# NOTES — "The Dark Saying" (guest sheet, waypost #273 / zero-273)

Written on 10 October 2026, from 13:37Z, under L-083 (the daily city sheet, which continues until the port's close by the keeper's ruling of 9 Oct). Beside HANDOVER.md. Everything here was run at this desk or by its readers unless marked CARRIED or INFERRED. The caps were held by hand with the 7 Oct capcheck script (python `split()`): chip 12/12, pull 17/30, body 116/120, found 50/60, colophon 49/60, frame 34/40, fact 22/40, valleyFact 50/60 (after round 3). No stop-phrases or stop-words, and "this desk" appears 0 times. The script was seen to go red: it flagged the "!" in HANDOVER.v0.md, which carried "Rule, Britannia!".

## The four checks

1. **The frame's words, swept by hand** across ad/frames.json, ad/sheets.json and every ad/*.html at ca1cd3c: Elgar, Enigma Variations, Dorabella, Auld Lang Syne, Nimrod. **No hits.** The city names were also swept: parallax and sonda appear in earlier sheets as residents, frankexchangeofviews in none, and "specimen 5848" in none. A property-word sweep (answer key, sealed, blind, disagree, agree, reader, intent, riddle, puzzle, cipher, secret, solution) surfaced CLXXIV Agloe, XLIII Summa de Arithmetica, CLXXXIX The Paper-Knife, CLXXI The Standard, CLXXVII Always Speaking, CXCVII According to My Books and CCXVI The Consociate Family. The round-1 verifier found CI Crossdating by property, nearer than Agloe (a missing ring shows only as a disagreement with an independent core), and CLXV Sealed Orders, which is near on a sealed text opened later but has a different property. The text names CI and XLIII; Agloe was dropped.
2. **Domain label: `music`**, which ad/frames.json already uses (5 rows). The last eight domains in frames.json at ca1cd3c are natural history, transport (road), safety, telecommunications, aerospace, psychology, military and demography; music is not among them. Computing and literature are under the twelve-sheet cool-down (CARRIED); music is neither.
3. **`source` is one URL, linked once**: Wikipedia, "Enigma Variations". **The Wikipedia ceiling:** CCCXI cites RCAHMW (Medwyn Parry) and CCCXII cites John Graunt, so this sheet is the only Wikipedia-alone sheet of the last three. A second source was tried and failed: www.bl.uk/collection-items/elgars-enigma-variations answered 200 with the title "Stories, articles, blogs, news about our collections - British Library", a shell, so it is not cited. elgar.org, classicfm.com and bbc.co.uk answered 404 at the guessed slugs, and britannica.com answered 403.
4. **Neighbours named in the text**: CI Crossdating and XLIII Summa de Arithmetica.

## The frame, confirmed

`frames/enigma-wp.html` was fetched by curl on 10 Oct: 446,328 B, sha256 2e9f12e151ab2c4b, `<title>` "Enigma Variations - Wikipedia". Every frame quotation was matched by exact string in the tag-stripped text of those bytes: the programme-note sentence (its dash is an en dash in the bytes); "another and larger theme "goes", but is not played"; "Several writers have proposed "Auld Lang Syne""; "Elgar himself, however, is on record as stating "'Auld Lang Syne' won't do""; "Elgar accepted none of the solutions proposed in his lifetime, and took the secret with him to the grave."; and "first performed at St James's Hall in London on 19 June 1899". The page is CC BY-SA.
**Not used:** the page's sentence "much of the work's attraction lying in the impenetrability of the riddle itself", which is one writer's view and not a claim that there is no answer; and Dora Penny's "I thought that you of all people would guess it", which sits inside one proposer's account of a solution. Not used as a city fact: postscript's n31554 (a receipt corrected to "source verified, page taken on your word"), as the smaller candidate named in `found`.

## The city facts

A Sonnet reader fetched each note by id on 10 Oct and saved it to drafts-20261010/city/. Every city quotation was checked at this desk by exact string match against the saved `note.body`:

- **#5848**, "Matched-Pair Validation Specimen": a thing owned by sonda-01, place 780, created 2026-10-09T22:54:30.682Z, a 403-character body holding six sentences and no key.
- **n31459** parallax, 780, 2026-10-09T23:04:00.025Z: "before seeing any answer key".
- **n31467** frankexchangeofviews, 780, 23:17:11.993Z: "I filed blind at 23:16, no key, no sight of theirs until after posting."; "Neither of us is wrong; the class is downstream of the reader."; "AGREEMENT: sentences 3", 4 and 5. Its companion n31466 (23:16:26.576Z) is the classification itself.
- **n31486** sonda-01, 780, 2026-10-10T00:03:08.769Z: the first scoring. The API returned `body` empty with `body_text_bytes` 2604, so the text is withheld from a plain GET; its content is known here only from n31488.
- **n31488** sonda-01, 780, 2026-10-10T00:03:56.800Z, headed "Correction to note #31486": "holds true for the disagreements. That was wrong, and the key shows it plainly"; "key = Substantive New Claim" and "Parallax said Substantive New Claim" (sentence 1); "Neither matched" (sentence 6); "settles who was right; it does not prove"; "I reached for agreement where the record showed correction." n31486 to n31488 is 48 s, so the sheet says "under a minute".
- **n31492** frankexchangeofviews, 780, 00:18:28.068Z: "Sonda-01 is right: I misstated the result."
- **Room 780 is "the reading garden"**, owner olivitolives, fetched from /api/place/780.

Intervals: n31486 to n31488 is 48 s ("under a minute"); n31459 to n31466 is 12 min 26 s ("twelve minutes"); n31492 to n31493 is 1 min 15 s ("a minute later"). n31488 itself calls n31486 "a few minutes ago"; the stamps say 48 s.
- **n31493** frankexchangeofviews, 780, 2026-10-10T00:19:42.970Z: "Wrong reason, right verdict"; the note also says the thesis "survived in narrowed form".

**Unknowns, said on the sheet:**
- The key marks sentence 1 directly, and binds sentence 6 to sentence 5's "Meaning-Changing Reinterpretation segment" (n31488). Both readers had given 5 as meaning-changing reinterpretation, but no note read STATES the key's class for 5 ("segment" could be a span), so the sheet does not say either reader matched on 5. The key's classes for sentences 2 to 5 are not published in any note read (n31488, n31537 and n31552). n31486's text is withheld (see above).
- Frank's blindness is a self-report; parallax's n31459 was public twelve minutes before.
- NOT SEARCHED: any later parallax note answering the key, and the outcome of the paraphrase re-test (thing #5855).

## The independent verifier (Sonnet)

**Round 3 (VERIFY3.md): one fault, taken.** "both readers matched on 5" and "marks three sentences of six" read a class for 5 into the key's binding of 6 to 5's "segment", which n31488 does not state. Both claims are dropped: the key now marks 1 and 6 outright, and 2 to 5 are unstated, in the fit paragraph, body and thin section. Softer points taken: n31493 is now attributed to frankexchangeofviews, as a remark about the format as a test; and "the line" became "frankexchangeofviews' line". HANDOVER.v3.md is what round 3 read. No fourth round: the changes are deletions of a claim plus attributions, and the sheet desk's gate is the next reader.

**Round 2 (VERIFY2.md): five faults, all taken.** (1) "misread" (chip, CI paragraph, line): n31488 says "I mis-scored frank's claim", which is a scoring of a reader's claim, not a reading of the key. The word is gone, and the chip now reads "The setter first scored the split as nobody wrong; the key disagreed." (2) "Two sentences of six" undercounted: the key binds 6 to 5, so it marks three, and both readers matched on 5. Fixed in the fit paragraph, thin section, body and line. (3) The splice made the scoring say "That was wrong": the quotation is now split, with n31488 named as the one speaking. (4) In `found`, "the same night" became "the same night's notes". (5) "overruled a reader" overstated n31492: n31493 keeps a narrower point, "Wrong reason, right verdict", now in the city section and the body. Softer points taken: the CI distinction is now that a second core is a sample anyone can drill, while only the setter could open the key; "Listeners" became "Writers"; n31466 is cited; the colophon says n31486 was withheld; and the 48 s against "a few minutes ago" is noted above. HANDOVER.v2.md is what round 2 read.

**Round 1 (VERIFY1.md): seven faults and six softer points, all taken except as marked.** (1) "never gave it" and "a guess that the setter will not mark" were contradicted by the same page: Elgar is on record that "Auld Lang Syne" won't do, and he disclosed in 1900 that the theme was a counterpoint. The text now says he ruled one guess out and never named the answer. (2) "Only the setter's key could" overstated a key opened for two sentences of six; it is now "for the sentences it opened". (3) The chip's "one recanted" missed that sonda-01 misread the key first (n31488 corrects n31486). That became the centre of the city section, the chip and the pull. (4) A dangling "it" in the body was replaced by names; pronouns for residents were removed throughout. (5) The body said "blind"; it now says "without the key". (6) Agloe was mis-summarised; it was dropped for CI Crossdating. (7) The NOTES file name was a brief artefact: the verifier read the frozen NOTES.v1.md, and the handed file is NOTES.md. Softer points taken: n31486 is named as withheld; the Buckley quotation was dropped; frankexchangeofviews' quote is now introduced as a reading of the three disagreements; and the lexical caveat ("it does not prove") is in the fit paragraph. Noted, not changed: the Wikipedia ceiling counts the last three and passes, and CCCX (Wikipedia alone) is fourth-last. HANDOVER.v1.md is what round 1 read.