Government: Quality Engineering (Step 4b)

Date: 2026-10-03. Inputs checked: _text/manifest.json, all 5 digests, facts.json, claims.md, physicist.md, mathematician.md, contrarian.md, arbiter.md, primer.md, example.py, and the extracted text in _text/ for the spot checks. I also rendered memo pages 2-3 as images (to the scratchpad, not to the source folder) and listed the zip's contents.

Verdict: MINOR. All five central claims are covered. The arbiter already corrected most of Step 1's errors, and example.py runs. The defects are: stale fields in facts.json; one false source tag; and the DIA reference list, which is now legible in the OCR text but was never read. That list cites peer-reviewed papers reporting neutron tracks, which the reviews do not mention, so one sentence in the arbiter and primer needs rewording. No final verdict is expected to change.

---

1. Coverage

• Zip contents (pipeline gap). Step 0 extracted only the zip's file listing. It did not extract the 3 images or the 3 STAC (California State Threat Assessment Center) PDFs, so none of them has a digest. Step 1 read all six directly, and facts.json has an entry for each. I confirmed the archive holds exactly those 6 files plus __MACOSX junk (unzip -l). The material is peripheral, so this is acceptable for this project. tools/textprep.py should still be fixed to extract zip contents.
• No deep/ folder exists. Step 1b triage never ran because the project was reviewed before digests existed. The documents are short (3-15 pages each), so a "deep read" here means reading the complete OCR text, and that text now exists.
• text_available is wrong for three entries. facts.json gives "none" for the memo, Feinberg and DIA. OCR text exists for every page of all three (manifest: ocr_pages 3, 9 and 8). The field should be "partial" (OCR) or "yes".

2. Shortcuts

• DIA report, pp.7-8 not read (defect, moderate). Claim 2 (LENR, low-energy nuclear reactions) is the folder's most open question. Every reviewer, the arbiter and the primer say that "nobody read the references". The reference list is now legible in _text/DIA-...txt lines 331-500. It is not just a list of conference talks. It cites peer-reviewed papers: Mosier-Boss et al., "Triple Tracks in CR-39 as the Result of Pd/D Co-deposition: Evidence of Energetic Neutrons", Naturwissenschaften 96 (2009) 135-142; Szpak et al., Naturwissenschaften 92 (2005) 394-397; Szpak et al., Thermochimica Acta 410 (2004) 101-107; Iwamura et al., Japanese Journal of Applied Physics 41 (2002) 4642-4650; Arata, Physics Letters A 373 (2009) 3109-3112; Miles et al., China Lake final report NAWCWPNS TP 8302 (1996); and Sinha and Meulenberg, Current Science 91 (2006) [read: DIA p.7-8, OCR]. It also cites "personal correspondence" with McKubre and Forsley, both of whom contributed to the report [read: p.8]. This shows that the report's sources partly come from its own contributors, a point the reviews make only from p.6.
• DIA "radiation not seen" is overstated (defect, minor). The arbiter (Claim 2, point 3) and the primer (evidence table row 2b) say "DIA itself says that radiation was not seen". The DIA text says this only about Fleischmann and Pons in 1989 [read: p.1, lines 25-29]. For SPAWAR (the US Navy's Space and Naval Warfare Systems Center) in 2009, DIA reports "nuclear particles", citing the CR-39 neutron-track paper [read: p.3, lines 98-101; p.7]. The "Refuted by calculation" verdict for watt-level ordinary d+d fusion still holds. The claimed track rates are, as far as I recall, many orders of magnitude below the 1.9e12 neutrons/s per watt that this would need [memory, medium; to be confirmed by the deep read]. Even so, the sentence misstates the source, and the particle evidence is missing from the reviews.
• Feinberg article read "in part" (defect, minor). facts.json still records only partial reading, and its summary has errors (section 4). The arbiter repaired the load-bearing p.8 result. The rest of the article is now available as full OCR text (9 pages, about 41k characters). One full read would fix facts.json.
• Holt and memo. Holt was read in full and the memo was read from page images. Neither needs more reading.

3. Missed material (from the digests and text, absent from or understated in facts, claims, reviews, arbiter and primer)

1. Evidence of energetic neutrons (CR-39 triple tracks) from SPAWAR, published in Naturwissenschaften in 2009 and cited by DIA (above). It is the main "nuclear particle" claim in the DIA report, and no review discusses it. The contrarian in particular should have weighed it. The fact that it was later contested is [memory, medium] and needs a web check.
2. Feinberg's photon-stability argument [read: p.9, lines 1108-1123]: if charged tachyons existed, photons would decay into tachyon pairs. Photons cross intergalactic space for billions of years, so any charged tachyon must have a charge many orders of magnitude below the electron's, or a mass squared near zero. This indirect limit is a good teaching point, and no review or the primer uses it.
3. Alvager-Kreisler limit verified. The arbiter flagged the physicist's figure "rate < 1/10,000 of e+e- pair production" as unchecked. It is in the text: "less than one ten-thousandth of the known rate for producing electron-positron pairs by photons of slightly higher energy" [read: p.7, lines 770-777]. The arbiter's caveat can be removed.
4. DIA details not carried forward: reported laboratory explosions (Zhang et al., China, 1991-92; Biberian, France, 2009, with a photograph on p.5); Energetics Technologies (Omer, Israel) "superwaves" and ultrasonic excitation; the SPAWAR co-deposition technique; the statement that "magnetic fields ... stimulate increased heat"; the suggested uses (neutralising weapons of mass destruction, transmuting nuclear waste, a "dial-a-mode" power source, new explosives); the French CEA (Atomic Energy Commission) programme of 1997-1999; and India's restart [read: pp.2-6]. None changes a verdict, but facts.json should list the explosions and the uses because they bear on the "military interest" theme.
5. DIA's "10 million times" energy ratio [read: p.6, line 283]. I checked it: D+D to He-4 gives 23.8 MeV per 4 u, which is 5.7e14 J/kg; gasoline is about 4.6e7 J/kg [memory, high]; the ratio is about 1.2e7. The contrarian mentions it. The primer could use it as a one-line exercise.
6. The memo's context. The memo goes from "DT-ACO" to "DT (Dr. Vorona)". The CREST (CIA Records Search Tool) series RDP96-00788/00789/00792 is, in my recollection, the STARGATE remote-viewing collection, and Dr. Jack Vorona was a DIA scientific and technical intelligence official linked to that programme [memory, medium; not web-checked]. If that is right, the memo was probably written inside DIA, not by the CIA, and is filed with the remote-viewing papers. The primer calls it a "1987 CIA analyst" memo, which is unverified. The memo also contains redacted passages marked SG1B in paragraph 2, and the "(C/NF)" classification of that paragraph [read: p.1]. Neither is mentioned.
7. The memo uses Gaussian units. The equations carry the 1/c factors: E = -grad(phi) - (1/c) dA/dt, and phi = -(1/c) dS/dt [read: page image p.2, checked by me]. The mathematician noted the unit slip. The primer converts to SI correctly. Not a defect, but the primer could say the memo is in Gaussian units.
8. Holt's "22,000 mph or more" with no sonic boom [read: p.7] and his use of the Grief-Tolhurst wind-tunnel study for the winged disc [read: p.14] are not discussed. They are peripheral.

4. Accuracy spot checks (against the extracted text)

5. Process

• Source tags. Tags are present throughout. Primer: 48 tags. Physicist: 39. Mathematician: 28. Contrarian: 33. Arbiter: on every number. The one false tag is the mathematician's "Brookhaven [read: OCR]". The facts.json sources field for Feinberg also carries Brookhaven and "no events with negative missing mass squared" without correction.
• Web checks. The arbiter used 5 of its 8 calls. The web-search tool was refused, so all checks are Wikipedia (secondary) sources, and the arbiter says so honestly. The solar-flare energy check failed and is reported as failed. The medium-confidence facts the verdicts depend on (the 2004 DOE review, the Frey effect) were retagged [web]. Not web-checked, and not verdict-critical: brain field strength (about 1e-13 T), seawater conductivity, and transcranial magnetic stimulation (TMS) strength. Acceptable.
• Verdict scale. Applied consistently. Split verdicts are used where needed. Arbitrations are reasoned, and the "tested / never tested / dismissed" distinction is stated for every claim. One quibble: Claim 2d labels DIA's sociological statement "Tested, not supported". "Not testable as stated", or a plain "not supported by independent reviews", fits a claim about acceptance better than the experimental-test label does. Optional.
• Primer against Step 4 and the reader profile. All 9 sections are present in order, with answers at the end. Symbols are defined. Gauge freedom is introduced from scratch. Acronyms are defined (ELF, LENR, DIA, VLF). The worked example uses real numbers, and the source gap is repeated. Textbook chapter numbers are given only for Griffiths, where they are well known. Good.
• example.py. It runs cleanly with .venv/bin/python (exit 0). Its output reproduces the arbiter's numbers: E = B = 0 symbolically; skin depth 25.16 m at 100 Hz; 2.622e11 He-4 per joule; 1.909e12 neutrons per joule; a 1.02e7 m cube; tachyon energies 0.577 and 0.101; Gamow factor about 1e-2727.
• facts.json not revised after the arbiter's corrections. It still says: memo date "early-mid 1980s"; memo equations "partly garbled"; Feinberg "no events" with negative mass squared; Brookhaven. claims.md Claim 1 still says the waves "carry energy", which the arbiter's corrected claim list fixes. These are stale data, and later steps (the Curator, the overview) read facts.json.

6. Remediation actions

Deep reads (from the extracted text; render page images only where the OCR is unclear):

1. DIA-2009DIA-08-0911-003-a8012e, pp.1-8, Opus. Read the whole report, including the reference list on pp.7-8. Map each claim (excess heat, particles, He-4, transmutation, explosions) to its cited source and classify each source: peer-reviewed journal, conference, briefing, or personal correspondence with a contributor. Pull out what the SPAWAR CR-39 "triple tracks / energetic neutrons" claim (Naturwissenschaften 2009) and the China Lake He-4 report (Miles 1996) actually assert. Then web-check whether those were contested or replicated (2-3 calls). Answer this: does the neutron-track claim change the arbiter's sentence "DIA itself says that radiation was not seen", or any Claim 2 verdict?
2. CIA-CIA-RDP96-00787R000100030001-9-7119a0, pp.1-9, Sonnet. Read Feinberg in full and correct facts.json. Remove Brookhaven unless the text supports it. Record the 101 to 23 to 0 result and the less-than-1-in-400 limit; the Alvager-Kreisler less-than-1/10,000 limit; the antiproton-proton annihilation search; the photon-stability argument (p.9); and the Bilaniuk-Deshpande-Sudarshan reinterpretation (p.5). Also note the speed of light printed as "300,060 km/s", which is an OCR or print error.
3. CIA-CIA-RDP96-00792R000500240001-6-cc9638, pp.1-3, Sonnet (low priority). Confirm from the images: the addressee (DT, Dr. Vorona); the originating office (DT-ACO); the redactions (SG1B, SG1J); the classification markings; the Gaussian units; and the 12/22/87 date. Do one web check on whether DT/Vorona means the DIA Directorate for Scientific and Technical Intelligence and on whether the RDP96-00792 series is the STARGATE collection.

File changes (MINOR route):

• facts.json: set text_available to "partial" (OCR) for the memo, Feinberg and DIA, and update how_read. Memo: date 12/22/87 (probable); equations legible; Gaussian units; addressee and series context once checked. Feinberg: correct the search results as in deep read 2 and remove Brookhaven. DIA: add the reference-list classification, the CR-39 neutron-track claim, the explosions, the Arata inconsistency, and the fact that sources include personal correspondence with contributors.
• claims.md: replace Claim 1's "carry energy" with the arbiter's corrected wording. Update the source-gap note: the DIA references have now been read, and the cited papers themselves still need to be obtained.
• arbiter.md: add a "QE revision" section. (a) Reword Claim 2 point 3: DIA reports no radiation for Fleischmann-Pons in 1989 but reports particle (CR-39) evidence from SPAWAR in 2009; state whether that rate is anywhere near 1.9e12 n/s per watt. Keep or revise the verdicts accordingly; the expected result is no change. (b) Remove the "unchecked" caveat on the Alvager-Kreisler 1/10,000 figure, which QE verified. (c) Record the false Brookhaven tag. (d) Optionally relabel 2d.
• primer.md: fix evidence table row 2b and the takeaways line on the same point. Add the CR-39 claim to the Claim 2 "Genuinely open" bullet. Change "CIA analyst" to a neutral "intelligence analyst (memo held in CIA files)" unless the check confirms CIA authorship. Optionally add the photon-stability argument to the tachyon discussion, and the "10 million times" energy ratio as a check-yourself item.
• example.py: optionally add the fusion-to-gasoline energy ratio (about 1.2e7). Nothing else needed.
• Pipeline (outside this project): make tools/textprep.py extract and OCR the contents of zip archives, not only the listing.

---

Recheck

Date: 2026-10-03. Checked against current facts.json, claims.md, arbiter.md (with its "QE revision" section), primer.md, example.py, the new deep/ reads, and tools/textprep.py.

Defects 1. DIA pp.7-8 references never read: FIXED. deep/DIA-2009DIA-08-0911-003-a8012e.md exists; facts.json records the endnote read; the arbiter classifies the sources and reads Miles 1996. The primary papers other than Miles remain unread, stated honestly as a source gap. 2. "DIA says radiation not seen" overstated: FIXED. Arbiter Claim 2 point 3 reworded (1989 only; CR-39 neutron tracks for SPAWAR 2009, about 9 orders of magnitude too weak). Primer row 2b matches. 3. Feinberg only partly read / stale facts: FIXED. deep read pp.1-9 done; facts.json has 101 to 23 to 0 and the 1-in-400 limit. 4. False "Brookhaven [read]" tag: FIXED, and QE's own finding corrected. The deep read found "Brook-/haven" split across lines in the p.3 caption; the arbiter records this. 5. text_available wrong (none) for memo, Feinberg, DIA: FIXED (now "partial"). 6. Stale facts.json (memo date, garbled equations, Feinberg "no events"): FIXED (date 12/22/87, Gaussian units, equations read from images). 7. claims.md Claim 1 "carry energy" and source-gap note: FIXED. 8. Alvager-Kreisler "unchecked" caveat: FIXED (removed, verified p.7). 9. Claim 2d label: FIXED (relabelled). 10. Primer "CIA analyst": FIXED (neutral wording; memo held in CIA files). 11. Missed material (CR-39 tracks, photon stability, 10 million times ratio, Gaussian units, memo context, DIA details): FIXED in arbiter and primer; example.py reproduces the 1.24e7 gasoline ratio. 12. example.py runs: CONFIRMED (exit 0).

Remediation actions - Deep read 1 (DIA, Opus): done. Deep read 2 (Feinberg, Sonnet): done. Deep read 3 (memo, Sonnet): done; DT/Vorona/STARGATE context is tagged [memory, medium] because the web check failed, which is honest and not verdict-critical. - facts.json, claims.md, arbiter.md QE revision, primer.md, example.py updates: done. - Pipeline: tools/textprep.py still handles .zip as a file listing only (line 146); it does not extract or OCR the contents. REMAINS. This is outside the project's files, so it is not a defect in the Government work product.

Remaining issues: 1 (the textprep.py zip-extraction pipeline gap, outside the project output).

Final: OPEN (1 issues)

---

Recheck 2

Date: 2026-10-03. Checked tools/textprep.py and _text/ for the one issue Recheck 1 left open.

1. Pipeline gap, tools/textprep.py extracts only a zip listing: REMAINS. Line 146 still reads entry['kind'] = 'zip listing'; pages = [run(['unzip', '-l', paths[0]])], and _text/EM-effects-on-human-body-237054.txt is still a 1.2K file listing with no extracted contents. The reason is that the fix is outside the Government work product, and the user paused work on remaining items for now. The six zip files were read directly in Step 1 and are in facts.json, so the Government deliverables are unaffected.

Final: OPEN (1 issues)

---

Step 0 re-run (2026-10-05)

The zip extraction gap (Recheck 2, issue 1) is closed. Step 0 now extracts "EM_effects_on_human_body.zip" member by member: three STAC PDFs (text layer) and the three images as one 3-page OCR sequence. Every manifest document has a digest and a facts.json entry; the old zip-listing entry is replaced by one entry per member (plus one entry for the OCR sequence), and the earlier direct-reading details (image summaries, Sharp 1974 note, MEG and Frey-effect [memory] tags) are carried over. facts.json validates.

What the extracted text added or changed: - STAC 96-hour report: now fully digested (earlier pass read only part and keyword-searched the rest); added items are in the entry. A grep of all three STAC extracts for electromagnetic, microwave, directed energy and mind control finds nothing, so the earlier "no physics content" finding stands. - Gif (page 3 of the sequence): the OCR lists more asserted effects than the earlier summary (transparent eyelids, genital-area itching and forced orgasm, "caffeine field", microwave burns, electric shocks "used in the Gulf War", toe-bending "demo", zero trace evidence, break-ins as psychological warfare research). Added to the entry and to claim 5's wording in claims.md; these are unsourced assertions, and no claim, verdict or arbiter row changes. - 9RNM.jpg: OCR (page 2) is unreadable except the title, and the Sharp 1974 note on the gif was not recovered by OCR, so those two items still rest on the earlier direct image reading and are tagged accordingly. - claims.md: source-gap note updated (zip no longer a Step 0 gap).

Nothing is left open in the Government work product.

Final: PASS

---

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), with claims.md and the Step 0 digest as cross-checks. The arbiter's Evidence-rule revision (items 1-16) holds in arbiter.md and primer.md: single-source items are quoted and sourced (DIA p.4 journal-reluctance statement, Sharp-Grove via Wikipedia/Justesen), stakes are stated (DIA contributors, Miles at China Lake, lenr-canr.org, ERAB, Wikipedia), the memo and Holt silences are analysed with competing explanations and deciding records, and evidence cutting both ways is kept (proprietary-aspects sentence, Miles air-helium admission, 2021 CR-39 follow-up, Nature "anomalous findings").

Remaining violations:

1. Labels that read as dismissal, left in claims.md (the arbiter replaced them in arbiter.md, items 1, 10 and 15, but not here). Rules: no suppression of evidence / interested parties. - claims.md line 3: "Miles 1996 was read via an advocacy archive, lenr-canr.org" (arbiter.md line 222 and primer.md line 15 describe it instead as an archive run by supporters of the field, hosting a Naval Air Warfare Center publication by the experimenters). Same line: "the three images are unsourced infographics" (credits are Supratik Saha, www.raven1.net watermark, one uncredited). - claims.md line 11 (claim 5): "three unsourced 'EM effects on human body' infographics, which also assert unsourced effects". Replace with the credits and "cite no references, methods or field strengths".

2. Keep pertinent information: the gif's asserted effects are incompletely recorded. The Step 0 digest and facts.json list "transparent eyelids", genital-area itching and forced orgasm, and a "caffeine field" among the gif's effects (the Step 0 re-run entry above says these were added to claim 5), but they appear in none of claims.md line 11, arbiter.md line 147 (Claim 5 "kept in full") or primer.md line 134 ("recorded in full"). Both of those passages say they record the list in full, so either add the three items or drop "in full". No verdict changes.

Final: OPEN (2 issues)

Recheck (2026-10-05)

1. claims.md labels: fixed. "advocacy archive" now reads "an archive run by LENR proponents"; "unsourced infographics" and "unsourced effects" now describe the images as citing no sources, with their credits.
2. Infographic effects recorded in full: fixed. "Transparent eyelids", genital-area itching and forced orgasm, and the "caffeine field" are added to claims.md claim 5, arbiter.md and primer.md.

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, sites.md and the eleven research/ notes against prompt.txt (STEP 4b Retention and Process, SOURCE HONESTY). research/files/ is empty, as sites.md and the arbiter both say; nothing downloaded, so every Investigator item is correctly tagged [web] at snippet, abstract or OCR level.

What holds: - Coverage of the Investigator's leads is nearly complete. Every sites.md lead (Holt NTRS record, Bearden's staff note 0002, Rein 0008, 2021 JCMNS CR-39, ENEA heat-and-helium, Sharp-Grove/Justesen, Texas Tech helium, ARPA-E, Kowalski comment, memo originator, full 8-item folder, Buzby, workshop origin and DIA Public Affairs email) is weighed in all three reviews and the arbiter, with a verdict effect or "unchanged", and none is dropped because a reviewer judged it negatively. Negative and positive items sit side by side (ENEA beside Texas Tech; Holt persistence beside "no equations"; Kowalski beside the authors' reply; Rein's result beside its p-value flag). - SOURCE HONESTY: single-source items (Justesen via personal communication, Rein, ENEA) are quoted and sourced, not labelled; stakes are stated (Krivit, lenr-canr.org, JCMNS, Mosier-Boss and Forsley as DIA contributors, Preparata and Del Giudice, Rein as a supporter, Bearden note as the proponent's side); the memo-silence and Holt-silence analyses give competing explanations and the deciding records; limits are admitted (failed arbiter fetches, page 1 only of Justesen, OCR table). Tags are present and honest; no [memory] item is presented as read. - Arithmetic spot checks, redone by me: 19.5% inhibition; Rein t = 2.71 (p 0.022 two-sided, 0.011 one-sided) if SD and t = 1.11 (p 0.29) if SEM, n = 6; 1 pmol = 6.02e11 atoms; 15-240 pmol = 34-551 J at 2.62e11 He/J; 1-day 0.5 W = 43,200 J = 1.13e16 atoms = 19 nmol; 0.08-1.3% of a day, 1.9-31% of an hour; carbon-12 threshold 7.275 x 13/12 = 7.88 MeV; Miles air leak 2e12 atoms = 7.6 J (stated about 8 J). All correct. example.py runs (Parts 1-6) and reproduces them. - Primer matches arbiter on every verdict of the update: Bearden two-beam interference Refuted by calculation; detection at the magnet pole Not testable as stated; non-EM carrier Not testable as stated; Rein Never properly tested; claim 2 parts unchanged; claim 3 unchanged; speech by pulse-timed microwaves Plausible, untested; Frey Established; memo originator a DIA office by inference; explanation (c) best fit, (b) open.

Defects:

1. Retention in arbiter.md: pertinent Investigator and reviewer material left out of the arbiter's Investigator section. (a) Rein's proposed mechanism, the "Crystalline Transduction Theory" (scalar energy transduced to ordinary EM energy by liquid crystals in cell membranes and solid crystals in blood and tissue), which research/2026-10-05-cia-readingroom.md reports and which appears in no review, the arbiter or the primer. It is the proponent's account of how the cell effect would work, so it belongs next to the experiment. (b) Helium and neutron leads that research/ notes or the physicist name but the arbiter drops: Gozzi et al. (1998) and Arata and Zhang (1998) heat-helium reports and De Ninno's "4He detection" paper (all named in the lenr-canr note; only the physicist mentions Gozzi and Arata-Zhang), and the 2010 EPJ AP "Comparison of Pd/D co-deposition and DT neutron generated triple tracks" and JCMNS "Characterization of Neutrons Emitted during Pd/D Co-deposition" (in the jcmns note; only the contrarian names the 2010 paper). These are unread leads, but the arbiter's "Still not read" list should carry them, with their stakes (lenr-canr.org host; Arata and Gozzi as proponent-side authors). Fix: add (a) to F1.2 and (b) to F2.4 and the "Still not read" list.

2. Retention in primer.md: items the arbiter keeps but the primer omits. (a) Rein's other cited results, which the arbiter (F1.2) records and the primer's Section 4 row and Section 9 entry do not: Byrd (EEG with the watch), Persinger (mast-cell degranulation under two cancelling 0.5 Hz, 10 uT fields; Rein notes the scalar field was not measured there) and the Puharich/Max Planck E. coli gene-activation result; and Rein's theory (defect 1a). (b) The DIA report's workshop presenters outside the contributor list (Tanzella, Swartz, Hagelstein) and the DIA Public Affairs email "This is a DIA document" (23 Nov 2009, as posted), which bears on whether the report is authentic and is in the arbiter F2.1; the primer keeps only the workshop origin and coordination line. Fix: add a sentence for each in the Section 4 Rein row and the Section 6 row 2d(ii) (or "What happened next"), tagged [web] with the stake stated.

3. Wording and consistency. (a) arbiter.md F1.2: "None gives p < 0.01 except a one-sided SD test at 0.0109, which is only just above the stated figure" contradicts itself (0.0109 is not below 0.01); say "no test gives p < 0.01; the nearest, a one-sided SD test, gives 0.011". (b) primer.md Section 7 ("does not come out of a standard t-test whichever way") omits that near-miss, though the Section 6 row 1f states 0.011 one-sided; the lesson reads more dismissively than the arbiter's "flag, not a dismissal". Add "(the nearest, one-sided, is 0.011)". (c) Seawater factors differ between files: arbiter F1.1 gives 0.018 and 4e-18 (delta 25 m), the primer row 1d gives 0.019 and about 6e-18 (delta 25.2 m, as in example.py). Either state delta in both or use the same figures.

Not counted as defects: facts.json and claims.md carry no Investigator findings (they record what the project's documents claim, and the Investigator's items are outside them); the physicist's 80 versus the earlier 267 mW/cm^2 for Frey work is flagged as unreconciled in arbiter and primer and no verdict depends on it.

Final: OPEN (3 issues)

Recheck (2026-10-05, after Investigator update)

Checked arbiter.md and primer.md directly (grep and reading the cited passages). 1. Retention in arbiter.md: fixed. Rein's "Crystalline Transduction Theory" is now in F1.2 with its source and stake. Gozzi et al. (1998), Arata and Zhang (1998), De Ninno "4He detection", the 2010 EPJ AP triple-tracks paper and JCMNS "Characterization of Neutrons Emitted during Pd/D Co-deposition" are in F2.4 (line 345) and in "Still not read" (line 387), with proponent-side stakes stated. 2. Retention in primer.md: fixed. Section 4 Rein row now carries the theory and the Byrd, Persinger and Puharich/Max Planck results, tagged [web] with Rein's stake and "not read by anyone"; row 2d(ii) now names Tanzella, Swartz, Hagelstein and McGinnis and the 23 November 2009 DIA Public Affairs email, with "as posted, not independently verified". The helium and neutron leads also appear in Section 9 (lines 303-304). 3. Wording and consistency: fixed. (a) arbiter F1.2 now reads "No test gives p < 0.01; the nearest, a one-sided SD test, gives 0.011". (b) primer Section 7 now says "the nearest, a one-sided test ... gives 0.011, just above the printed figure", and calls it a flag, not a dismissal. (c) Seawater factors agree: arbiter F1.1 gives delta = 25.16 m, 0.019 at 100 m, 5.5e-18 at 1 km, matching primer (25.2 m, 0.019, 5.5e-18) and example.py Part 6; the contrarian's 0.018 and 4e-18 are labelled as delta rounded to 25 m.

No remaining defects.

Final: PASS