{ "log": "academy-corrections", "codex": "watches", "schema_version": "0.1", "generated": "2026-10-03", "policy": "Every correction to a published claim or question is logged here with its date and its source. Claims this Codex decided NOT to publish after verification are logged too, because a withdrawal is part of the public record.", "entries": [ { "id": "corr-2026-09-25-001", "date": "2026-09-25", "date_ist": "2026-09-25 21:40 IST", "type": "withdrawn-before-publication", "affected": [ "WATCHES-CODEX-MASTER-SPEC.md Part 2 Q8 item 3", "draft question q-enthusiast-011" ], "claim_as_drafted": "A water-resistance rating such as '50 m' describes a laboratory test pressure under ISO 22810 rather than a safe depth for activity.", "finding": "The test-pressure framing could not be confirmed in ISO's published material for ISO 22810:2010, whose public scope states only that the standard establishes requirements and test methods to verify water resistance and the marking a manufacturer may apply. ISO's own 2010 news release uses depth language. The framing is widely asserted in enthusiast writing, which is a Never-Cite source here.", "action": "Claim withdrawn. The defensible sourced distinction, that ISO 6425 for divers' watches is a separate standard requiring resistance to at least 100 m and a secured measuring system visible in darkness, was published instead as question q-enthusiast-005.", "sources": [ { "tier": 1, "citation": "ISO 22810:2010, Horology, Water-resistant watches", "url": "https://www.iso.org/standard/45334.html", "accessed": "2026-09-25" } ] }, { "id": "corr-2026-09-25-002", "date": "2026-09-25", "date_ist": "2026-09-25 21:40 IST", "type": "spec-fact-confirmed-and-upgraded", "affected": [ "WATCHES-CODEX-MASTER-SPEC.md Part 2 Q8 item 2" ], "claim_as_drafted": "COSC announced a stricter 'Excellence Chronometer' tier on 12 February 2026; tolerance figures differ between reports, so mark 'verify'.", "finding": "Confirmed on COSC's own site: the news item is dated 12 February 2026 and COSC publishes the tolerance itself as −2 to +4 seconds per day, on the finished cased watch, over four days or more following the 15-day ISO 3159 movement test, in six semi-dynamic positions, with magnetic resistance tested at 200 gauss.", "action": "The 'verify' flag is cleared and the figures are publishable. COSC does not publish a temperature programme for the finished-watch phase beyond 23°C, so none is claimed.", "sources": [ { "tier": 1, "citation": "COSC, Cosc Excellence Chronometer Certified, 12 February 2026", "url": "https://www.cosc.swiss/news/cosc-excellence-chronometer-certified", "accessed": "2026-09-25" } ] }, { "id": "corr-2026-09-25-003", "date": "2026-09-25", "date_ist": "2026-09-25 21:40 IST", "type": "research-premise-corrected", "affected": [ "Seiko dossier, quartz section", "question q-enthusiast-007" ], "claim_as_drafted": "Seiko chose not to patent its quartz technology, which let quartz spread worldwide.", "finding": "Contradicted by Seiko's own museum, which states that Seiko disclosed a large part of the patents it had been granted during the development of its quartz watches.", "action": "Rewritten everywhere as 'obtained patents and then disclosed a large part of them'.", "sources": [ { "tier": 1, "citation": "The Seiko Museum Ginza, Stage 3 (1960s – 1980s)", "url": "https://museum.seiko.co.jp/en/seiko_history/company/company/003/", "accessed": "2026-09-25" } ] }, { "id": "corr-2026-09-26-004", "date": "2026-09-26", "date_ist": "2026-09-26 11:20 IST", "type": "error-caught-in-research", "affected": [ "Legacy page, the Longines section" ], "claim_as_drafted": "International registration number 306231 belongs to the Longines winged hourglass.", "finding": "306231 is the international registration for SUCHARD, named as such in WIPO's own podcast transcript. The number circulates attached to Longines and was nearly propagated here.", "action": "No international registration number is published for Longines. WIPO's Global Brand Database record ID MAD.0158574 is published as a record ID only, and the page states explicitly that 306231 is not Longines so this page cannot become a source of the error.", "sources": [ { "tier": 1, "citation": "WIPO, 'A History of Trademarks: An In Depth Look at the Madrid Agreement', podcast transcript", "url": "https://www.wipo.int/en/web/podcasts/madrid/transcripts/madrid-system-international-trademark-talks-episode6", "accessed": "2026-09-26" } ] }, { "id": "corr-2026-09-26-005", "date": "2026-09-26", "date_ist": "2026-09-26 11:20 IST", "type": "held-for-reverification", "affected": [ "Legacy page, the Cartier Tank section" ], "claim_as_drafted": "The Cartier Tank was created in 1917.", "finding": "Cartier's collection page documents only that Louis Cartier 'designed it after the tanks of WWI battlefields', with no year. The single Cartier page where 1917 surfaced returned a garbled, self-referential sentence consistent with an extraction artefact.", "action": "The year is not published. The sourced sentence is published without it, and the omission is stated on the page so a reader knows it was checked rather than missed.", "sources": [ { "tier": 1, "citation": "Cartier, 'Tank' collection page", "url": "https://www.cartier.com/en-us/watches/collections/tank/", "accessed": "2026-09-26" } ] }, { "id": "corr-2026-09-26-006", "date": "2026-09-26", "date_ist": "2026-09-26 16:19 IST", "type": "error-caught-in-qc", "affected": [ "Academy question q-enthusiast-003" ], "claim_as_drafted": "COSC states it tests every watch submitted rather than sampling.", "finding": "QC Gate 4 (Session 06, Part 27), a fresh agent given only this claim and its cited source, the COSC FAQ, found the page describes the test process and pass criteria but does not state whether COSC tests 100% of submissions or samples. Confirmed on a second direct fetch of the same page.", "action": "The sentence is removed from the explanation. The rest of the tested claim (the -4/+6 s/day tolerance and the test programme) is unaffected and remains fully sourced.", "sources": [ { "tier": 1, "citation": "COSC, \"COSC FAQ: All your questions answered\"", "url": "https://www.cosc.swiss/cosc-faq", "accessed": "2026-09-26" } ] }, { "id": "corr-2026-09-26-007", "date": "2026-09-26", "date_ist": "2026-09-26 16:19 IST", "type": "error-caught-in-qc", "affected": [ "Academy question q-enthusiast-004" ], "claim_as_drafted": "The Excellence Chronometer's finished-watch evaluation, six semi-dynamic positions and 200-gauss magnetic resistance, was described as being held to one shared tolerance of -2 to +4 seconds per day.", "finding": "QC Gate 4 (Session 06) found COSC's own certification page states two separate tolerances: -2/+4 s/day for the six-position simulated-wear test, and a wider +/-10 s/day for the magnetic-resistance test at 200 gauss. The original sentence structure read as if one tolerance covered both tests.", "action": "Rewritten to state the two tests and their two distinct tolerances separately.", "sources": [ { "tier": 1, "citation": "COSC, \"Certified Excellence Chronometer\"", "url": "https://www.cosc.swiss/certified-excellence-chronometer", "accessed": "2026-09-26" } ] }, { "id": "corr-2026-09-26-008", "date": "2026-09-26", "date_ist": "2026-09-26 16:19 IST", "type": "sourcing-gap-caught-in-qc", "affected": [ "Legacy page and sidecar, the G-Shock section (claim id gshock-1983)" ], "claim_as_drafted": "A single source record cited only Casio's general corporate-history page for a claim that bundled the 1983 release date together with the 'Triple 10' shortfall figures (seven years against a ten-year battery goal; one metre against a ten-metre drop goal), even though the record's own citation text named two separate Casio documents.", "finding": "QC Gate 4 found the cited URL (world.casio.com/corporate/history/chapter02/) states the release date but not the Triple 10 figures; those live only on a second Casio page (gshock.casio.com's Triple 10 history) that the record never carried as a URL. The published Legacy page prose already cited both pages correctly in its Sources list, the gap was confined to the machine-readable sidecar.", "action": "watches-legacy-claims.json's gshock-1983 entry changed from a single 'source' to a 'sources' array carrying both URLs, each annotated with which part of the claim it supports. Every other claim in the same file was also moved from 'source' to 'sources' for a consistent, extensible schema.", "sources": [ { "tier": 1, "citation": "Casio, 'History 1980-' (corporate history)", "url": "https://world.casio.com/corporate/history/chapter02/", "accessed": "2026-09-26" }, { "tier": 1, "citation": "Casio, 'The History of G-SHOCK's \"Triple 10\"'", "url": "https://gshock.casio.com/intl/cp/bc-one-japan-2025/triple10/", "accessed": "2026-09-26" } ] }, { "id": "corr-2026-09-26-009", "date": "2026-09-26", "date_ist": "2026-09-26 21:05 IST", "type": "published-error-corrected", "affected": [ "Glossary hub (/watches/glossary/), the closing section" ], "claim_as_drafted": "The glossary hub stated, in hand-written text, that \"Four entries could not be closed from a citable source.\"", "finding": "The number was typed once and never derived, so it was wrong when written (there were five such entries at the time, not four) and drifted further as the glossary grew, it still said four when the section listed 33. The same block also gathered two different statuses, 'no authority defines this' and 'verified with an open question', under a heading saying nobody had defined them, which misdescribed 18 of the 33 entries: those do have an identified authority, quoted as far as its published material goes.", "action": "The count is now derived from the records at build time and can no longer drift. The section was split into two lists under accurate headings, one for terms no authority defines and one for terms where an authority exists but part of the answer is not public, each carrying its own derived count. Found by reading the built page, not by the validator, which had nothing to compare the typed number against.", "sources": [] }, { "id": "corr-2026-09-26-010", "date": "2026-09-26", "date_ist": "2026-09-26 20:40 IST", "type": "drafting-defects-caught-before-publication", "affected": [ "60 glossary entries drafted in Session 09, before any of them was published" ], "claim_as_drafted": "Every claim in the 60 new entries carrying a number, date, standard reference or attribution, 187 sentences in all, was re-checked against its cited source by agents that were shown only the sentence and the URL, never the reasoning that produced it.", "finding": "36 of the 187 failed: 33 were right in substance but wrong in a detail, and 3 were not supported by the cited page at all. The recurring classes were: asserting that a standards text is sold when the cited page says nothing about price; writing a failed fetch as a fact about the world (one entry claimed ISO's site was unreachable when it is not); negative claims broader than what was checked ('no standards body defines this' when only one ISO committee's catalogue had been read); quotations stitched from non-contiguous fragments or left unclosed; the drafter's own gloss presented as sourced; and one arithmetic slip, 0.03059 days described as a little under 44 minutes when it is a little over.", "action": "All 36 were corrected before the entries were built, and the unsupported 'sold rather than published' wording was then searched for across all 60 entries and fixed in a further 13 places that no sampled claim had caught. Nothing in this group reached a reader; it is logged because the failure rate, roughly one claim in five, is a fact about how this Codex is drafted, and a reader is entitled to know the drafting pass alone is not what makes an entry trustworthy.", "sources": [] }, { "id": "corr-2026-09-26-011", "date": "2026-09-26", "date_ist": "2026-09-26 20:40 IST", "type": "attribution-corrected-before-publication", "affected": [ "Glossary entry: Tourbillon" ], "claim_as_drafted": "The draft reversed the makers of a British Museum pocket chronometer, crediting the watch to Breguet and the movement to John Arnold.", "finding": "The museum's own record states the opposite: the 1774 watch was made by John Arnold, and Breguet made the movement fitted to it between January 1808 and May 1809.", "action": "The attribution was reversed to match the record before the page was built. Singled out from the group above because a reversed attribution reads as confidently as a correct one and would not have been obvious to a reader.", "sources": [ { "tier": 1, "citation": "The British Museum, collection record for the Arnold pocket chronometer with Breguet tourbillon movement", "url": "https://www.britishmuseum.org/collection/object/H_1958-1201-1848", "accessed": "2026-09-26" } ] }, { "id": "corr-2026-09-27-012", "date": "2026-09-27", "date_ist": "2026-09-27 12:20 IST", "type": "drafting-defects-caught-before-publication", "affected": [ "History and How a watch works, both drafted in Session 10, before either was published" ], "claim_as_drafted": "Every sentence in the two new pages carrying a number, date, standard reference, attribution or direct quotation, 55 in all, was re-checked against its cited source by agents shown only the sentence and the URL, never the surrounding page or the reasoning behind it.", "finding": "16 of the 55 were wrong in a detail: a defect rate of about 29%, higher than the 19% measured on the previous session's drafting. None was a fabricated fact and none was unsupported outright, but the classes were consistent: quotations that silently dropped a parenthesis or presented a translation as the source's own words; unsourced claims of absence ('a Royal Society account could not be opened', 'Cartier's page could not be retrieved') where a failed search had been written as a fact about the world; mechanical assertions with no source behind them; a company's own dating generalised; and a quotation attributed to the wrong part of a patent.", "action": "All 16 corrected before either page was built. Logged because the rate is the point: this prose was written from fact records that already carried verbatim quotes and verified URLs, and it still needed 16 corrections. The independent re-check, not the care taken while drafting, is what makes these pages citable.", "sources": [] }, { "id": "corr-2026-09-27-013", "date": "2026-09-27", "date_ist": "2026-09-27 12:20 IST", "type": "misattributed-quotation-caught-before-publication", "affected": [ "History, the section on standardised time" ], "claim_as_drafted": "The draft quoted the 1884 International Meridian Conference as resolving to adopt the Greenwich meridian \"as the standard meridian for longitude\", presenting it as the conference's decision.", "finding": "The conference's own printed protocols show those words belong to Mr Rutherfurd's resolution offered at the session of 6 October, which was then temporarily withdrawn. The wording the conference actually adopted speaks of the initial meridian, not the standard meridian.", "action": "The sentence now names the resolution as Rutherfurd's proposal of 6 October, states that it was temporarily withdrawn, and notes that the adopted text differs. Singled out from the group above because quoting a withdrawn draft as an adopted resolution misstates the historical record while looking impeccably sourced.", "sources": [ { "tier": 1, "citation": "International Meridian Conference, Washington, October 1884, printed protocols of the proceedings", "url": "https://www.gutenberg.org/files/17759/17759-h/17759-h.htm", "accessed": "2026-09-27" } ] }, { "id": "corr-2026-09-27-014", "date": "2026-09-27", "date_ist": "2026-09-27 12:40 IST", "type": "published-error-corrected", "affected": [ "The index.md machine-readable twin of every content page, since Session 08" ], "claim_as_drafted": "Each page's index.md repeated its own section heading at the start of every section, preceded by the reading-level badge, and inserted a stray space before punctuation that followed an inline element, 11 occurrences on the Legacy page and 17 on the Seiko dossier, most of them inside direct quotations.", "finding": "Two faults in the extractor that generates the twins. The section heading was read into the record's heading field and then left in the HTML handed to the text extractor, so it came out twice. And a citation chip or glossary link immediately before a comma or full stop left a space stranded in front of it, so a quoted passage read 'a going train , a regulating organ .', altering quoted matter in the very files built for machines to read.", "action": "The extractor now removes the heading element and the level badge before extracting prose, and closes the space before punctuation, matching the rule the JSON fields already carried. Every twin on the site was regenerated: the stray-space count is now zero on all 87 pages. Found by reading a generated index.md rather than by any automated check, the validator tests that the twins exist and parse, not that their prose reads correctly.", "sources": [] }, { "id": "corr-2026-09-27-015", "date": "2026-09-27", "date_ist": "2026-09-27 12:00 PM IST", "type": "figure-corrected-before-publication", "affected": [ "The money side, the Titan Company Limited section, drafted in Session 11 before publication" ], "claim_as_drafted": "Titan Company Limited's consolidated revenue from operations for the year ended 31 March 2026 was drafted as ₹87,100 crore, in two places.", "finding": "The figure is ₹87,584 crore. Two independent re-verifiers, checking separately, both read it off the audited statement and both flagged the same error, and the document's own arithmetic settles it: revenue from operations of 87,584 plus other income of 552 gives the total income of 88,136 crore that the same sentence already cited correctly. The wrong figure came from the research pass, not from the source.", "action": "Corrected in both places before the page was built. Recorded because of how the error behaved: it was internally plausible, sat beside correctly-transcribed figures from the same table, and would have read as authoritative. The arithmetic cross-check, does this figure reconcile with the others on the page?, is what caught it, and is now the thing to run on any financial figure.", "sources": [ { "tier": 1, "citation": "Titan Company Limited, Statement of Consolidated Audited Financial Results for the quarter and year ended 31 March 2026", "url": "https://www.titancompany.in/sites/default/files/2026-05/Q4FY26%20-%20Financial%20results%20Uploaded.pdf", "accessed": "2026-09-27" } ] }, { "id": "corr-2026-09-27-016", "date": "2026-09-27", "date_ist": "2026-09-27 12:00 PM IST", "type": "drafting-defects-caught-before-publication", "affected": [ "The money side, drafted in Session 11, before publication" ], "claim_as_drafted": "Every sentence on the new financials page carrying a figure, a period or a claim of absence, 22 in all, was re-checked against the company document it cites by agents shown only the sentence and the URL.", "finding": "8 of the 22 were wrong in a detail: a 36% rate, the highest of the three sessions measured on this project (19%, then 29%, now 36%). Besides the Titan figure logged separately: a Seiko net-sales figure for one year was paired with an operating-profit figure for a different year off the same row; a German quotation was misspelt (Drittel for Dritteln); NOMOS was said to publish no employee figure when its own press site publishes one (240); Rolex's lack of an investor section was asserted for a whole website on the strength of one page; and the Swiss identifier register's recording authority was named as the commercial register when the record says the federal UID-Register.", "action": "All corrected before the page was built. The rising rate is not evidence that the work is getting worse, the material is getting harder, and financial figures are the least forgiving content this site publishes. It is evidence that the re-check is load-bearing: at a 36% rate, a page published without it would carry roughly one wrong detail in every three sentences that state a number.", "sources": [] }, { "id": "corr-2026-09-27-017", "date": "2026-09-27", "date_ist": "2026-09-27 4:40 PM IST", "type": "published-claims-not-supported-by-their-own-citations", "affected": [ "Glossary: Gauss, Mainspring, Complication", "all three published since Session 05" ], "claim_as_drafted": "Three live claims cited sources that do not support them. The Gauss entry opened by quoting a watch-marketing figure of \"15,000 gauss\" while citing only the BIPM SI Brochure and a NIST conversion table. The Mainspring entry carried a quoted sentence about the rotor winding the mainspring. The Complication entry quoted a manufacturer's operational test of what counts as a complication.", "finding": "QC gate 4 re-verification, run by agents shown the claim and the URL only. The SI Brochure does not contain the gauss at all and neither cited source mentions watches or 15,000 gauss. The quoted mainspring sentence appears on neither cited page, one is a history page about weights giving way to the mainspring, the other describes the escapement. The manufacturer's wording in the Complication entry had no citation whatever; the entry's only source is an ISO catalogue page that mentions neither that company nor the wording.", "action": "All three rewritten to state only what the cited sources support, with the unsupported quotations removed rather than re-attributed, re-attributing would have meant publishing a source this pass had not read. Each entry now says plainly what is not cited on it. These were live on the site, so they are logged as published errors rather than as drafting defects.", "sources": [] }, { "id": "corr-2026-09-27-018", "date": "2026-09-27", "date_ist": "2026-09-27 4:40 PM IST", "type": "published-claims-overstated", "affected": [ "Seiko dossier (two claims)", "Glossary: Jewel (two claims), Watch case" ], "claim_as_drafted": "Five further live claims went beyond their sources. The Seiko dossier said Kintarō Hattori's 1881 shop \"sold and repaired imported clocks and watches\", and that the Quartz Astron went on sale on 25 December at a time when \"a good mechanical watch drifted that much in a day\". The Jewel entry added that synthetic stones are \"not precious in any commercial sense\" and that a jewel count \"is not in itself a measure of quality\". The Watch case entry asserted flatly that ISO never defines the case.", "finding": "Same gate 4 pass. The Seiko museum page records the founding of K. Hattori & Co. in Uneme-cho, Kyobashi, never mentions repair, and dates the imported-timepiece trade to 1886, not 1881; the page cited for the Astron carries neither the 25 December date nor any comparison with mechanical drift. The jewels page says nothing about commercial value, and ties stone counts to how high-class a watch is as well as how complicated, so it does not support reading the count as unrelated to quality. And ISO 6426-2:2026 states that it defines the principal terms used in horology, so whether it defines the case is untested rather than settled.", "action": "The founding sentence now follows the museum's own wording and dates the imported-timepiece trade to 1886. The Astron's launch date is re-cited to the Seiko page that does carry it, and the unsourced comparison with mechanical drift is gone. Both jewel glosses are replaced with what the page actually says. The watch-case entry now says the question cannot be checked because only the standard's scope statement is public, instead of asserting an absence it cannot test.", "sources": [] }, { "id": "corr-2026-09-27-019", "date": "2026-09-27", "date_ist": "2026-09-27 4:40 PM IST", "type": "accessibility-defect-corrected", "affected": [ "The text-size control in the site navigation, on every page" ], "claim_as_drafted": "The three text-size buttons were 44px wide but only 38px tall, below the 44px minimum tap target this site sets for itself and that every other control on it meets.", "finding": "QC gate 6. The control that exists to help readers who need larger type was the one control too small to hit comfortably.", "action": "Raised to 44px. Found by auditing the live stylesheet against the site's own stated floor rather than by any per-page check.", "sources": [] }, { "id": "corr-2026-09-28-020", "date": "2026-09-28", "date_ist": "2026-09-28 2:05 AM IST", "type": "drafting-defects-caught-before-publication", "affected": [ "/watches/standards/ — the Standards core page, before it was built" ], "claim_as_drafted": "Nine quotations drawn for the Standards page were truncated without an ellipsis. Four of the nine changed the meaning of the text they quoted: Qualité Fleurier's requirement that all operations be carried out in Switzerland was quoted without its closing exception ‘à l'exception des matériaux non ouvragés’, turning a qualified rule into an absolute one; METAS-N001's statement that the order of cycles 2 to 10 is not critical was quoted without the sentence that immediately follows and reverses it; German Customs' summary of § 5 Nr. 1 of the Glashütteverordnung was quoted without the ‘und’ clause that makes it cumulative with § 5 Nr. 2; and the Poinçon de Genève's requirement of applicant firms was quoted as a bare ‘must be able to […]’ when the elided text was the whole operative requirement.", "finding": "Seven verification agents, each shown only a quoted string and the URL it was attributed to, never the page being drafted. 180 quotations were checked. None was fabricated and none was attributed to the wrong source; the failure mode was entirely truncation. The independent re-check is the only thing that found any of them.", "action": "All nine corrected to the full sentence before the page was built. Two further corrections followed from the same pass: METAS-N001's daily-rate criterion is stated per watch category (0 to 5, 0 to 6 and 0 to 7 seconds a day for categories 1a, 1b and 2), and quoting the 0-to-5 figure unqualified overstates what a smaller watch must meet; and METAS-N001's requirement that every movement and every watch be tested is an obligation on the applicant, with METAS itself performing sample checks under ISO 2859-1:2014, so ‘every watch is tested’ and ‘METAS tests every watch’ are different statements and only the first is in the document.", "sources": [] }, { "id": "corr-2026-09-28-021", "date": "2026-09-28", "date_ist": "2026-09-28 2:05 AM IST", "type": "build-defect-corrected", "affected": [ "The index.md and index.json twin of every page, and llms-full.txt" ], "claim_as_drafted": "The text extractor that generates each page's Markdown and JSON twin dropped the quotation marks around any element. A browser draws those marks; a text dump does not. Quoted statute and standards text therefore arrived in the twins reading as the Codex's own prose.", "finding": "Found by reading the built index.md rather than by any automated check, which is how the same class of defect was found in Session 10. The element was introduced on the Standards page this session, so nothing published before now was affected — but on a site whose entire claim is that quotations are quotations, this is the worst class of extraction defect available.", "action": "The extractor now emits the quotation marks itself. A second, smaller defect was fixed in the same pass: an inline element sitting immediately before a closing quotation mark stranded a space in front of it, giving ‘Swiss Made ”’ — the mirror image of the space-before-punctuation defect fixed in Session 10.", "sources": [] }, { "id": "corr-2026-09-28-022", "date": "2026-09-28", "date_ist": "2026-09-28 2:05 AM IST", "type": "build-defect-corrected", "affected": [ "/watches/standards/ — during the build, before publication" ], "claim_as_drafted": "A rule added to the citation composer this session, to collapse two identical citation chips printed back to back, silently deleted two sentences from the ISO section of the Standards page. The lazy quantifier in its pattern backtracked past the first closing tag and swallowed the prose between two chips.", "finding": "The validator passed with 4,251 checks and zero failures while the sentences were missing, because nothing it checks could know they were meant to be there. Found by looking at a screenshot of the rendered page.", "action": "The pattern was rewritten so it cannot span a closing tag, and the built page was then checked word by word against the hand-written source to confirm that nothing else had been lost. Recorded because the lesson generalises: a tool added to improve a page can remove content from it, and only reading the output catches that.", "sources": [] }, { "id": "corr-2026-09-28-023", "date": "2026-09-28", "date_ist": "2026-09-28 3:07 AM IST", "type": "published-claim-corrected", "affected": [ "/watches/make-your-own/, “Buying the movement” section, the “Who supplies watch movements to other brands?” FAQ item, the “What could not be sourced” list, and the corresponding entries in the FAQPage structured data" ], "claim_as_drafted": "Session 13's Make your own page stated, of the movement suppliers it examined (Time Module, Miyota, Ronda): “None of them publishes a price, a minimum order, or any statement of who may become a customer.”", "finding": "Session 15's research for the Citizen dossier found that Miyota's own product page for Caliber 9015 states “Minimum unit order : 1000pcs per each model” alongside a lead time, a published minimum order the earlier session's checked pages did not surface. Miyota's general FAQ page still says only that prices and minimums “vary” and does not itself state a figure, so the two pages are not contradictory, but the blanket claim that no movement supplier publishes any minimum order no longer holds.", "action": "The claim was narrowed rather than reversed: movement suppliers still publish no price and no eligibility rule, but the “none publishes a minimum order” clause was removed and replaced with the Miyota exception, cited to the Caliber 9015 page (new source 21). The “What could not be sourced” list and the FAQ answer were updated to match, and the aside's “Published minimums found” range was widened from “none, to 500 pieces” to “none, to 1,000 pieces”. Not fixed silently: this entry is the record that the earlier page was wrong on this point from 27 September 2026 (when it published) until this correction.", "sources": [ { "tier": 1, "citation": "MIYOTA (Citizen Watch Co., Ltd.), Caliber 9015 product page", "url": "https://miyotamovement.com/product/9015/", "accessed": "2026-09-28" } ] }, { "id": "corr-2026-09-28-025", "date": "2026-09-28", "date_ist": "2026-09-28 3:07 AM IST", "type": "build-defect-corrected", "affected": [ "/watches/standards/, the hover title text on all 174 citation chips (cosmetic only; no quoted text or link was affected)" ], "claim_as_drafted": "The citation composer builds each chip's hover title as “citation, period, read DATE”. Most Standards facts carry no “period” value (period applies to financial-statement-style facts, not statute or standard citations), so the joined string produced a bare double dash: ‘… preamble, — read 2026-09-28’.", "finding": "Found while building the Citizen dossier this session, when the same composer pattern produced the identical artifact on 96 new chips. Purely cosmetic, visible only in a hovered tooltip, never in the rendered prose or a quoted fact, but it is on every one of the page's 174 chips, so it was corrected rather than left as ‘just Standards, not this session's page’.", "action": "The composer now omits the period segment entirely when a fact carries none, instead of joining an empty string between two dashes. standards.html was rebuilt from the fixed composer with the source facts otherwise untouched, chip count (174) and source count (62) are unchanged, confirming nothing else moved.", "sources": [] }, { "id": "corr-2026-09-28-026", "date": "2026-09-28", "date_ist": "2026-09-28 3:07 AM IST", "type": "build-defect-corrected", "affected": [ "The index.md and index.json twin of every page, and llms-full.txt, site-wide, since the founding of these files, not only pages touched this session" ], "claim_as_drafted": "The text extractor's whitespace clean-up stripped the space after ANY double-quote character, curly or straight: ‘“’ (a curly opening quote) and the plain ASCII ‘\"’ alike. A straight quote is used for both opening and closing in this Codex's prose, so the rule welded every closing quote onto the word immediately following it whenever no punctuation separated them, ‘“founded”year’, ‘“CRYSTRON SOLAR CELL”the’, ‘“Citizen Watch Co., Ltd. established.”Yosaburo’, in the text dumps of every page carrying a mid-sentence quoted phrase, including the already-published Seiko dossier.", "finding": "Found while reading the Citizen dossier's built index.md this session, the same way the -element defect was found in Session 14: by reading the text extraction, not by any automated check, because the bug is invisible in the rendered HTML, browsers do not need a space between a quotation mark glyph and the next character for it to look correct, so the browser check and every prior visual pass missed it. Confirmed present in the already-published Seiko dossier's index.md too, meaning this defect predates this session and has been live since Seiko's index.md/index.json were first generated.", "action": "The rule was narrowed to the unambiguous curly opening quote only; the straight ASCII quote was removed from it. Rebuilt the full site and confirmed by grep across every built index.md that no closing-quote-welded-to-word pattern remains, that no new stray double-space-before-a-quote pattern was introduced, and that the validator still passes at 0 warnings and 0 failures.", "sources": [] }, { "id": "corr-2026-09-28-024", "date": "2026-09-28", "date_ist": "2026-09-28 3:07 AM IST", "type": "drafting-defects-caught-before-publication", "affected": [ "/watches/brands/citizen/, the Citizen dossier, before it was built" ], "claim_as_drafted": "Of 61 quotations drawn for the Citizen dossier, 7 had defects when checked against the live page, none of them fabrications or wrong attributions: two quotations from Citizen's own Japan and Europe group-company lists ran company names together with no ellipsis where the source page actually interposes each company's phone number and postal address; one quotation of a segment's scope in the 141st Annual Securities Report gave only 2 of the segment's 9 listed production companies, omitting (among others) two Swiss subsidiaries, Manufacture La Joux-Perret S.A. and Frederique Constant Holding SA, with no ellipsis marking the cut; one quotation added a full stop the source line does not have; one quotation used plain spaces where the source uses full-width Japanese commas between listed items; one quotation of a 2011 history entry did not reproduce a stray space the source itself has before a comma; and one figure (“Watches 177.1 billion yen”, from CITIZEN REPORT 2025) needed its exact chart context restated, since the report's “FY2024” period label and its 177.1bn figure sit close to, but not exactly matching, the tanshin's ¥179,211 million Watch-segment external sales for what appears to be the same fiscal year.", "finding": "Four verification agents, each shown only a quoted string and the URL it was attributed to, checked all 61 claims in four topic batches (corporate/founding history, technology and brand history, financials, and Miyota/OEM). Applying the terminal-punctuation-or-ellipsis rule adopted after Session 14, all seven defects above were caught as CLOSE verdicts before any of this material was composed into a page.", "action": "The two group-company lists were corrected to mark the omitted contact details with ellipses; the Watch-segment scope quotation was replaced with the full nine-company list, re-confirmed by two independent fetches of the same PDF table; the extra full stop was removed; the device-segment quotation's separators were corrected to the source's own full-width commas; the 2011 quotation was corrected to match the source's own stray space; and the Watches-177.1bn fact's meaning was rewritten to state the chart's own structure precisely and to publish both this figure and the tanshin's ¥179,211 million figure side by side, as a disclosed disagreement between two of Citizen's own documents, rather than picking one, per this Codex's standing reconciliation rule (Part 29.6).", "sources": [] }, { "id": "corr-2026-09-28-027", "date": "2026-09-28", "date_ist": "2026-09-28 10:35 AM IST", "type": "withheld-claim-now-published", "affected": [ "/watches/legacy/ — the “what could not be sourced” list", "data/watches-legacy-claims.json — the withheld list" ], "claim_as_drafted": "From Session 08 this Codex withheld any claim about Kikuo Ibe's role in the G-SHOCK, on the stated ground that he was not named on the Casio page documenting the “Triple 10” goal.", "finding": "Session 16's Casio research found a third Casio-owned page that does name him: “The author of these words was Kikuo Ibe, who was in charge of watch exterior design at the time.” The same page states that his vision “defied all conventional wisdom” and credits him with devising the five-stage shock-absorbing structure. An independent verifier confirmed all three sentences complete to their own terminal punctuation, and separately re-confirmed that he is still NOT named on either of the two pages that narrate the DW-5000C — so the original withholding was correct on the evidence then held.", "action": "The claim is now published in the Casio dossier, in Casio's own terms and no further: Casio describes Ibe by the job he held at the time and credits him with the proposal, the vision and the shock-absorbing structure, and it does NOT call him the inventor of the G-SHOCK — so neither does this Codex. The withheld entry was kept in the public withheld list rather than deleted, with a resolution note and the source, because what was withheld and why is part of the record.", "sources": [ { "tier": 1, "citation": "Casio, GMW-B5000 feature article", "url": "https://gshock.casio.com/intl/contents/mag/gmw-b5000/01/", "accessed": "2026-09-28" } ] }, { "id": "corr-2026-09-28-028", "date": "2026-09-28", "date_ist": "2026-09-28 10:35 AM IST", "type": "drafting-defects-caught-before-publication", "affected": [ "/watches/brands/casio/ — the Casio dossier, before it was built" ], "claim_as_drafted": "Of 93 quotations drawn for the Casio dossier, 9 had defects when checked against the live source, none of them fabrications or wrong attributions. Four were material: a quotation of the results summary's segment-table header dropped two columns without an ellipsis and thereby re-attached a footnote so that the pre-adjustment segment total read as the consolidated figure; the five-year employee row kept its own label promising average temporary-staff numbers and then omitted all five of them; a quotation of Casio's G-SHOCK technology index ran nine headings together when the page lists thirteen, and mislabelled one of them; and a quotation of the company-establishment sentence began mid-sentence with a dropped leading clause and a re-capitalised first word. Five were typographic: straight single quotation marks where the page prints curly doubles, ASCII parentheses where the source prints full-width CJK ones, an inserted ellipsis and space delimiters where a table has cell walls and Japanese commas, inserted slash separators where a document prints three separate lines, and a page title reproduced without the doubled space it actually carries.", "finding": "Four verification agents, each shown only a quoted string and the URL it was attributed to, checked all 93 claims in four topic batches (origin and entity data; the Casiotron, the G-SHOCK and Kikuo Ibe; brands, technology, production and contacts; financials). Zero fabricated, zero attributed to the wrong source, zero unreachable. The financial verifier also independently re-computed every figure: segment sales sum exactly to consolidated net sales, segment profit sums exactly to consolidated operating profit after the stated adjustment, every quarterly build-up ties, and all printed percentages recompute — with no rounding gap anywhere, which is worth recording because a previous session found one.", "action": "All nine corrected before the page was built. Three further framing fixes were made on records whose verdict was EXACT but whose presentation would have misled a reader: the quarterly segment row now carries its own printed column headers, because that row holds a next-year FORECAST column and a half-year subtotal beside the quarterly actuals and nothing in the bare figures distinguished them; the G-SHOCK unit-volume quote now carries the slide's own footnote giving the prior-year comparison, the only direction-of-travel signal on it; and a scoped negative about the absence of a service-centre locator had the word “global” removed, because four checked pages — one of them a single India-locale URL — cannot support a claim that broad.", "sources": [] }, { "id": "corr-2026-09-28-029", "date": "2026-09-28", "date_ist": "2026-09-28 10:35 AM IST", "type": "build-defect-corrected", "affected": [ "The index.md and index.json twin of every page, and llms-full.txt — site-wide, ten occurrences across eight pages" ], "claim_as_drafted": "An inline element sitting immediately before a closing PARENTHESIS stranded a space in front of it in the text twins: ‘certifies resistance to 1.5 T (15,000 gauss )’, ‘see Buying the movement )’. The mirror defect existed after an opening parenthesis: ‘The detent escapement ( chronometer’.", "finding": "Found by reading the Casio dossier's built index.md, the same way Session 15's quote-welding defect was found. Invisible in the rendered HTML, as that whole family of defects is. What makes this one worth a log entry beyond the fix itself is that it is the THIRD session to widen the same single rule one punctuation mark at a time — Session 10 added the sentence punctuation, Session 14 the closing quotation mark, Session 15 had to narrow the opening-quote case after it over-matched, and Session 16 found the parentheses.", "action": "The rule was closed rather than extended again: every closing mark now matches on the left and every opening mark on the right, in both the Markdown extractor and the JSON field cleaner so the two cannot drift apart. The square brackets are deliberately excluded, because the extractor renders a citation chip as ‘[ Casio ]’ and those spaces are its own format rather than a defect — confirmed after the fix by counting chip renders on the rebuilt page. Verified by grep across every built index.md: ten occurrences before, zero after, and no chip damaged.", "sources": [] }, { "id": "corr-2026-09-28-030", "date": "2026-09-28", "date_ist": "2026-09-28 11:36 AM IST", "type": "withheld-claim-now-published", "affected": [ "/watches/legacy/ — the “what could not be sourced” list", "data/watches-legacy-claims.json — the withheld list" ], "claim_as_drafted": "From Session 08 this Codex withheld the claim that a Speedmaster was worn on the lunar surface during the Apollo 11 EVA, on the stated ground that neither Smithsonian record supports it and that NASA's Apollo Lunar Surface Journal records Armstrong leaving his in the Lunar Module.", "finding": "Session 17 checked the withheld list before building the Omega dossier, as the rule added in Session 16 requires. The museum position is unchanged and the original reasoning was sound: the Smithsonian holds Armstrong's and Collins's Apollo 11 Speedmasters, holds no Aldrin chronograph, and its only statement about Apollo 11 wear concerns Collins wearing his outside his suit at launch and reentry while in lunar orbit. What resolved the question was a fuller passage of the same NASA-hosted Journal, in which Aldrin states that Armstrong left his inside and that Aldrin's own watch “was the only one out on the surface” — and, in the same breath, that the watch he shipped to the Smithsonian “turned up missing”, which explains the absence of a museum record. Confirmed by two independently-worded fetches of that page and separately surfaced by a verification agent approaching from a different direction.", "action": "The claim is published in the Omega dossier as Aldrin's own recollection, attributed to him, with this source's provenance stated as it is everywhere else on this site — individually authored and NASA-hosted rather than an agency statement, and with its withdrawn table of flown watches still not cited. The withheld entry was kept in the public list with a resolution note rather than deleted. The separate brand claim that the Speedmaster was the FIRST watch worn on the Moon is NOT thereby established and remains recorded as a brand claim. Second consecutive session in which checking the withheld list resolved a standing question.", "sources": [ { "tier": 1, "citation": "Apollo Lunar Surface Journal, “EVA Preparations”, Apollo 11 — NASA-hosted", "url": "https://www.nasa.gov/wp-content/uploads/static/history/alsj/a11/a11.evaprep.html", "accessed": "2026-09-28" } ] }, { "id": "corr-2026-09-28-031", "date": "2026-09-28", "date_ist": "2026-09-28 11:36 AM IST", "type": "drafting-defects-caught-before-publication", "affected": [ "/watches/brands/omega/ — the Omega dossier, before it was built" ], "claim_as_drafted": "Of 109 quotations drawn for the Omega dossier, 7 needed correction, none of them fabrications or wrong attributions. Two were quotation defects: a results-release sentence reproduced a footnote marker inside the closing parenthesis rather than after it, and a transcribed criteria table introduced a colon the source does not print. Two were quotations that were character-for-character correct but exist ONLY in a page's HTML meta description rather than in reader-visible body copy — a citation reading “the page says” would have sent a reader to a page where the sentence cannot be found. Three were scoped negatives that were accurate but claimed more than what had been checked could support.", "finding": "Five verification agents, each shown only a quoted string and the URL it was attributed to, checked all 109 claims, and were additionally asked to name the publisher of every page they fetched — because this is the first dossier whose citations are not all the company's own, and the independent/own distinction is only as good as the domains behind it. Zero fabricated, zero attributed to the wrong source, zero unreachable except one documented block. The most consequential single finding was not a defect at all: the Omega page a reader would naturally cite for the 1903 company renaming does not date that renaming, and only the press release does.", "action": "Both quotation defects corrected. Both metadata-only quotations re-labelled in their citations as page meta descriptions, and the page says so in prose where it matters. The three negatives narrowed: one asserting that no NASA page addresses Apollo 11 EVA watches was rewritten to say no NASA EDITORIAL page, because NASA hosts the Apollo Lunar Surface Journal which addresses it directly; one about undated employee figures was rescoped from the pages to the figures, since one page does carry dates against other metrics; and one about an absent India locale was given an explicit access date. The 1903 date is cited to the press release rather than to the Chronicle page.", "sources": [] }, { "id": "corr-2026-09-28-032", "date": "2026-09-28", "date_ist": "2026-09-28 11:36 AM IST", "type": "disclosure-added", "affected": [ "/watches/brands/omega/#financials" ], "claim_as_drafted": "The Swatch Group's 2025 results release states net sales of CHF 6 280 million, −5.9% at current exchange rates. Its own 2024 release states net sales of CHF 6 735 million.", "finding": "A verification agent recomputed the change from the two published figures and got −6.8%, not the −5.9% the group prints — and found the same is true of the constant-rate figure, so it is not a one-off in a single bracket. Both stated percentages reconcile only against a restated prior-year base, which the group does not publish as a figure. The group's own footnote explains why: an eyewear business was transferred at the end of 2024 and the comparatives exclude it. The operating-profit line, by contrast, ties exactly between the two releases.", "action": "Published as a disclosed non-reconciliation under Part 29.6 rather than smoothed over, with the group's own footnote quoted as the explanation, and with an explicit warning that the two published net-sales figures cannot be described as a 5.9% decline. The restated base itself is NOT published: it is a derived number, and this Codex does not publish estimates. Recorded separately: the group's Annual Report 2025, which would carry the audited statements and the segment note, is refused at the robots.txt stage and could not be read at all — the first dossier in this series whose parent's audited accounts were unavailable, which the page states plainly rather than implying a completeness it does not have.", "sources": [] }, { "id": "corr-2026-09-28-033", "date": "2026-09-28", "date_ist": "2026-09-28 1:06 PM IST", "type": "mis-citation-fixed", "affected": [ "/watches/legacy/#not-published" ], "claim_as_drafted": "The Legacy page's resolved-claim bullet for Buzz Aldrin's watch — quoting his own words that his watch “was the only one out on the surface” and that it “turned up missing” after he shipped it to the Smithsonian — cited source 14, the Apollo Lunar Surface Journal's Omega-chronograph history page.", "finding": "Session 18's link-and-citation sweep re-fetched source 14 directly. It does not contain Aldrin's quoted words at all. It gives a different, third-person account from its own authors: the watch “was lost in transit in or about 1971 whilst en route to the Smithsonian Air and Space Museum whilst Buzz was attempting to loan the item for display.” The actual source of Aldrin's own words, already correctly cited on the Omega dossier as source 30, is a different ALSJ page — the “EVA Preparations” transcript, read 2026-09-28. The Legacy bullet had reused the wrong source id for a claim that had moved to a different page during drafting. Also found: the Sources section's blanket “All read 26 September 2026” line was false for the sources added or re-verified in Sessions 15–17 (28 September), including this one.", "action": "Added source 18 (the EVA Preparations transcript, correctly dated) and repointed the Aldrin quotation to it. The bullet now also discloses, briefly, that the Journal's own Omega-history page (source 14) gives a materially different account of the same loss — lost in transit while being loaned, not shipped and turning up missing — from its authors rather than from Aldrin, published alongside rather than adjudicated. Source 14 remains correctly cited elsewhere on the page for the claim it does support (Armstrong leaving his watch in the Lunar Module). The false blanket read-date line was removed and replaced with a statement that each citation carries its own read date.", "sources": [ { "tier": 1, "citation": "Apollo Lunar Surface Journal, “EVA Preparations”, Apollo 11, NASA-hosted, corrected transcript and commentary © 1995 Eric M. Jones", "url": "https://www.nasa.gov/wp-content/uploads/static/history/alsj/a11/a11.evaprep.html", "accessed": "2026-09-28" }, { "tier": 1, "citation": "Omega Speedmaster Professional Chronographs, Apollo Lunar Surface Journal, NASA-hosted, Lee Bailham and Eric M. Jones", "url": "https://www.nasa.gov/wp-content/uploads/static/history/alsj/omega.html", "accessed": "2026-09-28" } ] }, { "id": "corr-2026-09-28-034", "date": "2026-09-28", "date_ist": "2026-09-28 1:06 PM IST", "type": "citation-clarity-fixed", "affected": [ "/watches/brands/omega/#deep", "/watches/brands/omega/#channels" ], "claim_as_drafted": "Two Omega dossier Sources-list entries (source 29, supporting ‘the museum holds no Aldrin Apollo 11 chronograph’; source 45, supporting ‘Omega operates no India locale’) each linked as their clickable ‘source’ a URL that returns HTTP 404 — a guessed Smithsonian object-record slug for the first, omegawatches.com/en-in/ for the second.", "finding": "Session 18's Gate 2 link-rot sweep flagged both as DEAD links. Both are dead by design: the drafting session deliberately fetched these exact URLs and recorded that they 404, using the absence itself as the evidence for a scoped negative claim (Part 31's ‘scope a negative to what was checked’ rule, applied correctly). But the published citation gave no sign of this — it read like an ordinary ‘source’ link, so a reader clicking through, or a future link-rot check, would see a plain 404 and reasonably conclude the citation had rotted, not that its absence was the point.", "action": "Rewrote both Sources-list entries to say plainly that the linked URL returns HTTP 404 and that this is cited as the evidence of absence, not as a working page. Source 29 now also names the sibling record types that were checked and links one of those (a live page) as an example of the record type, instead of the 404 itself, so the entry has a clickable working link alongside the documented negative. No change to either underlying claim: both scoped negatives still stand exactly as researched.", "sources": [] }, { "id": "corr-2026-09-28-035", "date": "2026-09-28", "date_ist": "2026-09-28 1:06 PM IST", "type": "stale-count-fixed", "affected": [ "/watches/#hubs" ], "claim_as_drafted": "The home page's country-hub cards showed a ‘brand pages live’ chip for Japan (kept current across Sessions 15 and 16 as Citizen and Casio shipped) but no equivalent chip for Switzerland, even though Session 17 gave Switzerland its own live brand page (Omega).", "finding": "Gate 6's hub-count sweep compared every country card against the brand sidecar. Japan's chip was correctly maintained at each brand session; Switzerland's card was not given the same treatment when Omega shipped, so a reader comparing the two cards would see Japan advertise its live coverage and Switzerland advertise none, though one Swiss dossier already exists.", "action": "Added a ‘1 brand page live’ chip to the Switzerland card, in the same format as Japan's. No other hub card is affected: no other country has a live dossier yet.", "sources": [] }, { "id": "corr-2026-09-28-036", "date": "2026-09-28", "date_ist": "2026-09-28 3:20 PM IST", "type": "mis-citation-fixed", "affected": [ "/watches/brands/seiko/#astron", "/watches/brands/seiko/#timeline" ], "claim_as_drafted": "Two sentences on the Seiko dossier, one under the Astron section, one in the timeline, stated that the Astron's 450,000-yen launch price was ‘about the price of a popular car at the time’, both citing source 9 (the Seiko Museum's Astron collection page, collect020).", "finding": "Gate 4's independent re-verification fetched source 9 directly and twice found no price or car comparison anywhere on that page. A follow-up search located the actual quotation, ‘The price was 450,000 yen, around the same price as some of the popular cars of the time’, on a different Seiko Museum page already cited elsewhere in the same dossier as source 10 (‘The birth of the quartz timepiece’). Both instances had been drafted against the wrong museum page.", "action": "Both citations repointed from source 9 to source 10, with the citation title text updated to match. Source 9 remains correctly cited elsewhere on the page for what it does support (the Astron object description). No change to the underlying fact, which is correct, only the citation was wrong.", "sources": [ { "tier": 1, "citation": "The birth of the quartz timepiece, The Seiko Museum Ginza", "url": "https://museum.seiko.co.jp/en/knowledge/Quartz01/", "accessed": "2026-09-27" } ] }, { "id": "corr-2026-09-28-037", "date": "2026-09-28", "date_ist": "2026-09-28 3:20 PM IST", "type": "citation-clarity-fixed", "affected": [ "/watches/brands/seiko/#museums" ], "claim_as_drafted": "The Seiko dossier's museums section gave the Seiko Museum Ginza's street address, ‘Ginza 4-3-13, Chuo-ku, Tokyo 104-0061’, citing source 17, which combines the museum's ‘Visiting’ and ‘About Us’ pages.", "finding": "Gate 4's independent re-verification fetched both URLs behind source 17 and found neither contains a street address: ‘Visiting’ gives opening hours and admission terms, and ‘About Us’ describes the museum's mission without a postal address. The address text does appear, verified by a further fetch, on the museum's separate ‘Contact’ page, which the dossier had not cited at all.", "action": "Added a new source 23 for the Contact page and repointed the address sentence's citation to it. Source 17 is unchanged and remains correctly cited for the opening-hours and mission-statement sentences it does support.", "sources": [ { "tier": 1, "citation": "Contact, The Seiko Museum Ginza", "url": "https://museum.seiko.co.jp/en/contact/", "accessed": "2026-09-28" } ] }, { "id": "corr-2026-09-28-038", "date": "2026-09-28", "date_ist": "2026-09-28 3:20 PM IST", "type": "citation-scope-narrowed", "affected": [ "/watches/brands/seiko/#what-to-know" ], "claim_as_drafted": "‘Seiko still makes watches at every level, from a few thousand rupees to genuinely expensive, and it is one of very few brands that makes both mechanical and quartz movements itself’ was cited as a single sentence to source 4, Seiko Watch Corporation's own Corporate Profile page.", "finding": "Gate 4's independent re-verification fetched source 4 directly. It confirms only that the company ‘designs and develops its own movements’; it says nothing about price range, rupee figures, or being one of few brands making both movement types, a comparative claim across the industry that no single company page could establish in any case.", "action": "The sentence was split: the price-range observation is now presented as general market context rather than a sourced claim, the own-movements fact stays cited to source 4 as it is supported, and the both-mechanical-and-quartz point is now grounded in the dossier's own following sentences about Spring Drive rather than carried by source 4's citation. No fact changed, the citation no longer claims more than its source states.", "sources": [] }, { "id": "corr-2026-09-28-039", "date": "2026-09-28", "date_ist": "2026-09-28 6:04 PM IST", "type": "drafting-defect-caught-before-publication", "affected": [ "/watches/brands/rolex/#certification, before the page was built" ], "claim_as_drafted": "A drafted sentence said Rolex's Superlative Chronometer design/manufacturing-stage criteria (magnetism resistance, reliability, sustainability) were ‘new as of 2026’, implying the certification page's own sentence carried that date.", "finding": "Two independent verification agents, each shown only a claim and its single source URL, checked all nine claims drafted for the Rolex dossier before it was built. Eight were confirmed EXACT against Rolex's own pages and its 2023 Sustainability Report, including the Hans Wilsdorf Foundation ownership statement, the 15,548-employee figure, the Oyster and Perpetual rotor claims, and the 1905/1919/1920 founding dates. One came back CLOSE: the certification page's own sentence says the new criteria are ‘now’ implemented, not ‘new as of 2026’, the 2026 date exists only in the page's own URL and section title (“New watches 2026”), not in the quoted sentence itself. Also confirmed by the same check: the certification page attributes the finished-watch criteria (precision, waterproofness, self-winding, power reserve) to the certification's 2015 redefinition, and gives no certifying-body name or numeric precision tolerance, both of which appear only in a separately dated 2016 Rolex release hosted on fhs.swiss, cited separately and labelled as such.", "action": "Rewritten before publication to attribute the 2026 date to the page's own section title rather than to the sentence describing the new criteria, and to state plainly that the finished-watch criteria are the page's own 2015 material. No page was ever published with the imprecise version. Logged as this Codex's practice is to log a drafting defect caught by verification whether or not it reached a reader, per the precedent set for the Omega dossier at Session 17 (corr-031).", "sources": [ { "tier": 1, "citation": "New Superlative Chronometer certification, Rolex", "url": "https://www.rolex.com/en-us/watches/new-watches/superlative-chronometer", "accessed": "2026-09-28" } ] }, { "id": "corr-2026-09-29-040", "date": "2026-09-29", "date_ist": "2026-09-29 8:40 AM IST", "type": "drafting-defects-caught-before-publication", "affected": [ "/watches/brands/titan/, the Titan dossier, before it was built" ], "claim_as_drafted": "Of roughly 45 assertions drawn for the Titan dossier, 11 needed correction before the page was built. None was a fabrication, a wrong attribution or a figure attached to the wrong period, every number checked to the digit, including the 27.88%/20.84% shareholding split, three published MRPs, the segment figures and the 2024 liquidation date.", "finding": "Three verification agents, each shown only a claim and the single URL it was attributed to, checked all of them and were additionally asked to name the publisher of every page fetched. The defects were: (1) a quotation of TIDCO's statutory purpose cropped before the qualifier ‘in association with the private sector’ with no ellipsis, precisely the truncation class Part 31.4 exists to catch, and the dropped words describe the joint-venture model the whole dossier is about; (2) a source labelled as a 2025 AGM notice which is in fact the 41st Annual Report 2024-25, the filename having misled the drafting; (3) the incorporation date of 26 July 1984 and the original name ‘Titan Watches Limited’ resting on a single page whose Titan ownership could not be established from the page itself, no copyright line, registered-office line or company number on it; (4) the 2011 Favre Leuba filing quoted as ‘by asset purchase mode’ where it prints ‘on an asset purchase mode’, and its stated purpose cropped before ‘with a Swiss heritage brand’; (5) the 2021 scale-down quotation cropped before the covid-19 reason Titan itself gave and the ‘year ended 31 March 2021’ frame; (6) the subsidiaries page quoted as ‘formerly Favre Leuba AG’ where it prints ‘earlier known as’; (7) the subsidiary's accounts printing the name hyphenated as ‘Favre-Leuba AG’; (8) the ₹5,267 crore Watches figure described as ‘revenue’ where the audited table heads the column ‘Revenue (including other income)’; (9) ₹88,136 crore less ₹81,235 crore leaving ₹6,902 crore against a printed profit before tax of ₹6,801 crore, unexplained until the statement's own ₹101 crore exceptional item is named; (10) ‘755 Stores’ published without its own footnote marker or its March 2026 date; (11) a scoped negative about Favre Leuba's absence that counted two documents, the FY26 annual report and the FY26 BRSR, which neither agent could read in full.", "action": "All eleven corrected before the page was built. The TIDCO sentence is now quoted whole, and the recovered qualifier is used rather than merely restored. The mislabelled source is cited as the annual report it is, with a note that its filename suggests otherwise. The thin-provenance citation is published with its weakness stated in a callout and in the source entry, alongside the observation that the company's own registry number independently corroborates 1984 and Tamil Nadu but not the day or the former name. Both Favre Leuba quotations now carry their full wording and the covid-19 reason. The ₹5,267 crore figure is no longer called revenue anywhere, and a callout gives both of Titan's own labels for it. The exceptional item is published as a disclosed reconciliation under Part 29.6. The store count carries its date. The scoped negative now names the four pages checked in full and explicitly excludes the two that could not be, with the read boundaries stated.", "sources": [] }, { "id": "corr-2026-09-29-041", "date": "2026-09-29", "date_ist": "2026-09-29 8:40 AM IST", "type": "published-figure-relabelled", "affected": [ "/watches/money/#brands, the Titan block, live since Session 11" ], "claim_as_drafted": "The money page stated: ‘Its Watches segment reported revenue of ₹5,267 crore against ₹4,598 crore’, and separately gave consolidated revenue ₹87,584 crore, total expenses ₹81,235 crore, profit before tax ₹6,801 crore and profit ₹5,073 crore as a sequence a reader could be expected to follow.", "finding": "Building the Titan dossier sent the same audited statement to independent verification, and two defects surfaced in the ALREADY-PUBLISHED money page rather than in the new draft. First, ₹5,267 crore is not the segment's revenue: the audited segment table heads that column ‘Revenue (including other income)’, and Titan's own presentation calls the figure total income. Calling it ‘revenue’ was more precise-sounding than the source allows, and it is exactly the error this Codex warns about elsewhere on that same page under ‘different things counted’. Second, the four group figures as published cannot be reconciled: total income ₹88,136 crore less expenses ₹81,235 crore leaves ₹6,902 crore, not the ₹6,801 crore printed as profit before tax. The missing line is an exceptional item of ₹101 crore, which the statement itself names as the incremental impact of the Labour Code changes. A reader doing the arithmetic would have found a hundred crore missing and no explanation on the page.", "action": "The word ‘revenue’ removed from that sentence, the column's own heading quoted, and the money page now says plainly that it carried the looser label until the dossier's verification caught it. The exceptional item added with its own note and figure, plus the observation that the prior year carries no such item so the two profit-before-tax lines are not like for like. Both fixes are also carried on the new dossier. Recorded because a figure that is right but labelled more precisely than its source supports is a defect of the same family as a wrong number, and because this is the first time verification for a new page has corrected an older one.", "sources": [ { "tier": 1, "citation": "Titan Company Limited, Statement of Consolidated and Standalone Audited Financial Results for the Quarter and Year Ended 31 March 2026", "url": "https://www.titancompany.in/sites/default/files/2026-05/Q4FY26%20-%20Financial%20results%20Uploaded.pdf", "accessed": "2026-09-29" } ] }, { "id": "corr-2026-09-29-042", "date": "2026-09-29", "date_ist": "2026-09-29 8:40 AM IST", "type": "build-defect-corrected", "affected": [ "/watches/standards/, 10 occurrences, live since Session 14", "/watches/brands/omega/, 1 occurrence, live since Session 17", "/watches/brands/rolex/, 1 occurrence, live since Session 19", "/watches/brands/titan/, 8 occurrences, caught before publication", "The index.md and index.json twin of each of those pages, and llms-full.txt" ], "claim_as_drafted": "In the text twins, a quotation marked up with the element and followed directly by a word welded shut against it: ‘the overall Titan ecosystem”and’, ‘“+14.5%”and’, ‘“Advertising”as its own named’, ‘“7 service centres”under the Watches’. Twenty occurrences across four pages.", "action": "Fixed at the source: the extractor's end-tag handler now emits its closing mark WITH a trailing space, and text() collapses any doubling and still removes a space before following punctuation, so quotations separate correctly whatever follows them. Verified by grep across every built index.md, twenty occurrences before, zero after, no double spaces introduced, and the one remaining match is a German ‚…“ pair on the Standards page where the space is correct as printed. Validator unchanged at 4,471 checks, 0 warnings, 0 failures.", "finding": "Found by grepping the new dossier's built index.md before publication, then widening the grep site-wide, which turned a drafting check into the discovery of a defect live on the Standards page for six sessions. THE PART WORTH RECORDING IS WHY IT SURVIVED SESSION 18's QC PASS: that pass swept every built index.md for the spacing families and reported zero instances site-wide, and it was right about the families it tested, the straight ASCII quote and the punctuation marks. It did not test the CURLY closing quote, which is the only mark the handler emits. So a class declared closed in Session 16 and confirmed clean in Session 18 was neither, in a glyph nobody had grepped for. This is the fifth session to meet one defect class: sentence punctuation (S10), the straight closing quote (S14), a narrowing of the opening-quote case (S15), the parentheses (S16), and now the element's own curly mark. The previous four were all regex rules in text(); this one was in the parser code, which is why extending the regexes never reached it. The rule is now stated in the code as a general one, any handler injecting a character the HTML did not contain also injects the whitespace around it, rather than as another glyph added to a list.", "sources": [] }, { "id": "corr-2026-09-29-043", "date": "2026-09-29", "date_ist": "2026-09-29 9:45 AM IST", "type": "live-metadata-corrected", "affected": [ "llms.txt — the generated-on and Last-Updated lines", "sitemap.xml — every lastmod", "Every page's JSON-LD datePublished and dateModified", "Every page's entry-last-reviewed meta", "academy-corrections.json — thirteen source read dates" ], "claim_as_drafted": "One constant, BUILD_DATE, hardcoded to “2026-09-26”, served three incompatible purposes: when the build ran, when a page was published, and when a cited source was read.", "finding": "The site went live at thecodex.expert/watches/ on 29 September 2026 — the first deployment in the project's history. A verification pass against the live URLs found the constant had become three public untruths at once: llms.txt told crawlers it was “GENERATED by build.py on 2026-09-26” when it was generated on the 29th, every sitemap lastmod and JSON-LD dateModified said the same, and every page's JSON-LD claimed it was FIRST PUBLISHED on a date before the site existed publicly. The dangerous part was the third use: thirteen “accessed” fields in the PUBLIC corrections log carried the same constant as the date a source was read. A single dynamic date would have fixed the first two problems and silently re-dated every one of those citations to whenever the site was last rebuilt — turning a record of evidence into a record of the last build.", "action": "Split into three constants with three different lifetimes. BUILD_DATE is now DERIVED from the clock at build time, so it cannot go stale again, and is used only for what should move on a rebuild: dateModified, sitemap lastmod, llms.txt, and the per-page last-reviewed stamp. FIRST_PUBLISHED is frozen at 2026-09-29, the day the site actually became publicly reachable — which is the honest value of schema.org datePublished for every page, including those drafted earlier, because a zip on a workstation is not publication. READ_DATE_EARLY is frozen at 2026-09-26 for the thirteen citation read dates, because a read date is a fact about the past. Verified after the fix: llms.txt and sitemap now carry 2026-09-29, and the corrections log's read dates still span 25 to 29 September at their real values.", "sources": [] }, { "id": "corr-2026-09-29-044", "date": "2026-09-29", "date_ist": "2026-09-29 9:45 AM IST", "type": "live-claim-scope-corrected", "affected": [ "/watches/brands/omega/#certification — live since Session 17" ], "claim_as_drafted": "The Omega dossier explained METAS “Master Chronometer” certification at length, quoting the regulator's own requirements document, and never stated which of Omega's movements carry it. A reader had no way to learn that it does not cover the whole mechanical range.", "finding": "The owner said the Omega page “feels less right” and asked what was missing. The audit that followed found the certification section was not merely incomplete but capable of misleading, and Omega's own pages say so three separate ways. Its certification page states the brand has subjected “almost all” of its mechanical timepieces to the process — not all. Calibre 321, which Omega itself describes as “The calibre that was worn on the Moon”, is in the current catalogue with no mention of “Master Chronometer” or “METAS” anywhere on its page. And most consequentially, Omega operates TWO designations distinguished only by word order: “Co-Axial Master Chronometer” is METAS-certified, while “Master Co-Axial Chronometer” — calibre 8511 in a De Ville Trésor — is certified only to ISO 3159 via COSC, with neither “Master Chronometer” nor “METAS” on the page. The Trésor exists in both versions, under titles that differ by that transposition alone.", "action": "Added a #calibres section naming the movements that hold the certification (3861, 8800, 8900, 9908) with Omega's exact wording, and setting the two word-order designations against each other in a callout, because reproducing either phrase carelessly would convert a COSC claim into a METAS one or the reverse. Omega's own “almost all” hedge is now quoted. The page states plainly that having the Co-Axial escapement does not entail the certification — calibre 8511 has the escapement without it, and the first Co-Axial watch of 1999 predates the certification by sixteen years. What Omega does NOT publish is any sentence naming a calibre as uncertified, so the page describes a documented pattern of omission rather than an explicit exclusion.", "sources": [ { "tier": 1, "citation": "OMEGA, The Master Chronometer Certification", "url": "https://www.omegawatches.com/world-of-omega/watchmaking/the-master-chronometer-certification", "accessed": "2026-09-29" }, { "tier": 1, "citation": "OMEGA, De Ville Trésor Master Co-Axial Chronometer 40 mm, ref. 432.53.40.21.04.001", "url": "https://www.omegawatches.com/en-us/watch-omega-de-ville-tresor-master-co-axial-chronometer-40-mm-43253402104001", "accessed": "2026-09-29" }, { "tier": 1, "citation": "OMEGA, Speedmaster Calibre 321 Chronograph 38.6 mm, ref. 311.50.39.30.01.001", "url": "https://www.omegawatches.com/en-us/watch-omega-speedmaster-calibre-321-chronograph-38-6-mm-31150393001001", "accessed": "2026-09-29" } ] }, { "id": "corr-2026-09-29-045", "date": "2026-09-29", "date_ist": "2026-09-29 9:45 AM IST", "type": "structural-defect-corrected", "affected": [ "/watches/brands/omega/ — rebuilt, live since Session 17", "All six brand dossiers — now audited on every build", "/watches/brands/seiko/ — a non-standard section anchor" ], "claim_as_drafted": "Six brand dossiers had drifted into six different shapes. The Omega page mentioned the Speedmaster eighteen times against two mentions of the Seamaster, which Omega itself lists FIRST; carried no manufacturing section at all, on a site whose central legal subject is what Swiss Made requires; never used the word “quartz” despite the quartz crisis being the reason it belongs to the Swatch Group; and did not mention the Olympic Games once, though Omega has been Official Timekeeper since 1932 and both Omega and the IOC document it.", "finding": "The owner's instinct that the page “feels less right” was correct, and measuring it showed the cause was systemic rather than Omega-specific: nothing in the spec said what a dossier must contain, so each page took the shape of whatever its session was chasing. Omega got the Moon because Session 17 was chasing the Apollo 11 claim. The audit also found that every existing brand-page check in the validator tested ONE page, brands/seiko, and therefore said nothing about the other five. Across the six, #where-made was missing from four, #prices from five, #collections from four, #certification from four. Seiko carried a #gaps anchor where every other page used #gaps-exploring, so no cross-page convention could hold.", "action": "Added Part 32, the dossier completeness standard, and made it machine-checked on every build across EVERY brand page rather than one: eleven spine sections as critical failures, five required-or-explain sections as tracked warnings, and a proportionality check that compares each dossier's coverage of the collections the brand itself names and warns when the most-covered outruns the least by more than eight times. That check independently reproduced the owner's instinct, flagging Omega at 18x, and also caught Seiko at 9x, which nobody had noticed. Omega was then rebuilt: 5,991 to 10,064 words, 46 to 68 sources, adding #the-quartz-crisis, #calibres, #where-made, #prices, #olympics and #chronology, and rebalancing the collections so the Moon material no longer stands in for the whole company. Omega now clears the rubric entirely and its proportionality warning is gone. Seiko's anchor was renamed to the site convention. The remaining 16 gaps across the other five dossiers are published as warnings on every build rather than tracked in prose, so the backlog cannot be quietly forgotten: it closes when those warnings are promoted to failures.", "sources": [] }, { "id": "corr-2026-09-29-046", "date": "2026-09-29", "date_ist": "2026-09-29 10:20 AM IST", "type": "build-defect-corrected", "affected": [ "/watches/how-it-works/, which word carries the glossary link, live since Session 09", "17 glossary entries whose term and an alias are the same length, same latent risk, most never actually co-occurring in one page's prose", "The full site, build reproducibility in general" ], "claim_as_drafted": "No single sentence was misquoted. The defect is that two separate builds of the identical source files could autolink two different words to the same glossary entry, with no code change and no content change between them, found only because this session's build.py reconstruction (see corr-2026-09-29-043 through -045) was verified by running the generator twice and diffing the output, and the two runs disagreed on one page.", "finding": "On How it works, the sentence “it has a name: isochronism” is followed later by a quotation containing “isochronous”, and both words are declared forms of the same glossary entry (term “Isochronism”, alias “isochronous”, both eleven characters). The autolinker builds its term list with `sorted(set(names), key=len, reverse=True)`. Sorting by length is fine, but feeding a `set` into it means that when two names TIE in length, which one sorts first depends on Python's per-process string-hash randomisation, not on anything in the source data, so which of the two words actually receives the link to /watches/glossary/isochronism/ could flip on every single rebuild. A search of glossary.json found sixteen more entries with the same latent tie (e.g. atomic-clock/“Atomic clock” vs “cesium clock”, gasket/“Gasket” vs “O-ring”, swiss-made/“Swiss made” vs “Swiss Made”); a full diff of two fresh builds against each other found How it works was the only page on the live site where both tied forms of one entry actually appear together in one page's prose today, so this is the only place the flip was ever visible, but the mechanism was latent everywhere and would have started silently relocating links on future dossiers the moment two tied forms of the same entry appeared together in new prose.", "action": "Replaced `sorted(set(names), key=len, reverse=True)` with `sorted(list(dict.fromkeys(names)), key=len, reverse=True)`, which de-duplicates without discarding declared order, so Python's stable sort now always breaks a length tie the same way: the entry's own term before its aliases, aliases in the order glossary.json lists them. On How it works this permanently settles the link onto “isochronism”, the term itself, rather than the alias buried inside a quotation. Verified by building the site five times in a row and diffing every run against every other, zero differences, where before the fix one page out of ninety-six differed roughly one run in four. Recorded as a build defect rather than fixed silently because it changed which word carries a live hyperlink on a published page, even though the destination and the underlying fact were never wrong.", "sources": [] }, { "id": "corr-2026-09-29-047", "date": "2026-09-29", "date_ist": "2026-09-29 1:15 PM IST", "type": "drafting-defects-caught-before-publication", "affected": [ "/watches/brands/longines/, the Longines dossier, before it was built" ], "claim_as_drafted": "Of the roughly twenty sourced claims drafted for the Longines dossier, three needed correction before the page was built, caught by independent verification agents shown only a claim sentence and its source URL.", "finding": "First, a draft sentence claimed FEI (the international equestrian federation) names Longines “Official Timekeeper” of FEI equestrian sport generally. The FEI page names Longines Official Timekeeper of specific series and events, the Nations Cup Jumping series, the three World Cup Leagues, and never of FEI or the sport as a whole; a forward-looking 2016 sentence about timing the 2018 and 2022 World Equestrian Games was also about to be published as a confirmed title rather than a stated intention. Second, a draft quotation reproduced Longines' own page text, “LONGINES watches name were adorned with a logo”, by silently deleting the site's own stray word “name” and inserting “[were]” without marking the edit, which would have presented a smoothed sentence as an exact quotation. Third, a price citation for the Record collection pointed at the international (non-US) product page while quoting the US-locale page's USD prices; the two pages show the same references at different amounts in different currencies (e.g. ref. L2.821.4.11.6: CHF2,150.00 internationally against $2,450.00 on the US-locale page), so the citation would have pointed a reader to a page that does not show the cited figure.", "action": "The FEI sentence was rewritten to quote only the specific titles the page actually grants, with a callout stating plainly that “Official Timekeeper of FEI” would overstate the source and that the 2018/2022 Games mention is a stated 2016 intention, not a confirmed title. The typo'd quotation was replaced with a paraphrase plus two independently verified facts from a second Longines page (the 1866 land purchase date, and the brand's own “as early as 1867” claim for the mark's first use) rather than repeating the flawed quote with a silent edit. The price citation now points at the US-locale page for the USD figures actually quoted, with the CHF figures from the international page recorded alongside as a currency difference, not a discrepancy. All three fixed before the page was built.", "sources": [] }, { "id": "corr-2026-09-29-048", "date": "2026-09-29", "date_ist": "2026-09-29 1:45 PM IST", "type": "build-defect-corrected", "affected": [ "/watches/brands/longines/, the Longines dossier, before its first build" ], "claim_as_drafted": "The Longines source fragment, wired into build.py for its first build, contained two HTML syntax defects: a stray extra closing tag in the chronology table's 1983-registry row, and a missing closing tag for the page's outermost content wrapper, left open from the header through the end of the file.", "finding": "validate.py's tag-balance check (§1), run before the first zip as required, failed twice in sequence while wiring the dossier in: first “ at line 311 was never opened”, tracing to a chronology-table cell that closed a citation superscript with an extra, unmatched before its own ; then, once that was fixed, “ closed while ['div'] still open”, tracing to the page's
wrapper (opened directly after the header, immediately before the on-page rail and
) never being closed at the end of the file, every other brand dossier closes it with
immediately after , and this one did not.", "action": "Removed the stray from the chronology table's Moneyhouse citation, and added the missing immediately after . Neither defect had been published: this was the dossier's first build, so no live page was ever malformed. Logged anyway, per the standing rule that a real defect is logged whether or not it ever reached a published page.", "sources": [] }, { "id": "corr-2026-09-29-049", "date": "2026-09-29", "date_ist": "2026-09-29 5:31 PM IST", "type": "gate-4-citation-scope-error", "affected": [ "/watches/legacy/, the Longines trademark section" ], "claim_as_drafted": "The page stated that the International Registry the Longines mark belongs to, “the Madrid System,” itself “only began in 1893,” citing WIPO's “125 Years of the Madrid System” as the source.", "finding": "Session 25's QC pass (Part 27, Gate 4) drew this sentence in its independent-verification sample and found the cited page does not support it: WIPO's own article states its founding treaty, the Madrid Agreement, “was adopted in Madrid on this day in 1891,” 125 years before the article's 2016 publication. 1893 is not the Madrid System's start date at all; it is the year WIPO gives for Longines' own mark being registered under that already-existing system. The sentence had conflated the object being dated (Longines' registration) with the system's own founding date, both cited to the same source that in fact distinguishes them.", "action": "Rewritten to state plainly that the Madrid System itself dates to 1891, the year its founding treaty was adopted, and that 1893 is when Longines' own mark was registered under it — both facts drawn from the same cited page, now correctly attributed. A separate link-rot check of the same URL in this pass returned a 404 on one fetch attempt; a direct re-fetch immediately after loaded the page normally with the content quoted above, so this is recorded as a transient failure, not a dead link, and no source was changed.", "sources": [] }, { "id": "corr-2026-09-29-050", "date": "2026-09-29", "date_ist": "2026-09-29 5:31 PM IST", "type": "gate-4-citation-scope-error", "affected": [ "/watches/brands/omega/, the SSIH/ASUAG merger section and chronology table" ], "claim_as_drafted": "The page stated that SSIH and ASUAG merged in 1983 “to form ASUAG-SSIH,” citing the Historical Dictionary of Switzerland (HLS) for that exact name, in both the callout prose and the chronology table.", "finding": "Session 25's QC pass (Part 27, Gate 4) checked this sentence against its cited HLS source and found HLS does not use that word order: its own text names the merged entity “SSIH-Asuag” (“...51% des Aktienkapitals der SSIH-Asuag”), the reverse of what was published. “ASUAG-SSIH” is a real name, but it belongs to a different source: the Swatch Group's own history page uses that order in its 1985 entry (“ASUAG-SSIH is renamed SMH”). The page had attributed one source's wording to the other.", "action": "Rewritten so each name-order is attributed to the source that actually uses it — ASUAG-SSIH to the Swatch Group's own page, SSIH-Asuag to HLS — with a sentence stating plainly that the two independent sources disagree on word order and that neither is preferred, per this Codex's standing rule for sources that conflict. The chronology table row was updated to match. The 1983 merger date itself was already correctly sourced to HLS alone and is unaffected.", "sources": [] }, { "id": "corr-2026-09-29-051", "date": "2026-09-29", "date_ist": "2026-09-29 5:31 PM IST", "type": "citation-hygiene", "affected": [ "/watches/brands/citizen/, source list entry src-2" ], "claim_as_drafted": "Citizen's “About CITIZEN” citation pointed at https://www.citizenwatch-global.com/about/index.html.", "finding": "Session 25's QC pass (Part 27, Gate 2) found this exact URL returns 403 to an automated fetch, in two independent checks; the directory form of the same page, https://www.citizenwatch-global.com/about/, loads normally and carries the same cited content, confirmed directly. This is a server-side quirk affecting the specific file path, not evidence the page or its content has changed or gone missing.", "action": "Citation URL repointed to the directory form, which is reachable and serves the identical page. No factual claim changed; this is a presentation fix so a future link-rot check does not raise a false BLOCKED flag against a page that is, in substance, live and unchanged.", "sources": [] }, { "id": "corr-2026-09-29-052", "date": "2026-09-29", "date_ist": "2026-09-29 6:00 PM IST", "type": "mobile-layout-defect", "affected": [ "/watches/brands/longines/", "/watches/brands/omega/", "/watches/brands/rolex/", "/watches/brands/titan/", "each dossier's chronology table; also Titan's shareholding-pattern and MRP price tables" ], "claim_as_drafted": "No claim was misstated; this is a presentation defect, not a factual one. These four dossiers' wide tables were published as bare elements.", "finding": "Session 25's QC pass (Part 27, Gate 5, browser pass Tier B) found horizontal page overflow at the 375px mobile breakpoint on four of the seven brand dossiers (Longines, Omega, Rolex, Titan), each caused by one or more
elements that had never been wrapped in this Codex's own established
mobile-scroll convention — the same convention Seiko, Casio, Citizen and the Standards page already apply to 100% of their tables. On a 375px viewport this forced the whole page to scroll sideways instead of just the table.", "action": "Wrapped every previously-bare
on the four affected pages in
...
: one table each on Longines and Rolex (chronology), and three on Titan (shareholding pattern, MRP price comparison, chronology), plus one on Omega (chronology). Re-ran the Tier B browser check afterward at 375px and confirmed zero remaining overflow on all four. No content or citation changed.", "sources": [] } ] }