformer.

Other Chapter 3 devices (new). - Klingelhoefer "Gabriel Device" (pp.169-170): 60 W in (0.5 A x 120 V) and 480 W out (4 A x 120 V), COP 8. These are RMS volts times RMS amps with no power factor or instrument: Never properly tested. - Tseung magnetic frame: "COP = 1.5" after a replicator cut three magnets to two (p.149): a second-hand anecdote, replicator unnamed, no values: Never properly tested. Kelly's mode (3) (magnet flux reinforced by the pulse) and the saturation caveat are his reading. - Markov EP 0844626 (pp.152-164): the patent claims voltage transformation ratios of 5 to 11 (examples) and "10 to 20 times" with one primary, not a power gain; Table 1 (the measurements) is not reproduced in the book. Markov's own statements (a transformer "20 to 30 times lighter", operation 10 kHz to 40 MHz) are those of the inventor and company head. As a power claim: Not testable as stated; as a voltage-ratio claim: Never properly tested (a parametric, time-varying-permeability transformer is not excluded by anything in the files, and the patent gives no power balance). - "Jack Noskills" resonant transformer (pp.164-169): bulbs dark in the primary, half bright in the secondaries; "I have no meters": Never properly tested. A dim 5 W filament is not zero input. - Alexkor zero-back-EMF toroids (pp.174-179): the claim that the input current does not change when the output is shorted contradicts ampere-turn balance in an ordinary transformer, so the claim is that the device is not an ordinary transformer; no measurement is given: Never properly tested. - Figuera (pp.170-173): the single contemporary item is a 1902 Daily Mail report as quoted by Kelly ("550 volts" and "a 20 horse-power motor", about 15 kW [calc: 20 x 746 W]), with no date or page; everything else is Kelly's reconstruction: Never properly tested (newspaper lead, not obtained). - Mains-rod circuits (Matchett, Ascanius, pp.204-211): Ascanius reports 2.6 V with the fuse out, 129 V (two rods) to 162 V (five rods) with the fuse in, and 61 V under a "7 W LED array" (p.207). Kelly's own doubts are kept in full: "I'm not at all convinced that the circuit shown above does actually draw net power from the mains. The mains meter charges you for power assessed by multiplying the average voltage by the average current" (p.211), and he does not recommend the circuits. Calculation (new): the rod is a capacitor plate at mains potential. A coil of 0.405 mm wire over 610 mm of 4.76 mm rod with thin insulation has a capacitance of order 250-560 pF (assuming a relative permittivity of 3 and a gap of 20 to 50 percent of the rod radius) and so can pass only about 15 to 33 microamps at 155 V peak and 60 Hz [calc]. A real 7 W load at 61 V needs about 0.11 A, which would take about 2 microfarads of coupling [calc: 0.115 A / (2 pi x 60 x 155 V)]. So either the array draws far less than its 7 W nameplate at 61 V (no wattage was measured), or the energy comes by a path other than stray capacitance; the book does not say which. Verdict (new): Never properly tested; the "7 W" is a nameplate and not a measurement. - Barbosa and Leal, WO 2013/104041-43 (pp.179-211): the figures conflict. The patent says "at least 1 to 100" (COP 100 minimum, p.192); the video is "22 watts producing 6 kilowatts" (p.182), a ratio of 273 [calc]; the "169 kilowatts of power flowing from the ground" is "104 times" (p.211) and "COP=102.4" in the p.37 table, with no date, place, witness or instrument. The mechanism ("an endless supply of electrons is drawn from the earth", current circulating "with or without voltage") names no source of electromotive force; the earth is an equipotential, and power is volts times amps, so a current with no voltage across it carries none. The "Clarence" 3 kW replication (pp.198-204) is a forum account relayed through Kelly, built from a battery, a 20 A charger, a 3,000 W inverter and many ground rods that Clarence says are essential; no input or run time is given, and with a battery-charger-inverter chain the energy source cannot be told from the text. Kelly states the difference from mains-rod circuits (the Barbosa/Leal test used "a battery-driven inverter and not the mains") but the battery is part of the circuit. Patentees are interested parties; the claim that the patents contain "misdirection" is Kelly's and Clarence's. Verdicts (new): the COP figures (100 / 273 / 104 / 102.4 for one family): Internally inconsistent as a set; "electrons drawn from the ground" as an energy source: Not testable as stated (no EMF named; if the source is a named reservoir, the Claim 1 budgets apply); the 169 kW and the Clarence 3 kW: Never properly tested.

Other Chapter 2 devices (new). - Tewari US 2015/0084467 A1: the patent text claims only "a partial nullification of the back torque reaction"; the COP 2 to 3 is Kelly's (p.101), and Kelly himself notes Tewari "does not mention anything about 20 kilowatts of self-powered output" (p.102): Never properly tested (the COP has no source; the patent claims an efficiency improvement). - Kawai (p.115, from the patent): 17 V x 1.15 A = 19.55 W in; 60.52 kg-cm at 100 rpm = 5.935 N m x 10.47 rad/s = 62.15 W out; COP 3.18 [calc, agrees]. How torque was measured and whether 17 V x 1.15 A is the average power of a pulsed input are not stated: Never properly tested (unchanged; now read in full). - RotoVerter: "at least 10 times" output over input, "reproduced by several independent researchers" (unnamed); the one conventional figure in the range, Chorister's "a one-horsepower motor will output about three-quarters of a horsepower", is below unity (p.123). Kelly warns against self-powering it (p.119): Never properly tested. - Coil shorting (pp.123-128): "can boost the output power by an estimated factor of 100". The measurement is of open-circuit pulse voltage, 16 V to 440 V (27.5 times) and an expected 1,600 V (100 times), on a Bedini Window Motor replica "just spun briefly by hand" (p.127). Voltage is not power, and an open circuit delivers none; a factor of 100 in voltage across a resistive load would give 10,000 in power, so Kelly's "100 times" is a voltage ratio relabelled as power: Never properly tested, and the power label Internally inconsistent. The energy released at switch-off is the energy the coil already held [calc: 0.5 L I^2, as for Newman]; no rotor torque or input power is reported with and without shorting. - Hatem (pp.128-129): "an energy gain ... of 20 times", one video, "144 watts of excess power"; the dates conflict ("1955" and FR 2,826,800 of January 2003): Never properly tested. - UFOpolitics rewound motors (pp.137-146): the evidence is terminal voltages of lead-acid batteries over minutes (e.g. 12.24 V rising 0.01 V per minute); a few hundredths of a volt after a load is removed reflects recovery, not energy added: Never properly tested. Kelly: "I have to admit that motor windings and operation tend to confuse me" (p.144), kept. - Infinity SAV, Tong wheel, Hardy pump (pp.116-118, 130-146): video-only, no numbers beyond a ratio: Never properly tested. - The South African builder's pulse generator (p.1369; deep-read earlier, not re-read by me): 0.89 W in (70 mA at 12.7 V), "40 watts or close to it", COP 45, three weeks continuous; the 40 W may be a bulb rating and the output battery chain is not stated: Never properly tested.

Q3. Kelly's own measurement and the book's true-power caveat (Claim 2)

Kelly's LM2587 converter, "COP = 1.5" (p.1259; deep read, page image checked). A £3.17 eBay boost module, October 2015, measured with three NiMH cells: "an input of 800 mA at 3.33 volts (2.664W) driving the 10-LED array displays with 389 mA at 10.3 volts (4.01W) which is COP=1.5" (4.007 / 2.664 = 1.504 [calc]). No meter type, no steady-state statement, no averaging of the pulsed battery current, no repeat. A boost converter is normally 80 to 95 percent efficient [memory, high], so a 90 percent converter would deliver 2.40 W, i.e. 233 mA at 10.3 V [calc]; the reported 389 mA is 67 percent more. Rounding cannot explain that. With a battery as the only source, a true mean DC input and output of those values would violate energy conservation, so the figures point to a measurement artefact (a switching converter draws pulsed current; a DC ammeter's reading of it and of the cell voltage under pulsing is unreliable) or to a slip, and the book gives no way to tell. It is the only COP above 1 in the book that Kelly says he measured himself, and he asserts it without the qualifications he gives elsewhere (below). The page also says the same module charges ten NiMH cells from 10.5 V to 13 V in 30 minutes "at 50 mA current draw"; 50 mA for half an hour is 25 mAh, far too little for ten cells [calc; typical NiMH capacity memory, medium], so that sentence is inconsistent as written or the 50 mA is not the input. Verdict (new): Kelly's LM2587 COP 1.5 as a measurement: Never properly tested (method unstated); as a power ratio between true mean DC input and output: Refuted by calculation (a battery-fed converter cannot exceed its source); the decisive check is cheap (a series shunt and a recording instrument on the battery, or a timed battery discharge, which Kelly names as the reliable method on p.544).

Kelly's own warning, pp.1089-1091 (Chapter 12; digest-level, quoted by the QE and not re-read by me). Volts times amps is not power for non-sine or reactive loads, a worked example gives "COP = 1.56" that "may be wrong", and true power needs a sampling wattmeter [read: qe.md section 3C, from digest 47]. Beside p.1259 this shows that the book contains the correct warning and a case where its author did not apply it. Both are kept; the primer should quote both.

Other Chapter 14 cases (new, from the deep read). - Perkins and Frenette heaters (pp.1161-1164): "some measurements indicate that it is at least 100% efficient and some people believe that it is well over the 100% efficiency, though they don't want [to] get drawn into long discussions on methods of measurement." A shaft-to-heat device is by construction near 100 percent of its shaft input; "at least 100%" is the ordinary resistive-heater value, so Kelly himself is not claiming a gain. The "well over" view is attributed to unnamed people, with no data. Verdict: efficiency near 100 percent of shaft input: Established; "well over 100%" relative to the electrical input: Never properly tested. - Kim's 30 L/min HHO heater (pp.1157-1159): Kelly's own check gives 4.2 kW of electricity for 30 L/min at Faraday efficiency and argues the stated flow is overstated; the gas carries at most about 3.3 kW of heat [calc]. No over-unity claim is made for it in this range, and none is credited. - Ram pump "COP of infinity" (p.1217): the input is a falling stream, about 2 W for 3 L/min dropping 4 m [calc]; the delivered output is 0.8-1.2 W, a ratio of 0.4-0.6 [calc]. "Infinity" means no purchased input; Kelly traces the energy to the sun himself. Established as ordinary hydraulics. Kelly's formula D = S x F x E / L lacks a factor of 1,440 minutes per day [calc]. - Bhargava bicycle (pp.1260-1262): "one hour of pedalling can produce 24 hours of (low-power) electricity" is 100 Wh spread over 24 h, about 4.2 W [calc], which is ordinary energy shifting. Kelly's remark that the flywheel "causes an energy gain" is Kelly's; a flywheel stores and returns energy and creates none: Refuted by calculation for the flywheel-gain statement, Established (ordinary) for the energy-shifting reading.

Claim 2 as it now stands. Added to the Never properly tested list: Adams, RotoVerter, coil shorting (voltage ratio), Tewari, Hatem, UFOpolitics, the South African COP 45, Kelly's own LM2587 (with the bound above). Added to Internally inconsistent: the "100 times power" for coil shorting. The Joule Thief lux test stays the only Tested, not supported item, now beside Kelly's LM2587 figure, which he did not test the same way.

Q4. Papp: the mechanism from the patent (Claim 6; replaces a statement in the Claim 6 text above)

What the arbiter wrote above ("plasma recombination returns at most the ionisation energy that the electrical drive supplied [memory, high]") answers the electrical-recombination part of Papp's account only. The patent (book pp.2196-2228) states a different mechanism, and the earlier text did not address it. The deep read gives it in the patent's own words; the patent hedges nearly every physical statement with "it is calculated" and shows no calculation. - Gas: 35.6 % helium, 26.3 % neon, 16.9 % argon, 12.7 % krypton, 8.5 % xenon by volume, about 100 cc at one atmosphere in the smallest cylinder (p.2218). The ratio is said to equalise "the total charge for each of the gases"; the calculation is not given. - Electrodes: about 4 g of thorium-232 in the cathode; the anode holds about 2 g of "rubidium-37" and 3 g of "phosphorus-15" in mineral oil (the numbers are atomic numbers written where mass numbers belong; the text means ordinary rubidium and phosphorus) (p.2214). Three coils of about 100 mH per cylinder, capacitors of 1.3 and 2.2 microfarads, 40,000 V pulses three times per cycle, a 24 V battery, and a high-frequency source (27.12 MHz at 3.4 to 8.4 A, voltage not stated, so its power cannot be computed) (pp.2213-2217). - The energy claim (p.2219, quoted): "an ignition discharge occurs in which the helium splits into hydrogen in a volume not larger than 2 or 3 x 10^-6 cubic millimetres at a temperature of approximately 100,000,000 degrees F ... there is a minute fusion reaction in the helium consisting of the energy conversion of a single helium atom, which releases sufficient energy to drive the piston ... with a force similar in magnitude to that generated in a cylinder of a conventional internal combustion engine." And (p.2220): "a cylinder can run at 1200 rpm approximately 1000 hours, if not more, on a single charge of gas." - Output: none. The pages contain no measured power, torque, rpm, temperature, radiation reading or run time. The 300 hp test, the 1968 detonation and the 1983 "certification" are Kelly's chapter text (p.643), not the patent. Kelly reproduced the specification without the numbered claims; claim 1, which I read on Google Patents [web], claims the engine with the ionised noble-gas mixture and does not use the word fusion.

Calculations [calc], using standard constants [memory, high]. 1. Even converting the entire rest mass of one helium-4 atom (6.65e-27 kg) to energy gives E = m c^2 = 6.0e-10 J (3.7 GeV). A conventional 100 cc cylinder stroke delivers of order 100 J (10 bar x 100 cc). The stated energy source is about 1.7e11 times too small, even at total mass conversion. 2. Splitting helium-4 into hydrogen absorbs energy (the helium-4 binding energy is about 28 MeV); it does not release it. Fusion of hydrogen into helium releases energy, but the gas mixture contains no hydrogen to fuse. 3. The radioactive material cannot be the fuel: 4 g of thorium-232 has an activity of about 1.6e4 Bq, an alpha power of about 1e-8 W, and about 1e-7 W with all its decay products in equilibrium [calc; half-life 1.4e10 years, memory, high]. A 300 hp output is 224 kW. 4. Chemical energy: noble gases have none; recombination of ionised atoms returns at most the ionisation energy the discharge put in (the earlier statement, which still stands for that part). 5. Time: 1,000 hours at 1,200 rpm is 7.2e7 revolutions; the charge holds about 9e20 helium atoms, so "one atom per stroke" lasts far longer than 1,000 hours, but only if one atom could drive a piston, which item 1 excludes. The patent's own 0.002 % duty (10^-6 s once per revolution) is arithmetically consistent [calc: 1e-6 / 0.05]. 6. Electrical input from the mixer source is internally unreconciled: "120 V AC, 60 Hz, 8.4 amp, 560 watt" (p.2222) is 1,008 VA; "27,120 MHz" is printed for what must be 27.120 MHz (a probable unit slip, flagged by the deep reader).

Verdicts (changed/new): - The mechanism as the patent states it (a minute fusion releasing the energy of one helium atom per stroke, in a mixture of noble gases with a few grams of thorium): Refuted by calculation (energy per event is about 11 orders of magnitude short; the proposed reaction is endothermic; the radioactive content supplies about 1e-8 W). This replaces the earlier "plasma recombination returns at most the ionisation energy" as the main argument; that argument stays as the answer to the electrical-recombination step. - The engine as a working 300 hp device (Volvo test, 1968 and 1983 events): Never properly tested by a neutral party, as before; and, if the 300 hp is taken as stated with only an alternator as the source, Refuted by calculation (unchanged). - The patent is a description of an intended design with no measurement; a granted patent is not a test. The patentee is an interested party. The inconsistency "300 HP for 40 min" (p.37 table) against 35 minutes (p.642) is added to the list of internal conflicts. - Britt US 3,977,191 (book pp.2229-2238, new): a six-cylinder "AEROPS" engine, a sealed mirror-walled sphere with 40,000 V and 500 V discharges into a gas of "xenon, krypton, helium and mercury", the photons said to cause a "chain reaction" that expands orbits by "centrifugal force". No proportions, power, or test. As a gas engine heated by an electrical discharge it can do work only from the energy the discharge deposits; the patent does not quantify any surplus. Verdict (new): Never properly tested; the "orbit expansion by centrifugal force" mechanism: Not testable as stated.

Q5. Water splitting and fuel economy (Claim 5): additions

Puharich, article and US 4,394,230 (pp.1902-1931; equations read from page images). Facts [read]: 0.9 % saline (0.154 M), 600 +/- 5 Hz carrier; 25 to 38 mA rms at 4.0 to 2.6 V rms, 98.8 to 100 mW for 30 minutes (180 J); "since we were interested in following the voltage and current separately, it was decided not to use the watt meter"; measured mean 10.80 cc H2 and 5.40 cc O2 by gas chromatography and mass spectrometry in two laboratories (Gollob Analytical Services, Berkeley Heights NJ, named in the article; the second not named), standard error +/- 10 %. The article claims 91.3 percent efficiency and a "surplus of approximately 15 percent" (26.85 J). I confirm the QE recomputation [calc]: - 10.80 cc H2 is 0.482 mmol of water split. At the Gibbs energy of 237.1 kJ/mol that took 114 J, so the Gibbs efficiency is 63.5 % (58 % if the gas volumes were at 25 C rather than 0 C, since the article uses both reference states). That is inside the 50 to 71 % range the article itself gives for conventional DC electrolysers (p.1908). - Burnt back, the gas returns 0.482 mmol x 285.8 kJ/mol = 137.7 J against the 180 J debit: a net loss of 42 J before any conversion loss. The printed 206.85 J credit is 1.50 times that. - The slip: 10.578 J/cc is the Gibbs energy per cc of H2 (22,400 cc per mole), but the article applies it per cc of mixed gas; likewise 12.77 J/cc is the enthalpy per cc of H2, applied to 16.20 cc of mixed gas. The article's own Faraday section a page later uses the correct 33,621 cc of mixed gas per mole of water, so the two parts disagree. The printed 91.3 %, 114.9 % and +26.85 J all carry the same factor of 1.5. - The textbook root of the "20 percent surplus" is the enthalpy-to-Gibbs ratio, 285.8 / 237.1 = 1.205. The extra 48.7 kJ/mol is heat that an ideal cell absorbs from its surroundings while splitting water; a fuel cell returns at most the Gibbs energy as work, and the heat comes back as heat. The patent quotes Penner and Icerman saying this itself ("with heat absorption from the surroundings ... energy-conversion efficiencies during electrolysis greater than unity", p.1928). So the surplus exists only if the heat is counted as work. This is the same "up to about 120 percent of 2.34 Wh/L" window recorded in Claim 5, now seen as the origin of the printed surplus. New point from the deep read (not in any earlier document) [calc]: by the article's own Faraday table, 25 mA for 30 minutes can make at most 5.23 cc of H2 and 38 mA at most 7.94 cc, but the reported mean is 10.80 cc, which is 2.07 and 1.36 times the charge limit. So the stated current, duration and gas volume cannot all be right: the gas was made at higher current than quoted (the table's first row, 100 mA, allows 20.9 cc), the rms ammeter under-read the charge in a non-sinusoidal waveform (the patent reports half-wave rectified, unipolar pulses), the volumes are not 30-minute totals, or the yield is genuinely super-Faradaic, which the article does not claim. The text cannot decide between these. Also, volts times amps from separate rms meters is apparent power, not real power, and at resonance the article says the net power is "reactive"; the true input could be lower (raising every efficiency) but it is not measured. The patent's pure-water Stage D gives 0.1 cc H2 per minute at 11.2 mW rising to 100 mW (p.1926): the Gibbs power of that gas is about 17 mW [calc], above the starting input and below the later one, and no measurement method is given. The motor-home story ("hundreds of thousands of miles" on water, p.1902) is Kelly's unsourced statement; neither the article nor the patent mentions a vehicle. It stays in the record, as one statement from one source. Verdicts (new): the printed 91.3 percent, 114.9 percent and 15 percent surplus: Internally inconsistent (the article's two sections use different gas-volume bases) and, on the measured gas, Refuted by calculation as a net surplus (138 J back for 180 J in; 114 J of work for 180 J). The gas yield itself: a measured Gibbs efficiency of 63 percent is an Established, ordinary result; the conflict with the Faraday limit for the stated current is unresolved. The motor-home claim: Never properly tested. The two-laboratory gas analysis is the most careful measurement in the water section of the book and is credited as such.

Hasebe US 4,105,528 (pp.1932-1937; table read from page image p.1936). 2.8 V, 30 A (84 W, pump not included), 116 cc/s of gas, "hydrogen 2.6 cc per coulomb", "over 20 times" the Faraday rate. Faraday gives 0.116 cc of H2 per coulomb at STP, so 2.6 cc/C is 22.4 times [calc: 22,414 / (2 x 96,485)] (20.3 times at the table's own 25 C and 1003 mbar). Energy: 84 W / 116 cc/s = 0.72 J per cc of mixed gas against a Gibbs floor of 7.05 J/cc at STP (6.40 J/cc at 25 C), so the claim is 9 to 10 times below the minimum work; 116 cc/s needs at least about 820 W [calc]. The voltage and current are ordinary (2.8 V at high current density), so the anomaly is entirely in the gas volume. The excerpt gives no measurement method, no gas dryness, no composition check, no run time; the per-coulomb rows equal the total rate divided by 30 A and split 2:1, so they look derived, not measured [inference from the arithmetic]. Competing readings: entrained liquid, foam, vapour or air counted as gas in a circulating gas-liquid loop; a units slip (cc/min for cc/s, which would put the rate at 37 % of Faraday, but needs two errors); or a genuine anomaly. The patent is the patentee's document; no independent test is cited. Verdict (new): Refuted by calculation if 116 cc/s of dry 2:1 gas was produced from 84 W; the measurement itself: Never properly tested (no method).

Chambers (Xogen) US 6,126,794 (granted 3 October 2000 [web]; the book's "16th July 2002" is wrong). 12 V at 300 mA (3.6 W) pulsed, 10:1 mark-space, 10-250 kHz; output "1 p.s.i. per minute"; "a small one cylinder gasoline engine ... has thus successfully run ... without any gasoline" (p.1952). The only output measurement is a pressure gauge; the head-space volume is not given. At the Gibbs minimum 3.6 W makes at most about 31 cc of mixed gas per minute, which a 0.068 atm/min rise reproduces only if the head-space is about 450 cc [calc], so the cell figure is neither anomalous nor informative. For the engine, 3.6 W of input can carry at most about 4.3 W of gas heating value even at 100 percent Gibbs efficiency [calc: 3.6 x 285.8 / 237.1], against hundreds of watts to about 1 kW a small engine burns [memory, medium]; but the patent does not say the engine run used 3.6 W (it implies "sufficient electrodes"), gives no run time, load, engine size or gas rate. The patentee's claim that orthohydrogen is "highly combustible" and parahydrogen "a slower burning form" is his characterisation; the mainstream view is that the two spin isomers have essentially the same chemistry and combustion behaviour [memory, high; not web-checked]. Verdicts (new): the engine run as a demonstration of net energy: Never properly tested (no input given for the run); if the run used 3.6 W: Refuted by calculation; ortho/para combustion difference: Not testable as stated (no measurement, and no mainstream mechanism).

Spiros WO 9528510 (pp.2003-2037; tables read from page images). The argument, as the patent states it: electrolysis at pressure produces "over-unity" energy because "pressure is a by-product" of containing the gas (1 L of water gives about 1,850 L of gas); voltage falls with temperature (Fig.17); a loop of electrolyser, expander, combustion or fuel cell feeds itself. Measured by the inventor (his own test, no witness): a transformer-fed cell, 20 runs from 1 atm to 2,450 psi (167 atm), 1.82 to 1.85 kW, 38.5 V falling to about 29 V, 88 to 100 L/h, with the gas volumes collected over 12 to 26 s (0.35 to 0.69 L per run). No run measures recovered energy. Calculated, not measured: the cell voltages from 150 to 2,000 C (attributed to Bockris, 1974); "1.49 kW from 1 kW"; 760.4 W at 150,000 atm; "1,235 Wh from 1,000 Wh"; break-even "at approximately 66 kAtmospheres". The pressure range of the loop (150,000 atm) is about 900 times the highest measured. My calculations [calc; thermodynamics memory, high]: 1. The "free" pressure is not free. Making gas at pressure P costs more than making it at 1 atm by the free-energy difference of the gas. For ideal gases and 1.5 mol of gas per mole of H2 that is 1.5 R T ln P per mole of H2: at 167 atm and 298 K it raises the reversible cell voltage by 0.099 V per cell (8 % of the 1.23 V reversible voltage); at 150,000 atm it would be 0.23 V (19 %). Expanding the gas back isothermally returns exactly that amount and no more (the Gibbs energy is a state function). Spiros's 760 W is the same quantity: 500 L/h at 1,273 K compressed 150,000-fold has an isothermal expansion work of 782 W [calc], which the cell must have supplied in addition to its electrolysis input. His finding that the voltage "does not rise" with pressure comes from a source-limited transformer supply (open-circuit 60 V, voltage falling as the current rises: 47 A at 38.5 V, 65 A at 29 V). The number of cells in the stack is not stated; if it was about 20 (29.4 V / 1.49 V, an inference), the expected Nernst rise from 300 psi to 2,430 psi (20 to 165 atm) is about 0.04 V per cell, i.e. about 0.8 V in all [calc], against an observed change of +0.3 V with scatter of 29.1 to 30.6 V. So the data cannot show the rise, and do not show its absence. The big drop (34.4 V to 29.1 V) happens between runs 3 and 4 (170 to 300 psi), and the voltage then stays flat or rises slightly. 2. The temperature leg counts the voltage and not the heat: the Fig.17 voltages above 100 C are Bockris's calculated values, not measurements; the 0.94 V at 1,000 C agrees with the reversible voltage of steam splitting (Gibbs energy about 181 kJ/mol at 1,273 K, [memory, medium]), and I did not check the others. Such voltages are thermodynamic minima, and the heat to hold the cell hot, the difference between enthalpy and Gibbs energy, is not charged to the input. The patent's "1.49 kW" divides hydrogen's heating value by the electrical input only (462 L x 11 BTU / 3,413 = 1.49 [calc, arithmetic right]) and is an energy ratio (1.49 kWh per kWh), not a power. Its constant of 11 BTU per litre is used for hydrogen on p.2032 and for the 2:1 mix on p.2037; both cannot be right. 3. The table "88 litres per kWh" on p.2035 misreads its own column: the units are litres per hour at about 1.85 kW, i.e. 48 to 54 L per kWh [calc]. The Fig.18 flow curve is the Faraday relation (flow = 471 / V) and restates Fig.17 rather than confirming it. Figs 27-28 are not reproducible from the stated formula (they start with an unexplained offset of 65 to 125 Wh). Figs 16, 29 and 30 are not in the book. Verdict (new): the closed loop producing "excess energy" from self-generated pressure and heat: Refuted by calculation (the pressure term is returned at best and was never uncounted; the heat is not charged); Spiros's 20-run pressure test as a measurement of voltage versus pressure: Never properly tested (a source-limited supply; no energy balance, no calorimetry, no independent witness); the engine and "implosion pump" claims: Not testable as stated (no data).

High-mpg carburettors and boosters (pp.2041-2087, 2280-2313). [read] Kelly puts the same preface over every patent: the carburettor "was able to produce very high mpg figures ... but which is no longer available as the oil industry does not want functional high mpg carburettors to be available to the public" (p.2041). That is the compiler's claim with no source; the patent texts say much less. Only three contain a number, each in one sentence by the patentee (an interested party), with no baseline, speed, distance or method: Shelton US 2,982,528, "approximately eight times the mileage" (the book prints 1940; Google Patents shows publication 2 May 1961 [web]); Schwartz US 3,294,381, "over 200%" for "the six-cylinder engine of a popular car"; Ogle US 4,177,779, "in excess of one hundred miles per US gallon" on "a conventional 1971 American-made car" V-8. Pogue (three patents, the first printed as "US 642,434, 1932" against "1,938,497, 5 December 1933" in his own later patents), Newberry and Tucker give no mpg figure. The booster reports: Hotsabi, 18 to 27 mpg on a 1992 Chevrolet Caprice at 15 A (about 207 W) with no gas rate or test method; Pearson / Saturn SL, 26 to 44-46 mpg over three months, one "carefully driven" 88 mpg run, a seller of programmable chips (commercial stake), the gain attributed to leaner ECU maps with 1 L/min of hydroxy at 8-10 A. Energy-bound check [calc; the deep reader's road-load assumptions, which I recomputed: small car 1,100 kg, CdA 0.65 m^2, rolling coefficient 0.010, steady 55 mph, level road, no wind; large 1971 V-8 car 2,000 kg, CdA 0.9 m^2, rolling coefficient 0.012; gasoline about 121 MJ per US gallon; all [memory, medium]]. Road load is about 0.55 MJ/mile for the small car and 0.90 MJ/mile for the large one. Fuel energy per mile: 88 mpg is 1.38 MJ, so tank-to-wheel efficiency 40 %; 100 mpg is 1.21 MJ, so 74 % for the large car; Shelton's eight-fold gain on an assumed 18 mpg conventional car (144 mpg, 0.84 MJ/mile; the patent gives no baseline) needs 107 % at the large car's road load. Ordinary spark-ignition engines peak around 30 to 38 % brake thermal efficiency, before driveline losses [memory, medium; not web-checked]. The hydroxy energy at 1 L/min is about 0.1 kW, about 0.3 % of the fuel power of the 45 mpg car (41 kW at 55 mph) and about 300 times smaller than the fuel power the Caprice's reported 18 to 27 mpg change would save (34 kW) [calc]; so whatever happened, the gas is not the energy source, which both Hotsabi ("an igniter") and Pearson (lean ECU tables) say themselves. Verdicts (new): Shelton's "eight times" on a conventional baseline: Refuted by calculation under typical road-load and baseline assumptions (it exceeds 100 % of the fuel's heating value; stated assumptions, not a measurement); Ogle's 100 mpg for a 1971 V-8 car: Never properly tested (it needs about 74 % tank-to-wheel efficiency at the assumed load, far above any practical engine, but it does not by itself break the first law; the figure is unspecified in speed and method); Schwartz "over 200%": Never properly tested (the text is ambiguous between a 2x and a 3x gain); Pogue, Newberry, Tucker: Not testable as stated (no number); the Hotsabi and Pearson booster reports: Never properly tested (operator-reported averages, no fuel-measurement method); the HHO booster sub-verdict above is unchanged. The oil-industry suppression statement: Never properly tested and unsourced (compiler's claim; Claim 7).

Q6. Medical and other non-energy claims in the book: not reviewed

The book contains health claims. The project has no medical reviewer and the arbiter gives no medical verdicts. The deep read (pp.1060-1071, 1203, 1210-1212, 1375-1417) inventories what the book says, so the reader knows it is there and is not left to assume the energy material is the only content. - Beck blood electrification. Chapter 11 introduces it; Chapter 12 designs the pulser (a 555 or LM358 circuit, 27 V from three PP3 batteries, about 4 Hz) as a worked electronics project. Beck's own text, quoted on pp.1069-1071, is titled "Expanded instructions for experimental/theoretical HIV blood neutralization - hypothetical protocols for experimental sessions" (revision 20 March 1997). It prescribes about two hours a day for about two months, 3 mA typical and 7 mA maximum tolerable at the skin, "50 to 100 uA through blood is essential", and claims the treatment neutralises "many other viruses, fungi, bacteria, parasites, and microbes in blood". Kelly adds that the precautions "presumably" apply to all treatments with the device. Source and stake: Beck is the inventor and copyright holder and cites his own patents; no trial or patient data are given. - Lakhovsky oscillator (Chapter 20, pp.1375-1417): an anonymous American contributor, with "eighteen other" replications and two years of personal use, all his own statements. Settings "based on very limited patient experience" for "brain tumour" (3,500 V, field intensity 10) and "liver or pancreas tumour" (4,000 V, 12) on p.1413; "surface cancer" is "a Liver problem" and treating it that way "always solves the problem" (p.1415); "results will be excellent every time" (p.1415). The book's own safety warnings, which are real and kept: mains and high voltage "liable to kill you"; capacitors can "hold lethal levels of charge for a very long time"; nobody with a pacemaker within ten metres; eye hazard as from a welder; Kelly's formal disclaimer on p.1408 ("must not be construed as being an encouragement to make or use any such equipment"). The pages contain no advice to see a doctor or to continue conventional treatment. - Other statements, each unsourced or sourced to an advocate: "autism can be quickly combatted by use of the 100-year old drug suramin" (p.1375; one sentence, no source, dose or trial); an audio recording "claimed to reverse Alzheimer's disease" (p.1416; Kelly: "Whether or not the recording is effective, I have no way of knowing"); colloidal silver "cures AIDS and Ebola" (p.1212) and "can cure both cancer and malaria" (p.1417), "completely harmless to humans and has no side effects of any kind" (p.1210), with an "interesting facts" list of named doctors and cases and no citations. The pages do not mention argyria, a skin discoloration documented with chronic silver ingestion [memory, high; not web-checked; this is the arbiter noting an omission, not a medical verdict]. - Essays on finance, media and climate (pp.993-1005) are likewise outside the physics review. The Keshe "GaNS" water claim (pp.1418-1423) was deep-read earlier and is also outside this review. Verdict: not reviewed (no medical reviewer; no controlled study is cited for any of these claims in the pages read). They are pertinent source material and stay in the record. Kelly is the operator of the sites that host the Beck and silver documents (interested party in promoting them; no commercial stake in a medical product is shown in the pages read).

Q7. Items restored to the record (retention defects found by the QE)

• Lyden US 11,588,421 B1 (granted 21 February 2023; priority 15 August 2019; "Receiver device of energy from the earth and its atmosphere" [web: Google Patents via sites.md]). A recent granted patent that says "As disclosed by Carlos Benitez in Great Britain Patents GB 121561A1, GB191514311A1, GB191417811A, and BG 191505591A, it is possible to generate electrical energy and to maintain the same in at least two batteries with relatively little or possibly no energy loss" and cites Moray's US 2,460,707. This is the applicant's characterisation of prior art, not a measurement, and the applicant has the stake of a patentee. It sits beside Benitez (Claim 2) because it relies on the same Benitez example that fails the stored-energy bound above (0.16 J per spark, at most 160 W at 1,000 sparks/s); a patent grant is not a test of that example. It belongs in the record next to Claim 6's Moray entry for the second citation.
• Sweet's earliest outputs. hyiq.org reports the first test of the Space Quanta Modulator on 19 May 1986, shown to Mark Goldes and Darryl Roberts on 24 May 1986, with Generation 1 output of 6 to 19 W and a second revision of 100 to 110 W, the lab notes taken by Roberts in May-June 1986 [web: https://www.hyiq.org/Page/Details?Name=Floyd%20Sweet's%20Vacuum%20Triode%20Amplifier%20-%20Generation%201, via research/2026-10-08-sweet-sources.md]. These are the earliest and lowest figures in the Sweet record and sit before the 1987 figures (500 W and 1 kW) and the 1988 5 kW. The ratio of output over the 1987-88 inputs is orders of magnitude larger than anything in 1986. Goldes funded Sweet from 1985 (an agreement dated 29 December 1985 is listed in the AboveUnity compilation): Mark Goldes (Aesop Institute, later Magnetic Power Inc) is an interested party. The 1986 notes are single-source as printed on that site; they are not corroborated by a second source I found.
• Maurice's further statements about Sweet (p.2243, as recorded in the deep read): a VTA said to have been passed to "Senator John Glen", then the Department of Energy, then General Motors; a planned $200 million suit; people "claiming to be from the FBI" removing equipment the morning after Sweet's death; and an MIT MSEE (1969, thesis "Dynamics of Magnetic Domains") for Sweet. These are an unnamed associate's statements, uncorroborated in the files; they sit beside the "two unknown visitors" in Claim 7. The checks that would decide them are the MIT thesis index, the National Archives and DOE records, and FOIA requests to DOE and the FBI.
• Kelly's own corrective statements are now in the record: Perendev "has not been confirmed" (p.78), Tewari "does not mention anything about 20 kilowatts" (p.102), Matchett circuits (p.211), the PM-only motors are "very difficult" (p.72), the Adams diagram "will never reach 100%" (p.90), and the true-power passage (pp.1089-1091).
• Papp's "300 HP for 40 min" (p.37) against 35 minutes (p.642): added to the internal conflicts.

Q8. Effect on the headline verdicts

None of the seven headline verdicts changes in kind. Changes of substance: - Claim 3: a new sub-class (magnet-only motors) with a refuted-by-calculation closed-cycle verdict; Heins moves from a digest-level "Never properly tested" to Not testable as stated for the COP figure (the printed input resembles the primary's own winding loss). - Claim 5: Puharich's surplus is refuted on his own measured gas, with the careful two-laboratory gas measurement credited and a new Faraday inconsistency noted; Hasebe and the Spiros loop are Refuted by calculation on their printed numbers; Shelton's eight-fold mpg is Refuted by calculation under stated assumptions; the other carburettor and booster claims stay Never properly tested. - Claim 6: the Papp mechanism is judged on the patent's own fusion statement and is Refuted by calculation (the earlier recombination argument is kept for the electrical step only); Britt and Barbosa/Leal added; Matchett added with Kelly's doubts. - Claim 2: Kelly's own LM2587 "COP 1.5" is added as the only measured COP above 1 by the author, with the instrument problem stated. - Verdicts that rested on digests for Chapters 1-3, 14 and 20 and for the appendix items Puharich, Hasebe, Chambers, Spiros, Papp, Britt and the carburettor patents now rest on deep reads. Still digest-level: Kelly's true-power passage (pp.1089-1091, quoted from the QE's digest read) and the Aguero, Horvath, Eccles and Blue items of the water appendix; Meyl's 0.5 mW to 6 mW source and the Chapter 3 Heins patent text itself (only Kelly's quotation of it, with page images p.150-151) remain unread. No verdict above covers them.

Corrections to earlier documents (QE revision)

(For the orchestrator, under CORRECTIONS ARE WELCOME. I edited only this file.) 1. facts.json PJKbook.pdf#ch1-2-pp42-146, summary: "one replication with 27 W electrical output" is wrong. The text is "input power is 27.6 watts and the output power is 33.78 kilowatts", COP 1,223 (p.93) and "27 watts ... 32 kilowatt output" (PDF p.28, not p.29). Also state that the 27 W / 32 kW and 27.6 W / 33.78 kW figures are probably one report; the text does not say. 2. facts.json #ch3a (Heins): add the patent test (0.29 W in, 11.51 W out, "COP of 39.6", p.150), the 3.77 variant (106.9 mW to 403.3 mW, p.151), and that "thirty times" (p.149) and 39.6 are unreconciled. Add Klingelhoefer (60 W to 480 W, COP 8, p.170), Barbosa and Leal (COP 100 minimum p.179; 22 W to 6 kW p.182; 169 kW / 104 p.211; p.37 table COP 102.4) and Clarence's 3 kW (pp.198-204). The earlier text "Barbosa and Leal COP about 100" should say the figures differ by page. 3. arbiter.md (this file), Claim 6 and the Corrected claim list item 6: the statement "plasma recombination returns at most the ionisation energy" is incomplete for Papp; the patent's stated mechanism is fusion of helium (see Q4). Corrected below in the Corrected claim list. 4. arbiter.md, Open question 10: "Chapter 14 and appendix pp.1901-2238 ... remain unread; no verdict here covers them" is out of date. Chapter 14, pp.1901-1960 (Puharich, Hasebe, Chambers, Garrett), pp.2003-2037 (Spiros), pp.2041-2087 and 2280-2313 (carburettors, boosters) and pp.2196-2238 (Papp, Britt) are now deep-read and covered above; Aguero, Horvath, Eccles, Blue and Meyl's source remain. 5. qe.md section D (Puharich): the deep read confirms the recomputation, with these changes. The source words are "it was decided not to use the watt meter" (p.1909), not "A wattmeter was not used". The 16.20 cc is 10.80 cc H2 plus 5.40 cc O2. The 63 percent assumes STP volumes (58 percent at 25 C). Add the Faraday inconsistency (2.07 and 1.36 times the charge limit). Hasebe: the Gibbs floor is 7.05 J/cc, not 7.06 (immaterial). Chambers: the cell figure gives no gas volume, so the energy test is possible only for the engine run, and the patent gives no input for it. 6. Dates in the book: Chambers US 6,126,794 was granted 3 October 2000, not 16 July 2002; Shelton US 2,982,528 was published 2 May 1961, not 1940 [web]; the first Pogue patent is printed as US 642,434 (12 November 1932) on p.2041 against No. 1,938,497 (5 December 1933) in Pogue's own later patents (pp.2049, 2054): a transcription error in the book. The Kawai, Johnson and Flynn dates are consistent. 7. Page labels: the Adams intro statement is PDF p.28 (printed I-18). The "Section 3.9" Barbosa and Leal patent numbering is Kelly's: the electron-trap text on p.190 carries application BR2013/000015. 8. The deep reader notes (Papp) that Kelly's reproduction of US 4,428,193 omits the numbered claims. Any file that says the Appendix reproduces the patent "in full" should say "the specification and drawings, without the claims". 9. Kelly's Hatem date "1955" against patent FR 2,826,800 of January 2003 is unexplained in the text; flag when quoted.