Morton-Wall: Quality Engineering (Step 4b)
QE agent (Opus), 2026-10-08. Inputs read in full: prompt.txt, _text/manifest.json, all 20 digest files, deep/ (one file), facts.json, claims.md, sites.md, all 8 research notes, physicist.md, mathematician.md, contrarian.md, arbiter.md, primer.md, example.py. I also read the source text myself: Houston 1871 (PDF pp.594-596, text layer) and Wahl 1871 (PDF pp.319-323, text layer, plus page images of pp.320-322 at 200 dpi and crops at 500 dpi rendered to /tmp/mwqe). I spot-checked the Investigator's files in research/files (Houston Jan 1876, Morton 1867, Woodbury 1944).
Tags: [read] = I read it in the extracted text; [read: page image p.N] = I read it on a page I rendered; [computed] = my own arithmetic.
Verdict: MINOR
The review reaches sound verdicts, and its primer is careful and honest. The defects are stale figures in the Step 1 files (claims.md, facts.json, deep read) that the arbiter has already identified but no one has yet applied, a few small page-image misreads by individual reviewers, and some minor retention and wording gaps. No central claim or verdict changes. All remediation fits the MINOR path: update facts.json and claims.md, add a "QE revision" section to arbiter.md, and make small edits to primer.md.
1. Coverage
- The manifest lists one document: a 623-page Google Books scan of the whole Journal of the Franklin Institute (JFI), Vol. LXI (Jan-June 1871). It has 20 chunks, and all 20 have digests (digest/…01.md to …20.md). No document was skipped. OCR covers 622 of the 623 pages and is good apart from fractions and tables.
- facts.json splits the bound volume into five sensible entries: Houston; Wahl; the editors' short items; other electrical items; and the rest of the volume. That meets the rule "If one PDF bundles several documents, list each" for the relevant articles.
- Coverage: no defects.
2. Shortcuts
- Houston 1871 (central): deep read of the text layer plus page images at 110 and 300 dpi. Acceptable.
- Wahl 1871 (claims 5 and 6): there is no deep/ file, but the article was read in full by the Step 1 agent, and the physicist, mathematician, contrarian and arbiter all read its page images. Acceptable in substance. The record is uneven, though: the only deep/ file still carries the OCR figures (see section 4).
- Peripheral editorial items (Highton over-unity claim, tinfoil electromagnet test, "Magnetic Engines" duty calculation, Sharples' bichromate battery note): digest only. Acceptable, because no central claim rests on them.
- Investigator's downloads (Houston Jan 1876, Houston and Thomson Apr 1876, Morton 1867, Woodbury, Atherton, Edmonds): read by the reviewers, with quotes and line references. Acceptable.
- Shortcuts: no defects.
3. Missed material (from my reading of the digests and sources)
- Houston's own 1876 wording widens the 1871 reach, and arbiter Correction 7 and primer §4 do not acknowledge it. The arbiter and primer call the free-energy retelling ("any insulated metal object anywhere in the building") "a retelling that grew", measured against the 1871 note ("within eight or nine feet"). Houston himself wrote in Jan 1876, about the 1871 experiments: "sparks could be drawn from any metallic objects in the neighborhood of the coil, and, indeed, in adjoining rooms" [read: research/files/JFI-1876-01…, OCR line ~4214]. For the Dec 1875 repetition he wrote "from all metallic objects in the same room with the coil, or in adjoining rooms" [read: same file, ~4308]. Houston also gave the boiler on the floor below and the gas pipes in adjoining rooms (the physicist, Correction 6, notes these). The retelling still differs in date (1872), person (Thomson) and "continuous". But part of its wider reach comes from the inventor's own later account and is not pure accretion. Under SOURCE HONESTY this should be stated.
- Houston's stated aim is "increasing the quantity of the spark of the induction coil without greatly diminishing its length" [read: p.594]. The primer (§1) and arbiter (claim 4) reflect it. facts.json and claims.md do not state it. Add it to facts.json, because it is what makes the "maximum" proposal self-defeating (arbiter claim 4).
- Same-volume context on battery EMF: Sharples' note (digest p.433) says a bichromate-nitric cell has the same EMF as a Bunsen cell, with internal resistance 50 percent higher and steady for twelve hours. This bears on Morton's three-cell bichromate battery and the open "E₀" question (arbiter Open questions; primer §5 step 7). It is not load-bearing; record it as context.
- Editors' handling of over-unity claims in the same volume (digest pp.256-257, 266-267, 448-449). The editors computed the ideal duty of zinc against coal. They treated Highton's doubled-wire claim as perpetual motion, and "H. M." tested the tinfoil-electromagnet claim with Mayer's apparatus and found "no improvement whatever". facts.json and the contrarian record this; the arbiter and primer omit it. It is pertinent context for readers who come through the "radiant energy" route: the same editors tested claims of extra energy experimentally. A one-line mention in the primer §7 would be enough.
- Unread leads worth carrying forward (they appear in research notes but not in the arbiter's "What would settle it"): - Susskind's snippet "...Thomson's 1871 experiments while his own paper was in proof" (research/2026-10-08-google-search.md, query 21) bears directly on the authorship question. - Compton (1939), snippet "tiny sparks could be obtained from metal objects wherever they were". - Coulson (1957, JFI 264), which cites Houston 1871 as part of Thomson's story. Add all three to the arbiter's records list.
- Peripheral and correctly omitted: Abbay's induction coil and Geissler tube used as a spectral reference at the 1870 eclipse (p.149); Morton's purchase of the Bancker collection for Stevens (p.553); John Wise's statistics on lightning rods as "incendiaries" (p.223); the editorial jibe at boiler-explosion "electricity theorists" (p.439). None of these bears on the claims.
4. Accuracy spot checks (5)
| # | Fact checked | Where used | Source check | Result |
|---|---|---|---|---|
| 1 | Houston's distances and figures: "eight or nine feet", "fully thirty feet" (engine only), "less than one square foot … about five inches" (two insulated conductors, neither earthed), "several hundred square feet" | deep, claims, arbiter, primer | [read: pp.594-596] | Correct in deep read, arbiter and primer. claims.md claim 3 still blends the two-conductor result with the earthed case (arbiter Correction 3, not yet applied). |
| 2 | Wahl's secondary: 44⅓ miles; weighs 44½ lb; condenser 325 sq ft; primary 0.1655 in | arbiter Corrections 1-2, primer §5 step 6 | [read: page image p.321]: "about 44⅓ miles", "weighs 44½ pounds", "325 square feet", "0·1655" | Arbiter, primer and mathematician are correct. The physicist misreads the weight as "44 1/4 pounds" (Correction 1); the contrarian reads "about 44½ miles" for the length. Both are small misreads; the arbiter's reading stands. |
| 3 | Wahl's jar: "leyden jar of 1⅓ square feet surface … sparks of 2¾ inches" | arbiter Correction 2, primer | [read: page image p.322, 500 dpi crop] | 1⅓ is confirmed. The spark fraction looks like ¾ (⅜ cannot be fully excluded at this resolution, as the arbiter says). Correct. |
| 4 | Houston 1876: engine "also in metallic connection with the water-pipes"; battery wording "One pole of the battery was placed in metallic connection with a gas pipe" and sparks "between the platinum points of the coil" | arbiter Correction 5, claim 2 | [read: research/files JFI-1876-01, lines ~4218-4225, 4300-4307] | Quoted correctly. The ambiguity (primary or secondary side) is honestly flagged. |
| 5 | Morton 1867: "sparks of but one inch … with a single jar … six jars arranged for 'cascade' increase the striking distance to 4 inches" | arbiter Correction 8, primer §3.1 | [read: research/files JFI-1867-04, ~line 3220] | Correct. |
| 5b | Woodbury: coil "90 feet below" the observatory; Snyder "Electric energy transmitted through space!" | arbiter claim 2, primer table | [read: Woodbury OCR ~lines 3455-3465] | Correct. Snyder also "tried the pencil for himself", so the primer's "Snyder drawing sparks" is fair. Note: research/2026-10-08-archive-org.md says Snyder witnessed sparks "90 feet below" the coil. The source says the coil was 90 ft below them. Wording slip in the research note. |
New small accuracy findings (page image p.321, my reading): - Core wire thickness: printed as a fraction with 1 over 18 or 16 ("about 1/18th [or 1/16th] inch"). The physicist (Correction 2) and the digest read "1/8". Not load-bearing; it should not be stated as 1/8. - Width of base: reads "13 inches". The text layer, the digest and facts.json have "18". The arbiter corrected only the bobbin sections (13 in). The width should be corrected too. - Wahl journal pages: PDF p.319 has running head "212" and p.323 has "216", so the article is j.212-216. The primer is correct. facts.json says "212-217" and the deep read says "213-216"; both are wrong. (The volume index on p.30 and p.34 gives "218" and "213" through OCR.)
5. Retention
- Kept, and stated fairly beside the verdicts in both arbiter and primer:
- Woodbury's authorship account and Snyder's testimony, described as one published source with Thomson's stake.
- The historians' radiation reading (Susskind, Atherton).
- The free-energy "radiant energy" reading, given as that community's characterization.
- Edison's etheric force and the Western Union explanation, with its stake.
- The editors' priority note, with its stake.
- The three silences, analysed with competing explanations.
- Wahl's interested-party status.
No pertinent material was dropped because a reviewer judged it negatively. Retention: pass. - Minor gaps (not driven by negative judgement): - Item 1 in section 3: the "retelling that grew" wording slightly understates the inventor's own 1876 breadth. - Item 4 in section 3: the editors' over-unity tests are absent from arbiter and primer. - The Compton and Coulson leads and the Susskind "in proof" lead are absent from the arbiter's records list.
6. Process
- Source tags: present and honest throughout. [computed] is used consistently for model numbers, and every assumed E₀, L and C is labelled. The primer's "Read this first" box tells the reader that every electrical number is modelled. Good.
- Web checks: the reviewers checked the load-bearing web facts (spark-voltage table, Hertz, etheric force). The arbiter used 2 of 8 web calls, adequate because no verdict hangs on an unchecked [memory] fact. The mathematician's breakdown-field citation is "relayed by a web search summary … did not open". The tag is honest, but it is a weak check and not load-bearing. Glass permittivity (about 6, [memory, medium]) is unchecked and moves only the illustrative jar energy.
- Verdict scale: applied consistently. The arbiter reconciled the differing labels (physicist "Plausible, untested" or "Not testable as stated" against the others' "Never properly tested") with stated reasons. The split verdicts are clear. The primer evidence table matches the arbiter's final verdicts exactly.
- Primer against STEP 4: all nine sections are present and in order. Every symbol is defined and acronyms are introduced. The textbook references avoid uncertain chapter numbers except Griffiths ch. 2, 7 and 11, which are correct [memory, high]. The worked example uses real numbers. The check-yourself section has five questions with answers. Further reading lists both mainstream sources and the inventors' works. The reader profile is respected (no tensors; the near and far field are explained in plain words). Good.
- example.py: runs cleanly with exit code 0 (
.venv/bin/python example.py), and every number it prints matches primer §5. - Corrections propagation: the arbiter's "Corrections to earlier documents" have not yet been applied to claims.md, facts.json or the deep read, so those files still state "443 miles", "443 lb", "825 sq ft", "0.1635 in", "14 sq ft jar", "2 in", "journal pp.212-217", and the blended claim 3. This is the main defect. Under CORRECTIONS ARE WELCOME it is the orchestrator's job, but until it is done the source of truth disagrees with the primer, and any STEP 5 entry built from facts.json would repeat the errors.
7. Defects and remediation
| # | Defect | Severity | Action |
|---|---|---|---|
| D1 | claims.md claims 1, 3, 5, 6 and its scope note still carry OCR errors and outdated statements (443 miles; 14 sq ft jar and 2 in; five-inch result blended with the earthed case; table described as insulated; Morton 1867 "missing") | Minor (source of truth wrong) | Apply arbiter Corrections 1-4 and 8. Claim 5 becomes "44⅓ miles … one third of App's 150 miles"; claim 6 becomes "1⅓ sq ft jar, about 2¾ in". |
| D2 | facts.json Wahl entry: 443 and 444 miles, 443 lb, 825 sq ft, 0.1635 in, 14 sq ft, 2 in, bobbin 18 in, base width 18 in, core "1/8"-style reading, journal pp.212-217 | Minor | Correct to 44⅓ miles, 44½ lb, 325 sq ft, 0.1655 in, 1⅓ sq ft, 2¾ in, bobbin 13 in, base width 13 in, core about 1/16 or 1/18 in, j.212-216. Reverse the inverted "more wire" contrast. |
| D3 | facts.json Houston entry lacks the stated aim, the 1876 pipe connection for the engine, and the battery caveat; the Thomson background is still [memory, high] although it has been web-checked | Minor | Add the aim quote (p.594), Houston 1876 "metallic connection with the water-pipes" (as a later account), and the primary/secondary caveat. Retag the Thomson note [web] https://en.wikipedia.org/wiki/Elihu_Thomson. |
| D4 | The deep read gives Wahl as j.213-216 | Trivial | Mark as superseded (j.212-216). Do not rewrite the deep read. |
| D5 | Physicist misreads: secondary weight "44 1/4 lb", core wire "1/8" in | Trivial | Note in the arbiter's QE revision; no verdict effect. |
| D6 | Arbiter Correction 7 and primer §4 call the retelling's breadth pure accretion, without Houston's own 1876 "any metallic objects … in adjoining rooms" | Minor (SOURCE HONESTY, fairness to the alternative reading) | Arbiter QE revision and primer §4: add the 1876 quote. Keep the date, attribution and "continuous" differences. |
| D7 | Arbiter and primer omit the same-volume editorial tests of over-unity claims (Highton, tinfoil "no improvement whatever", zinc duty) | Minor (retention of context) | Add one sentence in primer §7 (open questions or context) with digest pp.256-257, 266-267, 448-449. |
| D8 | The Susskind "in proof", Compton 1939 and Coulson 1957 leads are not in the arbiter's records list | Minor | Add them to arbiter "What would settle it" §4 (records). |
| D9 | Research note wording: Snyder "90 feet below" | Trivial | Orchestrator may correct to "the coil was 90 ft below the observatory" (Woodbury). |
No verdict changes. Remediation is MINOR: run the deep reads below; update claims.md and facts.json; add a "QE revision" section to arbiter.md (D5-D8, no changed verdicts); edit primer.md §4 and §7 (D6, D7). Then the QE recheck (Sonnet) appends "Recheck".
8. Deep-read requests (most important first)
- Wahl, "A Large Induction Coil": slug
on-a-new-connection-for-the-induction-coil-Journal-of-the-Franklin-Institute-47c675, PDF pp.319-323, Sonnet. Question: transcribe every number in the article from page images at 400 dpi or more (fractions especially), so that facts.json can be corrected in one pass: wire length and weight, core wire thickness, primary diameter, condenser area, dimensions (height, base, bobbin), jar area and spark length (2¾ or 2⅜), the 20-pane and 8-pane condensers, the 12 in quantity spark, the four half-gallon jars and the zoetrope distance. Record the journal page range from the running heads. - Editors' short items on energy claims and batteries: same slug, PDF pp.256-257, 266-267, 433, 448-449, Sonnet. Question: what exactly did the editors argue and test about the Highton doubled-wire claim, the Newark galvanic engine, the tinfoil electromagnet (method, apparatus, result) and the zinc and coal duty calculation? What does Sharples' note say about the EMF and internal resistance of bichromate cells? Quote with page tags, for primer §7 context and the arbiter's open E₀ question.
Corrections to earlier documents
- facts.json and claims.md: see D1-D3 (OCR figures; Houston aim; claim 3 blending).
- facts.json Wahl path and deep read: Wahl is journal pp.212-216 (running heads PDF p.319 "212", p.323 "216") [read].
- facts.json and digest part 11: the base width is 13 in, not 18 [read: page image p.321].
- physicist.md Correction 1 and 2: the secondary weighs 44½ lb (not 44¼); the core wires are about 1/18 (or 1/16) in, not 1/8 [read: page image p.321].
- arbiter.md Correction 7 and primer §4: Houston 1876 already extends the 1871 reach to "any metallic objects in the neighborhood of the coil, and, indeed, in adjoining rooms" [read: Houston 1876]. The retelling's breadth partly traces to the inventor.
- research/2026-10-08-archive-org.md: Woodbury places the coil "90 feet below" the observatory, not Snyder 90 ft below the coil.
Recheck
Recheck (Sonnet, 2026-10-08) of claims.md, facts.json, arbiter.md ("QE revision"), primer.md and the research note, read in their current state.
- D1 (claims.md): FIXED. Claim 5 is now 44 1/3 miles against App's 150 miles; claim 6 is a 1 1/3 sq ft jar, about 2 3/4 in, with 14 in for the 20-pane condenser; claim 3 separates the neither-earthed five-inch case; claim 1 says the table was in partial earth connection; the scope note now covers Morton 1867 and the 1876 papers.
- D2 (facts.json Wahl entry): FIXED. It now has 44 1/3 miles, 44 1/2 lb, 325 sq ft, 0.1655 in, 1 1/3 sq ft, 2 3/4 in, base width 13 in, core about 1/18 in and journal pp.212-216. The "more wire" contrast is reversed, and an explicit CORRECTION note is added. The old figures remain only inside the sources line that labels them superseded.
- D3 (facts.json Houston entry): FIXED. The aim quote, the engine/water-pipe connection (1876), the primary/secondary caveat on the battery wording and the [web] Thomson tag are all present.
- D4 (deep read page range): FIXED. The deep read now has a later section giving j.212-216 and flags the earlier 213-216 as superseded.
- D5 (physicist misreads): FIXED. Recorded in arbiter QE revision section 2; no verdict effect.
- D6 (Houston 1876 reach): FIXED. Arbiter QE revision section 1 amends Correction 7. Primer §4 and the evidence table carry the 1876 quotes, and the date, person and "continuous" differences are kept.
- D7 (editors' over-unity tests): FIXED. Arbiter QE revision section 3 and the primer §7 context block cover Highton, the tinfoil test and Sharples, with page tags.
- D8 (leads): FIXED. Susskind "in proof", Compton 1939 and Coulson 1957 are in the arbiter's records list.
- D9 (research note wording): FIXED. The archive-org note now says Snyder witnessed sparks in the observatory; the "90 feet below" slip is gone.
Residual, not a defect: the two stale-figure mentions left in the primer (the "443 miles" units check in Step 6 and the review question 5) are deliberate teaching examples that name the error.
Final: PASS