troupe-standup/recoveries/evidence-tier-audit.md

Are the evidence tiers earned? — Q12 audit of the 121 published badges

Run by Merlin, 2026-10-10, ~09:01–09:35 UTC. Independent of the build: every page read over HTTPS from this host with curl, nothing inferred from the repository's own claims about itself. Working files /root/crucible/q12/ (index.html, entry-*.html, src-*.html, skin-*.html, EVIDENCE.txt, parse_index.py, sample.py, fetch_entries.py, fetch_sources.py, check.py, excerpts3.py, dates.py, deep.py, skinner2.py, unlinked.py, evidence.py).

Verdict in one line: no badge in the sample overstates its evidence — 18/18 justified, 0 too strong, 0 too weak, 0 indeterminate — but three entries cite their evidence at the wrong URL or not at all, and the legend's fourth tier is defined and never used.

---

1. Census reproduced before any grading

https://alexanderwhitestone.com/stackchain/blocks/ → HTTP 200 / 34,488 B / sha256 4f764ef4dfb9c08b2654ff0783550d20ad1bb7a31cb0c32070974a5fe451557e.

$ grep -o 'badge documented'    index.html | wc -l   ->  115
$ grep -o 'badge testimony'     index.html | wc -l   ->    5
$ grep -o 'badge corroborated'  index.html | wc -l   ->    1
$ grep -o 'badge unrecovered'   index.html | wc -l   ->    0
$ grep -o '<li id='             index.html | wc -l   ->  121

115 + 5 + 1 + 0 = 121, and 121 <li id= elements. The 2026-09-26 census in the queue is reproduced to the entry. The page's own header reads "121 entries · 8 ledger blocks and receipts · 22 dated events · 91 magazine pieces".

The legend, as published on the same page (verbatim, detagged):

Documented — the archive holds the post, capture or receipt. Corroborated — established by a published or community account, with the source quoted on the page. An original screenshot is not required. Testimony only — a participant's recollection, recorded as recollection; where the practice it describes is corroborated, that is shown too. Unrecovered — known to exist, no copy and no independent account.

2. Sample rule, stated before any fetch

1. Every entry whose published badge is not documented — all 6 (corroborated 1, testimony_only 5). There is no unrecovered entry to include.
2. Every 9th documented entry in published index order, starting at the 9th. The 115 documented entries numbered 9, 18, 27, 36, 45, 54, 63, 72, 81, 90, 99, 108 → 12 entries.

Total sample 18 of 121. The rule is mechanical and re-runnable from index.html alone (parse_index.py → sample.py); it is not a judgement about which entries looked interesting. The six mandatory non-documented entries are event-2022-07-18-anthony-enters-at-16-…, block-25, block-6, block-88, block-89, block-90.

3. Method

Per entry: fetch the entry page, read the badge as published, take the strongest source the entry itself cites, fetch that, and grade the badge against the legend's own words.

Detag / normalisation rule, printed so it can be re-run. Strip <script> <style> <svg> <noscript> blocks, replace <br> and block-closing tags with a space, strip remaining tags, HTML-unescape, then NFKC-normalise and collapse runs of whitespace. Quotes are normalised (' ' " " → ' "), em/en dashes → -, … → ..., NBSP → space. A quoted fragment counts as verbatim when it is a substring of the detagged source under exactly that rule. Nothing is graded on a paraphrase.

Grading vocabulary, as the item defines it: justified / too strong / too weak / indeterminate, applied per entry (the badge is published per entry, not per clause); per-clause defects are recorded separately and named.

A dated reading, not an identifier: every byte count and hash below is what this host read at ~09:0x UTC on 2026-10-10. Third-party pages drift; the hashes are for re-checking this run, not for claiming a page is permanently that size.

4. Per-row table

Fixed column order: entry | badge as published | strongest cited source | verdict | settling evidence

entry	badge as published	strongest cited source	verdict	settling evidence
event-2022-07-18-anthony-enters-at-16-names-the-hill-at-24	Testimony only	https://www.stackchainmagazine.net/waht-stackchain-done-dided-the-king-of-stackchain/ — 200 / 153,199 B / sha 7ff55e52eafa…	justified	Entry's own words: "Treat the claim to first use of the name as participant testimony, not an independently exhaustive Twitter search." Cited article carries Block 16 ×2 and Block 24 ×1; its opening excerpt is verbatim on the page. The badge understates nothing and claims nothing held.
block-25	Testimony only	https://www.stackchainmagazine.net — 200 / 215,836 B / sha e7df6364180a…	justified	Entry states "The specific block numbers are the contributor's recollection and are not corroborated" and "This is testimony, not a held artifact." No post or capture is claimed. Citation defect, see §6.1.
block-6	Corroborated	https://bitcoinmagazine.com/culture/how-5-bitcoin-purchase-cascaded-into-1-5-million-viral-movement — 200 / 341,802 B / sha e1826c17863d…	justified	The legend's own test is met: a published account, quoted on the page. The quoted sentence "…followed it up with a $6 buy and Satskeeper followed Bob's with a $7 buy" is verbatim ×1 in the fetched article, as are Happyclowntime ×1 and Satskeeper ×1. Receipt panel says plainly "No receipt is held for this entry" — the legend's Corroborated tier explicitly waives the screenshot.
block-88	Testimony only	https://www.stackchainmagazine.net — 200 / 215,836 B / sha e7df6364180a…	justified	Same shape as block-25: "First-person oral history, 2026-09-21. The posts themselves are not held." Citation defect, see §6.1.
block-89	Testimony only	https://www.stackchainmagazine.net — 200 / 215,836 B / sha e7df6364180a…	justified	Same. "No post, capture or timestamp located." Citation defect, see §6.1.
block-90	Testimony only	https://www.stackchainmagazine.net — 200 / 215,836 B / sha e7df6364180a…	justified	Same, and the disputed material is quarantined in the Uncertainty panel: "Skinner attributes the naming to Anthony, 'most likely' - tentative and uncorroborated." Citation defect, see §6.1.
event-2022-08-18-the-bitcoin-beach-campaign-closes-and-hodlonaut	Documented	https://x.com/AnthonyDessauer/status/1560325688481026048 — 200 / 111,659 B / sha b03ba32bdbd4…; blockstream.info/api/tx/ae63c499155ce14064686fc7218af6630cf64fd58a601bba48368b13cf955aa1	justified	The held artifact is real: the quoted post text "The last of sats donated to #Stacks4BitcoinBeach are on their way" appears ×3 in the fetched page; the tx the entry prints is confirmed, block height 749999, block time 2022-08-18T17:45:12Z, one output of 851,979 sats. The entry's own date, 2022-08-18, matches the receipt to the day. One clause has no source at all, see §6.2.
event-2023-05-00-siggy-reaches-nostr	Documented	https://github.com/plebiolira/stackchainsiggy_nostr — 200 / 279,677 B / sha 020fc5625905…; commits API	justified	README line verbatim: "This is the version made for Nostr for the Stackchain mascot that monitors for #stackjoin buys." Commits API (301 → repositories/640699187): 48 commits, earliest 2023-05-14T23:37:38Z, latest 2023-08-23T12:12:05Z, months 2023-05 … 2023-08 — exactly the entry's "visible commit history from May to August 2023" and its stated basis label "visible repository commit dates".
magazine-012	Documented	https://www.stackchainmagazine.net/inflation-you/ — 200 / 141,125 B / sha 8f6efedbd052…	justified	3/3 quoted excerpts verbatim; entry date 2023-08-28 == the article's own article:published_time 2023-08-28T06:08:50+00:00.
magazine-017	Documented	https://www.stackchainmagazine.net/where-ideas-end-up-and-why-that-matters/ — 200 / 136,829 B / sha 1a98badabb88…	justified	3/3 verbatim, including "We currently have 841 active projects on Geyser, 40% of which are from parts of the world that were previously underserved"; 2023-10-26 == 2023-10-26T17:34:30+00:00.
magazine-026	Documented	https://www.stackchainmagazine.net/stackchain-for-newcomers-how-i-infiltrated-the-stackchain/ — 200 / 137,771 B / sha b1fba3b77840…	justified	3/3 verbatim; 2024-04-10 == 2024-04-10T00:17:53+00:00. Minor: the summary names @a_koby, which is not an anchor on the page (§6.3).
event-2024-05-20-one-community-multiple-living-chains	Documented	https://www.stackchainmagazine.net/two-years-of-stackchain/ — 200 / 136,607 B / sha bbb528f37241…	justified	Excerpt verbatim; the entry's basis label "explicit internal composition date; web upload" is honest — the article's own line is "As of the date of my writing this May 20th 2024…" against article:published_time 2025-07-22T14:33:48+00:00. Both halves verified (this row was Q11's row 8).
magazine-042	Documented	https://www.stackchainmagazine.net/meet-the-orange-pill-app/ — 200 / 143,271 B / sha 50ab3c791664…	justified	3/3 verbatim; 2025-01-25 == 2025-01-25T18:20:18+00:00.
magazine-050	Documented	https://www.stackchainmagazine.net/a-pleb-travels-to-el-salvador/ — 200 / 184,537 B / sha f9a4e4071123…	justified	3/3 verbatim; 2025-07-15 == 2025-07-15T00:19:14+00:00.
magazine-059	Documented	https://www.stackchainmagazine.net/plebs-priced-out-of-conference-tickets-after-blackrock-tokenizes-them-as-a-brc-20-token/ — 200 / 139,413 B / sha c778d10edd86…	justified	3/3 verbatim, including "A general admission ticket shot up from $399 to almost $5000, or close to the price of a whale pass overnight."; 2025-08-01 == 2025-08-01T14:15:10+00:00.
magazine-067	Documented	https://www.stackchainmagazine.net/what-is-bitcoin-art/ — 200 / 142,126 B / sha 3b31cc06403a…	justified	3/3 verbatim; 2026-01-12 == 2026-01-12T13:15:58+00:00.
magazine-076	Documented	https://www.stackchainmagazine.net/dsb-and-me-by-clumsy-dork/ — 200 / 173,819 B / sha 865855cc14d8…	justified	3/3 verbatim; 2026-01-20 == 2026-01-20T13:09:30+00:00.
magazine-085	Documented	https://www.stackchainmagazine.net/the-diy-signing-device-revolution-is-just-beginning/ — 200 / 162,809 B / sha 5d6b1a9801ca…	justified	3/3 verbatim; 2026-08-13 == 2026-08-13T14:00:25+00:00.

All 18 entry pages returned HTTP 200 on the first attempt (3,181–5,146 B).

5. Verdict counts

verdict	count
justified	18
too strong	0
too weak	0
indeterminate	0

By tier: Testimony only 5/5 justified, Corroborated 1/1 justified, Documented 12/12 justified. No badge in the sample misstates its evidence. The three findings below are citation and legend-policy defects, not badge overstatements — which is the distinction the item's own framing asked for.

6. Named findings

6.1 block-25, block-88, block-89, block-90 — the corroborating quote is cited at a URL that does not contain it

All four Provenance panels read: "The named practice is independently corroborated: a published Stackchain Magazine account records the term 'Skinner Stacking' and describes it, quoted below." The link attached to that quote is the bare site root, https://www.stackchainmagazine.net (200 / 215,836 B / sha e7df6364180a…), which detags to 5,036 characters and contains Skinner 0 times, Skinner Stacking 0 times. The quoted passage is therefore not verifiable from the URL the four pages print.

It is verifiable — elsewhere. The passage is verbatim in https://www.stackchainmagazine.net/a-conversation-with-burnchainers/ (200 / 155,217 B / sha 50a2e57e3ba6…), which is the only result the publication's own search returns for ?s=Skinner (200 / 107,021 B / sha 95f57a614ebc…; ?s=skinner+stacking returns the same single article). The passage, verbatim from that page:

"I've been watching Stackchain and they adopt new ideas over time. Like some used to call it Skinner Stacking when you did two blocks in one post. But who wants to have their stack named after someone else? So as the idea was separated from the person it got more palatable."

Effect: the claim holds, the citation does not point at it. Four pages should cite the a-conversation-with-burnchainers permalink instead of the site root. Badge unchanged (Testimony only is correct and appropriately modest); this is a link fix on four live pages, so it belongs to Alexander via Q13's channel, not to this run. Independently, the term also appears in x.com/rockachopa/status/1550583498397782017 (200 / 107,399 B / sha ff71ecb4ac768b…, article:published_time 2022-07-22T20:48:09.000Z, Skinner Stacking ×6) — a second corroborating source neither page cites.

6.2 event-2022-08-18… — the summary names an account it does not link, and that account 404s

The "What happened" panel reads: "…StackchainQuant's October account also embeds Yellow's August 18 donation appeal and reports several thousand dollars raised under #Stacks4Hodlonaut." StackchainQuant appears twice in the page and is an anchor nowhere — the page's only external anchors are the AnthonyDessauer post and the Bitcoin Magazine article. And the handle itself does not resolve: https://x.com/StackchainQuant → 404, 37,702 B, sha 2dfa62a242e6…, <title>User Profile Not Found - X | 404 Error</title>.

The Provenance panel says the summary is "Summarised by the archive from the sources linked below", which is not true of that clause. Effect: the entry's badge is earned by the parts that are linked and checked (post text verbatim ×3, tx confirmed at 2022-08-18T17:45:12Z, 851,979 sats); this one clause rests on an account that is unlinked and currently unreachable. Fix is either a link to a surviving capture or withdrawal of the clause. Alexander's call.

6.3 magazine-026 — a named handle with no anchor (minor)

The summary names "Koby (@a_koby)"; the page's only external anchor is the article permalink. No badge effect. Noted for completeness because it is the same defect class as 6.2 at lower stakes.

6.4 The legend's fourth tier is defined and never applied

Unrecovered — "known to exist, no copy and no independent account" — appears 0 times across 121 entries, while material that fits its definition is sitting inside lower-tier entries' Uncertainty panels: block-6 "The $6 post itself has not been produced"; block-25 "No post or capture has been located for either attempt"; block-88 "No post, capture or timestamp located". The archive is disclosing this honestly, but it is doing it in prose rather than in the tier vocabulary the legend publishes. This is a legend-design question for Alexander, not a per-entry defect.

6.5 Documented is a liveness claim, not a held-artifact claim, for the 91 magazine entries

The legend's first tier says "the archive holds the post, capture or receipt". All 12 sampled magazine entries answer that with a Receipt panel reading "No receipt is held for this entry." and a Provenance panel naming a live third-party URL. So the badge rests on a URL resolving today, which I measured: 12/12 at HTTP 200, 8/8 entry dates equal to the cited article's own article:published_time to the day. That is a strong result, but it is not the same claim the legend makes, and it is fragile in a way I hit myself: the Bitcoin Magazine URL returned 403 (4,545 B) to a Chrome user-agent and 200 / 341,802 B to a bare curl agent, in the same minute. One publisher, one UA difference, one wall.

Recommendation for Alexander, not for this run: either narrow the Documented legend's wording to "a published source the archive cites, resolving at read time" for the magazine class, or start holding captures. Nothing here is a badge defect today; it is a claim the badge makes that nobody has tested until now.

7. Measurement: are the excerpts the badge rests on actually in the source?

Across the 18 sampled entries, the "From the sources" panels carry 37 quoted fragments of ≥25 characters (split on each re-cited permalink so distinct excerpts stay distinct, then detagged and normalised per §3).

33 of 37 (89%) are verbatim in a source the entry itself cites. The 4 that are not are the same Skinner passage repeated on block-25, block-88, block-89 and block-90 — finding 6.1. No sampled entry prints an excerpt that its own cited source does not contain, other than that one passage, and no excerpt was found in the one off-citation control I ran (x.com/rockachopa/ status/1550583498397782017, which carries the term but not the passage). I did not sweep the whole publication for each excerpt, so "verbatim in a cited source" is a per-row result and "found only in an uncited source" is not a claim this run makes.

8. Evidence

Every URL read this run, with its byte count and SHA-256, in /root/crucible/q12/EVIDENCE.txt (18 entry pages, the index, 17 cited sources, 2 off-citation controls, 3 search pages). The full list is printed by python3 /root/crucible/q12/evidence.py.

Reads that mattered and their outputs:

index               200 / 34,488  sha 4f764ef4dfb9…   /stackchain/blocks/
block-6 source      200 / 341,802 sha e1826c17863d…   bitcoinmagazine.com (403 under Chrome UA, 200 under bare curl)
block-25 source     200 / 215,836 sha e7df6364180a…   stackchainmagazine.net  ("Skinner" 0 occurrences)
skinner permalink   200 / 155,217 sha 50a2e57e3ba6…   /a-conversation-with-burnchainers/  (passage verbatim ×1)
ev-08-18 post       200 / 111,659 sha b03ba32bdbd4…   x.com/AnthonyDessauer/status/1560325688481026048 (text ×3)
tx ae63c499…        confirmed, height 749999, 2022-08-18T17:45:12Z, 1 output, 851,979 sats
siggy repo          200 / 279,677 sha 020fc5625905…   48 commits, 2023-05-14 … 2023-08-23
StackchainQuant     404 / 37,702  sha 2dfa62a242e6…   "User Profile Not Found"

9. Limits

• Testimony only: every byte count and hash is a dated reading from this host at ~09:0x UTC 2026-10-10. Third-party pages drift; a later reader will get different sizes. The hashes let this run be re-checked, not the pages pinned.
• Sample, not census: 18 of 121. The 6 non-documented entries are a complete census of their tiers; the 12 documented entries are an every-9th sample and license no rate estimate for the other 103. A clean 18/18 is evidence that the tiering discipline holds where it is hardest (the entries with the least evidence), not that all 121 are clean.
• Not tested: the accuracy of the summaries' content (that is Q8/Q10's lane), the 103 unsampled documented entries, and whether any documented magazine entry cites an article that has since been taken down outside the sample.
• Not touched: no live page, no blocks.json, no other party's file. Nothing in this run changed the archive; the three link fixes in §6 wait on Alexander through Q13's channel.