Gray-EV: Quality Engineering (Step 4b)

QE date: 2026-10-03. Inputs checked: _text/manifest.json (10 documents), all 11 digest files, facts.json, claims.md, physicist.md, mathematician.md, contrarian.md, arbiter.md, primer.md, example.py. Spot checks used the extracted text in _text/ and three page images I rendered myself (crosby_letter.pdf p.1, static_generator.pdf p.1, various_articles.pdf pp.2-3), tagged [read: page image p.N].

Context: this project was reviewed before Step 0/1a existed. Step 1 worked from page images and poor text layers and said the four patents had "no text layer". Step 0 has since OCR'd every page (all 18 pages of US 3,890,548, all 7 of US 4,595,975 and US 4,661,747, all 5 of GB 2,030,801), and the digests cover the patent description pages that no reviewer read.

Verdict: MINOR. The energy-budget arithmetic, which carries the main physics verdicts, is correct and reproduced by example.py. But several load-bearing facts about who tested the machine and what they said are wrong or missing. They run through every review, the arbiter and the primer. The verdict labels will probably stand, but the reasoning behind Claims 2, 4 and 5 and the primer's evidence table must be corrected. Details below.

---

1. Coverage

No manifest document was skipped. No skipped file had text that should have been read.

2. Shortcuts

• Acceptable: digest-only treatment of ems_inverter.pdf, ems_electronic_power.pdf and the divisional US 4,661,747; reading only p.5 of GB 2,030,801 (the rest is drawings).
• Not acceptable: 1. Claim 1 (the motor itself) rests on the abstract, background and claims of US 3,890,548. The 8 pages of detailed description (pp.9-16, now OCR'd) were read by nobody. This is "sampling only part of a key document". 2. Claim 3's verdict on US 4,595,975 says "internal consistency unchecked" because only claims were read. The full text is now available. 3. Claims 2 and 5 rest on Step 1's summary of the garbled various_articles.pdf and gray_suppressed.pdf text. Three reviewers wrote "[read: Step 1 summary; not re-read]". That is a central experimental/historical claim resting on a summary of garbled OCR, and it led to the errors in section 4.
• Source-gap note is now out of date. claims.md, arbiter.md and primer.md all say the patents have no text layer and their descriptions were "read by nobody". Step 0 OCR now provides full text. The note should be rewritten: the remaining gaps are the missing test data, the missing court records, and the garbled tabloid page(s).

3. Missed material (from my reading of the digests, checked against text and page images)

1. Pan World Enterprises was a Japanese conglomerate, and a Japanese team walked away after a test. The TATTLER piece "Gasless Auto Gets $6-Million Backing" says J. A. Maize "conducted the intensive testing on behalf of Pan World Enterprises Company, Ltd., a Japanese conglomerate" [read: page image various_articles.pdf p.3]. The AP article quotes TRW's Donald Cronin: "A group of scientists from a big research firm in Japan had come with an elaborate amount of equipment - gauges, instruments, TV cameras - and were ready to sign a letter of intent. After the test, they packed it all up and went back to Japan." [read: page image p.2]. Gray himself told TATTLER, "We could have had all that Japanese money last year" [read: page image p.3]. So the party named in the Crosby letter as the tester was probably the instrumented group that declined to invest after testing. All reviews describe the testers only as "promoter-side" and the arbiter writes "no controlled test on record". The Japanese visit is the closest thing to an instrumented, interested-third-party test in the folder. Its outcome (they left without signing) is reported second-hand by a sceptic, and its data are unknown. This does not make it a controlled test, but the arbiter must weigh it under "tested / never tested / dismissed" for Claims 1-2.
2. Cronin's test ended in failure, not a pause. The AP text reads: "It ran for about 10 seconds and then everything blew," he said [read: page image p.2]. facts.json and the primer say it ran about 10 s and "wasn't convincing". That phrase is not in the article.
3. DA investigators invested in Gray themselves. Deputy DA Mitchell Harris: two investigators "went down to see and got hustled themselves - they invested money of their own for a while, but later got it back" [read: page image p.2]. This bears on Claim 5 (motive) and is missing from all files.
4. The charges were one count of conspiracy to commit grand theft, eight counts of grand theft and three counts of a securities violation [read: page image p.2]. facts.json says "three counts of fraud" (from the digest); the AP says securities violation. AP also gives "some 800 shareholders", Gray aged 50, selling shares from 1971.
5. The patent itself uses "static". US 3,890,548 says the delay coils 23, 24 form a "harness ... wound about the engine case" which is "believed to provide a static floating flux field" (coil 114), and Fig. 1 labels it "FLOATING STATIC FLUX FIELD" [read: gray3890548 text lines ~126, 780, 929; digest p.14]. claims.md and the arbiter say Claim 4 is "promoter and press claim, not in the patent text" and "patents do not use it". "Cold electricity" is indeed absent from the patent, but the static-field idea is in it. Claim 4's wording and the primer's section 4 table must be corrected.
6. The patent says "stores the excess energy". US 3,890,548 background: "Thus, the new engine produces torque and stores the excess energy for subsequent use" [read: text p.10]. facts.json says the patent "does not claim over-unity output". That is roughly fair (it also says "conservation of energy"), but the reviewers should quote "excess energy", which is the patent's own hint toward the stronger claim.
7. 3,500 V is in the patent drawing. Fig. 1 labels the capacitor banks "3500 VOLTS EACH" (OCR "3500 VeLrs Facn") [read: gray3890548 text p.2]. The arbiter calls the 3,500 V "unverified". The voltage is verified; the count of 18 capacitors still is not. The text also says there are nine stator electromagnet pairs, 40 degrees apart, and three rotor sets 120 degrees apart (digest p.10-11, p.16).
8. Further TATTLER claims not recorded: a "battery substation to supply 100 megawatts of electricity at peak periods while expending only 40 kilowatts"; funding from "a trio of wealthy U.S. geologists who made fortunes in oil exploration"; Italian government interest; Hackenberger's description that the motor "draws small amounts of direct current from a battery and blends it with static electricity" (a further Claim 4 source) [read: page image p.3]. The 100 MW / 40 kW claim is a clean second worked-example candidate (ratio 2,500:1).
9. Fascination car designer is Paul M. Lewis of Highway Aircraft Corp., Sidney, Nebraska, aged 77 (he designed the 1936 "Airomobile") [read: page image p.3]. The digests garble this ("Paul Innis", "Gray is now reported as 72"); facts.json has it mostly right but the primer does not use it. Peripheral.
10. Static generator sheet: "EvGray Static Generators", Generator-Only model about 45 W from a 12 V car battery to run 110 V lights and TVs, $50; Battery model 30 W for 6 hours, $75; "provides approximately 50% brighter illumination from a given light bulb wattage" and "no moving parts" [read: page image p.1]. The 50%-brighter claim is a small testable claim of the same kind as the inverter's "four times longer". Peripheral.
11. US 4,595,975 description (digest p.4-5) treats the counter-EMF during the power-delivery phase as "energy normally wasted" and also claims to harness "electrostatic or impulse energy" from the spark. Whether that is an over-unity claim needs a deep read before the Claim 3 caveat ("not read beyond claims pages") can be removed.

4. Accuracy spot checks

5. Process

• Source tags: present throughout and mostly honest; reviewers say clearly when they relied on "Step 1 summary; not re-read". That honesty is what made error 2 traceable. The primer tags the "deputy DA" quote as "[read: AP item, per contrarian]", but nobody had read the AP item; it should have been tagged as a second-hand summary.
• Web checks: budget was exhausted. Atmospheric-electricity values and patent dates were web-checked [web: en.wikipedia.org/wiki/Atmospheric_electricity; patents.google.com/patent/US3890548A/en]. Claim 5 legal facts failed web check and are labelled so. mathematician.md has no inline [web] tag; its Google Patents fetch is recorded only in a prose note (line 7). Acceptable, but inline tags would be clearer.
• Verdict scale: applied consistently. The arbiter correctly fixed the physicist's misuse of "Never properly tested" for a non-physical claim (Claim 5). Split verdicts are stated clearly.
• Primer vs STEP 4 and reader profile: all 9 sections present in order; symbols and acronyms defined (EMF, RLC, ODE, DA); worked example with real numbers; answers at the end; textbook chapters named only where safe. Good for the reader profile. Defects: the source-gap box is out of date; section 4's "what powers it" row and section 6's evidence table repeat errors 2, 3 and 5.
• example.py: runs cleanly with .venv/bin/python (sections A-E), and its numbers match arbiter and primer.

6. Verdict

MINOR. The physics verdicts (Refuted by calculation for the 10 hp / 26.8 W claim and the net-gain recovery claim; Established for ordinary flyback recovery; Plausible, untested for the mechanism) do not depend on the defective facts and should stand. The defects change the testing history and the dismissal narrative: there was an instrumented visit by the Japanese firm named as the tester, a sceptic eyewitness who saw the machine fail, and the "laws of physics" remark came from that eyewitness, not the prosecutor. Claim 4's "not in the patents" is wrong. These are fixable by updating facts, adding an arbiter QE revision, and editing the primer. If the deep read of various_articles.pdf shows that the Pan World / Japanese test produced recorded data or an explicit negative result, the arbiter should consider moving the "real machine" part of Claim 2 from "Never properly tested" to "Tested, not supported" (informal), and Claim 1 likewise.

7. Remediation actions

Deep reads (most important first; read page images for all press items):

1. various-articles-c05170, pp.1-6, Opus. Full read from page images. Questions: exact wording and attribution of every quote in the AP item (Cronin, Harris, the "energy engineering specialist"); what the Japanese group tested and when, and whether it was the Pan World testing named in the Crosby letter; all TATTLER numbers (26.8 W, 99%, 5,454 Wh, $6M, 100 MW / 40 kW substation, 500 miles between recharges); what Hackenberger says the motor draws; Paul Lewis and the "Fascination"; legal history items. Identify each separate article in the bundle and date it.
2. gray3890548-9b6412, pp.1-18, Opus. Full description. Questions: capacitor count, voltage and any capacitance values; the "static floating flux field" harness and what the patent says it does; the "excess energy" sentence in context; exactly how the overshoot reaches battery 10a and the safety-discharge return; Table I firing sequence; any statement that output exceeds input; the linear/hammer-blow embodiment.
3. US4595975-7db340, pp.1-7, Sonnet. Questions: does the description claim net energy gain or only reduced loss; what "harness electrostatic or impulse energy" and capture of CEMF "during the power delivery phase" mean in the text; any numbers, efficiencies or test results; role of the second voltage source and switch 48.
4. gray-suppressed-71bbc1, pp.1-2, Sonnet, from page images. Questions: dates and sequence of raid, charges and outcome; what TATTLER actually alleges about motive; any named officials, investors, test witnesses or numbers; separate the unrelated Lennon/Ono item.
5. ems-electronic-power-924df2, pp.1-3, Sonnet. Questions: demonstration details (battery voltage, 3,000 V reading, magnet catch); 788 stockholders and $2M; who Chalfin, Wester and Price were and what each said; legal outcome (plea, fine, date); magazine name/date if legible.
6. static-generator-124b3c, pp.1-1, Sonnet, page image rotated 90 degrees. Questions: full product claims (45 W, 30 W for 6 h, "50% brighter illumination", prices), any circuit or efficiency statement, and its link to the EMS inverter sheet.

facts.json: - Update how_read and text_available for all four patents (full OCR text now exists; text_available "yes", OCR quality noted). - various_articles.pdf: attribute "violate all the laws of physics" to Cronin (TRW) and the energy specialist; replace "wasn't convincing" with "then everything blew"; add the Japanese research firm's instrumented visit; add the DA investigators' investment; charges = conspiracy, 8 grand theft, 3 securities violation; add Pan World = "a Japanese conglomerate"; add the 100 MW / 40 kW claim; fix Lewis details. - crosby_letter.pdf: 99%, 26.8 W, 5,454 Wh, J. A. Maize; notary date 9/28/73; "test reports available at EvGray". - gray3890548.pdf: 3,500 V per capacitor bank confirmed from Fig. 1; nine stator pairs; "static floating flux field"; "stores the excess energy". - static_generator.pdf: replace the "could not be determined" entry with the product description.

claims.md: rewrite the source-gap note (patent text now available; remaining gaps are the test data, court records and garbled tabloid pages). Claim 2: name Pan World as a Japanese conglomerate per TATTLER. Claim 4: remove "not in the patent text"; say the patent describes a "static floating flux field" harness, and "cold electricity" and the atmospheric tap come from the press.

arbiter.md: add a "QE revision" section that: - corrects the "laws of physics" attribution and withdraws the "prosecutor dismissed it on theory" point (Claims 2 and 5); - adds the Japanese instrumented visit and Cronin's "everything blew" to the Claim 1-2 testing history, and decides whether "Never properly tested" still fits the real machine (likely yes, with the note "informal tests by an interested investor and a sceptic eyewitness reported as negative; no data published"); - corrects "3,500 V unverified" and Claim 4's "patents do not use it"; - adds the DA investigators' investment and the securities (not fraud) counts to Claim 5; - states the Claim 3 result of the US 4,595,975 deep read.

primer.md: update the source-gap box; section 4 table row "What powers it" (patent's static flux harness); section 6 evidence table rows 1, 2 and 5 and the "dismissed without testing?" paragraph; section 7's open questions (remove "unread description pages"). Optional: add the TATTLER 100 MW / 40 kW substation claim as a check-yourself question.

example.py: no change required. Optional: add the 100 MW / 40 kW ratio if the primer uses it.

Recheck

Recheck date: 2026-10-03. Checked the current claims.md, facts.json, arbiter.md (QE revision, items 1-10 and verdict table), primer.md and example.py against this QE's defect and remediation lists. example.py runs cleanly (sections A-F, including the new 100 MW / 40 kW case).

Defects (sections 1-2) - US 3,890,548 description unread: FIXED (Opus deep read pp.1-18; deep/gray3890548-9b6412.md; facts.json how_read updated). - US 4,595,975 description unread: FIXED (Sonnet deep read; arbiter QE item 7: no net-gain claim). - various_articles.pdf unread from images: FIXED (Opus deep reads from page images). - gray_suppressed.pdf not read from images: FIXED (Sonnet deep read). - static_generator.pdf digest "unreadable" / facts "could not be determined": FIXED (deep read from rotated image; facts.json updated). - ems_electronic_power and ems_inverter: FIXED (deep read for the former; "text layer sampled" wording gone). - US 4,661,747 digest-only: REMAINS, but accepted as stated in section 2 (divisional, near-identical text); facts.json says so openly. Not counted as an open issue. - Out-of-date source-gap note: FIXED in claims.md, arbiter.md and primer.md.

Missed material and accuracy errors (sections 3-4) - "Laws of physics" misattributed to the Deputy DA: FIXED in arbiter (QE item 1, superseded markers), facts.json and primer. The first-pass physicist.md, mathematician.md and contrarian.md still contain the old attribution; this is not on the remediation list and the arbiter flags it as superseded. Remains as a known stale statement in the raw reviews only. - Cronin "everything blew" (not "not convincing"): FIXED. - Japanese instrumented visit; Pan World as Japanese conglomerate; link left explicitly unresolved: FIXED. - DA investigators' investment; securities counts (1 conspiracy, 8 grand theft, 3 securities): FIXED. - Claim 4 "not in the patents" and "static floating flux field": FIXED (claims.md, arbiter, primer, facts.json). - "Stores the excess energy": FIXED (arbiter item 6). - 3,500 V confirmed, 18 capacitors (Fig. 2, not Fig. 1; the deep read corrected this QE): FIXED. - TATTLER 100 MW / 40 kW claim, Hackenberger, Lewis details: FIXED (facts.json, arbiter item 10, primer Step F, example.py section F, check-yourself question). - Crosby letter details (26.8 W, 5,454 Wh, notary 9/28/73, Maize): FIXED in facts.json. - Fascination, static generator sheet details: FIXED.

Remediation actions (section 7) - Deep reads 1-6: all done (deep/ holds all six files). - facts.json updates: done. - claims.md rewrite (source gap, Claims 2 and 4): done. - arbiter.md "QE revision" section with corrected reasoning and verdict table (labels unchanged): done. - primer.md (source-gap box, section 4 row, section 6 table, dismissal paragraph, open questions, optional 100 MW question): done. - example.py optional addition: done and runs.

Remaining items: none that affect a verdict, fact or the primer; only the stale first-pass wording in the three reviewer files (superseded by the arbiter's QE revision, not a required action).

Final: PASS