Eskridge-Amy: Quality Engineering (STEP 4b)

QE: Opus, 2026-10-07. Inputs read: _text/manifest.json and _textprep.log, all 11 digests, all 5 deep reads, facts.json, claims.md, sites.md and all 10 research/ notes (research/files/ is empty), physicist.md, mathematician.md, contrarian.md, arbiter.md, primer.md, example.py (run). I also checked passages against the extracted text in _text/ and listed the source folder (read only) to confirm the manifest is complete. Tags: [read] project document; [web] URL; [memory] with confidence; [calc] my own calculation (Python, .venv, 2026-10-07).

Verdict: MINOR

The work is complete and careful. Every document was digested, and every central document was deep-read. Source gaps, interested parties and absence-of-evidence analyses are handled well, and the primer is a strong study document. No central verdict needs to change. The defects can be fixed by updating facts, claims, arbiter and primer:

1. One arithmetic error in the arbiter, made with confidence (a sign error). Its "recomputation item 4" says that Table 1 of the NASA memorandum is not reproduced by eq. 28a. With the memorandum's own sign convention it is reproduced exactly. The mathematician was right, and the deep read (part 1, section 4), the physicist and the arbiter are wrong on this point. No verdict depends on it.
2. A hidden assumption that was never checked behind one "Refuted by calculation" verdict. The memorandum names four ways of aligning the bismuth nuclei. No reviewer evaluated two of them (Overhauser transfer and field-gradient flipping), and only Barnett was quantified. The memorandum's pages 25-32, which describe the rotor's design and mechanism, were covered by the digest only.
3. Retention gaps in the primer and arbiter. Several items that favour the claim, or bear on the historical record, were left out: the "pristine disc" argument, the memorandum's motor-phasing argument against a current-reversal artifact, the constant magnet-field measurement, Ning Li's documented fate, and the death ruling together with the resurfaced text.
4. Material in the digests and source text that no later file uses. This includes the Maxar plasma-thruster patent negotiation, the Austin investor, the identity of "Chantelle", the YBCO "Jim Ashworth" lead, the Tesla patent characterization and Robert Becker's role on the Institute staff.

---

1. Coverage

• The source folder holds exactly the 9 manifest documents. There are no skipped files and no unextracted text.
• There are 11 chunks and 11 digests. Coverage: PASS.

2. Shortcuts

• NASA TM pp.25-32 (orbiting-gyroscope experiments V1/V2; the Bi-209 nucleus model; the Einstein-de Haas and Barnett background; the V3 rotor design). These pages are covered by the digest and by spot reads of p.25 (arbiter) and pp.33-36 (fact-finder, physicist).
• Pages 31-32 state how the rotor was supposed to align nuclei: Barnett self-magnetization, Overhauser transfer from electrons to nuclei, a field gradient that "flips" a fraction of the nuclei, and the Wallace effect [read: digest p.31-32].
• The arbiter's verdict that the 0.3 g result is "Refuted by calculation" as "consistent with the theory" explicitly assumes that thermal polarization is the only alignment ("The calculation is conditional", arbiter claim 3). Only Barnett was quantified. Overhauser and gradient flipping were not evaluated by any reviewer (grep: "Overhauser" appears only once, in primer.md, without a number).
• This is a load-bearing step that rests on the digest alone. Defect (minor): deep read DR1.
• My own expectation, [memory: high], for the reviewers to check: Overhauser enhancement needs the electron spin resonance to be saturated, normally by microwaves, which the rotor does not have. A static gradient, and slow motion through it (rotation at about 10²-10³ Hz against a Bi-209 Larmor frequency of about 1.4 MHz at 0.2 T), keeps the populations at their Boltzmann values. So the verdict is unlikely to change, but it should be shown, not assumed.
• Interview ranges not deep-read: 00:00-00:30, 00:59-01:25, 02:58-03:15 (partly read by the fact-finder).
• These ranges hold the disclosure plan, the neutron-beam/DNO talk, the "six forces" passage, the Austin investor and "Hal" (00:47), the Maxar patent (00:57:56), the "prototypes scrapped" generalization (01:16) and the closing plans.
• The arbiter's attribution uncertainty for "six forces" and "German Air Force" rests on these unlabelled passages. The physics claims made there are peripheral, but several historical-record leads were missed (Section 3). Defect (minor): deep read DR2.
• Interview audio. Whose voice is whose in the interview, and who the first-hand voice in the lecture Q&A is, cannot be settled from captions. This is a record to obtain, not a deep read (already in arbiter "Open questions").
• Digest-only coverage of the program page, the two website captures and the README is acceptable: they are peripheral or promotional.

3. Missed material (in the digests or source text; absent from facts.json, claims.md, the reviews, arbiter and primer unless noted)

1. Maxar plasma-thruster patent (interview 00:57:56). Amy, reading her calendar: "i was negotiating with maxar about selling them a plasma thruster patent on the 17th of december" (2019) [read]. The Investigator found that US7808353B1 (the plasmoid-thruster coil patent) was assigned from NASA to Richard H. Eskridge on 2018-10-03 [web: https://patents.google.com/patent/US7808353B1/en]. - Together these are a documented commercial activity of the family business. - They bear on the "funding failure" and "legal/IP" explanations in arbiter "Absence of evidence A". - The fact is in facts.json; the link to the patent was never made.
2. The Austin investor and Hal (interview 00:47:14-00:47:44). "my dad got poached by this weird ass investor ... out of nasa ... who lives in austin texas ... this investor brings hal to our [house]" [read]. - The TM says Nelson's Space Act Agreements included partners in "Austin, Texas" [read: TM p.37]. - Quantum Machines LLC (the TM's funder, CEO Chris Milam) is the obvious candidate, but the TM gives no location for QM. This is a lead, not a finding. - It bears on claim 3 (who funded and holds the V5 data) and on claim 2.
3. "Chantelle" in the Tajmar confrontation (interview 01:25:34) is almost certainly Shantel Butler. She was the Institute's research director and Chief Research Officer of Geometric Energy Corporation [read: website 2018]. - The passage is first-person ("chantelle and i ... my dad tried to get me to replicate eugene's experiment when i was a ... junior in high school"), so the monolayer/bilayer argument is Amy's. - Only the "funded by the German Air Force" sentence follows an interviewer's interjection ("basically what you're trying to say is") and stays uncertain [read: 01:26:06-01:27:47]. - The arbiter's "Corrections" (claim 4: "The same applies to the German Air Force remark") and primer 4a ("speaker not certain" on the monolayer row) should separate the two. Butler is a possible second witness to the 2018 meeting.
4. The father's long interest in Podkletnov. "my dad tried to get me to replicate eugene's experiment when i was a junior in high school" [interview 01:26:06], and the digest's "2004 science fair" line. Context for claim 2, which none of the downstream files carries.
5. YBCO priority story, lecture 00:19:51-00:20:23 and 00:28:36. She says YBCO was "invented out of UAH by a grad student" and the patent "stolen by another university". In answer to "Guy that invented YBCO?", someone says "It was Jim Ashworth". - This is very probably J. R. Ashburn, the UAH co-author of Wu, Ashburn, Torng et al., Phys. Rev. Lett. 58, 908 (1987) [memory: medium-high; web check needed]. - The "stolen patent" claim is checkable against the published Wu/Chu priority history. The facts.json [memory] note names Wu and Chu but not the transcript's name.
6. Robert E. Becker. In the lecture (00:21:27-00:22:32) she says she photographed his UAH quantized-gravity thesis and that its loan record shows a 1999 "DIA" checkout. The 2018 website lists Becker as the Institute's "Aerospace Engineering and Quantum Gravity Technical Advisor" [read: website 2018]. - No later file mentions Becker. The loan claim is a checkable library record, and his being on her staff is relevant to her description.
7. Tesla patent 1,655,144 (slide p.12). The slide calls it a "space drive" or "anti-electromagnetic propulsion system". US 1,655,144 is Tesla's 1928 "Apparatus for aerial transportation", a VTOL aircraft [memory: high; web check needed]. - The slide's image source is William Lyne's "Occult Ether Physics" (deep read p.12), an interested popular source. - Nobody checked it. It is a factual error on her slide, and it should be recorded as hers next to the source she used.
8. "We told management about this" (lecture 00:51:12) and the first-hand voice's "we did see a whole lot of really strange effects. Like the tossing of the magnet and forces coming off the disc" (00:51:42) [read]. - The first is a record lead (MSFC management memos) for claim 2. Only contrarian.md mentions it. - The second is the speaker's own positive summary. It belongs beside his null on Poher's signals, which the arbiter and primer quote.
9. The PhD "court case" (interview 02:04:57-02:05:28). She says her biotech PhD was ended by "a sexual harassment thing ... a really horrible court case", which she calls a "fake vehicle of sabotage" [read]. A court record would exist. It bears on claim 6 and is listed in the deep read but in no later file.
10. The TM's own argument against a power-lead artifact (p.33). "A multiphase motor was used ... the direction is controlled by relative phasing of the three pulsed DC signals ... This kind of motor does not have a reversal of DC current which might explain a reversal of forces" [read: TM text, page 33].
    • The physicist lists "drive-current (power-lead) forces" as a candidate artifact without noting that the memorandum addressed it.
    • No file records this argument. See Retention.
11. Small slips in Step 1 files.
    • facts.json (HAL5 deck summary) says "slide p.27 says Torr's work 'never made public'". It is slide p.28 [read: slide text p.28].
    • The lecture deep read leaves "Boyer" unresolved. The slide deep read and the arbiter identify Poher (consistent with slide p.29), so the lecture deep read is the only file that keeps the doubt.

4. Accuracy spot checks (against _text/)

5. Retention check

The arbiter is mostly faithful. It keeps both sides of claims 2, 3 and 6, Tajmar's offer to test her samples, the Sokol positive-correlation claim, Robertson's 2025 replication, Ning Li's record, the death ruling, the resurfaced text and the father's statement. The primer condenses further and drops several pertinent items, some of which favour the claim. The rule "keep pertinent information" applies whatever the verdict.

No item was found that the arbiter or primer removed and explained as removed because a reviewer judged it negatively. The gaps above look like condensation. They are still defects under the rule.

6. Process

• Source tags. Present and mostly honest throughout. [read] tags carry page numbers or timestamps; [web] tags carry URLs; [calc] is used for the reviewers' own arithmetic.
• Exceptions: the "[read: slide p.25]" mis-locator (Section 4, item 6), and facts.json "p.27" for p.28.
• The arbiter's reliance on "search result summary" for the Mansfield statute and the NewsNation and resurfaced-text items is flagged as such. Acceptable, but the resurfaced text itself was never seen. The primer should not cite it without that caveat if it is added.
• Web checks.
• The arbiter used 7 of its 8 checks and the physicist several. The Physica C 1997, Hathaway 2003, Quinn-Picard and Gallerati items were verified.
• Still [memory: medium] and unverified, though a stated physics point uses it: the torsion-balance and comagnetometer bounds on spin-gravity couplings (physicist, contrarian, primer section 3). The arbiter says no verdict depends on it; the primer still uses it in the textbook view. Web-check one source (for example the Eöt-Wash spin-pendulum results) or soften the wording.
• Not checked by anyone: Tesla patent 1,655,144 (Section 3, item 7) and the identity of "Jim Ashworth" (item 5).
• The 1973 Mansfield revision is still unresolved. The arbiter states this openly.
• Verdict scale.
• The arbiter's verdicts use the Step 2 names exactly. Its splits are well argued (for example "Never properly tested" for original-condition Podkletnov against "Tested, not supported" for the reduced versions).
• The contrarian's "Plausible" and the mathematician's "partially documented" are not scale names, but the arbiter normalised them.
• One inconsistency: the primer's section 3 and its evidence-table row "Gyroscope weight change (Laithwaite, Hayasaka) ... Tested, not supported" puts Laithwaite under the 1990 nulls. Those nulls tested the Hayasaka-Takeuchi claim, not Laithwaite's demonstrations. The arbiter gives no Laithwaite verdict, and the physicist lists Laithwaite under "Never properly tested".
• The primer's section 1 also lists "Hayasaka" among the devices she surveyed; Hayasaka is not in her slides or lecture.
• Primer.
• It follows STEP 4's nine sections in order, with answers at the end. Symbols are defined at first use (the Curie law, the Boltzmann factor, fm and u), and the worked example is numerical and reproduced.
• Its section E3 (unaligned bismuth averaging to −0.6 × weight) is a new calculation. The primer flags it for QE; it supports an existing verdict and adds none. Acceptable as an illustration, provided the assumption about how eqs 28b and 31 combine stays stated (it does). A reviewer should look at it in the QE revision (remediation item A6).
• Reader-profile fit is good. "Brillouin sum" in the arbiter is not reader-facing.
• example.py runs and matches the primer.

7. Corrections to earlier documents (for the orchestrator)

1. arbiter.md, "Where I recomputed", item 4. - The statement "Neither normalisation gives 2.0000 above 10 Hz. The mathematician's correction is not confirmed" is wrong, and so is "solving the table's co- and anti-spin entries for the 'spin term' gives different values for the two branches (e.g. 32.1 N vs 50.2 N at 60 Hz)". - Correct statement: Table 1 is reproduced exactly by eq. 28a with F_N = 89.85 N (disc mass), spin term 39.12 N at 60 Hz, and Δf = F_zero-spin − F_table (Appendix B: Fmeas = Fgyro − DF). The identity holds to 1e-6 at all rows [calc, Section 4 item 4]. - Also remove the open question "What did the Appendix B sheet do to make Table 1 differ between co-spin and anti-spin branches?" and the claim-3 sentence "Table 1: ... not reproduced from the printed equation, unexplained". - What stays: Table 1 contradicts the centre-of-mass theorem, and it mixes disc and gyroscope masses (stated openly in the memorandum).
2. physicist.md, Correction 6 and the claim-3 bullet "Table 1 ... reproduced by eq. 28a only at 10 Hz", and deep/NASA-TM part 1, section 4: the same error. The mathematician's correction stands.
3. arbiter.md "Corrections", six forces / claim 4, and primer 4a: the monolayer/bilayer argument is Amy's (first person, with "Chantelle", probably Shantel Butler, the Institute's research director). Only "funded by the German Air Force to publish no results" has uncertain attribution [read: interview 01:25:34-01:27:47].
4. arbiter.md and primer.md "[read: slide p.25]" for "I don't know what happened" should read [read: lecture 00:18:16].
5. facts.json (HAL5 slides summary): "slide p.27 says Torr's work 'never made public'" should read slide p.28.
6. primer.md section 3 and its table: the 1990 nulls tested Hayasaka-Takeuchi, not Laithwaite. Move Laithwaite to "Never properly tested", as the physicist has it (her own slide says Newton's second law explains it). Drop "Hayasaka" from the list of things she surveyed in section 1, or mark it as the reviewers' addition.

8. Remediation (one round)

Deep reads (most important first)

Web checks to run with the remediation (no deep read needed)

• US 1,655,144: its title and subject.
• J. R. Ashburn and the authorship of Wu et al., PRL 58, 908 (1987), plus the UAH/Houston patent priority history.
• One citable source for spin-gravity coupling bounds (torsion-balance spin pendulum or comagnetometer).
• If possible, Quantum Machines LLC's location (Austin?).

What each later file must change

• facts.json
• HAL5 slides: correct "p.27" to "p.28". Add the Tesla patent check result, recorded as an error on her slide with the Lyne book as its source.
• Lecture: add the YBCO/"Jim Ashworth" (probably Ashburn) lead, the Becker thesis claim (Becker was an Institute adviser per the 2018 website), "we told management about this" and the "strange effects ... forces coming off the disc" quote.
• Interview: add the Maxar negotiation linked to US7808353B1, the Austin investor and Hal story, "Chantelle" = Shantel Butler (probable), the 2004 Podkletnov science-fair remark and the PhD "court case". Correct the attribution note: the monolayer passage is Amy's.
• NASA TM: add the motor-phasing argument, the unchanged field measurement and the DR1 findings on alignment mechanisms.
• claims.md
• Claim 4: state that the monolayer/bilayer argument is hers (with Butler present) and that only the German Air Force remark is of uncertain speaker.
• Claim 3: add the NMR-variant wording the arbiter asked for (if not yet applied), and the TM's motor and field statements.
• Claim 2: add "disc pristine" as the speaker's argument against a thermal cause (already partly there as "disc stayed undamaged").
• arbiter.md: add a "QE revision" section that:
• (A1) corrects recomputation item 4 and the related claim-3 sentence and open question (Section 7, item 1);
• (A2) evaluates Overhauser transfer and gradient flipping (from DR1) and states whether the "Refuted by calculation" verdict on "the 0.3 g result matched the theory" still holds unconditionally or with the stated condition (expected: holds);
• (A3) adds the pristine-disc argument, the "strange effects" summary and the motor-phasing and field-measurement statements to "For the claim, stated fairly" (claims 2 and 3);
• (A4) adds the Maxar/patent and Austin-investor leads to "Absence of evidence A" and "What would settle it";
• (A5) fixes the attribution note for claim 4 and the slide p.25 tag;
• (A6) has a reviewer check primer E3's assumption (eqs 28b and 31 combined orientation by orientation) and either endorses it as an illustration of "Internally inconsistent" or asks the primer to drop it.
• No verdict is expected to change.
• primer.md
• Fix the Laithwaite and Hayasaka grouping (section 1, section 3 and the evidence table) and the slide p.25 tag.
• In 4a, say the monolayer argument is hers, with only the funding remark of uncertain speaker.
• In section 5, add the TM's unchanged-field measurement and motor-phasing argument beside the fading and the magnet suspicion, and add the DR1 result on the other alignment mechanisms to Step C.
• In the claim-2 row and section 7, add the pristine-disc argument and the "strange effects" summary beside the Poher null.
• In 4a or 7, add Ning Li's documented fate (Investigator, [web]).
• In the claim-6 row, add the suicide ruling, the 2026 resurfaced text (as reported; text unseen) and the Shellenberger testimony beside the family's account.
• Optionally name Prutchi in the Alzofon row.
• Soften or source the spin-gravity-bounds sentence in section 3.
• example.py: no change needed unless A2 adds a number for an alignment mechanism. If it does, add it to part C.

Recheck

Rechecked 2026-10-07 against the current arbiter.md (with its "QE revision" section), primer.md, claims.md, facts.json, example.py and the deep/ files. Method: grep and reading of the passages named in Sections 2, 3, 5 and 7 above [read]. I did not re-run my own calculations; the arbiter's A1 recomputation matches my Section 4 item 4.

Remaining

1. physicist.md and the NASA-TM deep read still state that Table 1 is not reproduced by eq. 28a (a sign error). Add a one-line "Withdrawn, see arbiter QE revision A1" note at physicist.md Correction 6 and its claim-3 bullet, and at deep read section 4. Minor.
2. claims.md claim 4 says the German Air Force line has an uncertain speaker, but arbiter A5 and the primer attribute it to her (01:05:49). Align claims.md with the arbiter, or state that the attribution to her rests on the arbiter's reading at 01:05:49. Minor.

No verdict changes. Neither remaining item affects a verdict.

Final: OPEN (2 issues)

Orchestrator, 2026-10-07: both remaining items applied (withdrawal notes in physicist.md Correction 6 and its Table 1 bullet and in the NASA-TM deep read section 4; claims.md claim 4 aligned with arbiter A5). Final: PASS.