LeBon-G: Quality engineering (Step 4b)

QE: Opus, 2026-10-08. Inputs read: _text/manifest.json; all 27 digest files (Matter 01-13, Forces 01-13, Vassilatos 01), read in full; all three deep reads, read in full; facts.json; claims.md; sites.md; the research notes (borderlandsciences, minor-sites, henripoincarepapers, local-Morray-TH read in full; the others through sites.md and the reviews); physicist.md, mathematician.md and contrarian.md (by section, with grep across all three for each topic below); arbiter.md and primer.md in full. I ran example.py and checked passages against the extracted text (_text/*.txt, by "=== page N ===" marker). Tags: [read: p.N] = PDF page of the extracted text; [calc] = my arithmetic (.venv Python); [memory] with confidence.

Verdict: MINOR

The central verdicts hold. The arithmetic, the quotations I checked and the main lines of argument are accurate. The arbiter and primer are careful, and they keep most of the contested material. The defects are: - (a) one factual statement that is repeated through the deep read, arbiter and primer and is wrong in its unqualified form: "no blank leak rate anywhere in the books". Matter p.354 gives an instrument baseline. - (b) several of Vassilatos's physical claims about radium were never checked against the book. The book contradicts them. - (c) three of Le Bon's own experimental or predictive claims rest on digests only and have no verdict: electrification by influence, the Branly-Le Bon Hertzian-wave tests, and the chemical theory of radioactivity with its prediction about radium. - (d) Le Bon's own over-unity statements (Matter p.231, Forces p.38) were missed. The arbiter says the trigger-release idea is "not shown in the books". - (e) several of the Investigator's related-work leads and some pertinent digest material were dropped from the arbiter and primer.

None of these is likely to reverse a central verdict. All can be fixed by updating facts.json and claims.md, adding a QE revision to arbiter.md, and updating primer.md after the deep reads below.

1. Coverage

• The manifest lists 3 documents. The source folder holds exactly these 3 PDFs (listing checked, read-only). Nothing was skipped.
• All 27 chunks have digests. The digests cover the book back matter too (indexes, errata, publishers' advertisements, library slips).
• Matter is OCR on 488 of 489 pages. The digests mark garbled passages honestly (formulas on pp.67, 75, 77, 129, 154; rotated pp.362, 424).
• No defect.

2. Shortcuts

The deep reads cover the Vassilatos review (in full), Forces pp.317-368 (Black Light) and Matter pp.389-432 (experimental chapters V-XI). They are thorough. They used page images for every table and figure, and they record controls, gaps and internal slips. The four "deep-read gaps" that claims.md listed were partly closed by the reviewers' own reading:

Other central or experimental passages that rest on digests alone: - Matter pp.201-216 (Book IV ch.VI). This is Le Bon's chemical theory of radioactivity: "an absolutely pure body cannot be radio-active". He says the existence of metallic radium is "very problematical", that if it is ever isolated "it will have lost all the properties which render its combinations so interesting", and that a gram of radium lasts about 100 years [read: p.211-213]. This is a testable prediction, and it failed. Metallic radium was isolated in 1910 by Curie and Debierne and was radioactive [memory, high]. No document in the project mentions it. Defect (bears on claims 1 and 3). - Matter pp.227-243 and 444-447. These pages cover Le Bon's static-machine over-unity statement (p.231). They also cover his effluves driven through glass and ebonite with Dr Oudin's resonator (pp.235-237), X-rays from an electrodeless vacuum vessel (p.447), and Legge's repeat with a Tesla transformer (p.446). They are the primary-text anchor for Vassilatos's Tesla-coil and "particles through mica, ebonite" assertions. The arbiter itself notes that "the physicist's page was not examined for figures". The arbiter and primer keep the effluve experiment as "kept in the record", which is good. But no page images were read, and the voltages, frequencies and thicknesses are uncertain (half a millimetre or half a centimetre). Defect (bears on claim 6). - Matter pp.159-176 (radioactive substances: heated radium, emanation, recovery times). See §3 item 2. Digest only, and the Vassilatos claims depend on it.

3. Missed material (from the digests and the source text)

Listed most important first. Each item gives the page and whether I checked the text myself.

1. The electroscope baseline exists. Matter p.354 states the instrument criterion: an electroscope is usable only "when it does not give a loss greater than one angular degree in an hour after being covered with its cap" [read: p.354, checked]. This is not a per-experiment blank run, but it is a stated baseline leak. - His 8-day dark cylinder gives "1 to 2° per minute" (p.430). That is 60-120 times this baseline. - The warmed cage gives 5-6° per minute (p.430-431). - So "No blank leak rate anywhere" overstates the gap. These documents say it in that unqualified form: the arbiter (claims 1 and 3), the primer (§4 table row "Leak in sealed vessels", §7 "Without a blank run ..."), and deep/Matter §8.1 ("A keyword search of the whole extracted text found no baseline leak figure"). - The fair statement: an instrument baseline is given (≤1°/h, capped), but no blank run with the same open geometry, the same cylinder and the same warming is reported. - Verdicts are unchanged. Radon progeny on surfaces, desorption, and the different geometry of the cylinder all remain possible explanations. But the size of the effect relative to his baseline should be stated. - A further point: Le Bon says that "many observers" "have been unable to repeat my former experiments" because they used the small electroscope knob instead of a 10 cm receiving plate [read: p.381, checked]. This is a record of failed replications of the photo-leak work, with his explanation for them. It is missing from claim 3's "Tested? / Replicated?" entries.
2. Vassilatos's radium claims were not checked against the book, and the book contradicts them. facts.json lists this among the review's key claims: "radium heated red-hot loses activity for about a week (attributed to Le Bon)". The review also says Le Bon "discovered the means of altering the half-life of radium and measurably proved it" and "found" the emanation. - What the book actually says [read: p.168-169, checked]: heated radium loses most of its activity "but it regains it entirely in twenty days or so". After red heat or boiling it "still preserves a quarter of its primary activity", "solely due to the α particles". Radium emanation is "reduced by one-half in four days". These are reported as Rutherford's and Curie's results, not as Le Bon's discoveries. - Physics check [calc]: radon has T½ = 3.82 d, so ingrowth after driving it off is 1 − e^(−λt). That gives 72% recovered at 7 days and 97% at 20 days, which matches "twenty days or so". Ra-226 is one of four alpha emitters in equilibrium (Ra, Rn, Po-218, Po-214), so one quarter of the alpha activity remains, which matches "a quarter". - So the book reports correct mainstream radiochemistry. Vassilatos's "a week", "Le Bon found" and "altering the half-life" are misreadings: the half-life does not change; the emanation is removed and grows back. - Not one reviewer, nor the arbiter or primer, assesses these claims. This also makes a short worked example for the reader (exponential ingrowth), good material for the primer. - Vassilatos's "UV of the right frequency makes radium-like materials" is likewise unassessed. In the book it is the photo-leak (Matter p.180, 386-388: amalgamated tin "forty times" more active than uranium oxide). The arbiter's claim 3 verdicts cover the phenomenon, but the Vassilatos wording is never tied to them.
3. Le Bon's own over-unity statements. - Matter p.231 [read, checked]: "the creation, in the future, of such a machine—that is to say, of an apparatus giving forth a power extremely superior to that expended in setting it in motion—can be considered possible." The context is a hand-worked 50,000 V static machine giving 2 A, "100,000 watts, or 136 horse-power per second". 100,000 W / 735.5 = 136 metric hp [calc]; "per second" is his unit slip. - Forces p.38 [read, checked]: the energy matter emits under an excitant "may far exceed that which it has received". - The arbiter (claim 2, last bullet; claim 6) says the Tesla-coil trigger release is "not shown in the books". It should say: the books propose it as possible (Matter p.231; Forces p.38) but give no device, trial or energy accounting. - Verdict unchanged (Never properly tested; as a method, Not testable as stated). The reader of the Moray link needs to know that the idea is Le Bon's own and not only Vassilatos's.
4. The chemical theory of radioactivity and the radium prediction (Matter pp.207-216; see §2). Le Bon's own account of why he ran the hydration experiments on uranium, thorium and radium (Matter p.419) is this theory, so it ties claims 1 and 3 together. It should become a sub-claim with a verdict. A likely reading for the reviewers to confirm: Tested, not supported (radioactivity is independent of chemical state, Rutherford and Soddy 1902-03; metallic radium isolated 1910) [memory, high].
5. A controlled null test of a "hidden force", by the translator. In Forces p.401 [read, checked], Le Bon cites Georges Delbrück's speculation that soaring birds annul gravitation through intra-atomic energy ("not fundamentally absurd"). He also mentions needle attractions by living bodies and the spiritualists' levitation. Legge's footnote reports that he enclosed the needle apparatus in a vacuum jacket: "the needle remains motionless, no matter what bodies are brought near"; he concludes "convection currents formed by heat". - Forces p.402 (digest): Le Bon sat with Dastre at séances and found "nothing demonstrative", and he cites N-rays as a warning. - This bears on Vassilatos's "living energies" framing and on the Investigator's lead "electrogravitic experiment by Dr Le Bon ... about 1907" (Resines, BSRF). It is a documented test of a claim in this project's orbit, with a negative result. It is absent everywhere.
6. The Branly-Le Bon 1899 Hertzian-wave tests (Forces pp.180-196, digest). - A closed metal box, even tin 0.1 mm thick, is opaque. A 1 mm × 20 cm slit passes waves at all distances tested, while 100 round 1 cm holes do not. A slit perpendicular to the radiator axis passes waves at 6-8 times the distance of a parallel one. A 1 mm gauze acts as a solid sheet. Water increases absorption by walls, sand and stone. - In modern terms these are waveguide-cutoff and slot-antenna effects (Babinet's principle) and correct shielding results [memory, high]. The 1 April 1899 Comptes rendus note is named. - This is a facts.json key claim with no verdict. Also missing is the war proposal: parallel Hertzian beams to detonate ships and magazines at a distance, "rendering war impossible" (Forces p.179, digest).
7. Electrification by influence and the aluminium arrow (Forces pp.137-153, digest). - Le Bon's strongest experimental "matter to electricity" claim in Forces: an aluminium strip held "motionless in space" near a charged body, throwing out sparks, "the complete transformation of matter into energy". - Also here: an electroscope that changes sign after twelve minutes near a rod (Figs 3-5), and the claimed explanation of static-machine versus battery tension by atomic versus molecular energy. - Without a verdict, claim 2's "electricity is a product of dissociation" is judged only through the dynamo. The likely mainstream reading (corona and point discharge in a strong field, with energy from the charged body and its source) needs stating and checking against the page.
8. Reception record, both directions. - Legge's preface quotes P. D. Innes (Cavendish, Proc. Roy. Soc. 1907, p.442): "a great store of energy in the atom seems now beyond question" (Forces p.14-15, digest). He treats this as approval of Le Bon's position. - Matter p.86 F.L. note: Kelvin's 1904 statement that radium's heat could not come from stored energy, and his withdrawal at the British Association later that year. - Matter p.27: two Cambridge critics "admitted the theory was in the main correct" while challenging the experiments and originality (Athenaeum, Feb-Mar 1906). - These belong in claim 5's reception record beside the critics already listed. None appears in the arbiter or primer.
9. Peripheral experiments not mentioned anywhere (keep briefly in the record; no verdicts needed): - The infra-red plant experiments under black paper and red glass (Juvisy). Le Bon says himself they could not be repeated (Forces pp.369-378). - The "antagonism" experiments: a ZnS screen behind an ammoniacal copper-sulphate trough is about 14 times brighter (Forces pp.379-391). These bear on the arbiter's doubt that ammoniacal copper sulphate is "transparent" in the IR table: Le Bon here uses it as an IR absorber, which contradicts his own transparency table. This is an internal inconsistency to note. - Diamond phosphorescence used to tell Brazilian from Cape stones, about 200 stones examined (Forces pp.268-272). - Electric sparks restoring thermoluminescence after calcination (Forces pp.290-293); this is radiation-induced thermoluminescence [memory, medium]. - Solar-UV "disappearance" photographs dated 1901 (Matter pp.371-373). - Nova Persei read as atomic dissociation, with "nebulous masses moving at about the speed of light" (Forces p.118). This is the 1901 light echo [memory, high], an instructive misreading. - Dufour's radioactive exhaled air, and kola and caffeine as dissociation excitants (Forces p.419); theobromized caffeine (Matter p.314).
10. Le Bon's 29 Nov 1902 letter to Poincaré (Investigator lead, research note item 12): quinine sulphate in a closed box gives a discharge "énormément réduite. Les choses se passent comme s'il fallait que les ions puissent circuler pour agir." This is Le Bon's own evidence that the quinine effect is ion transport, not a penetrating radiation. It bears directly on claim 3 and the Rutherford "low penetration" dispute, and it was dropped.

4. Accuracy spot checks

All checked against _text (PDF page markers) or by calculation:

One metadata error: facts.json (Matter entry, "year") and deep/Matter say the year is "not on the extracted title page". It is there. PDF p.10 reads "1909" [read: p.10], so this copy is a 1909 printing. The first English edition was 1907, per Forces' preface [read: Forces p.13]. The primer's "1907" is right for the edition. facts.json should record that the copy is dated 1909.

5. Retention check

The rule: pertinent material from the sources or the Investigator must not be left out of the arbiter or primer because a reviewer judged it negatively.

Retained well. The arbiter and primer keep: - the Izarn and Gariel favourable reports; - the red-darkroom-lamp explanation of failures; - Berget's reproduction; - Kalähne, Gates, Ramsay and Spencer; - the Martins polarisation priority; - the 1903 Nobel nomination (labelled unconfirmed, not dismissed); - the John E. Moray FBI and library account, with its stake and competing readings; - the "Girard Le Beau" hearsay, with the decisive record named; - the Oudin effluve experiment with an untested mundane reading; - the reception context (IMEC, Moatti), stated as context, not verdict; - the absence-of-evidence analysis for black light after 1908.

These are good practice.

Defects (pertinent material absent from both the arbiter and primer): 1. Investigator leads (sites.md "Leads for review", priority 3, related work) dropped without mention: - J. Resines (BSRF booklet B0086): "an electrogravitic experiment by Dr Le Bon carried on by about 1907". - Vassilatos, "Induced Radioactivity by Solar Light" (JBR 53(3), 1997), on a magnesium plate under focused sunlight; it includes the W. G. Lehr anecdote. - Vassilatos, "Black Waves" (JBR 48(4)): Dan Winter's galena-and-carnotite experiment, and Paul Brown's claim that Moray's device was a nuclear battery. - W. R. Lyne, Pentagon Aliens; R. A. Ford, Homemade Lightning (the CaS experiment). - Oliver Lodge's congruent ether views (Kragh).

These are interested-party or secondary, but they are the alternative-science lineage of Le Bon's ideas. The 1997 solar-radioactivity article and the electrogravitic lead bear directly on claims 3 and 6. They belong in the arbiter's evidence for claims 3 and 6 and in the primer's "Kept in the record" list, labelled as their sources' characterisations. 2. Le Bon's 29 Nov 1902 letter (quinine discharge "énormément réduite" in a closed box). It is in the research notes but absent from all three reviews, the arbiter and the primer (§3 item 10). 3. The 1896 US press statement that "no one apparently has thought it worth while to take the trouble" to test the claims. The contrarian uses it to say "dismissed without testing". The arbiter adopts "Never properly tested" for the 1896 form but drops the evidence. It also omits what the STEP 2 scale requires for that verdict: "say whether it was dismissed without testing". The primer keeps only the light-leak suspicion. 4. Le Bon's own report of failed replications (Matter p.381) and Legge's vacuum-jacket null test (Forces p.401). These are evidence in both directions and are in the source text, not in any review file.

None of these omissions appears to come from a reviewer's negative judgement; they look like oversights. But the rule is about the outcome: the material is pertinent and absent.

6. Process

• Source tags: present and honest throughout. Interested parties are named with their stake (BSRF; John E. Moray; Le Bon himself as a party to the priority dispute; Moatti as a critic). Single-source evidence is described, not labelled (Izarn, the FBI story, the Minister of Marine).
• Web checks:
• Physicist: 3. Contrarian: 3.
• Mathematician: none recorded, although its proton-decay bound was [memory, medium] and fed a verdict. The arbiter repaired this with a web check (Super-Kamiokande, arXiv 0903.0676).
• Arbiter: 5 of 8. The arbiter flags the 2020 SK figure as "URL not confirmed"; the primer uses only the confirmed 2009 bound. Good.
• Minor gap: the mathematician should have spent its budget.
• Verdict scale:
• Mostly consistent. Split verdicts are used well.
• Two misuses. "Tested, not supported" is applied to a reading the text does not support: arbiter claim 6, "Bombardment/transmutation lineage: Tested, not supported by the Le Bon text"; and claim 3, "transmutation reading". The scale reserves "Tested, not supported" for controlled experiments run. Here no experiment tested the reading; the text simply contains no transmutation. "Not testable as stated", or a plain "no support in the text" outside the scale, fits better.
• "Plausible, untested" is applied to a priority question (claim 5, 1897 leak priority). The arbiter itself says the scale "fits a historical claim poorly".
• The 1896 black-light verdict omits the required statement on dismissal without testing (see §5.3).
• Primer:
• It follows STEP 4 order 1-9, with answers at the end and a corrections section.
• It defines symbols and units on first use (kgm, metric hp, γ, Q-value, u, MeV, φ, λ_c, ε₀). It suits the reader profile, and its worked example uses real numbers.
• Section 1 says the experiments ran "from 1896". Le Bon dates the start of the black-light work to about 1894, "studying for two years" before 1896 [read: Forces p.317]. This is minor; say "from about 1894-96".
• example.py: runs cleanly (exit 0) with the venv. Its outputs match the primer: 6.798e9 metric hp, 5.46e12 J relativistic, 35.33 MeV, 1.51e10 J, 1.4 million years, the dynamo balance, 427 kgm/kcal, 6.77e6 V, afterglow exponents 1.52 and 1.24, and the cut-off wavelengths.

7. Corrections to earlier documents

1. deep/The-evolution-of-matter-la-bon-2c7bb7.md §8.1, arbiter claims 1 and 3, primer §4 ("Leak in sealed vessels") and §7: "No blank leak rate anywhere" is wrong as an unqualified statement. Matter p.354 gives the instrument baseline (≤1 angular degree per hour, capped). The 8-day cylinder (1-2°/min) is 60-120 times that, and the warmed cage (5-6°/min) about 300-360 times. Reword: "an instrument baseline is stated, but no blank run in the same geometry". Verdicts are unchanged.
2. arbiter claim 2 (last bullet) and claim 6: "Tesla-coil trigger release ... not shown in the books" should read: the books propose such a machine as possible (Matter p.231; Forces p.38) but give no device, trial or energy accounting.
3. facts.json, Matter entry "year", and deep/Matter header: the PDF p.10 title page is dated 1909 (this printing). The 1907 first English edition is per Forces p.13.
4. digest/The-evolution-of-matter-la-bon-2c7bb7.11.md p.411: "Le Bon asks for an independent check" should read: Legge (F.L.) asks for it. The deep read is correct.
5. facts.json, Vassilatos entry, key claim "radium heated red-hot loses activity for about a week (attributed to Le Bon)": add that the book says "regains it entirely in twenty days or so" and that this is reported from Rutherford's and Curie's work (Matter p.168-169). "A week" and "Le Bon found" are the reviewer's.
6. arbiter claims 3 and 6: replace "Tested, not supported" for the transmutation and lineage readings with a scale term that does not imply an experiment (§6).
7. arbiter claim 4, 1896 form: add whether it was dismissed without testing (the 1896 press statement supports "dismissed without testing" in the US press, May-June 1896; Poincaré's protocol questions show engagement in France).
8. primer §1: "From 1896" should read "from about 1894-96" (Forces p.317).
9. claims.md claim 3 and/or claim 1: add the chemical theory of radioactivity and the radium prediction (Matter pp.207-216).

8. Remediation actions

Deep reads (most important first)

What each later file must change

• facts.json: Matter year (1909 printing); Vassilatos radium key claim (book text and source of the data); add Matter key claims for the chemical theory of radioactivity and radium prediction, and the over-unity machine statement (p.231); Forces how_read to record the new deep reads; Forces key claim on electrification by influence with page range 137-153.
• claims.md: add to claim 3 (or 1) the chemical theory of radioactivity and the failed radium prediction. Add to claim 2 that Le Bon himself proposed over-unity machines as "possible" (Matter p.231; Forces p.38). Apply the arbiter's §6 corrections not yet applied.
• arbiter.md, a "QE revision" section:
• the baseline correction (§7.1), with its effect on claims 1 and 3 (verdicts unchanged unless the deep read finds more);
• verdicts for: the chemical theory of radioactivity and the radium prediction (expected Tested, not supported); electrification by influence and the arrow (expected Refuted by calculation or Tested, not supported, pending deep read 3); the Branly-Le Bon shielding results (expected Established); Vassilatos's radium claims (the book's account Established; Vassilatos's "altering the half-life" not supported by the text);
• the over-unity wording fix;
• the scale-term fixes (§6);
• "dismissed without testing" for the 1896 form;
• the retention items in §5 (related-work leads, the 1902 letter, Le Bon's failed-replication remark, Legge's null test, Innes and Kelvin).
• primer.md:
• §4 and §7: reword "no blank run" as in §7.1.
• §6 evidence table: add rows for the chemical theory and radium prediction, electrification by influence, and the Hertzian shielding tests. Add to 3a and 3b the p.381 failed replications. Add to 6c the over-unity statement.
• "Kept in the record": add the related-work leads (Resines, Vassilatos 1997 and "Black Waves", Lyne, Ford), Legge's vacuum-jacket null test, the 1902 quinine letter, and Innes and Kelvin.
• Consider a short second worked example: radon ingrowth after heating radium (72% at 7 days, 97% at 20 days, one quarter of the alpha activity left), which tests Vassilatos against the book with one exponential. example.py would gain a Part F.
• §1: "from about 1894-96".

Recheck

QE recheck, 2026-10-08. I re-read the current arbiter.md (Section 8 R1-R11 and the summary table), primer.md, claims.md and facts.json by grep and section, and ran example.py with the project venv (exit 0, Part F present). Deep reads for the eight remediation actions now exist (deep/ files cover Matter pp.159-176, 201-216, 227-243, 350-432, 444-447 and Forces pp.137-153, 178-196, 317-402, 415-420, per facts.json how_read and claims.md status).

Remaining issues: 1. The digest p.411 misattribution above (one line). 2. Minor: Lyne, Ford and Lodge are not listed by name in the arbiter or primer related-work entries. I treat this as a nit, since the Resines, Vassilatos 1997 and "Black Waves" leads are named and the rest are labelled as aggregated leads. I do not count it as an issue.

Final: OPEN (1 issues)

Orchestrator fix (2026-10-09)

• digest/The-evolution-of-matter-la-bon-2c7bb7.11.md:59: old "Footnote 2: Le Bon asks for an independent check"; new "Footnote 2: Legge (the translator) asks for an independent check" with inline correction note.
• arbiter.md:283 (nit): Lyne (Pentagon Aliens), Ford (Homemade Lightning) and Lodge (Kragh) are already named in the related-work lead list; no edit needed.

Final: PASS