Bearden: Quality Engineering (Step 4b)

QE date: 2026-10-03. Inputs checked: _text/manifest.json, all 5 digests (digest/), facts.json, claims.md, physicist.md, mathematician.md, contrarian.md, arbiter.md, primer.md, example.py, with spot checks against _text/.txt and one rendered patent page. Context: this project was reviewed before digests existed. Step 1 read the 1993 paper in full but only sampled* the two 1994 updates and the patent. There is no deep/ folder, so no deep reads were run.

Verdict: MINOR. No central verdict is likely to change. All the material that was missed is either a further device claim that the same energy-budget argument already rules out, or evidence that makes the existing verdicts stronger. facts.json, claims.md, the arbiter (as a "QE revision" section) and the primer need updating, after two targeted deep reads.

---

1. Coverage

Every manifest document is digested or skipped for an acceptable reason. No coverage defects.

There is one bookkeeping inconsistency (minor). facts.json lists the content under the filename that the manifest marks as the duplicate ("...free energy.pdf"), and gives the manifest's primary file ("...(1993).pdf") how_read: "filename only". Because the files are byte-identical this does no harm, but the paths should be swapped so that facts.json matches the manifest.

2. Shortcuts

One shortcut was fixed downstream. The arbiter's main open question ("what do Figs. 6-8 show?") was answered in the primer: the largest measured COP is about 2.6 at 75 V, and 3.43 comes from the fitted curve extended to 100 V. I confirmed this from a rendered image of patent sheet 2 (Figs. 7 and 8): the plotted points run from COP about 1.7 at 10 V to about 2.6 at about 75 V, and output power reaches about 22 W at 75 V [read: page image p.3]. This result has not been carried back into facts.json, claims.md or the arbiter. The arbiter still says the figures were unread.

3. Missed material (from the digests and spot reads)

1. The Fogal charge-blocking-semiconductor (CBS) overunity device [read: pp.46-49, digest 03]. No file mentions it: Fogal, "charge blocking" and "negative resistor" get zero hits across facts, claims, all three reviews, the arbiter and the primer. The device works like this. A CBS blocks dm/dt but passes dPhi/dt into a step-up transformer primary (Fig. 1). An ammeter reads dPhi/dt as ordinary current. Bearden claims "very high variable power gain". Fig. 2 closes the loop, so that after start-up the power supply is switched out and the device runs as a self-powered "negative resistor". It is a second concrete self-running device claim, beside the MEG. It falls under the same "Refuted by calculation" (closed-cycle energy budget) as claims 3 and 5, but it should be listed and given a verdict.
2. The MEG patent's own statement of its energy source [read: patent p.12, col. 11]. The patent calls the device "an open dissipative system ... receiving, collecting, and dissipating energy from ... the magnetic flux stored within the permanent magnet". It says "continued operation ... causes demagnetization of the permanent magnet", recommends rare-earth magnets for "a relatively long life", and compares the device to a nuclear reactor's fuel rods. This makes the mathematician's magnet-energy bound (11-22 J stored, used up in under 1 s at 82 W excess) directly on point, not merely "supporting", as the arbiter rated it. The patent itself names the magnet as the fuel. Only the arbiter mentions demagnetisation, and only in its own words. No file quotes the patent's concession.
3. Internal inconsistency in the patent's pulse width [read]. The description gives 11.45 microseconds per FET pulse (consistent with 87.5 kHz). Claims 5 and 17 (text lines ~877, ~902) give "approximately 11.5 milliseconds", a factor of 1000 different. No file notes this. It is a small point, but it matters for a reader judging how carefully the patent was prepared.
4. Bearden's own concession on ordinary capacitors [read, p.45]: "With ordinary capacitors ... one can readily demonstrate that the efficiency can approach 1.0 rather than 0.50 as expected." This is the standard adiabatic (step) charging result, and example.py PART B reproduces it exactly. The arbiter quotes the related "rigidized" passage but not this sentence. Quoting it would show the reader that Bearden's own claimed observation is just the textbook loss reduction, with no gain above 1.
5. Energy formula typo/error [read, p.45]: "K = ½(CV)²" (it should be ½CV²; as written the units are C², not J). The mathematician, who checked units carefully elsewhere, does not list it. It should be added to the "Internally inconsistent" evidence, or noted as a likely typo.
6. SQUID detection of displacement-current magnetic fields [read, p.45]. Bearden cites this as proof that dPhi/dt is real. The magnetic field between charging capacitor plates is real and has been measured [memory, medium; needs a web check before use]. It confirms Maxwell's displacement current, not Bearden's split of charge into massless and mass parts. A fair point to raise in the contrarian and primer section on claim 2.
7. Other unrecorded items, all peripheral: - Operational efficiency alpha = (average load power)/(average power dissipated inside the source) (p.31-32). This is a non-standard definition that can exceed 1 while the true efficiency is below 1. It is a useful teaching point for the primer. - The claim that a MOSFET with "25,000 internal components" shows phase-conjugate behaviour (p.35-36). - The claim that rising EM signal density will be biologically harmful, plus a promised cancer cure (p.10-11; facts mentions the cancer paper only briefly). - The claim of 90% utility-bill savings and home units for "a few hundred dollars" (p.18). - The digest gives Kenny's name as missing from the inventors. facts.json correctly includes James L. Kenny. The OCR shows "James king ...", so the digest is the one in error.

4. Accuracy spot checks (against _text/)

No load-bearing number was found to be wrong.

5. Process

• Source tags: present throughout, and honest. facts.json labels the sampled sections "text layer sampled" and the arbiter repeats this. [memory] claims carry confidence levels. The arbiter marks search-summary web results with "page not opened by me", which is good practice.
• Web checks: the mathematician used 5 of 6 and the arbiter 5 of 8, with URLs given. The medium-confidence facts that verdicts depend on (charge neutrality, Maxwell 1865 gravity, the Naudin MEG test, Sweet VTA) were upgraded to [web]. The NdFeB (BH)max (300 kJ/m³) is still [memory, medium] and supports the magnet-drain bound. It now matters more because the patent names the magnet as the energy source (section 3.2), so it should be web-checked.
• Verdict scale: applied consistently, with split verdicts clearly separated. One label is off-scale: the arbiter's "Refuted by the historical record" for Maxwell-gravity is not on the Step 2 scale. Restate it as "Tested, not supported" (historical evidence) or "Not testable as stated", with the historical evidence as reasoning.
• Primer: follows all 9 STEP 4 sections in order plus answers. It suits the reader profile: symbols are defined, worked numbers are given, and it adds the [read: page image] Fig. 6-8 reading, which is the most valuable new evidence in the project.
• example.py: runs cleanly with .venv/bin/python (exit 0). PARTS B-E reproduce the collector budget (COP ≤ 0.5 for a single step, approaching 1 with N steps), the relaxation times, the MEG arithmetic and the phase-angle power demonstration. Its numbers match the arbiter and the primer.

6. Verdict

MINOR. The verdicts for claims 1-5 hold. The missed Fogal/CBS device needs its own entry and verdict, but it will fall under the existing "Refuted by calculation" for self-powered closed loops. The patent's demagnetisation passage strengthens the claim 5 refutation rather than changing it. The defects are: incomplete reading of the 1994 updates; Fig. 6-8 findings not carried back from the primer; the patent's ms/µs inconsistency; the ½(CV)² error; and one verdict label that is not on the scale.

7. Remediation actions

Deep reads 1. Bearden-Final-secret-of-free-energy-1993-d9aa85, pp. 42-49 (Opus): read the 13 May 1994 update in full. Cover the Phi(m) charge decomposition, the rigidized-capacitor and SQUID arguments, the K = ½(CV)² statement, and the Fogal CBS device of Figs. 1-2 (render pp. 47-49 as images). What exactly is claimed for power gain and self-powered closed-loop operation, what measurements (if any) are reported, and how does the dPhi/dt-versus-dq/dt argument hold up dimensionally? 2. Bearden-Final-secret-of-free-energy-1993-d9aa85, pp. 28-36 (Sonnet): read the 15 Feb 1994 update in full. Cover the step-charged capacitor collector, the definition of operational efficiency alpha and how it relates to true efficiency, the claims about patent filings, and any reported measurement. Does Bearden report any experimental result for step charging beyond "efficiency approaches 1.0"? 3. US6362718-383b35, pp. 11-15 (Sonnet): read the experimental section and claims 1-29 in full. Does the 14 W input include the oscillator, flip-flop and one-shot supply? What is the exact wording of the self-switching and demagnetisation passage? How does the pulse duration in the claims (ms) differ from the description (µs)? Is Fig. 7's output power per coil or total?

facts.json - Swap the paths so that the primary file follows the manifest. - Update how_read for the two 1994 updates and the patent after the deep reads. - Add the Fogal CBS device to the 13 May 1994 entry's key_claims. - Add to the patent entry: the measured COP of about 1.7-2.6 [read: page image p.3]; the magnet named as the energy source, with demagnetisation conceded; the ms/µs inconsistency.

claims.md - Claim 5: add "measured COP up to about 2.6 at 75 V; 3.44 is a curve extrapolation; the patent names the magnet's stored flux as the energy source". - Claim 3: add the CBS self-powered "negative resistor" variant (or make it a sub-claim).

arbiter.md: add a "QE revision" section that: - records the Fig. 6-8 results and closes the "unread figures" open question; - moves the magnet-energy bound from "supporting" to direct evidence, citing the patent's demagnetisation passage (web-check NdFeB (BH)max first); - gives the Fogal CBS device a verdict (expected: self-powered net output Refuted by calculation; reported gain Never properly tested, since no data are given); - notes the ms/µs inconsistency and the ½(CV)² unit error; - restates "Refuted by the historical record" on the Step 2 scale.

primer.md - Add the CBS device to the evidence table. - Quote the patent's own "energy comes from the magnet; it demagnetises" passage in 5b. This is the clearest teaching moment: the magnet stores about 11-22 J, so how long could it supply 82 W? - Quote Bearden's "efficiency approaches 1.0 rather than 0.50" next to example.py PART B. - Add the alpha-versus-efficiency distinction to section 7.

Recheck (Sonnet): after remediation, append a "Recheck" section here.

---

Recheck

Recheck date: 2026-10-03. Checked against the current facts.json, claims.md, arbiter.md, primer.md, example.py and deep/. File timestamps: primer.md was last written 2 Oct 15:15, before qe.md (3 Oct 09:25), so it has not been touched since the QE.

Defects (section 6)

Remediation actions (section 7)

Open issues

1. The 15 Feb 1994 update (pp.28-36) was never deep-read. The alpha definition rests on sampling and earlier reviewers' quotes. No verdict depends on it, but the QE required the read.
2. primer.md was not updated at all. It lacks the CBS device, the patent's demagnetisation quote, the 1.0-vs-0.50 quote, the alpha distinction, the ms/us and (CV)^2 notes, and the corrected verdict label.
3. NdFeB (BH)max is still [memory, medium] and unverified on the web, although the arbiter says the magnet bound is now direct evidence.

Final: OPEN (3 issues)

---

Recheck 2

Recheck 2 date: 2026-10-03. Checked against the current facts.json, claims.md, arbiter.md, primer.md and deep/. Timestamps: deep/ 1993 file 11:15, facts.json 11:19, claims.md 11:20, arbiter.md 11:21, primer.md 11:23.

Cosmetic leftovers, not counted as issues: facts.json line 8 and the claims.md caveat still say "no separate deep/ file" for the 1993 paper pp.1-27, which is now stale for the 15 Feb 1994 update.

Final: OPEN (1 issues)

Evidence-rule check (2026-10-05)

Checked arbiter.md and primer.md against the four SOURCE HONESTY rules (no suppression of evidence, interested parties, absence of evidence, keep pertinent information). A grep for dismissive labels (uncorroborated, hearsay, single source, unverified, conspiracy, advocate, proponent, not neutral) finds them only in arbiter.md lines 225-241, where the revision log describes their removal, and in "not dismissed without testing" verdict wording, which is not evidence labelling.

• No suppression: single-source items are quoted with source and status (Sweet VTA, arbiter 34 and 90; Bearden's COP 17-20 and >100, arbiter 39 and 113; the university/DoD outcome, arbiter 40 and 115; the "suicide circuit" quote, arbiter 204). Evidence against the claims is also stated (Naudin's failed loop closure and coil failure, arbiter 38 and 114).
• Interested parties: Bearden (co-inventor, company spokesman), the patentees, Naudin, Fogal, ZPEnergy and the archive sites are each identified with their stake (arbiter 10, 110, 165; primer 9, 251). The USPTO grant and the Evans et al. paper are placed correctly (arbiter 110, 116).
• Absence of evidence: the silences (MEG after 2001, Sweet VTA, Fogal CBS, the collector) each have competing explanations, evidence for and against, and the records that would decide (arbiter 79, 90, 118, 168; primer 255-268). Suppression is listed as one explanation and weighed, not assumed or excluded.
• Keep pertinent information: Naudin's 1.75 COP, coil destruction above 3 kV and failed loop closure, Bearden's COP claims, Fogal's own patents and the stalled replication are all retained beside the verdicts.

Non-blocking notes (no evidence-rule breach; counted as 0 issues): 1. arbiter.md line 52 states "nobody dismissed it" with no search behind it; it is an unsourced statement of absence. Could read "no dismissal found in the sources searched". 2. arbiter.md lines 143-148 ("Open questions") still say nobody has read Figs. 6-8 as images and ask whether the 14 W includes the logic supply. Both are closed in the QE revision (lines 159-160, 216) and the primer, so the early list is stale, though the closure is stated. 3. arbiter.md line 113 attributes the physicist's "COP 5 and 10" to "the same family of Bearden accounts" without a source.

Final: PASS

Investigator update check (2026-10-05)

Checked: the "Investigator findings (2026-10-05)" sections of physicist.md, mathematician.md, contrarian.md and arbiter.md; primer.md (sections 5c, 6, "Where the record goes silent"); sites.md; the six research/ notes; example.py Part F (runs under .venv/bin/python, Part F prints 2.08, 2.3e-16 / 2.5e-13 ohm, 4.1 kW/m^2, f >= 0.59/0.50/0.38, matching the arbiter and primer).

What holds: - Retention of the main finds: all eight Investigator items (2004 file wrapper and mechanism, demonstration request and abandonment, Naudin v2.1, Flickinger, Tektronix, EP grant and removed COP plot, assignment/Colmac trail, 15 Aug 2002 letter, Poynting remark, prior-art list, Alabama search) appear in all three reviews, in the arbiter's table "What each new finding adds" and in the primer. Negative or contrary judgements (physicist's "Tested, not supported" split, contrarian's relabel of claim 4) are stated and the arbiter's reasons for not adopting them are given. - SOURCE HONESTY: single-source items are quoted with source and stake (Flickinger, Naudin, Bearden 2002, Tektronix reply relayed by a builder); no dismissive labels; the second-hand status of every [web] item (nothing downloaded, pages not re-opened, 50,000 of 85,447 characters read) is stated in arbiter, reviews and primer; the absence-of-evidence analysis for claim 5 is updated with for/against per explanation and the deciding records; arbiter-run web checks are disclosed (2 of 8, one refused connection). The NdFeB retag to [web] rests on search summaries with pages not opened, and says so (accepted). - Primer matches arbiter: verdict table rows 4f-4h, 5a-5d, the 2.08 ratio, the three "1.75" values, the 7-22 J bound, the two-descriptions-of-one-fuel tension, the silence section and the "who is speaking" paragraph all agree with the arbiter's final table.

Defects: 1. Untagged claim. arbiter.md (item 7, USPTO continuation) and primer.md (section 5c, USPTO bullet) state "The parent patent was granted in 2002 without such a demand." No source supports this: the parent file wrapper 09/656,313 was not found (archive.org "Item cannot be found") and was not read by anyone. Fix: say "no such demand is known; the parent file was not obtained", or tag it and cite. 2. Stale and conflicting figures in arbiter.md. Check list item 5 (line 35) still says NdFeB (BH)max about 300 kJ/m^3 is "Still [memory] only", and the energy-budget bullet (line 105) gives 11.6-22 J and 0.14-0.27 s with [memory, medium]; the Investigator section (line 253 on) retags the figure [web] and uses 7-22 J (primer: 0.09-0.27 s). The older lines carry no pointer to the update, so the arbiter contradicts itself on tag and range. Fix: add "(updated 2026-10-05, see Investigator findings, web check 11)" or edit the two lines. 3. Retention gaps (pertinent Investigator material not carried into arbiter or primer; none was dropped on a negative judgement, but the rule covers the primer and arbiter). From research/ and sites.md: (a) Bearden's 2004 remarks also invoke negative energy from Dirac-sea holes and Bedini, "time-energy" E = tc^2 converted to spatial energy, Prigogine dissipative structure, Hilbert/Logunov local versus global conservation and Halliday and Resnick's COP 9.22 heat-pump maximum: absent from arbiter and primer (physicist and contrarian F1 list some, none of the three reviews mentions Dirac/Bedini or time-energy; the primer omits Prigogine); (b) the examiner's other grounds (35 USC 112, double patenting over claims 1-29 of the parent, title seen as "a transformer which relies on ... perpetual motion", missing page 2, Newman v. Quigg) are omitted; the double-patenting ground is what supports the "redundant continuation" explanation for abandonment, which both documents use; (c) Bearden's 15 Aug 2002 "vertical cliff" claim that every legitimate COP > 1 system was defeated at the scale-up funding stage is in the arbiter table only, not in the primer, which cites the letter without it; also omitted: the 10% of sales for poor countries, his 2002 book, and the "two others in various stages" patents link; (d) company-records leads: the Manta "noncommercial research organizations" listing (in the arbiter only), the Facebook page, the pdfcoffee P.O. Box 1472 address and the ZPEnergy "Letter to JLN Laboratories" offering to develop the MEG with Magnetic Energy Ltd are in neither primer nor arbiter. Fix: add one sentence each (or a list under the Investigator section) so the record keeps them.

Non-counted notes: the physicist's section keeps the NdFeB estimate as [memory, medium] and its own 3-4 J, which the arbiter supersedes with the patent's own magnet volume (acceptable, stated by the arbiter); research/files/ is empty, so no primary page was archived (disclosed everywhere).

Final: OPEN (3 issues)

Recheck (2026-10-05)

Checked against the current arbiter.md and primer.md (grep for the old wording, then read of the replaced passages).

1. Untagged claim "granted in 2002 without such a demand": Fixed. arbiter.md item 7 (line 271) and primer.md 5c USPTO bullet (line 254) now say the parent was granted 26 March 2002 [read: front page], that whether a utility objection or demonstration demand was made is unknown, that "no such demand is known; the parent file was not obtained" (archive.org "Item cannot be found"), with the source and a correction note. No remaining hit for the old wording in either file.
2. Stale NdFeB tag and range in arbiter.md: Fixed. Check-list item 5 (line 35) and the energy-budget bullet (line 105) now carry "[updated 2026-10-05 ... web check 11]" pointers, the range 190-450 kJ/m^3 giving 7-22 J and 0.09-0.27 s, and the recomputation (7.35 / 17.4 / 22.2 J). The old 300 kJ/m^3 and 11.6 J figures are kept as the middle of the range, so the file no longer contradicts itself. Primer already uses 7-22 J (lines 228, 340).
3. Retention gaps: Fixed. (a) Dirac-sea/Bedini, time-energy (with a units check), Prigogine, Hilbert/Logunov, Halliday and Resnick COP 9.22: in arbiter 297-301 and primer 117-121. (b) The examiner's other grounds (35 USC 112, double patenting, perpetual-motion title, missing page 2, Newman v. Quigg), with double patenting tied to the "redundant continuation" explanation without excluding others: arbiter 303, primer 254. (c) "Vertical cliff" (marked as Bearden's characterization), 10% of sales, 2002 book, "two others in various stages": arbiter 304, primer 258. (d) Manta, ChamberofCommerce, Facebook page, pdfcoffee P.O. Box 1472, ZPEnergy "Letter to JLN Laboratories": arbiter 305, primer 260, flagged as directory traces, not registration records. All carry source tags and stakes; nothing is dismissed.

No new SOURCE HONESTY breach found in the replaced passages.

Final: PASS