stackchain/troupe-standup
2
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
67b892d7bb35c46ee1c85ae7d969833c12602741
troupe-standup/checkins/2026-10-10-merlin.md
T
Merlin 67b892d7bb Q12 DONE: are the evidence tiers earned? (recoveries/evidence-tier-audit.md) - 18/18 justified, 0 too strong / 0 too weak / 0 indeterminate, no badge in the sample misstates its evidence; sample rule stated before any fetch (every non-documented entry = all 6: 5 testimony_only + 1 corroborated + 0 unrecovered, plus every 9th documented entry in index order = 12, total 18 of 121); census reproduced from the live index (200 / 34,488 B / sha256 4f764ef4dfb9, grep 'badge documented' 115 / testimony 5 / corroborated 1 / unrecovered 0 = 121 <li id=); all 18 entry pages and 17 cited sources fetched live this run, 33 of 37 quoted excerpts >=25 chars verbatim in a source the entry cites, and 8 of 12 sampled magazine entries carry a displayed date equal to the cited article's own article:published_time to the day; five named findings, none a badge overstatement - (1) block-25/88/89/90 quote the 'Skinner Stacking' passage and link the bare site root which contains 'Skinner' 0 times while the passage is verbatim at /a-conversation-with-burnchainers/ (200 / 155,217 B / sha 50a2e57e3ba6), the publication's only ?s=Skinner result; (2) event-2022-08-18 names StackchainQuant's October account twice, links it nowhere, and x.com/StackchainQuant returns 404 'User Profile Not Found', its badge still earned by the post text verbatim x3 and tx ae63c499 confirmed at height 749999 / 2022-08-18T17:45:12Z / 851,979 sats; (3) magazine-026 names @a_koby with no anchor; (4) the legend's Unrecovered tier is defined and applied 0/121 times while material fitting it sits in lower-tier Uncertainty panels; (5) Documented on the 91 magazine entries is a liveness claim not a held-artifact claim (Receipt panel says 'No receipt is held', and bitcoinmagazine.com answered 403 to a Chrome UA and 200 / 341,802 B to a bare curl UA in the same minute); limits: 18 of 121 is a sample, the 6 non-documented are a complete census of their tiers and the 12 documented license no rate for the other 103, all byte counts and hashes are dated readings from this host ~09:0x UTC, nothing live edited; check-in 2026-10-10-merlin.md; Q12 status block appended; Q20 next, Q9 blocked on the audio handover, Q13/Q19 gated on Alexander, depth not thin so no new item opened
2026-10-10 09:08:47 +00:00

11 KiB
Raw Blame History

2026-10-10 — merlin

Screen. Every claim below carries a path, a hash, a URL, or a command with its output, or the word unverified. Nothing secret is printed anywhere in this check-in or in the files it lands.

Did

Claimed and finished Q12 — are the evidence tiers earned?, the top unclaimed executable item (WORK-QUEUE.md: Q1–Q11 closed or claimed, Q13/Q19 gated on Alexander, Q9 blocked on the audio handover, Q20 open behind it, and my own 2026-10-09 check-in named Q12 as my next pick).

Deliverable: recoveries/evidence-tier-audit.md (19,516 B, sha256 625bfb1e9afd9ba3541dc83c09fab39032c7e77095dea616dea60fa70c4ca1dd) — sample rule, legend, method, 18-row table in the fixed column order entry | badge as published | strongest cited source | verdict | settling evidence, verdict counts, five named findings, and the full read list.

Result: 18/18 justified, 0 too strong, 0 too weak, 0 indeterminate. No badge in the sample misstates its evidence. What the audit did find is three citation defects and two legend-policy problems — which is the distinction the item's own framing asked for, and the reason a clean badge result is not a clean report.

Sample rule, stated before any fetch: every non-documented entry (all 6 — 5 testimony_only, 1 corroborated, 0 unrecovered) plus every 9th documented entry in published index order (12) = 18 of 121. Census reproduced from the live index (https://alexanderwhitestone.com/stackchain/blocks/, HTTP 200 / 34,488 B / sha256 4f764ef4dfb9c08b2654ff0783550d20ad1bb7a31cb0c32070974a5fe451557e): grep -o 'badge documented' → 115, testimony → 5, corroborated → 1, unrecovered → 0, <li id= → 121. The 2026-09-26 census in the queue reproduces to the entry.

All 18 entry pages and 17 cited sources fetched live from this host, ~09:01–09:35 UTC; 18/18 entry pages HTTP 200 on the first attempt (3,181–5,146 B). 33 of 37 quoted excerpts ≥25 characters are verbatim in a source the entry cites; the 4 that are not are one passage repeated on four pages (finding 1). For 8 of the 12 sampled magazine entries the entry's displayed date equals the cited article's own article:published_time to the day — e.g. magazine-085 shows 2026-08-13 and the-diy-signing-device-revolution-is-just-beginning/ (200 / 162,809 B / sha 5d6b1a9801ca1792…) carries article:published_time 2026-08-13T14:00:25+00:00.

Five findings, none of them a badge overstatement:

  1. block-25, block-88, block-89, block-90 cite their corroborating quote at a URL that does not contain it. All four Provenance panels say the practice "is independently corroborated: a published Stackchain Magazine account records the term 'Skinner Stacking' and describes it, quoted below" — and link the bare site root, https://www.stackchainmagazine.net (200 / 215,836 B / sha e7df6364180a…), which detags to 5,036 characters containing Skinner 0 times. The passage is verbatim at https://www.stackchainmagazine.net/a-conversation-with-burnchainers/ (200 / 155,217 B / sha 50a2e57e3ba661a6…), the publication's only result for ?s=Skinner (200 / 107,021 B / sha 95f57a614ebc…). The claim holds; four pages need the permalink.
  2. event-2022-08-18… names a source it does not link, and that source 404s. The summary says "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, and https://x.com/StackchainQuant returns 404 / 37,702 B / sha 2dfa62a242e6…, <title>User Profile Not Found - X | 404 Error</title>. The badge is still earned by what is linked and checked: the quoted post text is verbatim ×3 in x.com/AnthonyDessauer/status/1560325688481026048 (200 / 111,659 B / sha b03ba32bdbd4…), and the printed tx ae63c499155ce14064686fc7218af6630cf64fd58a601bba48368b13cf955aa1 is confirmed, height 749999, block time 2022-08-18T17:45:12Z, one output of 851,979 sats (https://blockstream.info/api/tx/ae63c499…).
  3. magazine-026 names @a_koby with no anchor (minor, same defect class, no badge effect).
  4. The legend's Unrecovered tier is defined and applied 0/121 times, while material fitting its definition ("known to exist, no copy and no independent account") sits in lower-tier Uncertainty panels — block-6: "The $6 post itself has not been produced"; block-88: "No post, capture or timestamp located."
  5. Documented on the 91 magazine entries is a liveness claim, not a held-artifact claim. The legend's first tier says "the archive holds the post, capture or receipt"; all 12 sampled magazine entries answer with a Receipt panel reading "No receipt is held for this entry." and a live third-party URL. 12/12 resolved 200 today, but the fragility is measurable: the Bitcoin Magazine URL returned 403 / 4,545 B to a Chrome user-agent and 200 / 341,802 B / sha e1826c17863d… to a bare curl agent, in the same minute.

Method note worth keeping: bitcoinmagazine.com 403s a browser UA and answers curl, so a 403 there is a wall, not a dead link — the same class of false negative Q2's record already warns about for the Wayback availability API.

Blocked

Nothing blocked the run. Two things deliberately unverified, both recorded as such in the deliverable:

  • Whether any of the other 103 unsampled documented entries has the same defect class. The 6 non-documented entries are a complete census of their tiers; the 12 documented are an every-9th sample and license no rate for the rest. A clean 18/18 is evidence that the discipline holds where the evidence is thinnest, not that all 121 are clean.
  • Whether magazine-012's "a year where official CPI was 6% while food, rent, electricity and gas ran higher" is true of the article's figures. The article is live (200 / 141,125 B / sha 8f6efedbd052…) and carries 6% once; whether the surrounding sentence says what the summary says is Q8/Q10's lane, not a badge question, so I left it alone rather than half-grade it.

Next

Q20 — excerpt fidelity and placement for the 91 magazine entries (open and unclaimed, opened by the pipe 2026-10-08). It is the natural successor: Q12 just measured that 33/37 sampled excerpts are verbatim in their cited source, and Q20 asks the question this audit did not — whether an excerpt is a literal quotation or a paraphrase, and where inside its source it sits, with offsets and fractions. Its finish line is recoveries/excerpt-placement-audit.md under a stated sample rule.

Q13 stays gated on Alexander: findings 1 and 2 above are link/provenance fixes to live pages and belong in that channel, not in this run. Nothing in this tick executed on decisions/.

Ask

Three decisions that are yours, none of them blockers:

  1. The four Skinner links — repoint block-25/88/89/90 from stackchainmagazine.net to /a-conversation-with-burnchainers/? It is a link change on four live pages, so it waits for you through Q13.
  2. event-2022-08-18's StackchainQuant clause — link a surviving capture of the October account, or withdraw the clause? The account's handle 404s today, so the clause currently rests on an unreachable source.
  3. The legend's Documented wording — either narrow it for the magazine class to "a published source the archive cites, resolving at read time", or start holding captures. As written the badge claims an artifact the archive says it does not hold.

No new queue item opened — depth is not thin (Q9 blocked on the audio handover, Q13/Q19 gated on you, Q20 open).

Run detail — Q12, claimed and finished

Clock and host. All fetches from this host, 2026-10-10 ~09:01–09:35 UTC. Working tree at /root/crucible/q12/; repo at /root/troupe-standup, branch main, HEAD before this run aab23e5, working tree clean.

Scripts, all re-runnable. parse_index.py (index → 121 entries with badge + external hrefs), sample.py (the sample rule), fetch_entries.py (18 entry pages), fetch_sources.py (17 cited sources), check.py, deep.py (bitcoinmagazine retry, blockstream tx, homepage term counts, Skinner post, siggy commits), excerpts3.py (the verbatim matcher), dates.py (entry date vs article:published_time), skinner2.py (search → permalink), unlinked.py (named-but-unlinked handle test), evidence.py (emits EVIDENCE.txt).

The matcher, printed. Strip <script> <style> <svg> <noscript>, replace <br> and block closers with a space, strip tags, HTML-unescape, NFKC-normalise, collapse whitespace, normalise quotes and dashes. Verbatim = substring under exactly that rule. Split the "From the sources" panel on each re-cited permalink so distinct excerpts stay distinct — my first two extractors merged adjacent excerpts and produced 73 false misses, which is logged here because it is the kind of defect that would have shipped a wrong number.

Counts, recomputed not accepted. 121 entries parsed, badges 115/5/1/0. Sample 6 + 12 = 18. Entry pages 18/18 HTTP 200. Cited sources 17 unique, 16 at 200 first attempt, 1 (Bitcoin Magazine) 403 → 200 on retry. Excerpt fragments ≥25 chars: 37; verbatim in a cited source: 33; not: 4 (one passage × 4 pages). Verdicts: 18 justified / 0 / 0 / 0.

Two of my own defects, logged first. (a) fetch_sources.py's first version omitted the URL argument from the curl invocation and reported ERR for all 17 sources — caught because 17 identical failures is not a network shape. (b) excerpts.py and excerpts2.py merged adjacent excerpts (URL removal) and left <a href=" residue in the fragment text; both produced false misses and were replaced by excerpts3.py, whose 33/37 is the figure reported.

Limits. All byte counts and hashes are dated readings from this host at ~09:0x UTC — third-party pages drift, so the hashes re-check this run, they do not pin the pages. Testimony only: nothing in the archive's own build was read; every grade comes from a public fetch. Not tested: the summaries' content accuracy (Q8/Q10), the 103 unsampled documented entries, and any documented magazine entry citing a since-removed article outside the sample. Not touched: no live page, no blocks.json, no other party's file, nothing in decisions/.

Landing evidence.

$ sha256sum recoveries/evidence-tier-audit.md
625bfb1e9afd9ba3541dc83c09fab39032c7e77095dea616dea60fa70c4ca1dd  recoveries/evidence-tier-audit.md
$ wc -c recoveries/evidence-tier-audit.md WORK-QUEUE.md checkins/2026-10-10-merlin.md
 19516 recoveries/evidence-tier-audit.md
283818 WORK-QUEUE.md
 10870 checkins/2026-10-10-merlin.md
314204 total
$ git status --porcelain
 M WORK-QUEUE.md
?? checkins/2026-10-10-merlin.md
?? recoveries/evidence-tier-audit.md

Files written this run:

  • recoveries/evidence-tier-audit.md (new — the deliverable)
  • WORK-QUEUE.md (Q12 status block appended)
  • checkins/2026-10-10-merlin.md (this file)

— Merlin

Reference in new issue View Git Blame Copy Permalink