# Later-artifact status audit

## Baatar Kerei pointer (2017)

- The PGP-signed Tengri message points through `bit.ly/baatarkerei` to YouTube ID `XV2AlgPBZPE`.
- A preserved 2022 YouTube payload identifies it as *The secret history of Baatar Kerei and future 3* by `astromankorea`, uploaded `2013-01-13`.
- The preserved payload reports `377` seconds and no caption track.
- The research record's 2026 live probe observed the target as **private**, not deleted. It also found 15 of 18 videos from the same channel private, so this is consistent with channel-wide privating rather than Tengri-specific removal.
- **Boundary:** the video file, audio, 24 historical YouTube comments, and a transcript have not been recovered. This is an external ideological reference identified by its metadata, not a new Tengri cipher solve.

## Vumarii image (2019)

- The PGP-signed pointer names the Imgur Vumarii image; the preserved PNG SHA-256 is `bfbdd8b5847e8ae87cbfbdb4cb1894e69fd2a40a90894ab4db875d1c410160a4`.
- Its seven-line body contains `154` non-separator glyph slots and `25` explicit word separators. The two-line signature contains `17` slots.
- The complete transcription uses `28` classes: `23` letters, `4` numerals, and one word separator. `15` values are fixed by the image's own red labels; `7` more are contemporary secondary readings; `5` classes remain unassigned.
- Even after accepting the secondary readings, only `134` of `154` body positions have a Latin transliteration. The remaining `20` positions use `C16, C17, C18, C20, C21`.
- A 5,000-control, eight-orientation positional test against Page 23 finds no unusual direct indexed correspondence: the best glyph-class mapping has empirical `p >= 0.7313` and the best direct transliteration equality has `p >= 0.5855`. This rejects that small model family only.
- The upper diagram has a stable raster-geometry record (`29041` ink pixels, `36` compact enclosed-circle candidates) and is now strongly attributable to a cleaned, re-inked north-polar detail of the **Dunhuang Star Map** (British Library Or.8210/S.3326). A labelled side-by-side comparison preserves nine distinctive asterism/wall/chain regions; in the same 500-crop bounded source comparison, the ordinary orientation has local-darkness contrast `3.1935`, above horizontal mirror `2.5729`, vertical mirror `2.3113`, and 180-degree rotation `2.0162`.
- The attribution identifies the visual reference, not a payload: it does not yet name every individual asterism, select a traversal, translate the glyph body, or connect the chart to Page 23.
- Map 13's scholarly Table 5 does list `34` asterisms, while the complete image has `34` whitespace-separated groups only when the two visible signature groups are counted with the seven-line `32`-group body. That is a source-led comparison worth recording, but not a key.
- A fixed, generous count-control tests the natural full text order against Table 5's north-to-south/east-to-west order and its reverse. The best route has only `5` exact token-length/star-count matches; `100,000` equally allowed permutations match or exceed it with `p = 0.5315`. Absolute correlation (`p = 0.4615`) and total-difference (`p = 0.3857`) controls agree: the count does not select a traversal, mapping, or plaintext.
- A separate conditional Turkish-completion test exhausts all `6,720` assignments of the five unassigned classes after accepting the seven contemporary secondary readings only as a test condition. Its best score beats `0/512` glyph-position controls (empirical `p = 0.0019`) and all but `3/512` secondary-key controls (`p = 0.0078`).
- That result is still `8.70` standard deviations below held-out Turkish and has no full word segmentation. It is a conditional phonotactic lead, not a source-selected alphabet or plaintext.
- A follow-up exhausts all `312` invertible A–Z affine mappings (including Caesar shifts and Atbash-like maps) across those same `6,720` completions: `2,096,640` candidates. The identity map is the global Turkish optimum; the best non-identity map scores worse (`-6.6174` versus `-5.4116` bits/character). This closes that simple classical-cipher family for the conditional Turkish lead, not other languages or cipher types.
- The preserved community `Solver key 1.0` visibly names `Esperanto` and `swahili`. Each was tested, not assumed: under the same seven conditional community readings, every one of the `6,720` residual five-glyph completions was scored against a separate held-out, language-tagged public corpus. The best Esperanto surface is `18.05` standard deviations below held-out Esperanto; the best Swahili surface is `13.91` standard deviations below held-out Swahili. Both rank at the `0.00` held-out percentile and yield no readable translation.
- This does not turn the surprising conditional phonotactic scores into a solve. Esperanto's best completion beats all `256` glyph-position and secondary-reading controls, while Swahili's score is matched or exceeded by `13/256` secondary-key controls; in either case the entire candidate remains far below natural prose. The labels/readings are community evidence only, and the test does not exclude a different language, transliteration, keyed cipher, or diagrammatic layer.
- A 5 September 2018 Reddit thread reports a Patreon URL as an official Tengri 137 account and says a puzzle might arrive the following year. Its author later characterises the Patreon description as a charitable-purpose statement. Those are contemporary community assertions, not recovered Patreon text or author-signed instructions; they provide chronology and claimed context, **not a Vumarii decryption key**.
- A bounded public-recovery audit records the source URL, a current Cloudflare block, and the exact archive-query outcomes. It deliberately treats timeouts and gateway failures as inconclusive rather than evidence that no historic capture exists.
- **Boundary:** the 2019 body has a reproducible glyph transcription and its upper panel now has a strong historical-chart source attribution, but no verified language translation, chart traversal, or demonstrated relationship to page 23. Its shared 154-letter count is not a decode.

## Why this distinction matters

A source can be recovered without its media bytes, and a text can be transcribed without being translated. Treating either intermediate result as a solved hidden message would make the public chronology less reliable.

## Local evidence

- Baatar metadata: `archive/gaps/tr1-07-files/yt-XV2AlgPBZPE-METADATA-VERBATIM.txt` (SHA-256 `016cdaf78a7f3c1f4bca620f461f747404366e3a5d3aa14aa5eb11313de11b3c`)
- Baatar research record: `archive/gaps/tr1-07-youtube-XV2AlgPBZPE-recovered-and-wiki-comment-loss-bounded.md`
- Vumarii transcription: `archive/gaps/tr1-00-files/vumarii-2019-TRANSCRIPTION.txt` (SHA-256 `2c6eb4492a3975c034e9fb4ba75ae676acad0fc979b76e7a0463bd77170ad6b3`)
- Vumarii research record: `archive/gaps/tr1-00-vumarii-script-corpus-transcribed.md`
- Vumarii/Page-23 positional audit: `solution/page23-vumarii-alignment-audit.json` (SHA-256 `08c7c0d07eebc744bef2ec8399df59264678ecc0338154e9005a43933efcbcf4`)
- Vumarii upper-diagram geometry audit: `solution/vumarii-2019-upper-diagram-audit.json` (SHA-256 `66d306e01f37a9a9723cb5df8a58d8889c5a1ddfe8881e52bf5d9ce4a03b0427`)
- Vumarii/Dunhuang visual-attribution audit: `solution/vumarii-2019-dunhuang-attribution-audit.json` (SHA-256 `109dde413175bc1a0b753ad32c47379c8a53683d4eaba153bbc965107e73696f`); labelled comparison: `_evidence/vumarii-2019-dunhuang-attribution.png`
- Vumarii/Dunhuang 34-unit count-control: `solution/vumarii-2019-dunhuang-traversal-audit.json` (SHA-256 `38f6e66a0a2fe1bf86e758796ffb9bdb6588a296d9518f0cc211ef127e77698f`); chart: `_evidence/vumarii-2019-dunhuang-traversal-audit.png`
- Vumarii conditional Turkish-completion audit: `solution/vumarii-2019-turkish-completion-audit.json` (SHA-256 `926b32fbda8bb7377300e9db10e2473095fc9b254e59c4833e072404dfa17ef0`)
- Vumarii affine-Turkish control: `solution/vumarii-2019-turkish-affine-audit.json` (SHA-256 `2d8a306a94c0a25bfe643a1b774a9b15e367056d9f4e4de106bf06dede2a77d3`)
- Vumarii Defango-listed Esperanto/Swahili control: `solution/vumarii-2019-defango-language-hypotheses-audit.json` (SHA-256 `c55c9cf1a0bbddf0fbcd3e84a0bb3c6f7303644a2df49c865dff0dda57bae331`); chart: `_evidence/vumarii-2019-defango-language-hypotheses.png`
- Vumarii 2018 Patreon context/recovery record: `solution/vumarii-patreon-recovery-audit.md` / `.json` (time-bound network observations; the contemporary Reddit source hash is recorded inside)