# VERIFICATION — artifact `pages-17-22-plaintext`

**Verifier posture:** adversarial / skeptical.
**Verified:** 2026-07-24.
**Target file:** `<research-workspace>/artifacts/pages-17-22-plaintext--MERGED.md` (+ `--A.md`, `--B.md`).
**Verdict:** `agrees = false` — the six-page plaintext itself is faithfully transcribed and NOT fabricated, but I found genuine (minor / whitespace / editorial) verbatim discrepancies and one **inverted adjudication** in the MERGED. Under the "any discrepancy however small ⇒ agrees=false" rule, this fails.

---

## SUMMARY

The core deliverable — the decoded English of pages 17–22, in both the PRIMARY blog form (VERSION 1) and the SECONDARY wiki form (VERSION 2) — **checks out**. I could not find any fabricated or memory-reconstructed letters. Three independent lines of evidence confirm the text, the dinome→element decode reproduces exactly by recomputation, and the odd spellings are genuine source features.

However, the MERGED's handling of ONE whitespace character is **backwards**: an independent capture of the actual wiki shows the trailing space in `THOSE ARE EVIL ` is really in the source, so Reconstruction A (which kept it) is the faithful one, B (which dropped it) deviated, and the MERGED wrongly declared "canonical = no trailing space." I also found a 2-word editorial insertion (`(element symbols)`) inside a block the MERGED presents as verbatim. Details below.

---

## WHAT I VERIFIED POSITIVELY (no fabrication found)

### 1. Internal diff claims recompute exactly
Using `difflib`-style character diffs on the extracted code blocks:
- **A-VERSION2 (wiki) vs B (wiki):** identical on all six pages EXCEPT `p18 L16` trailing space (`'THOSE ARE EVIL '` vs `'THOSE ARE EVIL'`). Matches the MERGED's claim. ✔
- **A-VERSION1 (blog) vs A-VERSION2 (wiki):** after stripping the cosmetic single leading space + trailing space, exactly the six documented content divergences [V1]–[V6] appear, no more, no fewer. ✔
- **MERGED spine vs A-VERSION1:** the authoritative spine reproduces VERSION 1 character-for-character (after removing `[V#]` tags / leading space). No introduced errors. ✔

### 2. The dinome→element→plaintext decode reproduces
I built a periodic table (Z→symbol) in Python and fed it the page-17 first-calculation dinome string
`43 77 25 63 87 76 37 22 80 63 43 37 92 22 72 …`:
- Element chain came out `Tc Ir Mn Eu Fr Os Rb Ti Hg Eu Tc Rb U Ti Hf …` — **identical** to the element string in all three files.
- First letters → `TIMEFORTHETRUTH…` ✔
- Schmeh's stated example `33 = As = A` checks (As is Z=33). ✔
- The gibberish tail the source prints, `…NPKIAKVGPPPFBIR`, does **not** perfectly match a naive recompute (`…GPPPFP…`, one extra `P` from Pu=94). This mismatch is **in the primary source itself** (both the blog and the independent wiki capture print `GPPPFBIR`). The reconstruction faithfully copied the source's typo rather than "correcting" it — this is **evidence of fidelity, not fabrication** (a fabricator recomputing would have written `GPPPFP…`).

### 3. VERSION 1 (blog, PRIMARY) confirmed live
Live fetch of `scienceblogs.de/klausis-krypto-kolumne/2017/03/08/how-a-blog-reader-solved-the-tengri-137-mystery/` (web.archive.org is blocked from this environment, same as prior session) confirmed all six contested VERSION 1 readings verbatim:
`THE REARE PEOPLE AMONG YOU`; `BASED ON AMATHEMA`; `KEEP DISTANCE` (no parenthetical); `STRAIGHT FOR WAR`; `THIS IS INEVITABLE (+?)`; `MANY CIVILIZATIONS FAIL (+?)`. The blog's comment #12 (by "Jim") interpreting AMATHEMA → "a mathematical truth" is also confirmed. ✔

### 4. VERSION 2 (wiki, SECONDARY) confirmed against an UNCITED independent capture
I located a **browser-saved HTML copy of the wiki "Tengri 137 Translation" page** inside the GitHub repo `bartman081523/tengri137-selenoprotein`
(`consecutive_reading/Tengri 137 Translation _ Tengri 137 Wikia _ FANDOM powered by Wikia.html`). The reconstruction cites this repo **only** for "page 21 in p23.json"; this HTML file is **not** cited. Pulled it via `gh api` blobs (Fandom returns HTTP 402 and Wayback is blocked here, so this was the only way to see the actual wiki bytes).

Char-by-char diff of its `<pre>` blocks (pages 17–22) vs the reconstruction:
- **vs A-VERSION2:** identical on all six pages, **including** the `THOSE ARE EVIL ` trailing space. ✔
- **vs B:** identical on all six pages EXCEPT B dropped that trailing space. ✔ (see Discrepancy D1)
- All odd spellings present in the source: `RECIEVE`, `TEACHER`, `THIS MESSENGERS`, `MEASURMENT`, `KNOWLEDGS`, `OURR`, `UNICUE`, `PRECENT`, `INT HE`, `LENGHT OEND`. ✔

> Provenance caveat: this saved HTML has its internal links rewritten to `web.archive.org/web/20171001170618/…` — i.e. it is a browser-save of the **same 1 Oct 2017 Wayback snapshot** the reconstruction cites for VERSION 2. So it is not a fully *independent origin* for the wiki; it is an **independent retrieval path** to the same snapshot. It still definitively settles the trailing-space question and exposes the `(element symbols)` insertion.

---

## DISCREPANCIES (why `agrees = false`)

### D1 — Trailing space `THOSE ARE EVIL `: adjudication is INVERTED (verbatim-fidelity)
- **What the MERGED says:** "A renders `THOSE ARE EVIL ` (trailing space), B renders `THOSE ARE EVIL` … Canonical form: **no** trailing space … a `<pre>`-block whitespace artifact." (MERGED §0 table, [W1]).
- **What the source actually is:** the independent wiki capture's `<pre>` block for page 18 ends `…\nTHOSE ARE EVIL \n` — the trailing space **is genuinely in the wiki source**.
- **Therefore:** A (kept the space) is the **faithful** wiki transcription; **B silently deleted a real character**; and the MERGED chose B's (less-faithful) rendering as "canonical" while framing the space as if A had added it. The provenance characterization is backwards.
- **Impact:** whitespace only, zero letters, not cryptographically meaningful — but for an artifact whose "entire point is verbatim fidelity," the delivered canonical wiki line drops a source character and the adjudication is wrong. The correct canonical VERSION-2 line is `THOSE ARE EVIL ` (with trailing space). (The VERSION-1/blog spine line `THOSE ARE EVIL` with no trailing space is independently correct — the two channels legitimately differ by this one space.)

### D2 — `(element symbols)` inserted into a "verbatim" worked-example block (fidelity)
- **What the MERGED says (§4, presented inside a fenced block as source text):** `Translation first calculation (element symbols): TC IR MN EU …`
- **What the wiki source actually says (independent capture):** `Translation first calculation: TC IR MN EU …` — **no** `(element symbols)`.
- The 2-word parenthetical originates in Reconstruction B (`--B.md` line 40) and was propagated into the MERGED's verbatim block. It is an editorial label, not source text. (Also: the wiki wraps the symbol string across four physical lines; both A and B reflow it — cosmetic.)
- **Impact:** contextual only (this is the decoding-method illustration, not the six-page plaintext), but it is a non-source string sitting inside a block labelled verbatim.

### D3 — VERSION 1 "leading single space on every line of pages 17/18/19 + p21 L1" is unverified this session (cosmetic)
- The MERGED [F1] states the blog prints a one-space block-quote indent on those lines. The live blog fetch (markdown-converted) rendered the plaintext with **no** leading spaces / no blockquote, and markdown conversion can strip leading whitespace, so I could **neither confirm nor refute** this from the live page. Immaterial (it is declared cosmetic and stripped from the spine), but the specific formatting claim is not independently corroborated this pass.

### D4 — Tier label `PRIMARY` is slightly generous (caveat, disclosed)
- The overall artifact is graded **tier PRIMARY**. The decoded English is fundamentally a **reader's reconstruction** (Klaus Tappeiner's decode, one line by Norbert Biermann) — a contemporary *secondary decipherment* of the PRIMARY page images/calculations. The blog reporting it is a genuine 2017 contemporary artifact and the decode is reproducible, so "PRIMARY" is defensible, but the plaintext is one interpretive layer removed from ground truth. The reconstruction **does disclose** this ("not a puzzle-maker ground-truth file"), so this is a caveat, not a hidden error.

---

## CLAIMS I COULD NOT INDEPENDENTLY RE-VERIFY THIS SESSION
- `web.archive.org` — blocked from this environment ("unable to fetch"). Could not pull the cited `…id_/…` raw captures directly. (Worked around via live blog + the GitHub-hosted wiki save.)
- Fandom (`tengri137.fandom.com`) — HTTP **402 Payment Required** (matches prior session).
- Reddit (`www.reddit.com`, `libredd.it`) — blocked / 404. Could not re-pull thread `5xsno3`.
- The wiki's independent origin: as noted, the GitHub HTML save traces to the **same** `20171001170618` Wayback snapshot as the cited VERSION 2, so it corroborates the *bytes of that snapshot* but is not a second independent origin for the wiki text. VERSION 1 (blog) is independently corroborated (live blog + a web-search hit that reproduced page 18 raw verbatim: `THE REARE … KEEP DISTANCE SUCH … BASED ON AMATHEMA … THOSE ARE EVIL`).

---

## BOTTOM LINE
No fabrication, no memory-reconstruction, no smoothed-over gap in the six-page plaintext — the letters are real and reproduce. But the MERGED (a) delivers a "canonical" wiki line that drops a genuine source trailing space and inverts that variant's provenance (D1), and (b) embeds a non-source `(element symbols)` label inside a verbatim block (D2). Both are small; both are real. **agrees = false.**