Gunderson: Arbiter (Step 3)

Inputs weighed: facts.json, claims.md, deep/ reads, physicist.md, mathematician.md, contrarian.md, sites.md and research/ notes. Tags: [read] document, [web] URL, [memory] confidence.

Source gap note

Neither patent contains a measurement, a prototype description or an energy budget [read: deep reads; all three reviewers]. US 2006/0163971 has one unquantified sentence (para 0040, "Experiments show ..."). Every number about performance comes from outside the patents: the inventor's conference descriptions (ESTC 2015/2016), one forum witness thread, and assignee web pages (Magnetic Power Inc. 2008, Chava). The 2016 "Magnetic Implosion Transformer" is a different device (resonant H-bridge, ferrite C-cores) from the 2006 toroid and from the 2008 magnetostrictive bar. It is related evidence, not a test of either patent. The Investigator read thread 13945 pages 1 and 5 only, and the archived MPI PDFs only in the browser. Conclusions on the 2016 device therefore rest on one forum witness and the inventor's own text, not on raw data. The patent text, the sentences quoted in the reviews and the claim language were checked by me against [web] (see Web checks).

Where the three reviewers stand

They agree on almost everything. The disagreements are about labels, not physics: - Claim 1 strong form: Physicist folds it into claim 2; Mathematician says "Never properly tested"; Contrarian says "Refuted by calculation" (magnet as energy source). The mechanism (EMF from modulated flux, no moving parts) is accepted by all as Established. - Claim 4 energy part: Mathematician adds "Not testable as stated" for the bare sentence; Contrarian adds "Internally inconsistent". - Claim 5: Contrarian says "Internally inconsistent", the others "Refuted by calculation". Contrarian's point (squares a sum where an average of squares is needed, and the patent squares a sum of "energy") is sound, but the instantaneous statement is true as stated. See claim 5. - Claim 6: Physicist "Established (geometry)"; Contrarian "Plausible, untested / Not testable as stated". The disagreement is over whether a design choice is a physics claim. - Claim 2 experimental reports: all three say Never properly tested. None says "dismissed without testing".

Recomputed numbers (reviewers differed or rounded)

Run in the project venv.

1. Magnet energy. Physicist used 330 kJ/m^3 for NdFeB and ~25 kJ/m^3 for ferrite; Mathematician used 400 kJ/m^3; Contrarian 200-400 and 25-35. The 2016 device has 20 magnets of 2 x 3 x 0.5 in: 20 x 49.16 cm^3 = 983 cm^3 [web: McKay post, via Investigator's note]. Upper bounds on total (BH)max energy, which is not all extractable: - ferrite 25-35 kJ/m^3: 25-34 J - NdFeB 330-400 kJ/m^3: 324-393 J - 9.5 W for 1 h = 34,200 J. Ratio: 100-105x (NdFeB at 330-400 kJ/m^3 gives 87x-105x) to ~1000-1400x (ferrite). Physicist's "100 to 1000 times" and Contrarian's "3 to 40 seconds" agree (9.5 W: 25 J lasts 2.6 s, 393 J lasts 41 s). - Web check [web: https://en.wikipedia.org/wiki/Neodymium_magnet]: the page gives sintered NdFeB (BH)max up to about 512 kJ/m^3 (theoretical, 64 MG·Oe) and a table range of 200-440 kJ/m^3, and says about 18 times ordinary ferrite. Using even 512 kJ/m^3 the bound is ~500 J, still ~68x short. So the magnet-as-fuel refutation holds with any value in the reviewers' range. Upgrade: these were [memory, high/medium], now retagged [web]. - Open detail: McKay writes "Manganese Ferrite PM's". Manganese-zinc ferrite is normally a soft magnetic material, not a permanent magnet; the magnet material is therefore unspecified in the only source [read via research note]. The bound above is generous for any ferrite and for NdFeB, so no verdict turns on it.
2. Phase-error sensitivity. Mathematician: 500 VA, 0.18 degree gives 1.5-1.6 W; 100 VA, 0.88 degree gives 1.5 W. Reproduced: 1.57 W and 1.54 W. Contrarian: 100 VA at 89 degrees gives 1.75 W, at 88 degrees 3.49 W; 1 degree at 50 kHz is 55.6 ns. Reproduced. These are illustrative (apparent power is assumed, not measured). They show that a small timing skew can produce a watt-scale error in resonant circuits; they do not show an error occurred.
3. Square wave and Parseval. Mean-square of a unit square wave's harmonics: fundamental 0.8106 (81.1% of total), n=1,3: 0.9006 (90.1%), n=1,3,5: 0.9331 (93.3%), total to n=399: 0.9990. Ratio of total to fundamental-only: 1.232. [Corrected in QE revision: this item originally said "0.9015" for n=1,3,5 and "corrected" the Mathematician's 0.9006; the Mathematician was right.] No verdict changes. Cross-terms average to zero, so average power is the sum of squares.
4. Lumped acoustic model (Mathematician): reproduced exactly. At resonance (5.03 kHz), input power 10.0, 5.69, 2.09, 1.43 W for R = open, 100, 10, 1 ohm; load power 0, 2.40, 0.88, 0.11 W; sign-reversed coil at R=100: 21.4 W in, 57.1 W load, 78.6 W mechanical loss. Input falls with load at fixed force, load power stays below input. A negative real part is needed for any "aiding" gain. Physicist's chain model (Re Z >= 0 at all frequencies) gives the same conclusion by a different route; I did not rerun it (script is in /tmp and not part of the record).
5. Battery: 12.7 Ah x 12 V = 548.6 kJ; one hour at 9.5 W is 34.2 kJ (6.2%). So a one-hour run cannot separate stored from free energy unless charge is counted (Mathematician and Physicist agree).

Web checks (budget 8; 3 used)

• US 7,830,065 B2 [web: https://patents.google.com/patent/US7830065B2/en]: grant 9 Nov 2010; both quoted sentences are present ("decreasing amounts of energy are generally required to operate it"; "Experiments show that high-quality permanent magnets can be magnetically 'pushed against' in this way for very long periods of time"); claim 11 ends "improves the efficiency of the conversion device". Status expired for non-payment of fees; the fetch summary gave 12 Dec 2022, while the Investigator recorded an effective date of 9 Nov 2022 (12th-year fee). I treat the exact date as unresolved (a fee-event date versus an effective date); the lapse itself is confirmed.
• Neodymium magnet energy product: above.
• Manley-Rowe relations [web: https://en.wikipedia.org/wiki/Manley%E2%80%93Rowe_relations]: page confirms they are conserved quantities for nonlinear wave interactions with several frequencies, with power exchanged between frequency components via the pump. Used by Physicist and Mathematician as [memory, high]; confirmed in general terms. I did not obtain the exact power-flow statement from the page, so the specific claim "a static bias magnet does not pump" is the reviewers' application, not a quotation.
• Not re-checked: the energeticforum thread, ESTC pages, Wayback MPI report. They rest on the Investigator's reads. The 9-10 W figures and the "0 W for an hour" statement are therefore documented in one forum source; I have no second source and did not find one.

Corrected claim list (final verdicts)

Claim 1. Flux "virtually moves" across the core holes, inducing EMF without mechanical motion (US 2006/0163971; granted as US 7,830,065 B2)

• Agreement: all three accept the mechanism; it is a variable-reluctance / flux-switching / magnetic-amplifier action [memory, high]. Nothing in the document is quantified.
• Disagreement: only the label for the strong form. The strong form in claim 1 ("the magnet's field moves, the magnet is the source") is really claim 2, so I put it there.
• Final: Mechanism (modulated flux induces EMF, no moving parts): Established. Practical output, efficiency or advantage over other flux-switch designs: Never properly tested (no neutral test; no data in the patent; not shown to have been dismissed). Granted claim 11 recites the geometry and "improves the efficiency"; that is an examiner accepting structure, not a test.
• Settles it: build the geometry with and without the bias magnets (or with non-magnetic spacers), measure input and output real power by a method that does not depend on phase.

Claim 2. The Lenz reaction aids the input; more output needs less input; magnets are the energy source until demagnetised

• Agreement: no equations; the argument is flux-arrow bookkeeping, not an energy budget (all three). Passive-network and parametric (pump) analysis say the extra power comes from the modulating source, and a static magnet is not a pump. Magnet energy bound refutes magnets as fuel at useful power (recomputed above). All three call the instrument reports "never properly tested".
• Why the stronger arguments win: the reflection of the output load into the input is positive for any passive reciprocal network (Mathematician's Z_in = Z11 - Z12^2/(Z22+Z_L), real input power >= real load power). The energy-product bound uses the witness's own numbers. Counter-arguments from the inventor's side (witness reading; Contrarian's strongest case) are instrument reports, which can be tested directly, so they do not overturn a calculation; they set up the test.
• Evidence kept in the record: inventor's ESTC 2016 text ("570% more energy than could be accounted for", input dialable to zero, "still in the process of validating", "not quite sure why it works") [web: https://emediapress.com/shop/magnetic-implosion-transformer/; seller has a stake; text signed by Gunderson]. Mark McKay, PE, friend of ~10 years, not a commercial party: two analyzers and a Tektronix scope agreed, 1.53 W in / 9.43 W out (COP 6.16); an hour later input 0 W with output 9.5 W [web: https://www.energeticforum.com/forum/energetic-forum-discussion/renewable-energy/13945-magnetic-implosion-transformer-replication]. "Reiyuki": ~3.5 W in / 8.5 W out; phase-dependent (110 degrees: 4 W in, 8 W out; 140 degrees: -1 W in, 3 W out, as recalled). 2015 device COP ~1.02-1.03. MPI 2008 report ("reaction-less behavior ... soundly demonstrated") and Chava ("more than 100 times the input" at "very low power level"): interested-party statements, no data. An anonymous "Physics Review Board" calls Chava's solicitations fraud: a critic's characterisation, no court finding found. These are documented in single sources each; I weigh them as reported instrument readings from an engineer, not as raw data.
• Inconsistencies on the inventor's side that bear on weight: the patent says "no consumption of natural resources" (0036) and magnets are "spent" (0040); the assignee 2008 site names zero-point energy; the inventor's 2018 site names ambient energy [web: https://web.archive.org/web/20180115005940/http://grahamgunderson.com/ou]. The grant removed "persists until the magnets are no longer magnetized" and "reduction of required input power as loads are applied" but kept 0044 and 0040 [web: Google Patents, confirmed for 0044 and 0040 by me; the deletions are the Investigator's comparison]. The inventor's own proposed test (instantaneous v x i, para 0048) measures reactive power at quadrature [read: deep read], a methodological hole.
• Final, split (QE revision refines the basis for patent 1's geometry, labels unchanged):
• Strong form, magnets as the source (the patent's own account): Refuted by calculation (energy-product bound; cyclic-energy bound; the linear passive-network bound applies to the acoustic patent and the resonant 2016 circuit, not directly to the holed core, see QE revision).
• The broader, source-agnostic reading ("some device returns more than it takes", 2016 reports): Never properly tested. Not tested and failed; not dismissed without testing either (forum engineers looked at it, no refutation published), but no independent lab, no published waveforms, no completed replication on the pages read.
• Weak form (input shows little real-power loading at some tuning because the circuit is resonant and mostly reactive; the modulating coil is a pump): Plausible, untested.
• Settles it: closed loop (output rectified and fed back, battery removed, run for hours); or an integrating coulomb counter on the DC battery for hours, repeated with magnets replaced by non-magnetic spacers; plus calorimetry of the lamp. A phase-independent input measurement decides between artefact and effect.

Claim 3. Acoustic Lenz reaction returns with 90 + 180 + 90 = 360 degrees and aids the drive (US 8,093,869)

• Agreement: arithmetic and 1 cm / ~600 m numbers are right (c/125 kHz = 2398 m, quarter 600 m, as the reviewers computed). The 180 degree step is asserted, and the patent itself says phase depends on inductance and tuning capacitor [read: deep read]. For a resistive (power-consuming) coil circuit, reaction is in quadrature (90 degrees) with the flux, not 180. A reactive 180-degree reaction draws no net power. All three reach "refuted" for the energy-saving form.
• Which argument is stronger: Physicist's impedance argument (Re Z of a passive load >= 0) and Mathematician's lumped table are the same result by two routes, and my rerun matches. Contrarian adds that 90 + 90 of travel plus 180 is the condition for resonance of the bar by itself.
• Mathematician and Physicist both note the patent's "less acceleration input than before current flowed" is true in a limited, conventional sense: at fixed force, loading lowers resonant amplitude and drive power, but output stays below input (table above). I keep that distinction in the verdict.
• Final: Phase arithmetic: correct. 180 degree step: Internally inconsistent with the patent's own phase caveat (assumed, not derived). Strong form (net drive energy saving beyond delivered output): Refuted by calculation. Conventional form (loading changes drive impedance and power at fixed force): Established. Any in-phase return benefit in a real bar: Never properly tested (assignee claims 2007 prototypes, no data).
• Settles it: bench test of a magnetostrictive bar with calibrated force and velocity at the source; measure Re(Z) at the source with the pickup coil open versus loaded at several resistances. Prediction: Re(Z) >= 0, efficiency < 100%.

Claim 4. Opposite currents in neighbouring coils add magnetic opposition, raising delivered energy and cancelling reactance

• Agreement: reactance cancellation (L1 + L2 - 2M) is textbook; the "more opposition means more energy delivered" step is not supported, and the patent's own transformer analogy has the loaded secondary raising input current.
• Final: Reactance cancellation / impedance matching: Established (circuit theory; benefit small, no data). Increase in total delivered energy beyond the acoustic input: Refuted by calculation, and the text says both that fields "tend to cancel" and that they increase delivered energy [read: deep read], so the strong form is also Internally inconsistent. The bare sentence with no input-output accounting: Not testable as stated.
• Settles it: measure delivered electrical power and acoustic input power for one coil versus the opposed pair at the same drive.

Claim 5. Square-wave drive makes odd-harmonic peaks coincide; power is the square of the sum; "substantially increased"

• Agreement: peak coincidence at the centre node of a free-free bar is true for odd modes; average power follows Parseval (sum of squares). Recomputed: square-wave drive gives 1.23 times the fundamental-only power, which the drive's extra harmonic content pays for.
• Disagreement on label: the patent sentence is about instantaneous power [read], which is true as stated; the error is concluding gain. I side with Physicist and Mathematician: instantaneous statement true; average-power gain Refuted by calculation. The patent also squares "the sum of the energy" (a unit slip, Contrarian), which supports "Internally inconsistent" for the wording only.
• Final: Peak-coincidence fact: Established. "Substantially increased" power relative to input: Refuted by calculation. Conventional version (a square drive delivers 23% more average power than a sine of the same fundamental amplitude): Established, with no gain.
• Settles it: Fourier-analyse the measured drive current and compare average input to output; a spreadsheet is enough.

Claim 6. Spacing >= 1/8 (preferably 1/4) acoustic wavelength, bias magnet, separate source distinguish the device from gyrators and harvesters

• Agreement: geometry is consistent (quarter wave = 1 cm at the patent's example numbers); no performance comparison in the patent; novelty is a patent-law matter. Claim 1 ends with a "nonlinear time-changing resistance" load not defined in the description [read: deep read]. OCR renders 1/8 as "%"; page images resolved this [read: deep read, page images].
• Final: Workable transducer arrangement: Established. Any advantage from the spacing: Never properly tested. As a distinction from prior art: Not testable as stated in physics terms (a legal question). The patent also calls FIG. 7 an embodiment "designed to harvest power from a source of vibratory energy" [read via Contrarian, p.22], which blurs the distinction.
• Settles it: build a bar with the pickup at 1/8 versus at the end or at other positions with otherwise identical coils and drive; compare power.

Corrections to Step 1 (adopted)

1. facts.json says it is not stated whether US 2006/0163971 was granted. It was, as US 7,830,065 B2 (9 Nov 2010), 11 claims, claim 1 narrowed to a resonant circuit with a square-wave source, new claim 11 reciting "improves the efficiency"; lapsed for non-payment of fees in 2022 [web]. The claim list should say "no claim recites over-unity or Lenz aiding"; the aiding language is in the description only, in both patents [web, read].
2. claims.md claim 2 should note the contradiction between "no consumption of natural resources" (0036) and magnets "spent" (0040), and the later change of stated source (zero-point, ambient).
3. claims.md claim 5 paraphrases "square of the summed amplitudes"; the patent says "square of the sum of the energy induced by each acoustic frequency".
4. The claims list does not mention the later non-patent claims (ESTC 2015-16, witness posts, MPI/Chava web claims). They are included here.

Absence-of-evidence analysis

Where the record goes dark: Chava announced "several prototypes of more advanced devices", an "additional, very broad" patent application and 30 more patents in three years. The inventor filed a provisional in July 2016 (McKay, one source). Since then there is no published family under Gunderson or Chava, no data, no paper, no independent test, and US 7,830,065 lapsed in 2022 [web: Google Patents]. 8,093,869 is active (fees paid late in 2019, expiry 2029). Gunderson's later public trail (2019 video, a 2025 post per the Investigator) was not opened.

Competing explanations, with evidence: - (a) The effects did not survive validation. For: inventor said "still validating" in 2016 and the 2015 device was COP 1.02-1.03; the grant text was softened to "believed to"; the physics of claims 2-5 conflicts with passivity; a phase-sensitive resonant measurement is exactly where artefacts occur. Against: this explanation cannot be shown by silence alone; a person who found an artefact might have said so publicly, and none did on record. - (b) Filings lapsed for cost. For: 12th-year fee unpaid; the inventor describes building at night alongside a full-time job. Against: fees on 8,093,869 were still paid (late), so cost is selective, not general. - (c) Trade secrecy / NDA. For: MPI offered NDAs; provisionals are never published. Against: a company seeking investment would be expected to show some data. - (d) Funding loss or internal dispute. For: one commenter says Goldes was "ejected" from Chava (single anonymous comment); assignment chain shows repeated Chava re-registrations (2010, 2015, 2018). Against: no court or corporate record found. - (e) Suppression. No evidence for or against found. The Patent Office classed the family H02K53/00 ("alleged dynamo-electric perpetua mobilia") yet granted structural claims; that is classification and examination, not suppression.

I judge (a) the explanation most compatible with known physics, and (b)-(d) not excluded. The records that would decide between them: USPTO file wrappers for 11/336,337 (the 2006 application) and the application behind US 8,093,869 (12/327,719, see QE revision) (examiner rejections and amendments; not reached), the status and conversion of the 2016 provisional, Chava Energy LLC filings in Florida and Nevada, the 2009 FOIA entry on Magnetic Power Inc (agency unidentified), forum pages 2-4 and thread 13946, the archived MPI PDFs, and Gunderson's 2019 video and any statement from McKay after 2016. Gunderson's own account of a Sweet-VTA-like system that charged an EV battery bank in a week (undated, unnamed) is a lead for the inventor's statements; I have no second source.

What would settle it (summary)

1. Closed loop on the 2016 device: output rectified, fed to the drive, battery removed, run for hours.
2. Integrating coulomb counter on the DC battery for hours, repeated with magnets replaced by non-magnetic spacers; lamp calorimetry.
3. Phase-independent input measurement (DC side) and a bandwidth-verified output measurement.
4. For the acoustic patent: Re(Z) at the source, pickup coil open versus loaded, and position scan of the pickup.
5. Obtain USPTO file wrappers and Chava filings to resolve the dark period.

Open questions

• What is the magnet material in the 2016 device (the only source says "manganese ferrite PM")? Does the device still work with magnets removed or demagnetised?
• Was the zero-input reading a phase/bandwidth artefact, genuine energy return from stored reactive energy, or neither? No one has published waveforms.
• Did the 2016 provisional lead to an application, and what do the file wrappers show about why claims were narrowed?
• The lapse date of US 7,830,065 (9 Nov versus 12 Dec 2022) is unresolved; the lapse itself is confirmed.
• Was there any later test by Gunderson, McKay or others after October 2016?

Verdict summary

QE revision (2026-10-04)

Inputs: qe.md (verdict MINOR), the two QE deep reads appended to deep/generator-US20060163971-7e4776.md (circulation of reaction flux (c) around the holes) and deep/us8093869-8ca547.md (FIG. 1 lattice, "multiplier"), facts.json (already updated for the grant and the shared switched-load element), and QE's own web check of the forum thread. No verdict label changes. Three bases change, two numbers are corrected, and retained material is restored. Tags: [read] document, [web] URL, [memory] confidence, [analysis] reasoning from the documents.

1. Corrections to this file

• Parseval shares (Recomputed item 3, corrected in place). I reran it: unit square wave, share of mean-square carried by harmonics 1: 0.8106; 1 and 3: 0.9006; 1, 3 and 5: 0.9331. My earlier "0.9015" was wrong and the Mathematician's 0.9006 was right. No verdict depends on it (claim 5 uses the 1.23 ratio, unchanged).
• Application number behind US 8,093,869: 12/327,719, filed 3 Dec 2008, provisional 60/992,009 of 3 Dec 2007 [read: p.1 of the patent text; facts.json]. The file wrapper to request is therefore 12/327,719 (and 11/336,337 for patent 1).

2. Claim 2 (patent 1): which energy bound fits the holed-core geometry

The deep read of FIG. 3 [read: page image p.4; paras 0037-0044] settles what the drawing shows: - The reaction flux (c) is drawn as a closed loop round each hole. At all eight drawn hole positions it runs against the input flux (a) on the magnet side of the hole and along (a) on the magnet-free side. The text assigns the magnet side to "(b) and (c) cancel" and counts only the free side as "(c) aids (a)" [read: para 0043]. The Mathematician's point (aids on one side, opposes on the other) is confirmed. The document gives no argument that the two sides are unequal in their effect on the input coil. - The patent contradicts itself on back-action: para 0039 says the reaction is directed against the magnets "not the input flux source (6, a)" (no back-action on coil 6), while paras 0043-0044 say (c) "aids the input flux (a)" with the effect ranging from "idling" to "strengthening and overall acceleration of the electrical input signal". It never mentions flux linkage, mutual inductance or reflected impedance. No number or measurement supports either statement, and claims 1-10 recite none of it [read: pp.8-10]. I add this to the "Internally inconsistent" list as a statement-level inconsistency in the description (para 0039 versus 0043-0044). It does not change the claim-2 verdict. - Why the linear reflected-impedance argument does not carry over. QE proposed that mutual inductance between the input coil and the output wire is about zero by symmetry, and gave the reason as the alternating wire direction and staggered magnets. The deep read confirms the conclusion but corrects the reason: the holes sit on the mid-plane of the core [read: para 0029], so a (c) loop crosses any cross-section linked by coil 6 either not at all or twice in opposite directions, and the output circuit's surface lies in a plane the circumferential flux (a) runs along. In a linear, mirror-symmetric core the coupling is zero. The staggered magnets make the bias (b) asymmetric top to bottom, which is what lets an output appear at all. The coupling is thus parametric: the output comes from a cross term (bias field x input field x slope of permeability), which needs core saturation, since in a linear core (a) and (b) would superpose and nothing would be induced [analysis, [memory] high for the magnetics]. - What parametric does not mean. The deep read warns, and I adopt, that "parametric" is not "no back-action". At the biased operating point there is a nonzero small-signal coupling, and for a lossless, single-valued core both directions come from second derivatives of one magnetic energy function, so load current reflects into coil 6 through the same coupling. The patent's claim of reaction only against the magnets would require that reciprocity to fail, and nothing offered shows it does. Whether the reflection loads or "aids" the input depends on phase and is not decided by the figures [analysis]. - Energy bound for this geometry. I replace the Mathematician's two-port argument, for patent 1 only, with the cyclic form. Over one steady cycle the electrical work into all windings, the sum over windings of the integral of i dlambda (i = winding current, lambda = flux linkage), equals the core's hysteresis and eddy loss, which is not negative, because the stored magnetic energy of core plus magnets returns to its starting value [memory, high; valid for nonlinear cores; QE derivation, not in the sources]. So input real power >= output real power + loss, with the magnets unchanged over a cycle. This is the same conclusion as before by a route that does not assume linearity. The Mathematician's Z_in = Z11 - Z12^2/(Z22 + Z_L) form still applies to the acoustic patent and to a linearised small-signal view of the 2016 resonant circuit. - The magnet-as-fuel bound (one-off stored energy of the magnets, recomputed above) is unaffected: the cyclic bound says magnets cannot supply energy in steady state, and the energy-product bound says even a depleting magnet could not supply 9.5 W for more than seconds. - Claim 2 verdict: unchanged. Strong form (magnets as source) Refuted by calculation, now resting on the energy-product bound and the cyclic-energy bound; source-agnostic 2016 reading Never properly tested; weak form Plausible, untested. Counter-view kept: someone who reads para 0043 as true of the free side alone is reading the patent text literally, and the drawing does not contradict that reading. It fails only because the opposed half of each loop is dropped.

3. Claims 2 and 6: the shared switched load, and the measurement-artefact question

Claim 1 of US 8,093,869 ends with a load of "nonlinear time-changing resistance, by including semiconductor switching elements for purposes of periodically modulating the current". The description never supports it [read: deep read; page image p.22; claim 1]. The 2016 demonstration output was a SiC-FET synchronous rectifier opened briefly near the phase transition, and Reiyuki's reported COP depends on the phase of that interruption [web: research/2026-10-04-energeticforum.md]. facts.json now records this. My weighing: - It is a continuity between the 2012 granted claim and the 2016 demonstration, so my Source gap note statement that the 2016 device is "a different device" from both patents stands for the magnetics (resonant H-bridge, ferrite C-cores versus toroid or bar) but not for the output stage. I amend it: related evidence, with one shared element. - It bears on measurement. Switching at a chosen phase is where phase and bandwidth artefacts concentrate, which supports the phase-independent DC-side measurement I named as the settling test. A time-varying resistor with R(t) >= 0 still only absorbs energy, so the shared element does not itself imply gain. - It is also what the inventor tuned, and the only source for the dependence is a forum witness. I record it as a lead on why the claim was amended, not as a finding about the physics. The file wrapper for 12/327,719 would show when and why the element was added. - Claim 6 verdict: unchanged (the load element is outside the spacing question). I add that claim 1 is not purely a spacing claim, and its last element is undescribed.

4. Claim 3 and the acoustic patent: FIG. 1 lattice and "multiplier"

The second deep read [read: pp.10, 14-16] gives: - FIG. 1 is a free, lossless chain of masses and springs given one push. The text says the initial acceleration "has never been dissipated or relieved, and reappears in the model as if applied for a second time", that "the net action is an inchworm motion successively toward the right", and that "final absorption of the original kinetic energy 120 requires intervention beyond the scope of this model" [read: p.15, col.12]. The deep reader reads this as conservation-consistent: one impulse, a centre of mass drifting at constant velocity (momentum J, speed J over total mass, [memory] high) plus an internal vibration. I agree. The text attributes the motion to the single initial push, and no numbers, experiment, thrust or reaction-less drive appear in pp.10-16. Page 16 then switches to resonators fixed in space with an alternating force, where "net motion ... is restrained and in balance". - The wording is loose, and "reappears ... as if applied for a second time" is the conceptual seed of the later claim (p.20) that the reflected wave is "an independent source of mechanical acceleration" so that the drive requires "less energy". Only the p.20 statement is an energy claim, and it is the one I already rule on (strong form refuted; conventional loading effect established). A returning wave reinforcing the drive in a resonator is ordinary resonance (stored energy, quality factor), and becomes a gain claim only if output is said to exceed input, which the patent never says. - "Energy transfer and multiplier element" [read: p.10, col.2] appears once, is undefined, and carries no gain figure. Four readings are available (resonant build-up; reflected wave adding to the drive; harmonic addition at the nodes; amplitude magnification by horns). The first and fourth are amplitude or stored-energy multiplication, not energy creation. It is the applicant's own wording in a patent summary, so it is the applicant's characterisation. I record it as a term that invites an over-unity reading and that the patent does not cash out. - Not yet read: pp.18-19 use "increase in total momentum carried" by the acoustic wave, and p.13 has a "forward-thrusting" phrase [read, per the deep reader; not deep-read]. A momentum claim could arise there. I have not weighed it; it is listed under open questions. Claim 3 verdict: unchanged.

5. Web check by QE adopted

QE fetched the energeticforum thread (page 1) on 2026-10-04 and confirmed these quotes by "Spokane1", 11 Jul 2016: "1.53 watts in and 9.43 watts out or a initial COP of 6.16"; "input wattage was down to zero (with four places of accuracy) with the output still at 9.5"; "-.25 watts in"; "12.7 A-hr lead-acid gel battery"; "about 20 Manganese Ferrite PM's (2" x 3" x 0.5"" ; peak excitation around 800 V at 50 kHz; two power analyzers and a Tektronix 4 channel scope; "loop the system and have a self runner" [web: https://www.energeticforum.com/forum/energetic-forum-discussion/renewable-energy/13945-magnetic-implosion-transformer-replication]. This upgrades my Web checks list: the figures I use are now confirmed against the source for page 1 (they were previously only the Investigator's reading). It does not add a second source: they remain one witness. Note that a peak of about 800 V matters to the measurement: it shows a high-reactive-power circuit where small phase errors dominate (see my phase-error numbers; apparent power was not reported).

6. Material restored to the record (dropped between Step I and Step 3)

Each is a lead or an interested-party statement, kept with its stake. None changes a verdict. - OverUnity Research thread "Graham Gunderson Energy conference High COP demonstration" (Chet K, 11 Jul 2016): the poster asks why Gunderson "abandon[ed] a device that claimed high COP" [web: sites.md, snippet; page blocked by a login]. For the absence-of-evidence analysis it is one poster's question, not an answer. It fits (a) (validation failed or was dropped) and (c)-(d), and does not exclude (b). The answer would be in the thread replies, the unread forum pages 2-4, and Gunderson's 2019 video. - Related work, previously in the reviews and sites.md but not in this file: - Tae Chang N.E.T., "Sound wave resonance generator", US20110233941A1, priority 3 Dec 2008, the same day as Gunderson's non-provisional [web: research notes; sites.md]. A same-day filing in the same area. I have no evidence of contact or copying either way. - Patents citing US 7,830,065: Southeast University's "time-varying magductance" generator (US12456936B2 / CN115173664B, 2022), an academic group building on flux modulation; Lovshin's Earth-field energy patents (compare published claim 7 / granted claim 8, where the Earth's field replaces the input coil as modulator); Shaikh GB2446656A [web: sites.md]. The first is the strongest mainstream-adjacent follow-on found. I have not read these documents; the inference that they share the flux-modulation idea is from the Investigator's notes. - E.V. Gray and the TK / Kapanadze / Akula circuits are named by the forum witnesses as comparable devices [web: sites.md]. They are comparisons made by interested enthusiasts, not tests. I take no view on those devices. - Interested-party statements (assignee): the MPI 2008 report describes a "constant voltage" output signature and plans for 1 kW modules and 75-150 kW car units; Goldes' 26 Mar 2007 proposal set two 1 kW Magnetic Power Modules (5" x 5" x 16") against the GM Volt [web: research/2026-10-04-wayback-magneticpowerinc.md]. These are the assignee's statements with a commercial stake and no data. They are claims about intended products, and no test of such modules is on record that I found. - Inventor background: an aboveunity.com snippet says "Graham Gunderson worked for Mark Goldes for many years", which links the "paid verifier" statement to the MPI/Chava circle [web: research/2026-10-04-naturalphilosophy-wiki.md]. One source, not checked further. Gunderson sat on the 2016 ESTC panel with Lindemann, Babcock and Moray King. Hidden Energy (Manning and Manewich, 2019) has a Gunderson section; it is a secondary source I have not read. - Published claim 7 (Earth's field modulates the core in place of the input coil) is another energy-source statement in patent 1's application. It is a fourth stated source (magnets, zero-point, ambient, Earth's field). It is covered by the same cyclic bound if the modulator is the only source of change and the Earth's field is static, and no test is on record. I did not give it a separate verdict.

7. Absence-of-evidence analysis: additions

• The "abandoned" thread (above) is added as one piece of evidence for (a).
• Gunderson's 2012 granted claim already contains the switched load; if the 2016 tuning involved the same element, the dark period is not a dark period for the idea, only for published data. That favours (c) or (a) over a simple loss of interest, but does not decide between them.
• The file wrappers for 11/336,337 and 12/327,719 remain the single most useful unread record.

8. Updated "What would settle it" and open questions

Additions (earlier items stand): - For patent 1: a bench test of the toroid with the magnets staggered versus the magnets removed or non-magnetic spacers, measuring the small-signal mutual inductance from output wire to input coil at the bias point, and the loaded input real power on the DC side. Prediction under my verdict: a small but nonzero coupling, and input real power >= output real power + core loss. - For the 2016 device: repeat the phase-error experiment with the SiC rectifier timing varied while measuring DC-side input and output energy with an integrating method. - Open: do pp.18-19 of US 8,093,869 ("increase in total momentum carried") or p.13 ("forward-thrusting") make a momentum or thrust claim? Needs a deep read of those pages. - Open: was the switched-load element in claim 1 of 8,093,869 added in response to an examiner, and does the 2016 device's behaviour depend on it? Needs the file wrapper.

9. Net verdict changes

None to a verdict label. Basis changes: claim 2 strong form (cyclic-energy bound for the holed core; linear reflection argument restricted to the acoustic patent and the linearised resonant circuit); claim 2 (patent text) adds a statement-level inconsistency (0039 versus 0043-0044). Numbers corrected: Parseval shares; application number supplied. Corrected claim list entries and the verdict table above carry these edits.