Gunderson: QE (Step 4b)

Inputs checked: _text/manifest.json and both extracted texts, digest/ (all 3 files), deep/ (both), facts.json, claims.md, sites.md, research/ (8 notes; research/files/ is empty), physicist.md, mathematician.md, contrarian.md, arbiter.md, primer.md, example.py. Page image p.22 of us8093869.pdf was rendered to /tmp/qe_gund. One web check was made (below).

Verdict: MINOR. Coverage is complete and both documents had full deep reads. The verdicts rest on calculations that I reproduced. None of the defects below would change a central verdict. They are retention gaps (Investigator leads that were dropped), one physics point that the primer teaches using a model that does not fit patent 1's geometry, two small numeric or record errors in arbiter.md, and facts.json/claims.md not yet updated with the corrections that all reviewers agreed.

---

1. Coverage

No document was skipped. Every manifest entry has a facts.json entry. The project folder holds only these two patents, so there is a source gap: every performance number comes from the Investigator's web finds. claims.md, the arbiter and the primer all state this gap.

Small digest errors, caught downstream and harmless: - Digest 02 p.23 says claim 1 sets the conductor spacing at "≥ 1/4 acoustic wavelength". The claim says 1/8. I checked this on page image p.22 [read: page image p.22]. The deep read and all later files use 1/8 correctly. - Digest 01 places para 0040 ("Experiments show") on p.9. It is on p.8 (extracted text, line 528). The deep read has p.8.

2. Shortcuts

• Documents. None. Both central documents had full deep reads, with page images where the OCR mangled fractions. This is acceptable.
• Investigator material. The Investigator read the archived MPI PDFs in the browser and downloaded none, so research/files/ is empty. About 10 archived MPI PDFs, forum pages 2-4, thread 13946 and the 2019 YouTube video were not read. sites.md records these as "not reached" with Recheck = yes. This is acceptable for a capped first run, but it matters: the strongest pro-claim evidence (the McKay account) and the worked example rest on one forum page.
• Arbiter web budget. The arbiter used 3 of its 8 web checks. It did not re-check the energeticforum figures on which primer Section 5 depends, although it said so openly. QE closed this gap. I fetched https://www.energeticforum.com/forum/energetic-forum-discussion/renewable-energy/13945-magnetic-implosion-transformer-replication on 2026-10-04 and confirmed these quotes from "Spokane1", 11 Jul 2016 [web]:
• "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 volts", 50 kHz
• "Two Power Analyzers", "Tektronix (2000) 4 channel scope"
• "loop the system and have a self runner"
• operational "7/6/2016"

The figures the primer uses therefore match the source.

3. Accuracy spot checks (against the extracted text)

example.py runs cleanly in the venv. Parts A-E reproduce every number in primer Section 5, including the energy-balance check P_in = P_load + P_mech.

4. Missed material

From the digests, deep reads and Investigator notes, these items are not carried by facts/claims, the reviews, the arbiter or the primer, or are carried only partly.

1. Claim 1 of US 8,093,869 and the 2016 device share the same switched load. No file connects them. Claim 1's final element is a load with "nonlinear time-changing resistance, by including semiconductor switching elements for purposes of periodically modulating the current". The description does not support this element anywhere (deep read). The 2016 "Magnetic Implosion Transformer" output is a SiC-FET synchronous rectifier "opened briefly near the phase transition", and Reiyuki reports that the COP depends on the phase of that interruption [web: research/2026-10-04-energeticforum.md]. - So the granted claim of 2012 already recites the element the inventor's later demo was tuned on. This is a reader's note, not a verdict. - It bears on the continuity between the patents and the 2016 claims (arbiter: "a different device"), and on why the claim was amended to add that element. - It also bears on the measurement question. Switching that happens at a chosen phase is exactly where phase and bandwidth artefacts concentrate. A time-varying resistor with R(t) ≥ 0 still only absorbs energy.
2. The physics model for patent 1's geometry. - Primer 3.2 (and physicist claim 2) explain the input loading with the linear reflected impedance (ωM)²/(Z22+ZL). In US 2006/0163971 the output wire passes alternately through radial holes, and the reaction flux (c) circulates around each hole (para 0037-0038). - Mathematician claim 2 point 2 makes the related Ampère observation: (c) aids (a) on one side of a hole and opposes it on the other. In a symmetric, unsaturated core, that circulating flux gives no net tangential flux through the core cross-section that the input coil links. Alternating the wire direction also cancels the linkage of uniform input flux hole by hole. So the linear mutual inductance M between input coil and output wire is about zero by design, and the coupling is only parametric, through the magnet-biased, nonlinear permeability [memory, high; QE derivation, not in the sources]. - The correct energy bound for this geometry is the cyclic one. Over one cycle, Σ_k ∮ i_k dλ_k (the electrical work into all windings, λ = flux linkage) equals the hysteresis and eddy loss, which is ≥ 0, because the magnet-plus-core magnetic energy is a state function. This holds for nonlinear cores. It gives the same conclusion as the reviews (input ≥ output + loss), so no verdict changes. - The primer currently teaches a model that does not describe this device. The arbiter also did not take up the mathematician's hole-geometry point.
3. Overunityresearch.com thread. "Graham Gunderson Energy conference High COP demonstration" (Chet K, 11 Jul 2016) asks why Gunderson "abandon[ed] a device that claimed high COP" [web: sites.md, snippet; blocked by login]. It is pertinent to the absence-of-evidence analysis (explanation (a) versus (b)-(d)) and appears in no review, the arbiter or the primer.
4. Related work dropped after Step 2. - Tae Chang N.E.T. "Sound wave resonance generator" (US20110233941A1, priority 3 Dec 2008, the same day as Gunderson's non-provisional) appears in all three reviews but in neither the arbiter nor the primer. - Patents citing US 7,830,065 appear nowhere after sites.md:
   • Southeast University's "time-varying magductance" generator (US12456936B2 / CN115173664B, 2022), an academic group building on the same flux-modulation idea;
   • Lovshin's Earth-field energy patents (compare the published claim 7 / granted claim 8 Earth-field modulation);
   • Shaikh GB2446656A.
   • E.V. Gray and TK/Kapanadze/Akula (named by the witnesses) are not in the arbiter or the primer.
5. Assignee statements. - 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 sets two 1 kW Magnetic Power Modules (5"x5"x16") against the GM Volt [web: research/2026-10-04-wayback-magneticpowerinc.md]. - These are interested-party claims and should stay in the record. They are not in the arbiter or the primer.
6. Inventor background. - "Graham Gunderson worked for Mark Goldes for many years" (aboveunity.com snippet) ties his "paid verifier" statement to MPI/Chava [web: research/2026-10-04-naturalphilosophy-wiki.md]. - Gunderson sat on the 2016 ESTC panel with Lindemann, Babcock and Moray King. - Hidden Energy (Manning & Manewich, 2019) has a Gunderson section. - None of these is in the arbiter or the primer, including Further reading.
7. Patent text items not discussed. - The US 8,093,869 summary calls the bar an "energy transfer and multiplier element" (p.10). - The FIG. 1 lattice model has the "inchworm motion successively toward the right" in free space, with the initial push "never ... dissipated" (p.15). - Both appear only in facts.json or the deep read. No reviewer comments on them, though the second reads like a momentum claim. - Published claim 7 (the Earth's field modulates the core in place of the input coil) is another energy-source statement that no reviewer weighs.

5. Accuracy defects in later files

1. facts.json / claims.md have not been updated. They still say the grant status of US 2006/0163971 is unknown. All three reviewers and the arbiter adopted the correction: granted as US 7,830,065 B2 on 9 Nov 2010, 11 claims, claim 1 amended, new claim 11 "improves the efficiency", lapsed 2022. The arbiter's other adopted corrections are also missing: the 0036/0040 contradiction, the "square of the sum of the energy" wording, and a note on the later non-patent claims.
2. arbiter.md, absence-of-evidence section, says the application number behind US 8,093,869 is "not recorded here". It is 12/327,719 [read: p.1; facts.json, deep read].
3. arbiter.md, Recomputed numbers item 3: - It gives "n=1,3,5: 0.9015 (90.2%)" and "corrects" the mathematician's 0.9006 to 0.9015. The correct values are 0.9006 for n=1,3 and 0.9331 for n=1,3,5, so the mathematician was right. - Primer Step E flags the slip but adds its own slip: "0.9015 is the share of harmonics 1 and 3 only". That share is 0.9006. - No verdict depends on this.
4. mathematician.md, correction 6, says the deep read "does not count the intermediate 270 degrees". The deep read's table does include "Running total at node | 270 deg". This is trivial and needs no change.

6. Retention check

Rule: pertinent material from the sources or the Investigator must not be dropped from the arbiter or the primer because a reviewer judged it negatively.

• Retained well. Retention of the evidence favourable to the inventor is good:
• McKay's readings, quoted with figures and checked above;
• Reiyuki's phase table;
• Gunderson's 570% statement, with "still validating";
• the 2015 COP 1.02-1.03;
• the MPI "reaction-less" and Chava "100 times" claims, with their stake;
• the anonymous fraud allegation, labelled as a critic's characterisation;
• the grant-text changes;
• the five competing absence-of-evidence explanations;
• the FOIA log;
• the EV-charger verification story.

Interested parties are labelled throughout, and single-source items are quoted and weighed, not dismissed. The primer's "What is genuinely open" section treats the 2016 readings fairly ("not refuted by any experiment"). - Defects (minor). - Items 4.3, 4.4, 4.5 and 4.6 above were dropped between Step I/Step 2 and the arbiter/primer. Nothing shows they were dropped for a negative reason; they look like condensation losses. Still, 4.3 (the "abandoned" thread) and 4.4 (the Tae Chang same-day filing; the Southeast University follow-on) are pertinent and should be restored. - The primer's "Related" line keeps Sweet's VTA and the MEG but drops Gray, TK/Kapanadze and the mainstream and academic follow-ons to patent 1.

7. Process

• Source tags. Present and honest in every file. The contrarian's [memory, medium] energy products were web-verified by the arbiter and retagged. Interested-party labelling is correct.
• Web checks.
• Physicist: 2 checks (Google Patents B2, Wikipedia NdFeB). Manley-Rowe was left as [memory, high]; acceptable.
• Mathematician: none, with the reason stated (all [memory] facts are high confidence); acceptable.
• Contrarian: one attempt, which failed and is reported.
• Arbiter: 3 of 8. Forum figures were not re-checked; QE has now verified them.
• The lapse date of US 7,830,065 (9 Nov versus 12 Dec 2022) is still open.
• Verdict scale. Applied consistently. Split verdicts are stated part by part, and the arbiter explains the label disagreements on claims 1, 4, 5 and 6. "Never properly tested" is correctly separated from "dismissed without testing".
• Primer.
• It follows STEP 4 order 1-9, with answers at the end.
• It suits the reader profile: symbols and acronyms (EMF, COP, PE, ESTC, MEG, VTA, j, Z, (BH)max) are defined, Manley-Rowe is introduced in plain words, and the worked numbers are real.
• Textbook chapter references (Griffiths 6-7, Taylor 5, Boas 7) are correct.
• The one teaching flaw is 3.2 (see 4.2).
• example.py runs, and its output matches the primer.

8. Remediation actions

Deep reads (most important first)

1. generator-US20060163971-7e4776, pp.4-5 (page images of FIG. 3 and FIG. 2/4) and paras 0032-0044 (pp.8-9). Opus. Question: - In the drawn geometry, which way does the reaction flux (c) circulate around each radial hole, relative to the input flux (a), above and below the hole? - Does the patent anywhere claim a net linkage of (c) with input coil 6? - Confirm or refute the mathematician's point that (c) aids (a) on one side and opposes it on the other, and QE's point that the linear output-to-input mutual inductance is about zero by symmetry (magnet staggering, alternating wire direction).
2. us8093869-8ca547, pp.14-15 (cols 9-12: harvester distinction, FIG. 1 lattice and "inchworm") and p.10 col.2 ("energy transfer and multiplier element"). Sonnet. Question: - What exactly does the patent claim for the FIG. 1 lattice: net momentum or translation in free space, and energy that is "never dissipated"? - What does "multiplier" mean in context? - Is any of it an energy or momentum claim that reviewers should weigh?

File changes

• facts.json.
• Add the grant of US 2006/0163971 as US 7,830,065 B2 (9 Nov 2010; 11 claims; claim 1 amended to a resonant circuit with a square-wave source; claim 11 "improves the efficiency"; CPC H02K53/00; lapsed 2022, exact date unresolved) [web].
• Add the changed text at grant.
• Note that claim 1 of 8,093,869 has the switched-load element and that the 2016 device uses one.
• claims.md.
• Record the grant.
• Note the 0036/0040 contradiction and the later change of stated energy source (magnets, then zero-point per the assignee, then ambient).
• Correct the claim 5 wording ("square of the sum of the energy").
• Add an evidence note on the non-patent performance claims (ESTC 2015/2016, McKay and Reiyuki, MPI/Chava), each with its source and stake.
• arbiter.md, add a "QE revision" section:
• correct 0.9015 to 0.9006 (n=1,3) and 0.9331 (n=1,3,5);
• give 12/327,719 as the US 8,093,869 application number;
• address the hole-geometry and cyclic-energy point (4.2). The verdict is unchanged, and the bound is now valid for nonlinear cores;
• add the switched-load link (4.1) to the claim 2 and claim 6 discussions, and to the measurement-artefact discussion;
• add the overunityresearch "abandoned a device" lead to the absence-of-evidence analysis;
• restore the related work: Tae Chang same-day filing, Southeast University magductance generator, Lovshin, Gray, TK/Kapanadze;
• note the QE verification of the McKay figures [web, URL above];
• weigh the deep-read results above.
• primer.md.
• In 3.2, add a paragraph saying that for patent 1's holed-core geometry the linear reflected-impedance picture does not apply (M ≈ 0 by symmetry). Give the cyclic bound Σ∮ i dλ = loss ≥ 0 in plain words, with symbols defined.
• Fix the Step E note (0.9006, not 0.9015).
• Add the switched-load / claim 1 link in Section 7 "genuinely open".
• Add the "abandoned" forum lead to the dark-period paragraph.
• Add Tae Chang, Southeast University, Gray and TK/Kapanadze to the related work.
• Add Hidden Energy (Manning & Manewich 2019) to Further reading as a secondary source.
• Add the MPI module and Volt claims as interested-party statements.
• sites.md (next Investigator run, not part of this remediation).
• Mark energeticforum page 1 as verified by QE on 2026-10-04.
• Priorities: the USPTO file wrappers 11/336,337 and 12/327,719; forum pages 2-4 and thread 13946; the archived MPI PDFs (download them to research/files/); the overunityresearch thread; the Gunderson 2019 video.

Recheck

Rechecked 2026-10-04 against the current facts.json, claims.md, arbiter.md, primer.md and example.py.

Defects (section 5): 1. facts.json / claims.md not updated: fixed. Grant as US 7,830,065 B2 is recorded in both; the 0036/0040 contradiction, the "square of the sum" wording note and the non-patent performance claims (ESTC, McKay) are in claims.md; the switched-load element is in facts.json. 2. arbiter.md application number: fixed. 12/327,719 is given (QE revision, absence-of-evidence section). 3. 0.9015 slip: fixed. arbiter.md gives 0.9006 and 0.9331 in place and in the QE revision; primer Step E now gives 0.9006 and 0.9331 and mentions 0.9015 only as a corrected slip. 4. mathematician.md correction 6: trivial, no change required (left as is).

Missed material and retention (sections 4 and 6): - 4.1 switched-load link to the 2016 device: fixed (arbiter QE revision; facts.json; primer). - 4.2 holed-core geometry and cyclic bound: fixed (arbiter QE revision item 2; primer 3.2b; deep reads appended). - 4.3 "abandoned" overunityresearch lead: fixed (arbiter, primer). - 4.4 related work (Tae Chang, Southeast University, Lovshin, Gray, TK/Kapanadze): fixed (arbiter, primer). - 4.5 assignee statements (MPI 2008, Goldes proposal): fixed, labelled interested-party. - 4.6 inventor background / Hidden Energy: fixed (primer Further reading). - 4.7 "multiplier element", FIG. 1 inchworm, published claim 7: fixed (arbiter QE revision items 4 and following; primer).

Remaining, not defects: lapse date of US 7,830,065 (9 Nov versus 12 Dec 2022) still unresolved and flagged as open; sites.md was not updated (assigned to the next Investigator run).

Final: PASS