Bedini-John: 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 spot checks. No deep/ folder exists; this project was reviewed before digests existed, so Step 1 read the documents directly (full text for the manual and web page, page images for some drawings).

Verdict: MINOR

The central verdicts hold. The defects are: (1) arbiter.md still presents as its "single most useful finding" a reading of Sriphan et al. Table 2 that the primer later showed to be wrong, with no QE revision section; (2) facts.json describes eight of the 17 drawings as "not opened / inferred from filename", but the OCR text and digest now show what they say, including one builder's sweet-spot reading (2.2 A draw, 0.3 A charge) that bears on Claim 2; (3) several manual passages were missed: the manual's own conventional-charger control test and the 1-ohm-test claim. None of these changes a final verdict. All can be fixed by updating facts.json, adding a QE revision to the arbiter, and making small primer edits.

1. Coverage

Every manifest document has a digest or a stated, acceptable skip reason. One text defect: the image sequence -5058f1 has OCR text on pages that facts.json says were never opened.

Mapping used below. The OCR pages follow the manifest's source order (checked against content: p.1 = 800px-Self-runner1, p.3 = Circuit diagram1 "Drawn by FFF", p.5 = JB-Circuit, p.6 = N-Pole motor 84, p.14 = dualbatt, p.17 = photo-249588 all match). So p.2 = B&C N-Pole motor update, p.7/p.8 = N-Pole-Motor-6a/6b, p.9 = North Pole Motor 2 batts, p.10 = SCH-illust. 2coil 4inch, p.11 = SG North Pole Motor, p.12 = SSG.jpg, p.15 = north1, p.16 = npole.

2. Shortcuts

• The central claims (1, 2, 4, 5) rest on the BM3 manual, which Step 1 read in full. Claim 3b rests on the web page, also read in full. Claim 3a rests on the Self-runner schematic, which was viewed as a page image. Claim 6 rests on dualbatt.jpg and N-Pole motor 84.jpg, both viewed. No central claim rests on a digest alone. That is acceptable.
• CBA manual sampling: acceptable because the document is peripheral (see above).
• Unopened drawings: a shortcut on secondary material, but one of these drawings (SCH-illust, p.10) holds the only builder-recorded current numbers in the folder. It should have been opened. This is a defect (minor).
• The Sriphan et al. 2018 paper, the only published data set, is outside the folder. The arbiter and primer read it from the web. Its weight in the verdicts is reasonable, and the primer's re-read of it is the most careful piece of work in the project.

3. Missed material

From the digests and the OCR text, not reflected in facts.json, claims.md, the reviews, the arbiter or the primer:

1. A builder's current readings (SCH-illust. 2coil 4inch.jpg, OCR p.10). "John Bedini SSG REPLICATION", 150 ft of twisted #18 AWG, 2 coils, "SWEET SPOT ... 2.2 AMP ... CHAR: 0.3 AMP", with a leehouse.net/john34 URL [read: OCR p.10, needs page-image confirmation]. If this reads as 2.2 A drawn from the primary and 0.3 A into the charge battery, it is the only measured primary-side figure in the folder, and it points the mainstream way. Nobody noticed it. It should be confirmed and added to the Claim 2/3a evidence.
2. SSG.jpg is not an "SSG photo" (facts.json guess). It is "SSG NORTH POLE MOTOR (c) John Bedini, 'Charging off collector'", charging small batteries through an SCR, 6-12 V, #23 and #26 wire, 300 ft each, dated August 20, 2005 [read: OCR p.12]. That is a further variant (SCR/capacitor discharge into the charge battery) that the Claim 6 lineage omits.
3. B&C N-Pole motor update.jpg (OCR p.2) and N-Pole-Motor-6a/6b (p.7, p.8). These are Badertscher redraws (copy-claim Sept 30, 2005 and Oct 28, 2005) of the Bedini/Cole N-pole motor, dated "1979-1984", with an SCR/capacitor stage added [read: OCR]. facts.json dates the lineage only to 1984.
4. The Self-runner backpopping note. "Backpopping of primary takes place once every rotation or more to keep it in recharging mode always", plus note 7: a switch S2 that disconnects the primary while pulsing "improves rotation and charging significantly" [read: OCR p.13]. Claim 3a cites the general assertion but not this mechanism statement. That statement is testable: a primary disconnected each cycle and "recharged" by its own pulses is a switched-battery arrangement close to the "Tesla switch" also linked from the web page.
5. The manual prescribes its own control experiment (p.24): two new batteries with the same date code, discharged to about 12.2 V on the same load, one charged by the SSG and one by a conventional charger, cycled to compare load times [read: manual p.24; digest p.24]. The arbiter's "What would settle it" and the primer's "Genuinely open" both propose a plain-charger control as if it were new. They should say that the community's own protocol already includes one, and explain why it still fails as a test: it compares chargers, not the whole system's energy, and it does not count the primary battery.
6. The 1-ohm resistor test claim (manual p.19-20): the manual says the test "demonstrates two kinds of energy" going to the charging battery and that conventional energy alone is insufficient to account for charging [digest p.19-20]. This is a distinct, checkable claim (a voltage under 1 V across 1 ohm caps conventional current at under 1 A) and no review addresses it. It belongs under Claim 1.
7. The matched-bulb power method (manual p.21) is mentioned in facts.json but not analysed by any reviewer. It is a weak measure of the power in a pulsed waveform: a bulb responds to RMS-like heating, not to the average current that charges a battery. This is a useful teaching point for primer section 7.
8. Minor: the digest calls the author of OCR p.9/p.14 "Terry Bedini"; that is an OCR error for "John Bedini". No later file used it, so no action is needed.

4. Accuracy spot checks (against _text/)

No numerical errors were found in the load-bearing figures.

5. Process

• Source tags: present throughout (each of the five reader files has tens of [read] tags; [web] carries URLs; [memory] carries confidence levels). Honest. facts.json correctly marks the filename-only descriptions as [memory] low.
• Web checks: the arbiter checked the patent, the 9 V capacity and the Sriphan paper, and listed which [memory] items it left unchecked and why. Acceptable. The primer's "1949-2016 [memory: medium]" for Bedini's dates is not web-checked, but no verdict depends on it.
• Verdict scale: applied consistently. The arbiter correctly refused "Refuted by calculation" for the physicist's invented numbers and for the anecdotal 3b. Split verdicts are clearly stated.
• Arbiter defect (the main one): arbiter.md, under "Recomputed numbers / System-level check" and in "Corrected claim list" item 2 and "What would settle it", reads Sriphan Table 2's "running battery 12.71 W" as the primary battery and derives a whole-system ratio of about 0.63, calling it "the single most useful finding". The primer and example.py show that 12.71 W and 4.94 W are exactly the means of Table 3's discharge and charge powers on the charge battery, so 0.63 is roughly T_out/T_in and not a system efficiency. The primer flags this in a callout, but arbiter.md itself is uncorrected. The verdicts do not change (Claim 2's whole-system part was already "Never properly tested").
• Second Sriphan inconsistency: example.py shows the replication design's COP column (2.13, 2.19, 1.73, 1.44) does not match its own joule columns (2.90, 3.82, 3.35, 3.34), and its Table 2 powers do not match Table 3. The primer reports this. The arbiter should too, and so should the contrarian's case for "real arithmetic", which holds only for the original design.
• Primer vs STEP 4 and the reader profile: all nine sections are present, in order. Symbols are defined (P, V, I, L, Q, ħ, ω, E_in/E_out), acronyms are defined (SSG, COP, SOC), the worked example uses real numbers, the evidence table has the required columns, there are 5 check questions with answers at the end, and further reading names mainstream and researcher works. Chapter numbers are given only for Griffiths. Good.
• facts.json wording: the CBA entry says "text layer sampled"; it should now read "digest (full, 29 pp.)". Several image entries need how_read and summary updates (section 3).

6. Remediation actions

Deep reads (most important first):

1. -5058f1, pages 1-17, Sonnet. Render each page image (and the source JPGs) for the eight files never opened: B&C update (p.2), N-Pole-Motor-6a/6b (p.7-8), North Pole Motor 2 batts (p.9), SCH-illust. 2coil 4inch (p.10), SG North Pole Motor (p.11), SSG.jpg (p.12), north1 (p.15), npole (p.16). Re-read p.1/p.13 (Self-runner notes 1-9). Questions: (a) On SCH-illust, what exactly do "SWEET SPOT ... 2.2 AMP" and "CHAR: 0.3 AMP" refer to (primary draw vs charge current), and is any other measurement recorded? (b) What do SSG.jpg and the Badertscher redraws add (SCR/capacitor "charging off collector" stage, dates)? (c) Transcribe Self-runner notes 1-9 exactly, including the "backpopping once every rotation" note and the S2 primary-disconnect note.
2. Bedini-Monopole-3-Group-Experiment-V1-01-c67ca7, pages 19-24, Sonnet. Questions: (a) What exactly does the manual claim the 1-ohm resistor test demonstrates, and what numerical bound does it imply on conventional charge current? (b) State the matched-bulb power method and the conventional-charger comparison protocol (p.24) exactly, and say whether either measures the primary battery's energy.

Changes to later files:

• facts.json: update the -5058f1 image entries with what the deep read finds (correct how_read to "page images" and drop the [memory]-low filename guesses for SSG.jpg, SCH-illust, the N-pole files, SG North Pole Motor, B&C update; add dates 1979-1984 and the 2005 redraws; add the 2.2 A / 0.3 A sweet-spot reading if confirmed). Change CBA how_read to "digest (full text, 29 pp.)". Add key_claims to the manual entry for the 1-ohm test and the conventional-charger comparison.
• claims.md: under Claim 1, add the manual's 1-ohm-test assertion as supporting text. Under Claim 3a, add the "backpopping once every rotation / S2 disconnect" mechanism. The claim list itself does not need to change.
• arbiter.md: add a "QE revision" section that (a) withdraws the Table 2 "whole-system ratio 0.63" finding (12.71 W and 4.94 W are charge-battery means; 0.63 ≈ T_out/T_in), (b) notes the replication table's internal inconsistency, (c) records the SCH-illust current reading (if confirmed) as the only primary-side number in the folder, (d) notes that the manual's p.24 already prescribes a conventional-charger control and explains why it still cannot test over-unity, and (e) addresses the 1-ohm test under Claim 1. No final verdict is expected to change; state that explicitly.
• primer.md: in section 5 or 7, add the SCH-illust reading (if confirmed) as a worked contrast (2.2 A out of the primary vs 0.3 A into the charge battery). In section 7 "Genuinely open", say that the manual's own charger comparison exists but compares chargers, not the system. Optionally add a sentence in section 3 on why matched-bulb brightness mis-measures pulsed power. Keep the existing arbiter-correction callout, but point it to the arbiter's QE revision once that is written.
• physicist.md / mathematician.md / contrarian.md: no rewrite is needed (MINOR). The contrarian's "real arithmetic" point should be qualified in the arbiter's QE revision as applying to the original design only.

Recheck

Date: 2026-10-03. Checked against current arbiter.md, facts.json, claims.md, primer.md, example.py and the new deep/ reads.

Defects 1. Arbiter "single most useful finding" (Table 2 ratio 0.63): FIXED. Marked WITHDRAWN in the body and explained in a new "QE revision" section (arbiter.md, section 1); Claim 2 text and the corrected claim list no longer use it. 2. facts.json: eight drawings "not opened / inferred from filename": FIXED. Entries now say "digest + deep read / page images" and describe SSG.jpg, SCH-illust, the N-pole files and the B&C update from the page images (dates 1979-1984 and the 2005 redraws included). Only the three Trevor bench photos remain unopened, which was an accepted skip. 3. Missed manual material (own conventional-charger control, 1-ohm test, matched-bulb method): FIXED. Deep read of manual pp.19-24 exists; facts.json, claims.md, arbiter.md (QE revision sections 6) and primer.md (section 5 step 8, section 7) now cover them. 4. Missed material: builder's 2.2 A / 0.3 A reading: FIXED. Confirmed from the page image as 15 V in at 2.2 A from a power supply vs 0.3 A charging out. This corrects QE's own guess (primary battery draw). It is recorded in facts.json, claims.md, arbiter.md and primer.md (step 7). 5. Missed material: backpopping and S2 note: FIXED (claims.md Claim 3, arbiter QE revision section 5, facts.json, primer). 6. Sriphan replication table inconsistency and the "real arithmetic" qualification: FIXED (arbiter QE revision section 2). 7. CBA manual how_read: FIXED ("digest (full 29 pp.)").

Remediation actions - Deep read 1 (-5058f1, Sonnet): done, deep/-5058f1.md exists. - Deep read 2 (manual pp.19-24, Sonnet): done, deep/Bedini-Monopole-3-Group-Experiment-V1-01-c67ca7.md exists. - facts.json, claims.md, arbiter.md QE revision, primer.md updates: all done. The primer callout now points to the arbiter's QE revision. - physicist/mathematician/contrarian: no change required (MINOR).

Checks: example.py runs, exit 0. No final verdict changed, as QE expected.

Final: PASS