Rankings Primer Arbiter Physicist Mathematician Contrarian QE Sites log Claims Sources Research notes

Radiant_One_Youtube: Quality Engineering (Step 4b)

QE inputs read: _text/manifest.json, _text/ (the one extracted text), digest/ (all: one file), facts.json, claims.md, sites.md, research/ (all three notes; research/files/ is empty), physicist.md, mathematician.md, contrarian.md, arbiter.md, primer.md, example.py (run). deep/ does not exist. I also viewed the source PNG myself and zoomed in on all four 1N4148 symbols (scratch crops; source folder not modified) [read: page image p.1].

Verdict: MINOR

The coverage is complete for a one-image project, and the arbiter's verdicts stand. One load-bearing number is wrong, though. All three reviews, the arbiter, the primer's worked example and example.py read the author's "each led is only drawing 4mW from a battery" (GndNrgEx_9, 2016) as LED output power. They then divide it by an input figure from a different video a year earlier (30 mA, GndNrgEx_3, 2015) to get an efficiency of "58 %". The sentence more naturally reports battery input per LED. If so, the record has no output measurement at all, and "on the author's own numbers there is no surplus" is not supported. The verdict categories do not change: the strong form is still "Never properly tested" and none of the verdicts rested on the 58 %. What has to change is the reasoning and the worked example. That can be fixed in the arbiter and the primer, so the verdict is MINOR.

1. Coverage

Manifest document	Digested?	Read further?	Status
Earth-energy-extra-schematic.png (slug Earth-energy-extra-schematic-33a2ab), 1 page, OCR	yes (4 lines)	Step 1b, all three reviewers, the arbiter and QE viewed the image directly	Covered

The manifest has one document and nothing was skipped. Step 0's OCR text is badly garbled ("Exjjaction", "41.13 L", "26.71"). Step 1b correctly went to the image for the component values, and the digest marked the OCR problems. No defect.
The source folder holds only this one file (321 KB PNG) [checked by directory listing]. No file was missed.

2. Shortcuts

No deep read. This is acceptable for a one-page schematic, since everyone who relied on it viewed the image directly.
All measurement data are second-hand. Every number (30 mA, +1.2 K / +0.7 K, 4 mW per LED, 12 V 42 Ah battery) comes from the Investigator's paraphrase of YouTube descriptions and auto-generated (ASR) transcripts. Not all of the key videos were read: the transcripts of GndNrgEx_1, GrnNrgEx_5 and GndNrgEx_10 were not read, and GndNrgEx_9 has no captions, so it is known only from its description and thumbnail. No reviewer watched a video. Every file states this honestly. It is acceptable as a source gap, but it is the direct cause of the defect in section 4. Nobody checked what the 4 mW figure is (how it was obtained, and whether it is input or output) against the video.
The Investigator ran short. The run used 7.6 of its 10 minutes and did not reach the overunity.com 2015 thread, Stiffler's SEC documents or the Wayback Machine. The arbiter and primer say so. This is acceptable for run 1, and those sites are logged with "Recheck? yes".

3. Missed material

Checked against the digest, the extracted text and the research notes:

Which supply gave the 30 mA reading. GndNrgEx_1 (2015) says "The power supply registers no change in the current" [web: https://www.youtube.com/watch?v=JAxUJODVMfY, via Investigator]. The supply type is not stated for the 2015 videos (GndNrgEx_1 and _3). The battery and the "should not be powered by a mains power supply" note come from GndNrgEx_9 in 2016. The physicist (correction 5), the primer §1 ("the battery current is the same, about 30 mA") and primer §3 combine the 2016 battery with the 2015 30 mA reading. This matters for the mechanism. If the 2015 runs used a mains bench supply, the earth-rod loop could close conductively through the supply's protective earth or its filter capacitors, not through ~25 pF of stray capacitance. That would undermine the "25 pF vs 150 pF" frequency test proposed in the arbiter and primer. The author's later warning against mains supplies may be his own recognition of this. No file raises it, apart from a passing line in the contrarian review ("removes mains-ground coupling ... as confounders").
The GndNrgEx_9 thumbnail meters. The thumbnail shows two meters reading "12.00" and "0.000"; the quantity on the second is not stated [web: K0YO3o4Tu9o, via Investigator]. If the second is a current meter on an amps range, 0.000 is not consistent with 17-30 mA, which suggests RF upset or a different quantity. This is recorded in the research notes but not in the reviews, arbiter or primer.
One Wire Energy power statement. "The power consumption in the driver does not alter with the earth connection" (xKaYttqzHcQ) is the same "no change in input" pattern as claim 3, in a second setup. The primer keeps the video but not this statement.
GrnNrgEx_5. The description says "a real ground connection is important if higher output levels are to be obtained", and the thumbnail says "First insert the diode into the connection to ground". The primer lists only the video ID. The statement bears on the "rod vs floating plate" control and should be quoted.
Google legal removals. The Investigator's Stiffler query showed "3 results removed on a legal request (LumenDatabase)" [research/2026-10-04-energeticforum.md]. No review mentions it. Under the absence-of-evidence rule it should be recorded with its competing explanations: most likely a copyright notice on kit documents, or something else. The Lumen Database notice would decide.
Forum figures from others. The Energetic Forum post by kcarring says "12V / 75 mA - 15 Watt 18" Fluoro". This is a third party's report of a 0.9 W input driving a tube rated 15 W, a sister claim from the same lineage. There is also jonnydavro's tuning with a magnet on L2. The primer summarises this only as "lighting LEDs and fluorescent tubes". It is peripheral, but it should be kept as a quoted item.

4. Accuracy spot checks

#	Fact	Checked against	Result
1	Component values V1 11-13, R1 1Meg, C1 470pf, Lb 1.4 µH, L1 8.6 µH, L2 26.7µ, Q1 MPSA06	PNG [read: page image p.1]	Correct in facts.json, the arbiter and the primer. The OCR "41.13 L" and "26.71" are garbage, as Step 1b said.
2	Orientation of the two 1N4148 diodes across the LED group	PNG, enlarged crops of both symbols	Arbiter correct: both point the same way, with the anode on the L2 side and the cathode on the earth-rod side. The physicist ("in opposite directions"), the mathematician ("opposite orientations") and the contrarian/facts.json ("each with an anti-parallel 1N4148", one per string) were wrong. The primer carries the correction. I also checked the clip diodes: two in series, both with the anode toward ground and the cathode toward the base (the usual base-reverse-voltage clamp of a Slayer exciter). No file states their orientation; the primer could add it.
3	Printed warning and tuning text	PNG	Verbatim correct in facts.json, the arbiter and the primer. The orange blocks do cut the title ("Ex▮action", "Yo▮Tube"), so reading them as an overlay is right.
4	"Each led is only drawing 4mW from a battery despite being brightly lit!!" = LED output power	research/2026-10-04-youtube.md (GndNrgEx_9 description)	Defect. The quote is exact, but the reading is not supported. "Drawing ... from a battery" describes input: 52 × 4 mW = 208 mW, which is 17.3 mA at 12 V (16-19 mA over 11-13 V). "Despite being brightly lit" fits that reading, because the author is stressing a low draw. The figure is also from 2016, while the 30 mA is from 2015. Dividing one by the other mixes two videos, possibly two builds and possibly two supplies. On this reading there is no LED-output measurement in the record, and the "58 % out/in" (physicist, mathematician, contrarian, arbiter table and verdicts, primer §1, §5 Step 5, the evidence table, §7 and example.py Step 5) is an artefact. The output-power reading is still possible, so the files should give both readings, not pick one.
5	Heat-test numbers (+1.2 K, +0.7 K over 22.5 °C; 6 mW and 2.5 mW at 200 K/W)	research notes; example.py output	Correct (23.7 − 22.5 = 1.2; 23.2 − 22.5 = 0.7). Arithmetic confirmed: 6.0 / 3.5 / 2.5 mW, the "42 K" counterfactual, and "factor of about 80" (208/2.5 = 83). The 42 K counterfactual and the "factor of 80" also assume 208 mW of LED output (see #4). They should be reworded as "if ~200 mW had moved ...".
6	Fair-weather air-earth current	arbiter web check	The physicist wrote "microamperes per square metre" [memory, high]. That is wrong by about six orders of magnitude, since the true value is about 2 pA/m². The contrarian's "~pA/m²" and the arbiter's web check (2 pA/m²) are right, and the primer uses the corrected value. Fixed downstream, so no remediation is needed beyond noting it.

The other primer arithmetic was spot-checked and is correct: 6.20 / 2.50 / 2.32 MHz; 24.6 and 151 pF; sphere capacitances of 11.1 / 27.8 / 55.6 pF; 59 pF at 4 MHz (answer 2); the 1.4×10⁷ shortfall; 504 Wh and 1400 h.

5. Retention

The arbiter and the primer keep the Investigator's findings even where the reviewers judged them negatively. Kept items: the author's Moray framing and his "much larger response from the ground electrons", placed beside his "no overunity" disclaimer; the heat test, credited as good practice; the "1N4148 stops ground energy" observation; the fibreglass-table and "asymmetrical pathways" statements; the One Wire and Full Metal Jacket videos; the HV "electro-radiant" videos; the Joule thief; the related YouTube demonstrations; the "Henry Murry" commenter story (kept as the commenter's claim); the "You didn't show the circuit working" comment; and the missing videos _4/_6/_7/_8, with a full absence-of-evidence analysis of competing explanations. No pertinent item was dropped because of a negative judgement.
Minor omissions, which are not negative-judgement removals, are the items in §3 points 2-6. They should be added to the arbiter's "retained" list and the primer's "Other material kept in the record".
The interested-party labelling is done correctly for the author and for Stiffler. However, the claim that Stiffler "sells exciter kits" is tagged [memory, low; not verified] and is used to label him an interested party. The Investigator's Google result "SEC Exciter 15/20 Kit Assembly Guide" (Scribd) partly supports the existence of kits. Retag it as [web: Google snippet, kit assembly guide exists; seller not confirmed], or web-check it.

6. Process

Source tags. These are present and honest throughout. Everything that comes from YouTube is labelled "[web, via Investigator]", and every file says that no one watched the videos. Calculations are labelled as the reviewer's own.
Web checks. The physicist did none ("no verdict depends on a medium/low-confidence [memory] fact"), yet relied on a wrong [memory, high] atmospheric-current figure. The contrarian left the fair-weather figure as [memory, medium], unchecked. The mathematician checked the MPSA06 datasheet. The arbiter checked the fair-weather field and current, the MPSA06 and the LED forward voltage. This is adequate after the arbiter's pass. Telluric current magnitude stays [memory, medium] and is flagged as such in the primer.
Verdict scale. The arbiter applies it consistently. The contrarian's non-scale phrasing on claim 3 ("observation reported by one source") and the physicist's "Established" for claims that rest on theory were normalised by the arbiter, with reasons given.
Primer structure and reader profile. The primer follows STEP 4 sections 1-9, with the answers at the end. Every symbol is defined (ω, Q, θ_JA, ε₀, displacement current), and there are worked numbers throughout. It is written at the right level. The source gap is repeated at the top. No new verdicts appear.
example.py runs cleanly (exit 0) with the venv Python, and its output matches primer §5 line by line. Its Step 4 and Step 5 inherit the "4 mW = LED output" reading (#4).
Absence of evidence. This is handled well: the arbiter gives four competing explanations for the missing videos, the records that would decide between them, and notes that the search was partial.

7. Remediation actions

Deep reads: none. The manifest has one document, a single-page image. I re-read the image for this QE and the diode orientations are now settled. The open questions need the videos, which are outside the manifest, so they go to the Investigator, not to a deep reader.

Investigator (next incremental run, Chrome), most important first: 1. Watch GndNrgEx_9 (K0YO3o4Tu9o). Establish what "4mW" means (input per LED or measured LED power), what each meter in the thumbnail reads ("12.00", "0.000"), and the supply current in that build. 2. Find out the supply type in GndNrgEx_1 and _3 (bench supply, mains-earthed or not, or battery), and whether the 30 mA was read on the supply's own meter. 3. Read the transcripts of GndNrgEx_1, GndNrgEx_10 and GrnNrgEx_5 (diode in the ground lead; "real ground connection"). 4. Check the Wayback Machine for the playlist and for _4/_6/_7/_8; search overunity.com for the 2015 thread; read the Stiffler SEC documents (counterpoise passage, and whether Stiffler sells kits); look up the Lumen Database notice behind the 3 removed Google results.

facts.json / claims.md: - facts.json: correct "two parallel strings ... each shunted by an anti-parallel 1N4148" to "two 1N4148 in parallel, same direction (anode at L2, cathode at earth rod) across the whole group; LED orientation not drawn". Correct the orange-bar description to "overlay; no electrical meaning". - claims.md claim 2: replace "with anti-parallel 1N4148 diodes" with the same wording. Optionally add a 5th line recording the author's figures as reported, giving the date and video of each (30 mA, 2015, supply type unknown; 4 mW per LED "drawn from a battery", 2016).

arbiter.md: add a "QE revision" section: - State the two readings of "4 mW per LED drawing from a battery". (a) Input per LED: total about 208 mW, i.e. about 17 mA at 12 V, in the 2016 build; then the record has no output measurement. (b) Measured LED power: then 58 % only if the 2015 input of 30 mA applied to the 2016 build. Withdraw "own numbers show 58 %, no surplus" as a finding. Replace it with: "The record contains no measured LED output power. The author's figures neither show nor exclude a surplus." This strengthens the contrarian's point already adopted by the arbiter. - Verdicts: claim 1 strong form stays Never properly tested, with the supporting reason changed. Claims 2-4 are unchanged. - Add the supply-type question (§3 point 1) to "What would settle it" and "Open questions". A mains-earthed bench supply would let the loop close conductively, which changes the capacitance and frequency test. - Add §3 points 2-6 to the retained list. Retag the Stiffler "sells kits" claim.

primer.md: - §1: replace "the battery current is the same, about 30 mA" with "the supply current" (the 2015 supply type is not stated). Replace "On the author's own numbers the LEDs get about 58%..." with the two-readings statement above. - §3 Supply bullet: say that the battery is documented for the 2016 build (GndNrgEx_9) and the 30 mA for the 2015 build (GndNrgEx_3). Add the clip-diode orientation (cathode toward the base). - §4 "Energy ratio" row and the closing paragraph: the "measured numbers fall on the no-overunity side" statement depends on reading (b); reword it. - §5 Step 4: keep it as an illustration ("if ~200 mW reached the LEDs ..."). Step 5: rewrite the table as a conditional under reading (b), add the reading (a) arithmetic (208 mW ÷ 12 V = 17 mA, with no output figure), and say the two figures come from different videos a year apart. Step 6: reword the "42 K" and "factor of 80" lines as conditionals. "What the example shows": drop "at about 58% efficiency". - §6 evidence table, claim 1 strong form, Result column: replace "208 mW out of 360 mW in (58%), so no surplus" with "No LED output power was measured. The 4 mW-per-LED figure is most naturally battery draw (2016). Neither a surplus nor its absence is shown." - §7: under "Genuinely open", change "How was '4 mW per LED' measured?" to "Is '4 mW per LED' input or output, and how was it measured?". Add "Was the 2015 supply mains-earthed?". - "Other material kept in the record": add §3 points 2-6.

example.py: in Step 5, print both readings: (a) 52 × 4 mW as input, giving the implied supply current at 11-13 V; (b) the current out/in table, labelled as conditional and cross-video. Label the Step 4 and Step 6 counterfactuals "if 208 mW reached the LEDs".

Recheck

Checked against the current arbiter.md, primer.md, example.py, facts.json and claims.md. example.py was re-run with the project venv Python (exit 0).

4 mW ambiguity (the main defect): FIXED. arbiter.md has a "QE revision" section (R1) that withdraws "58 %" as a finding and states both readings. The verdict table is updated. primer.md §1, §5 (Steps 4-6), §6 and §7 present reading (a) and (b), with (b) as conditional and cross-video. example.py Step 5 prints both readings, and the Step 4 and Step 6 counterfactuals are conditional. The remaining "58" in primer.md line 113 is the conditional table and line 116 explains it, and "58 days" is unrelated.
facts.json / claims.md diode wording and orange overlay: FIXED (same direction, anode at L2, cathode at the earth rod; overlay with no electrical meaning). claims.md has the new line 5 with dates.
Supply-type question (2015 supply unknown): FIXED in arbiter R2 and in primer §3, §7 and the open questions.
Missed material §3 points 2-6 (thumbnail meters, GrnNrgEx_5, Lumen removals, kcarring/jonnydavro; One Wire power statement): FIXED, in arbiter.md and in primer's "kept in the record" list. The One Wire statement is present in primer line 141.
Stiffler "sells kits" retag: FIXED (arbiter and primer).
Clip-diode orientation in primer: FIXED (arbiter notes it; primer line 40 describes the two series diodes).
Investigator and deep-read items (watch GndNrgEx_9, supply type, transcripts, Wayback, overunity.com): REMAINING as a source gap, but these are for the next Investigator run, not a document defect. They are stated openly in the files.

No remaining defect in the documents.

Final: PASS