# ADVERSARIAL VERIFICATION — pages-1-16-plaintext-and-rune-alphabet - **Artifact under test:** `/artifacts/pages-1-16-plaintext--MERGED.md` (plus `--A.md` and `--B.md` beside it). - **Verifier posture:** skeptical; goal was to REFUTE. - **Verdict:** **agrees = FALSE.** The decoded plaintext payload itself is sound and byte-faithful, but the merge contains a **confirmed false provenance claim** (V4 / §5 P014) about the primary source, plus a secondary provenance-tier / ledger-completeness issue. --- ## WHAT I CONFIRMED (the payload is genuine, not fabricated) 1. **Byte-faithful re-derivation — CONFIRMED against the cited primary.** I extracted the 23 `
` blocks from the actual raw HTML of the primary capture on disk
   (`harvest/raw-012.html` = oldid=634 = Wayback `20170929225902id_`; oldid=634 embedded twice, no
   toolbar, consistent with the `id_` raw capture). Blocks 0–14 = pages 1–16 (block 4 = pages 5+6).
   All **15** MERGED §4 page blocks match the primary `
` blocks **exactly**, including interior
   trailing grid-padding spaces (e.g. `NO ONE COULD DECIPHER ` with the trailing space), the four
   666 cubes cell-by-cell, the `(=666)` tags, the `// Info:` lines, the `{curly-brace}` figures, the
   CJK `天`, and `π7π^7`. Zero character differences. No sign of memory-reconstructed or smoothed-over
   text in the plaintext body.

2. **Independent capture diff — CONFIRMED.** I diffed against a *different* Wayback capture not used
   for the §4 derivation: `harvest/raw-082.html` = Wayback `20171001170618id_` (2017-10-01, the one
   Reconstruction B cited). Its 23 `
` blocks are **byte-identical** to raw-012 for pages 1–16
   (0 differing blocks). Two independent contemporary captures agree exactly. (The non-Wikia
   steemit/hive channel `harvest/073` only paraphrases page-1 prose in its title, so it is not a
   clean verbatim channel — not usable for a character diff.)

3. **Numeric recomputation — ALL CONFIRMED** (`fractions` + `mpmath`, 180 dps):
   - Repunit `11111111111111111111111` = **23 ones** ✓
   - P015 fraction `(2·23·499·19214759967251·55150662460749672076915609)/(3·11·47·139·2531·549797184491917·[23 ones])`
     = `0.(0072973525613766677788831415921618033299792458)`, **period exactly 46** ✓ (matches gloss)
   - P010 `2^9·3^-1·5^9·197^-1·5563^-1·41681^-1` = `0.00729735256454443…` and **exactly** `1/137.035999173` ✓
   - P010 "last ten of 50 000 places = 1515009125": the run `1515009125` sits at decimal places
     **49990–49999**; places 49991–50000 = `5150091252`. The MERGED file's boundary note states this
     correctly ✓ (honest, accurate annotation)
   - P013 `(π7)/(π^7)` and `(π^7)/(π7)` — both 114/116-digit gloss strings reproduce ✓
   - P014 `((7^π)/(7π))·6.67 = 137.0350666248123480977772779259697505406927854787723951833…` and
     `((7π)/(7^π))/6.67 = 0.007297402224336455467525344147398860004528664132146759443…` — both reproduce ✓
   - All four 666-cube grids: **36/36 rows and 36/36 layer-columns sum to 666** ✓ (transcribed digits
     internally consistent)

4. **Rune rules — CONFIRMED verbatim** in `harvest/017-...for-beginners.md` (the five "For beginners"
   rules, the Job 15:2 rough read `SKOOLD - A - ?ISA - MAN …`, `T?ITTAR → Twitter @666ab731`,
   the `3I68n.png` image link). The GAP (no per-glyph rune→Latin chart transcribed; `3I68n.png` dead
   with no CDX capture) is honestly disclosed and NOT fabricated. Confirmed genuine gap.

---

## DISCREPANCIES FOUND

### D1 — FALSE provenance claim: the P014 `137.0350666…` decimal IS in the oldid=634 capture (CONFIRMED)

The MERGED file states in **three** places that this decimal was absent from the primary capture:

- §5 P014 NOTE: *"the `137.0350666…` form … was NOT in the oldid=634 capture's commentary (which
  listed only the `0.00729740222…` form); it is mathematically exact and appears in later revisions."*
- §7 ledger row **V4**: *"ABSENT from the oldid=634 capture's commentary (only the `0.00729740222…`
  form there); present in later revisions … RESOLVED (verified)."*
- Reconstruction JSON `variants_seen` V4: *"ABSENT from the oldid=634 capture's commentary … from a
  later revision."*

**This is false.** The string
`137.0350666248123480977772779259697505406927854787723951833...` is present **verbatim in the
oldid=634 capture itself** — the exact primary the merge says it used for §4. In `harvest/raw-012.html`
it renders as `((7^π) / (7π)) * 6.67 = 137.0350666248123480977772779259697505406927854787723951833...`
(the `666` is wrapped in `…` highlight tags, which is why a *literal* string search misses it).
It is also present in the harvested markdown of that same capture (`harvest/012-...full.md`, line 582),
and in the **independent** 2017-10-01 capture (`harvest/raw-082.html`).

Two further errors in the same claim:
- The oldid=634 commentary lists **BOTH** forms consecutively (`137.0350666…` immediately followed by
  `and … 0.007297402224336455…`), not "only the `0.00729740222…` form."
- The "appears in later revisions / from a later revision" provenance attribution is **unsupported**
  and misleading — the value is already in oldid=634 and in the 2017-10-01 capture.

**Impact:** The plaintext/number is correct and IS primary-sourced, so no data is corrupted. But this
is a fabricated provenance rationale laundered into a "RESOLVED (verified)" ledger entry, and it means
the §8 VERIFICATION LOG's claim that the primary `
`/commentary bytes were "re-fetched and parsed"
did not actually catch what the primary contains. Almost certainly caused by a naive literal grep
defeated by the inline `666` tags — i.e. the merge asserted an absence it never actually
verified. Per the task's "any discrepancy → agrees=false," this alone flips the verdict.

### D2 — A-vs-B TIER disagreement omitted from the ledger; top-level tier overstated (provenance)

§7 asserts it enumerates *"every A-vs-B difference"* and that they are *"confined to transcription
hygiene … plus differing provenance citations,"* with *"No A-vs-B conflict remains unresolved."*
But there is a substantive provenance-**tier** disagreement not in the V1–V8 table:

- Reconstruction **A** grades the decoded-plaintext channel (Wikia "Translation", oldid=634)
  **TIER: PRIMARY**.
- Reconstruction **B** grades the same channel (its S1) **TIER: SECONDARY** ("contemporary community
  decipherment").

The reconstruction JSON adopts A's grade at the top level (`tier: PRIMARY`). Strictly, the decoded
**English** text is a community decipherment/translation layered on the genuinely primary rune rasters
(`P01.png…P16.png`); `harvest/012` itself flags the wiki's reading as carrying a SECONDARY commentary
layer. The MERGED §2 partially hedges ("PRIMARY for the page rasters/original book text it reproduces;
… notes are SECONDARY"), but B's position — that the translation *itself* is SECONDARY — is neither
reflected in the top-line tier nor logged as an A-vs-B conflict. This is a real, unlisted A/B
provenance difference and a mild overstatement of tier, though not a payload error.

---

## SUMMARY

- The decoded plaintext of pages 1–16, the rune reading-rules, the four 666 cubes, every equation,
  and all five long decimals are **faithfully reproduced and independently corroborated** — no
  fabrication in the payload.
- **D1 is a confirmed factual error** in the provenance/variants/verification metadata: the claim
  that `137.0350666…` is absent from the oldid=634 primary capture is false (it is present there,
  and in a second independent capture, with both P014 forms listed).
- **D2** is a lesser provenance-tier / ledger-completeness issue.
- **agrees = FALSE** on account of D1 (and D2).

### Evidence files
- `harvest/raw-012.html` (oldid=634 / 2017-09-29 primary), `harvest/012-...full.md` line 582
- `harvest/raw-082.html` (2017-10-01 independent capture) — byte-identical pages 1–16
- `harvest/017-tengri137-wikia-for-beginners.md` (rune rules verbatim)