GxPlex

Corrections

Every correction made to a published factual claim, with its date. Factual claims are never silently edited.

A tracker whose past statements can change without record cannot be cited: a reader who quoted it last month has no way to know whether the quotation still stands. Corrections are therefore additive and permanent.

Corrections, most recent first
DateSubject CorrectionCategory AffectedSource
record-tier (every document on this axis) correction: the core label said this archive undertakes to put an entry's status to a person. The closure rules now run nightly, and the label states what they undertake. was “Filed core. For that label this archive undertakes to put an entry's status to a person.” → “Filed core. For that label this archive undertakes to close an entry's open status question by rule on a clock, with its reason, and to re-read a ratification on record against its observation every night.” none: no part of the rule holds for this entry’s fields
109 routes recorded
 link (opens in new tab) · retrieved
fda-ai-guidance correction: the subject quotation on this row ran to 15 words, one over the cap, and stopped one word short of the title. A shorter span was selected. was “a 15-word quotation of the title, over the 14-word cap, cut before its last word” → “a 13-word whole-word span of the same title, ending where the title ends” wording
row named: /r/fda-ai-guidance/
 link (opens in new tab) · retrieved
nist-csf correction: this entry's page marked its text rung held on prose that came only from records of another page. No record of this document yielded prose. was “the evidence ladder's text rung marked held: prose was extracted from the retrieved bytes, so a quotation from this document can be checked against what this record retained” → “the text rung marked nothing held: the prose this record retains for this entry was extracted only from records classified as something other than the document it names” identity/URL
row named: /r/nist-csf/
 link (opens in new tab) · retrieved
fda-device-cybersecurity correction: this entry's page marked its text and diff rungs read, and not there, from bytes of another document. A read of another document measures no absence of this one. was “the evidence ladder's text and diff rungs marked read, and not there, the mark the legend defines as an absence this record measured” → “both rungs marked nothing held, an absence of this record's reach: bytes of another document are no reading of this one, so they measure no absence of its text or of a change” identity/URL
row named: /r/fda-device-cybersecurity/
 link (opens in new tab) · retrieved
verification (every document on this axis) correction: 16 rows whose status is unknown were marked awaiting, not yet checked: a check worded as still to come. Retired; they read unresolved, with their status-term rule outcome. was “awaiting — not yet checked” → “unresolved — no status recorded, on a row that is re-checked and records no person's search of the source; each row states what its status-term rule found” wording
143 routes recorded
 link (opens in new tab) · retrieved
iso-81001-1 correction: from 2026-09-07 this entry's page marked its text rung held in part, though no request for it has returned bytes with a digest. Nothing of the page or the text is held. was “the evidence ladder's text rung marked held in part, beside the sentence that what was retrieved is the publisher's own metadata page” → “the text rung marked nothing held, an absence of this record's reach: no request for the metadata page has returned bytes with a digest, so nothing of the page or of the text is held” presentation
row named: /r/iso-81001-1/
 link (opens in new tab) · retrieved
eu-ai-act correction: /how/ says each tracked document is read every day. Counted from 2026-08-17 to any day up to 2026-10-02, fewer than half of this entry's days were digest-bearing; 11 of 47 in all. was “Each one is fetched, read and dated every day.” → “digest-bearing on 11 of the 47 days from 2026-08-17 to 2026-10-02, fewer than half counted from 2026-08-17 to any day in it; on the other 36 days every observation answered HTTP 202 (35 days) or HTTP 403 (1 day)” none: no part of the rule holds for this entry’s fields
row named: /r/eu-ai-act/
 link (opens in new tab) · retrieved
eu-cyber-resilience-act correction: /how/ says each tracked document is read every day. Counted from 2026-08-17 to any day up to 2026-10-02, fewer than half of this entry's days were digest-bearing; 11 of 47 in all. was “Each one is fetched, read and dated every day.” → “digest-bearing on 11 of the 47 days from 2026-08-17 to 2026-10-02, fewer than half counted from 2026-08-17 to any day in it; on the other 36 days every observation answered HTTP 202 (35 days) or HTTP 403 (1 day)” none: no part of the rule holds for this entry’s fields
row named: /r/eu-cyber-resilience-act/
 link (opens in new tab) · retrieved
eu-cybersecurity-act correction: /how/ says each tracked document is read every day. Counted from 2026-08-17 to any day up to 2026-10-02, fewer than half of this entry's days were digest-bearing; 11 of 47 in all. was “Each one is fetched, read and dated every day.” → “digest-bearing on 11 of the 47 days from 2026-08-17 to 2026-10-02, fewer than half counted from 2026-08-17 to any day in it; on the other 36 days every observation answered HTTP 202 (35 days) or HTTP 403 (1 day)” none: no part of the rule holds for this entry’s fields
row named: /r/eu-cybersecurity-act/
 link (opens in new tab) · retrieved
eu-digital-omnibus-ai correction: /how/ says each tracked document is read every day. Counted from 2026-08-17 to any day up to 2026-10-02, fewer than half of this entry's days were digest-bearing; 11 of 47 in all. was “Each one is fetched, read and dated every day.” → “digest-bearing on 11 of the 47 days from 2026-08-17 to 2026-10-02, fewer than half counted from 2026-08-17 to any day in it; on the other 36 days every observation answered HTTP 202 (35 days) or HTTP 403 (1 day)” none: no part of the rule holds for this entry’s fields
row named: /r/eu-digital-omnibus-ai/
 link (opens in new tab) · retrieved
eu-ivdr correction: /how/ says each tracked document is read every day. Counted from 2026-08-17 to any day up to 2026-10-02, fewer than half of this entry's days were digest-bearing; 12 of 47 in all. was “Each one is fetched, read and dated every day.” → “digest-bearing on 12 of the 47 days from 2026-08-17 to 2026-10-02, fewer than half counted from 2026-08-17 to any day in it; on the other 35 days every observation answered HTTP 202 (34 days) or HTTP 403 (1 day)” none: no part of the rule holds for this entry’s fields
row named: /r/eu-ivdr/
 link (opens in new tab) · retrieved
eu-machinery-regulation correction: /how/ says each tracked document is read every day. Counted from 2026-08-17 to any day up to 2026-10-02, fewer than half of this entry's days were digest-bearing; 12 of 47 in all. was “Each one is fetched, read and dated every day.” → “digest-bearing on 12 of the 47 days from 2026-08-17 to 2026-10-02, fewer than half counted from 2026-08-17 to any day in it; on the other 35 days every observation answered HTTP 202 (34 days) or HTTP 403 (1 day)” none: no part of the rule holds for this entry’s fields
row named: /r/eu-machinery-regulation/
 link (opens in new tab) · retrieved
eu-mdr correction: /how/ says each tracked document is read every day. Counted from 2026-08-17 to any day up to 2026-10-02, fewer than half of this entry's days were digest-bearing; 12 of 47 in all. was “Each one is fetched, read and dated every day.” → “digest-bearing on 12 of the 47 days from 2026-08-17 to 2026-10-02, fewer than half counted from 2026-08-17 to any day in it; on the other 35 days every observation answered HTTP 202 (34 days) or HTTP 403 (1 day)” none: no part of the rule holds for this entry’s fields
row named: /r/eu-mdr/
 link (opens in new tab) · retrieved
eu-nis2 correction: /how/ says each tracked document is read every day. Counted from 2026-08-17 to any day up to 2026-10-02, fewer than half of this entry's days were digest-bearing; 12 of 47 in all. was “Each one is fetched, read and dated every day.” → “digest-bearing on 12 of the 47 days from 2026-08-17 to 2026-10-02, fewer than half counted from 2026-08-17 to any day in it; on the other 35 days every observation answered HTTP 202 (34 days) or HTTP 403 (1 day)” none: no part of the rule holds for this entry’s fields
row named: /r/eu-nis2/
 link (opens in new tab) · retrieved
status (every document on this axis) correction: 12 status entries that announced again a move already recorded, for four documents, were published as status changes on /changes, in the events API and on their own pages. was “16 status entries for four documents over four days, each published as this record's status field moving” → “4 transitions, and 12 entries announcing again a move already recorded, published as re-announcements and not as changes” wording · presentation
20 routes recorded
 link (opens in new tab) · retrieved
ich-e6r3 correction: this entry's address is a page listing guidelines, not the E6(R3) document, and none of its 157 records to 2026-09-28 holds a digest. The row's own account was empty. was “no account of why this entry's records hold no digest, beside fields declaring its address the E6(R3) document” → “an account: the address is a page listing guidelines, it answered a 687-byte page rendered client-side on 2026-09-28, and no reading of the document is held” identity/URL
row named: /r/ich-e6r3/
 link (opens in new tab) · retrieved
record-tier (every document on this axis) correction: the entry of 2026-08-30 records a /limits retitle that no commit holds; the page listed 23 register rows under Awaiting verification until it was withdrawn on 2026-09-05. was “Awaiting verification (41): has not yet been checked against it by a person” → “the page withdrawn on 2026-09-05; /limits/ answers with a redirect to /how/” wording
route recorded: /limits/ (redirected, 301, to /how/)
 link (opens in new tab) · retrieved
verification (every document on this axis) correction: 25 register rows carried the tracker's awaiting mark, a check owed and not yet made, though their label promised none. Changed 2026-08-15 at afe451131, with no entry then. was “awaiting — not yet checked” → “register — observed daily; no person-verification is promised” none: no part of the rule holds for this entry’s fields
routes recorded: / · /filter/verification/awaiting/ (redirected, 301, to /)
 link (opens in new tab) · retrieved
record-tier (every document on this axis) correction: register share cards read Last verified: not yet verified on 24, and 23 carried a NOT YET VERIFIED chip, though the label promises no check. Changed 2026-08-30 at 58ec8d584. was “NOT YET VERIFIED · Last verified: not yet verified” → “the card carries the entry's name, title, authority and address; no status, no date” wording
24 routes recorded
 link (opens in new tab) · retrieved
record-tier (every document on this axis) correction: 24 register row pages read Last verified: not yet verified, below a note saying none is promised for this tier. Changed 2026-08-25 at 0b8c76c82, with no entry then. was “Last verified: not yet verified” → “Last verified: no person is asked to confirm this entry” wording
24 routes recorded
 link (opens in new tab) · retrieved
record-tier (every document on this axis) correction: all 25 register row pages carried a note headed Not yet verified, a check still to come, though their label promises none. Changed 2026-08-16 at 0b978ffbb, with no entry then. was “Not yet verified.” → “Observation register. No person has verified it, and none is promised for this tier” wording
25 routes recorded
 link (opens in new tab) · retrieved
record-tier (every document on this axis) correction: the homepage search drew unknown as not yet verified on 23 register rows, a check still to come, though their label promises none. Changed 2026-09-29. was “? not yet verified” → “? unknown” wording
route recorded: /
 link (opens in new tab) · retrieved
record-tier (every document on this axis) correction: the status badge drew unknown as not yet verified on all 25 register rows, in the tracker and on each row page, though their label promises no check. Changed 2026-08-15 at 4ff0ab6e2. was “Status: not yet verified” → “Status: unknown” wording
27 routes recorded
 link (opens in new tab) · retrieved
iso-iec-sc42-programme correction: the verification date 2026-07-31 is withdrawn — no person's act and no stored machine read is dated to it was “2026-07-31” → “no verification date” status/date
row named: /r/iso-iec-sc42-programme/
 link (opens in new tab) · retrieved
mhra-ai-airlock correction: the verification date 2026-07-31 is withdrawn — no person's act and no stored machine read is dated to it was “2026-07-31” → “no verification date” status/date
row named: /r/mhra-ai-airlock/
 link (opens in new tab) · retrieved
status-unknown (every document on this axis) correction: /how defined 'unmeasured with cause' by a spent request budget. The class counts entries that carry a written account of why they are unresolved and meet none of the four earlier classes. was “No request was made — the host’s request budget was spent” → “The entry’s record carries a written account of why it is unresolved” none: no part of the rule holds for this entry’s fields
route recorded: /how/
 link (opens in new tab) · retrieved
eu-digital-omnibus-ai correction: the entry-into-force date was carried as this record's own derivation. The publisher states it, in a machine field naming this act's own identifier, in bytes already retained. was “2026-07-27 offered as publisher-derived — this record's arithmetic on the publisher's rule” → “2026-07-27 as the publisher states it, in a machine field whose subject is this act's own identifier” status/date · wording
row named: /r/eu-digital-omnibus-ai/ · route recorded: /calendar
 link (opens in new tab) · retrieved
record-tier (every document on this axis) correction: the published tier-1 figures were measured against the declared list alone while the build gate measured the ratified rule. Both were right about their own population; neither named it. was “tier-1 figures measured against the declared list alone: 20 undeclared, 35 of 37 tracked” → “tier-1 figures derived through tier_identity from the ratified rule: 0 unadmitted, 55 of 57 assigned” none: no part of the rule holds for this entry’s fields
route recorded: /findings
 link (opens in new tab) · retrieved
eu-digital-omnibus-ai correction: this entry publishes a status date of 2024-08-01 whose banked evidence names Regulation (EU) 2024/1689 — a different act. Borrowed authority, person-verified. was “status_as_of 2024-08-01, person-verified 2026-08-04” → “withdrawn as a claim about Regulation (EU) 2026/1744; the field awaits a keypress” status/date
row named: /r/eu-digital-omnibus-ai/ · route recorded: /calendar
 link (opens in new tab) · retrieved
correction: `topics` held this record's own authored vocabulary, an API field since 2026-07-27 and a page tree since 2026-09-09. 77 of its 96 identifiers had no publisher-stated twin. Retired. was “`topics`: this record's own vocabulary, typed per entry, cited nowhere — 96 identifiers” → “`topics`: derived by rule from each entry's own subject_as_stated — 24 identifiers, every value a publisher states” none: no part of the rule holds for this entry’s fields
routes recorded: /topic/ · /
 link (opens in new tab) · retrieved
correction: from 2026-09-12 to 2026-09-18 the front line took one figure from the run ledger and the rest from the observation store, so a nightly build named two nights in one sentence. Repaired. was “not-retrieved from the run ledger's last row; digests-moved from the observation store” → “every front-line figure from the observation store, for the night the sentence names” none: no part of the rule holds for this entry’s fields
route recorded: /
 link (opens in new tab) · retrieved
correction: 131 row pages marked an unchanged run of digests `stable`, while this record's own vocabulary page says a run of unchanged digests is not a statement that a document is stable. was “a window with no digest change is marked `stable`” → “a window with no digest change is marked `unchanged`” none: no part of the rule holds for this entry’s fields
route recorded: /
 link (opens in new tab) · retrieved
correction: the front page offered one candidate as a change to a document while the keystroke door refused it. The door was right; the row hashes a hub page. Withdrawn. was “1 offered as a change to a document, 1 as a change to a page about one” → “0 offered as a change to a document, 0 as a change to a page about one” none: no part of the rule holds for this entry’s fields
route recorded: /
 link (opens in new tab) · retrieved
coverage (every document on this axis) correction: five rows have hashed a page about their document, not the document, for their whole series. The address each actually reads is named here. was “five rows whose digests were held as readings of the document they name” → “five rows whose digests are of the publisher's page ABOUT that document, stated on each” identity/URL
routes recorded: /coverage/ · /changes/
 link (opens in new tab) · retrieved
health-canada-ml correction: this row names a machine-learning guidance document and its digests are of Health Canada's medical-devices hub page, which carries 77 dated list entries. was “a row declaring a guidance document and reading Health Canada's medical-devices hub” → “the same reading, stated as what it is: a hub page carrying a dated list” identity/URL
row named: /r/health-canada-ml/
 link (opens in new tab) · retrieved
coverage (every document on this axis) correction: 762 ich.org records answered HTTP 200 and contained no document. They were held as observations; they are now classed as responses that were not the artifact. was “913 ich.org records held as a series of readings” → “25 readings, 762 responses that were not the document, 126 transport failures” figure
routes recorded: /coverage/ · /changes/
 link (opens in new tab) · retrieved
status (every document on this axis) correction: the 55 status-log entries record this record's own status field moving, not a publisher moving a document's status. None is a movement observed between two readings. was “55 entries published as changes to documents' statuses” → “55 entries published as this record's own status field moving, with the publisher's passage shown beside each” status/date · presentation
routes recorded: /changes/ · /status-change/
 link (opens in new tab) · retrieved
fda-ai-guidance correction: this entry's document-series count leaves out 2 records that landed on a not-found page, and no page said so. Nothing is rewritten; both are now stated as held and excluded. presentation
row named: /r/fda-ai-guidance/
 link (opens in new tab) · retrieved
fda-csa correction: this entry's document-series count leaves 28 records out, and no page said so or on whose word. Nothing is rewritten; the 28 are now stated as held and excluded. presentation
row named: /r/fda-csa/
 link (opens in new tab) · retrieved
fda-device-cybersecurity-2026 correction: this entry's document-series count leaves 24 records out, every one of them hashed, and no page said so. Nothing is rewritten; they are now stated as held and excluded. presentation
row named: /r/fda-device-cybersecurity-2026/
 link (opens in new tab) · retrieved
correction: the published '43 detected candidates' figure was computed against the pages these rows hashed, not the documents they name. Withdrawn; the method is repaired and the re-screen published. was “43 detected change candidates” → “43 re-screened: 21 chrome, 16 oscillation, 3 attribute residue, 1 witness, 1 listing, 1 survivor” none: no part of the rule holds for this entry’s fields
routes recorded: /findings · /changes
 link (opens in new tab) · retrieved
fda-device-cybersecurity correction: this row's status is superseded and its document series is ended. Every observation it made was of the successor document; the successor row holds the one series. was “unknown” → “superseded” status/date
row named: /r/fda-device-cybersecurity/
 link (opens in new tab) · retrieved
astm-e2363 correction: this row declared an address its series has never answered at. The row now records the answered URL and keeps the prior one. Series continuous — same document, no splice. was “https://www.astm.org/e2363-23.html” → “https://store.astm.org/e2363-23.html” identity/URL
row named: /r/astm-e2363/
 link (opens in new tab) · retrieved
astm-e2500 correction: this row declared an address its series has never answered at. The row now records the answered URL and keeps the prior one. Series continuous — same document, no splice. was “https://www.astm.org/e2500-25.html” → “https://store.astm.org/e2500-25.html” identity/URL
row named: /r/astm-e2500/
 link (opens in new tab) · retrieved
ema-ai-reflection correction: this row declared an address its series has never answered at. The row now records the answered URL and keeps the prior one. Series continuous — same document, no splice. was “https://www.ema.europa.eu/en/use-artificial-intelligence-ai-medicinal-product-lifecycle” → “https://www.ema.europa.eu/en/use-artificial-intelligence-ai-medicinal-product-lifecycle-scientific-guideline” identity/URL
row named: /r/ema-ai-reflection/
 link (opens in new tab) · retrieved
eu-gmp-annex-21 correction: this row declared an address its series has never answered at. The row now records the answered URL and keeps the prior one. Series continuous — same document, no splice. was “https://ec.europa.eu/health/document/download/e2ddfe65-7b4e-4765-b71b-4681772d2949_en” → “https://health.ec.europa.eu/document/download/e2ddfe65-7b4e-4765-b71b-4681772d2949_en” identity/URL
row named: /r/eu-gmp-annex-21/
 link (opens in new tab) · retrieved
fda-aiml-device-list correction: this row declared an address its series has never answered at. The row now records the answered URL and keeps the prior one. Series continuous — same document, no splice. was “https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-aiml-enabl” → “https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-enabled-medical-devices” identity/URL
row named: /r/fda-aiml-device-list/
 link (opens in new tab) · retrieved
us-21-cfr-part-11 correction: this row declared an address its series has never answered at. The row now records the answered URL and keeps the prior one. Series continuous — same document, no splice. was “https://www.ecfr.gov/current/title-21/part-11” → “https://www.ecfr.gov/current/title-21/chapter-I/subchapter-A/part-11” identity/URL
row named: /r/us-21-cfr-part-11/
 link (opens in new tab) · retrieved
fda-ai-guidance correction: this row's resolved_url named a page ABOUT the document. It now names the document, derived from that page's own designator-labeled link. The witness series is unchanged. was “https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considerations-use-artificial-intelligence-supp” → “https://www.fda.gov/media/184830/download” identity/URL
row named: /r/fda-ai-guidance/
 link (opens in new tab) · retrieved
ich-q12 correction: this row's resolved_url named a page ABOUT the document. It now names the document, derived from that page's own designator-labeled link. The witness series is unchanged. was “https://www.fda.gov/regulatory-information/search-fda-guidance-documents/q12-technical-and-regulatory-considerations-pha” → “https://www.fda.gov/media/148476/download” identity/URL
row named: /r/ich-q12/
 link (opens in new tab) · retrieved
ich-q13 correction: this row's resolved_url named a page ABOUT the document. It now names the document, derived from that page's own designator-labeled link. The witness series is unchanged. was “https://www.fda.gov/regulatory-information/search-fda-guidance-documents/q13-continuous-manufacturing-drug-substances-an” → “https://www.fda.gov/media/165775/download” identity/URL
row named: /r/ich-q13/
 link (opens in new tab) · retrieved
ich-q14 correction: this row's resolved_url named a page ABOUT the document. It now names the document, derived from that page's own designator-labeled link. The witness series is unchanged. was “https://www.ema.europa.eu/en/ich-q14-analytical-procedure-development-scientific-guideline” → “https://www.ema.europa.eu/en/documents/scientific-guideline/ich-q14-guideline-analytical-procedure-development-step-5-re” identity/URL
row named: /r/ich-q14/
 link (opens in new tab) · retrieved
ich-q2r2 correction: this row's resolved_url named a page ABOUT the document. It now names the document, derived from that page's own designator-labeled link. The witness series is unchanged. was “https://www.ema.europa.eu/en/ich-q2r2-validation-analytical-procedures-scientific-guideline” → “https://www.ema.europa.eu/en/documents/scientific-guideline/ich-q2r2-guideline-validation-analytical-procedures-step-5-r” identity/URL
row named: /r/ich-q2r2/
 link (opens in new tab) · retrieved
ich-q9r1 correction: this row's resolved_url named a page ABOUT the document. It now names the document, derived from that page's own designator-labeled link. The witness series is unchanged. was “https://www.ema.europa.eu/en/ich-q9-quality-risk-management-scientific-guideline” → “https://www.ema.europa.eu/en/documents/scientific-guideline/international-conference-harmonisation-technical-requirement” identity/URL
row named: /r/ich-q9r1/
 link (opens in new tab) · retrieved
correction: text_retained was derived from the row's resolution_strategy, not measured. 31 rows published it true while holding cleaned PDF file bytes, not text. Now measured from the store. was “121 rows publish text_retained: true” → “90 rows publish text_retained: true” none: no part of the rule holds for this entry’s fields
routes recorded: /developers · /api/v1/instruments.json
 link (opens in new tab) · retrieved
correction: 47 rows carried tier 1 without appearing in the declared list, the approved canon or a documented exception. Reassigned to tier 2 in one act; the membership rule is now published. was “87 rows assigned tier 1” → “55 rows assigned tier 1” none: no part of the rule holds for this entry’s fields
routes recorded: /how · /
 link (opens in new tab) · retrieved
fda-ai-guidance correction: the corroboration count counted kinds of surface, not surfaces. This row has three distinct publisher addresses stating its status and reported one. was “corroboration_count 1” → “corroboration_count 3” none: no part of the rule holds for this entry’s fields
row named: /r/fda-ai-guidance/
 link (opens in new tab) · retrieved
eu-gmp-annex-15 correction: this row's summary made a scope claim in this site's own voice with no citation. It now reports the publisher's own title instead. was “a summary asserting scope in this site's own voice” → “a summary reporting the European Commission's own published title” none: no part of the rule holds for this entry’s fields
row named: /r/eu-gmp-annex-15/
 link (opens in new tab) · retrieved
eu-mdr correction: this row's summary made a scope claim in this site's own voice with no citation. It now reports the publisher's own title instead. was “a summary asserting scope in this site's own voice” → “a summary reporting the act's own published title” none: no part of the rule holds for this entry’s fields
row named: /r/eu-mdr/
 link (opens in new tab) · retrieved
iso-iec-27017 correction: this row's summary made a scope claim in this site's own voice with no citation. It now reports the publisher's own title instead. was “a summary asserting scope and a relation to two other standards” → “a summary reporting the published title and this row's licence limit” none: no part of the rule holds for this entry’s fields
row named: /r/iso-iec-27017/
 link (opens in new tab) · retrieved
astm-e2363 correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 50-word quotation, over the 14-word cap” → “a 7-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/astm-e2363/
 link (opens in new tab) · retrieved
astm-e2500 correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 50-word quotation, over the 14-word cap” → “a 7-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/astm-e2500/
 link (opens in new tab) · retrieved
ema-ai-reflection correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 33-word quotation, over the 14-word cap” → “a 12-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/ema-ai-reflection/
 link (opens in new tab) · retrieved
ema-ct-computerised-systems correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 48-word quotation, over the 14-word cap” → “a 10-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/ema-ct-computerised-systems/
 link (opens in new tab) · retrieved
ema-data-quality-framework correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 46-word quotation, over the 14-word cap” → “a 6-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/ema-data-quality-framework/
 link (opens in new tab) · retrieved
ema-work-programme correction: the quotation here exceeded the 14-word cap and no compliant span exists in the banked evidence, so none is published. Withheld rather than shortened. was “a 49-word quotation, over the 14-word cap” → “no quotation published; the reason is stated on the row” none: no part of the rule holds for this entry’s fields
row named: /r/ema-work-programme/
 link (opens in new tab) · retrieved
eu-ai-act correction: the quotation here exceeded the 14-word cap and no compliant span exists in the banked evidence, so none is published. Withheld rather than shortened. was “a 52-word quotation, over the 14-word cap” → “no quotation published; the reason is stated on the row” none: no part of the rule holds for this entry’s fields
row named: /r/eu-ai-act/
 link (opens in new tab) · retrieved
eu-cyber-resilience-act correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 42-word quotation, over the 14-word cap” → “a 11-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/eu-cyber-resilience-act/
 link (opens in new tab) · retrieved
eu-cybersecurity-act correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 36-word quotation, over the 14-word cap” → “a 11-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/eu-cybersecurity-act/
 link (opens in new tab) · retrieved
eu-digital-omnibus-ai correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 33-word quotation, over the 14-word cap” → “a 4-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/eu-digital-omnibus-ai/
 link (opens in new tab) · retrieved
eu-gmp-annex-11 correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 34-word quotation, over the 14-word cap” → “a 8-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/eu-gmp-annex-11/
 link (opens in new tab) · retrieved
eu-ivdr correction: the quotation here exceeded the 14-word cap and no compliant span exists in the banked evidence, so none is published. Withheld rather than shortened. was “a 51-word quotation, over the 14-word cap” → “no quotation published; the reason is stated on the row” none: no part of the rule holds for this entry’s fields
row named: /r/eu-ivdr/
 link (opens in new tab) · retrieved
eu-machinery-regulation correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 36-word quotation, over the 14-word cap” → “a 11-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/eu-machinery-regulation/
 link (opens in new tab) · retrieved
eu-mdr correction: the quotation here exceeded the 14-word cap and no compliant span exists in the banked evidence, so none is published. Withheld rather than shortened. was “a 49-word quotation, over the 14-word cap” → “no quotation published; the reason is stated on the row” none: no part of the rule holds for this entry’s fields
row named: /r/eu-mdr/
 link (opens in new tab) · retrieved
eu-nis2 correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 47-word quotation, over the 14-word cap” → “a 11-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/eu-nis2/
 link (opens in new tab) · retrieved
fda-csa correction: the quotation here exceeded the 14-word cap and no compliant span exists in the banked evidence, so none is published. Withheld rather than shortened. was “a 27-word quotation, over the 14-word cap” → “no quotation published; the reason is stated on the row” none: no part of the rule holds for this entry’s fields
row named: /r/fda-csa/
 link (opens in new tab) · retrieved
ich-e6r3 correction: the quotation here exceeded the 14-word cap and no compliant span exists in the banked evidence, so none is published. Withheld rather than shortened. was “a 46-word quotation, over the 14-word cap” → “no quotation published; the reason is stated on the row” none: no part of the rule holds for this entry’s fields
row named: /r/ich-e6r3/
 link (opens in new tab) · retrieved
ich-q2r2 correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 26-word quotation, over the 14-word cap” → “a 11-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/ich-q2r2/
 link (opens in new tab) · retrieved
ieee-7000 correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 25-word quotation, over the 14-word cap” → “a 12-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/ieee-7000/
 link (opens in new tab) · retrieved
iso-iec-22989 correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 24-word quotation, over the 14-word cap” → “a 6-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/iso-iec-22989/
 link (opens in new tab) · retrieved
iso-iec-23894 correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 25-word quotation, over the 14-word cap” → “a 6-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/iso-iec-23894/
 link (opens in new tab) · retrieved
iso-iec-42001 correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 46-word quotation, over the 14-word cap” → “a 6-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/iso-iec-42001/
 link (opens in new tab) · retrieved
mhra-ai correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 15-word quotation, over the 14-word cap” → “a 10-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/mhra-ai/
 link (opens in new tab) · retrieved
nist-sp-800-82 correction: the quotation here exceeded the 14-word cap. A shorter span was SELECTED from the same banked evidence, not trimmed. The evidence record is unchanged. was “a 30-word quotation, over the 14-word cap” → “a 6-word verbatim sub-span of the same banked evidence” none: no part of the rule holds for this entry’s fields
row named: /r/nist-sp-800-82/
 link (opens in new tab) · retrieved
pics-pi041 correction: the quotation here exceeded the 14-word cap and no compliant span exists in the banked evidence, so none is published. Withheld rather than shortened. was “a 20-word quotation, over the 14-word cap” → “no quotation published; the reason is stated on the row” none: no part of the rule holds for this entry’s fields
row named: /r/pics-pi041/
 link (opens in new tab) · retrieved
eu-ai-act correction: 2 August 2028 sat in status_as_of, the day a status is stated as of, so the page dated a present status two years ahead. It has moved to effective_date. The status itself is unchanged. was “status_as_of: 2028-08-02 · effective_date: null” → “status_as_of: null · effective_date: 2028-08-02” none: no part of the rule holds for this entry’s fields
row named: /r/eu-ai-act/
 link (opens in new tab) · retrieved
fda-csa correction: this row's account said fda.gov drops automated clients and that its status awaited a person. Both were false when the sentence was written. The account is withheld, not rewritten. was “fda.gov drops automated clients, so the guidance page cannot be read directly. The Federal Register records its issuance on 24 September 2025; that date is evidence and is waiting on a person to confirm the status it implies.” → “no account. The field is withheld: the row resolves to a document, hashes it, and its status carries a recorded human confirmation, so there is nothing unresolved for it to account for.” none: no part of the rule holds for this entry’s fields
row named: /r/fda-csa/
 link (opens in new tab) · retrieved
status-unknown (every document on this axis) correction: the four shares of `unknown` were published on one built page, far from the number they decompose, and the fourth rendered only above zero. All four now print under the chip they sum to. was “(the four shares of `unknown` on 1 built page of 722, below the ten reasons, with the fourth member rendered only above zero)” → “All four shares print under the active `unknown` chip, on every built page where `unknown` is the selected status, summing to the number that chip shows.” none: no part of the rule holds for this entry’s fields
route recorded: /filter/status/unknown/ (redirected, 301, to /)
 link (opens in new tab) · retrieved
record-tier (every document on this axis) correction: tier sentences stated what had happened to an entry and were false for part of each tier. They now state only what the label undertakes; a duplicate register note carrying one was cut. was “This entry is retrieved and hashed every night, and its status is one a person is asked to confirm.” → “Filed core. For that label this archive undertakes to put an entry's status to a person.” none: no part of the rule holds for this entry’s fields
routes recorded: /machine/ (redirected, 301, to /developers/) · /r/eu-gmp-annex-14/ · /filter/status/unknown/ (redirected, 301, to /)
 link (opens in new tab) · retrieved
verification (every document on this axis) correction: the homepage said all 44 frozen entries have no public document to verify against. 38 of 44 record another reason. The five reasons the rows record now render, and /limits lists each. was “no public document exists to verify them against, and each states which” → “the reason each one records is: 23 paywalled — progression tracked; 9 listing only; 6 not a document; 3 blocks automated requests; 3 register states no status” none: no part of the rule holds for this entry’s fields
routes recorded: / · /limits/ (redirected, 301, to /how/) · /filter/verification/frozen/ (redirected, 301, to /)
 link (opens in new tab) · retrieved
status (every document on this axis) correction: the tracker's Status column carried the document's status and this record's verification mark under one header, on 132 of 132 rows. Each axis now has its own column. was “one Status column over the document's status and this record's verification mark” → “Status and Verification as separate columns, each sorted by the axis it shows” none: no part of the rule holds for this entry’s fields
route recorded: /
 link (opens in new tab) · retrieved
fda-ai-guidance correction: this entry hashes FDA's guidance search index, not the guidance. It now carries an account saying so, and saying that its two recorded text changes are one excursion and its return. was “no account — a digest rendered under the guidance's title with nothing to place it” → “an account naming what the series hashes, and what the two recorded text changes were” none: no part of the rule holds for this entry’s fields
row named: /r/fda-ai-guidance/
 link (opens in new tab) · retrieved
fda-device-cybersecurity-2026 correction: this entry's account said no status had been confirmed by a keypress while the entry published final. The status is act-backed; the account was the stale half and is rewritten. was “an account ending 'the value stays unknown', rendered two sentences from a published status of final” → “an account naming the act the status traces to, and recording that the earlier sentence was stale” none: no part of the rule holds for this entry’s fields
row named: /r/fda-device-cybersecurity-2026/
 link (opens in new tab) · retrieved
fda-device-cybersecurity correction: this entry hashes the February 2026 successor's page, not its own. It now carries an account of what it tracks; the referent question stays open and is not settled here. was “no account — the successor's digest rendered under the June 2025 edition's title” → “an account naming what the series hashes and recording that the referent question stays open” none: no part of the rule holds for this entry’s fields
row named: /r/fda-device-cybersecurity/
 link (opens in new tab) · retrieved
nmpa-ai correction: the account said one identical hash across 95 observations while the same page counted 109 and one change. The store holds 109 records and 2 digests, the second from 2026-08-21. was “an account reading 'one identical hash across all 95 observations', one screen above a generated block reading '109 observation(s) recorded' and '1 change(s)'” → “an account stating 109 records and two digests, naming the same population the generated block counts, dating its figures, and recording the second digest as first observed 2026-08-21” none: no part of the rule holds for this entry’s fields
row named: /r/nmpa-ai/
 link (opens in new tab) · retrieved
status-unknown (every document on this axis) correction: five rows carried a publisher-share reason typed from a 404 or 403 returned to an address this archive constructed. Those codes no longer type a reason; no observation changed. was “the address returns not-found: 3 · the request was refused: 2” → “the address returns not-found: 0 · the request was refused: 0” none: no part of the rule holds for this entry’s fields
route recorded: /filter/status/unknown (redirected, 301, to /)
 link (opens in new tab) · retrieved
eu-gmp-annex-14 correction: the status moved unknown to in-force at commit 02a18d2d on 2026-08-24 and no changelog entry was written at the time. Recorded 2026-09-01 from git history. was “unknown” → “in-force” status/date
row named: /r/eu-gmp-annex-14/
 link (opens in new tab) · retrieved
record-tier (every document on this axis) correction: the tier labels no longer carry a claim the holdings contradict; each entry now states its promise in words was “register-observed” → “register” none: no part of the rule holds for this entry’s fields
route recorded: /
 link (opens in new tab) · retrieved
status-unknown (every document on this axis) correction: 40 rows moved from 'not yet read' to 'unmeasured with cause' — a reading error, not a data change was “not yet read: 66” → “not yet read: 26” none: no part of the rule holds for this entry’s fields
route recorded: /filter/status/unknown (redirected, 301, to /)
 link (opens in new tab) · retrieved
eu-gmp-annex-17 correction: the status moved unknown to in-force at commit cb52e496 on 2026-08-16 and no changelog entry was written at the time. Recorded 2026-09-01 from git history. was “unknown” → “in-force” status/date
row named: /r/eu-gmp-annex-17/
 link (opens in new tab) · retrieved
fda-device-cybersecurity-2026 correction: an entry's title and its evidence never referred to the same document. 232 observation records, including all 24 content hashes, are re-scoped to the February 2026 edition. none: no part of the rule holds for this entry’s fields
row named: /r/fda-device-cybersecurity-2026/
 link (opens in new tab) · retrieved
imdrf-aiml-wg correction: a body record asserted a kind no retrieved source states was “recorded in two collections under one id” → “one id, one record — the instrument row” none: no part of the rule holds for this entry’s fields
row named: /r/imdrf-aiml-wg/
 link (opens in new tab) · retrieved
ema-ct-computerised-systems correction: the status moved unknown to final at commit 237b4a5b on 2026-08-01 and no changelog entry was written at the time. Recorded 2026-09-01 from git history. was “unknown” → “final” status/date
row named: /r/ema-ct-computerised-systems/
 link (opens in new tab) · retrieved
eu-gmp-annex-22 correction: observations dated 28-30 July hashed a different document was “797a882ba0eafe08c329001260e2dabf7c05db82f8a5766d7f75df7569f2e85d” → “07dc6a9975d123f2aa8ca9a2c0d5ba28b09b743d847d136814b6dfa555531a26” status/date
row named: /r/eu-gmp-annex-22/
 link (opens in new tab) · retrieved
eu-gmp-annex-11 correction: status withdrawn — the source is the EudraLex Volume 4 index, which cannot state one document's status status/date
row named: /r/eu-gmp-annex-11/
 link (opens in new tab) · retrieved
eu-gmp-annex-22 correction: status withdrawn — the source is the EudraLex Volume 4 index, which cannot state one document's status status/date
row named: /r/eu-gmp-annex-22/
 link (opens in new tab) · retrieved
fda-ai-guidance correction: status withdrawn — the Federal Register announces this guidance as a draft status/date
row named: /r/fda-ai-guidance/
 link (opens in new tab) · retrieved
fda-ai-guidance correction — `final` was read from FDA's search-page help text; the entry also pointed at the search page rather than at the guidance was “final” → “unknown” status/date
row named: /r/fda-ai-guidance/
 link (opens in new tab) · retrieved
fda-cdrh-aiml correction: the status moved unknown to draft at commit 9c0a7368 on 2026-07-30 and no changelog entry was written at the time. Recorded 2026-09-01 from git history. was “unknown” → “draft” status/date
row named: /r/fda-cdrh-aiml/
 link (opens in new tab) · retrieved
fda-csa correction: the status moved unknown to final at commit 0d8fe64c on 2026-07-30 and no changelog entry was written at the time. Recorded 2026-09-01 from git history. was “unknown” → “final” status/date
row named: /r/fda-csa/
 link (opens in new tab) · retrieved
fda-guidance-agenda correction — `draft` was read from a sentence about the guidances the agenda lists, not about the agenda; an agenda has no status was “draft” → “unknown” status/date
row named: /r/fda-guidance-agenda/
 link (opens in new tab) · retrieved
ich-q2r2 correction: status withdrawn — the source is the ICH quality-guidelines index, shared with eight other entries status/date
row named: /r/ich-q2r2/
 link (opens in new tab) · retrieved
ieee-7000 correction: status withdrawn — the page states "active standard" three times, including in a labelled Status field status/date
row named: /r/ieee-7000/
 link (opens in new tab) · retrieved
nist-ai-rmf correction: status withdrawn — every occurrence of "draft" on the page describes a PRIOR version, under a heading reading "prior documents" status/date
row named: /r/nist-ai-rmf/
 link (opens in new tab) · retrieved
pic-s correction: status withdrawn — the source is the PIC/S publications listing, which cannot state one document's status status/date
row named: /r/pic-s/
 link (opens in new tab) · retrieved
pics-pi041 correction: status withdrawn — the source is the PIC/S publications listing, which cannot state one document's status status/date
row named: /r/pics-pi041/
 link (opens in new tab) · retrieved
astm-e2363 correction: source URL repointed at the edition the publisher marks current was “https://www.astm.org/e2363-14.html” → “https://www.astm.org/e2363-23.html” identity/URL
row named: /r/astm-e2363/
 link (opens in new tab) · retrieved
astm-e2500 correction: source URL repointed at the edition the publisher marks current was “https://www.astm.org/e2500-20.html” → “https://www.astm.org/e2500-25.html” identity/URL
row named: /r/astm-e2500/
 link (opens in new tab) · retrieved
cen-jtc21-programme correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/cen-jtc21-programme/
 link (opens in new tab) · retrieved
eu-gmp-annex-11 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/eu-gmp-annex-11/
 link (opens in new tab) · retrieved
eu-gmp-annex-22 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/eu-gmp-annex-22/
 link (opens in new tab) · retrieved
fda-cdrh-aiml correction: primary source URL no longer resolves was “resolving” → “error page” status/date
row named: /r/fda-cdrh-aiml/
 link (opens in new tab) · retrieved
fda-csa correction: primary source URL no longer resolves was “resolving” → “error page” status/date
row named: /r/fda-csa/
 link (opens in new tab) · retrieved
fda-guidance-agenda correction: primary source URL no longer resolves was “resolving” → “error page” status/date
row named: /r/fda-guidance-agenda/
 link (opens in new tab) · retrieved
ich-q12 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/ich-q12/
 link (opens in new tab) · retrieved
ich-q13 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/ich-q13/
 link (opens in new tab) · retrieved
ich-q14 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/ich-q14/
 link (opens in new tab) · retrieved
ich-q2r2 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/ich-q2r2/
 link (opens in new tab) · retrieved
ich-q9r1 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/ich-q9r1/
 link (opens in new tab) · retrieved
iso-iec-22989 correction: the status moved unknown to final at commit 249742e9 on 2026-07-29 and no changelog entry was written at the time. Recorded 2026-09-01 from git history. was “unknown” → “final” status/date
row named: /r/iso-iec-22989/
 link (opens in new tab) · retrieved
iso-iec-24028 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/iso-iec-24028/
 link (opens in new tab) · retrieved
iso-iec-38507 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/iso-iec-38507/
 link (opens in new tab) · retrieved
iso-iec-42005 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/iso-iec-42005/
 link (opens in new tab) · retrieved
iso-iec-5259 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/iso-iec-5259/
 link (opens in new tab) · retrieved
iso-iec-sc42-programme correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/iso-iec-sc42-programme/
 link (opens in new tab) · retrieved
mhra-ai-airlock correction: primary source URL no longer resolves was “resolving” → “error page” status/date
row named: /r/mhra-ai-airlock/
 link (opens in new tab) · retrieved
mhra-ai correction: primary source URL no longer resolves was “resolving” → “error page” status/date
row named: /r/mhra-ai/
 link (opens in new tab) · retrieved
pic-s correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/pic-s/
 link (opens in new tab) · retrieved
pics-pi041 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/pics-pi041/
 link (opens in new tab) · retrieved
pren-18229-1 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/pren-18229-1/
 link (opens in new tab) · retrieved
pren-18229-2 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/pren-18229-2/
 link (opens in new tab) · retrieved
pren-18281 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/pren-18281/
 link (opens in new tab) · retrieved
pren-18282 correction: status_as_of removed — the date belonged to a shared index page status/date
row named: /r/pren-18282/
 link (opens in new tab) · retrieved

This page carries every correction to a published claim about a tracked document, or about one of the axes this record publishes over all of them. Corrections to this record’s own machinery — counting rules, page templates, quoted spans — are kept in the repository history instead, because they change nothing this site has said about any document. Where a page’s published figures change what they COUNT, that is listed below instead of in the history: a reader who cited a figure cannot be expected to read commits.

Corrections to this site’s own surfaces

A page on this site whose figures changed what they count, with the before and after. No document and no publisher moved in any of these.

Surface corrections, most recent first
DateSurface CorrectionCategory Affected
/how/ correction: the definition of the publisher-stated basis named one of the two acts that now carry it. The second is added. publisher-stated defined as marking a revision candidate confirmed by a change note the publisher itself dated inside the candidate's window → publisher-stated defined as marking a revision candidate confirmed by such a change note, or an entry's status where the publisher's own status field prints a word that spells it wording
routes recorded: /how/ · /llms.txt · /llms-full.txt
/ correction: the definition of the tool-read state gave a reason that is not a property of every row in it. The reason is dropped. tool-read defined as read from the source by tooling, because the publisher prints no status a machine can read → tool-read defined as read from the source by tooling wording
7 routes recorded
/concordance/ correction: the concordance read texts the clause-anchor admission refuses; it now reads only admitted texts, and one span under a term is withdrawn the concordance read the newest retained text of every openly licensed row, asking nothing of the host's terms filing: 94 texts, among them one span shown under a term from a host with no terms filing → the concordance reads a row's newest retained text only where the clause-anchor admission (state/clause-anchors.jsonl, under its current version) admits it: 79 texts; 15 open rows whose text it refuses or holds no line for are counted apart and read as not measured none: no part of the rule holds for this entry’s fields
routes recorded: /concordance/ · /api/v1/concordance.json · /r/cen-jtc21-programme/
/card/ correction: every card printed this record's status under a label /card/ and /observatory/ called what the publisher asserts. The row pages were corrected on 2026-10-04. Document status, described as what the publisher asserts, printing this record's own status → Recorded status, with whether a publisher surface was read beside it wording
134 routes recorded
/graph/ correction: a NIST AI RMF quote citing ISO/IEC 22989 ran its words together, so the cap counted it as 6 words. The edge is withheld. a citation edge whose quotation was text extracted with its spaces lost, counted as 6 words → the edge withheld: its only occurrence offers no span whose words can be counted wording
4 routes recorded
/r/eu-ai-act/ correction: the Dates table printed each event's own date as the day its source was retrieved; Part 11 read retrieved 20 Aug 1997. 17 cells on 10 row pages. each Source link in the Dates table followed by retrieved and the event's own date → the Source link with no retrieval date, since a calendar entry records none presentation
10 routes recorded
/developers/ correction: /developers/ said a change of a value's meaning had happened twice. No field selects such a change, so the count is derived from nothing. that has happened twice, and both entries are public on corrections → each such correction is public on corrections figure
route recorded: /developers/
/api/v1/events.json correction: /api/v1/events.json said RC-1, the head writing a confirmation's basis, had not landed. It landed on 2026-10-10. The measured reason is stated. RC-1 has not landed, so the record holds no basis field to carry → RC-1 landed on 2026-10-10 and writes the basis on rows and revision entries; the act log's acts carrying one are counted none: no part of the rule holds for this entry’s fields
route recorded: /api/v1/events.json
/api/v1/findings.json correction: two findings headlines in the API and search data led with 0 where no agent request was sent, publishing an unmeasured count as a negative. two headlines opening 0 source(s) over a population of no agent requests → the same headlines ending: not measured, with no leading numeral figure
4 routes recorded
/how/ correction: /how/ said each tracked document is fetched, read and dated every day. Of 132 rows, 69 hold fewer digest days than days observed, and 6 hold none. Each one is fetched, read and dated every day. → each one is collected on a daily schedule, and not every day yields a reading none: no part of the rule holds for this entry’s fields
route recorded: /how/
/llms.txt correction: /llms.txt carried sentences of its own that /how does not say, among them a forecast, a ranking of errors and three instructions to its reader. prose of the file's own: a checking cadence, a list of qualifiers addressed to its reader, a terms section and a description of what the site is not → sentences /how renders, the board's figures in the forms already held to the board, and each address with its own page's first sentence or its own document's title none: no part of the rule holds for this entry’s fields
route recorded: /llms.txt
/r/eu-gmp-annex-17/ correction: two rows whose status rests on a machine read said it was recorded against the primary source on a dash, while the record held the day of the read. Recorded as the status against the primary source on — → Recorded as the status from a read of the primary source by tooling on its day; no person confirmed it presentation
routes recorded: /r/eu-gmp-annex-17/ · /r/fda-device-cybersecurity/
/papers/ correction: /papers/ said the collection code is published with the dataset. No route serves code, and SUCCESSION.md states the software is private. a line saying the collection code is published with the dataset, and the same claim in the lede and the page description → no claim about code; the lede and description name the data alone none: no part of the rule holds for this entry’s fields
route recorded: /papers/
/how/refusals/ correction: /how/refusals/ said a confirmed revision is written only by a keypress, while the front page counted 7 confirmed revisions, all confirmed by analysis. A confirmed revision is written only by a keypress → a rule's confirmation is recorded as derived; a revision entry is written only by a keypress none: no part of the rule holds for this entry’s fields
route recorded: /how/refusals/
/r/anvisa-ai/ correction: each entry page described itself to machines as an article about the document; it is this record's series for the entry, and is now typed a dataset. each entry page's schema.org node typed Article, its headline the document's title → each entry page's schema.org node typed Dataset, named by the document's title, with the entry's id, its page address, its source and its observation bundle as the one distribution none: no part of the rule holds for this entry’s fields
132 routes recorded
/r/astm-e2537/ correction: 8 rows with status unknown read unknown as of a day, a date belonging to the status they used to hold. An unknown is dated by none. Recorded status: unknown as of a day, on rows whose status is unknown → Recorded status: unknown, with no as-of day presentation
8 routes recorded
/developers/ correction: /developers/ said its exact-match box was on that page only; the same index and client are now also on /search/, and result rows link canonical addresses. a note under the exact-match index on /developers/ reading "This box is on this page only", and result rows linking all four kinds of target without the slash the target page names as its canonical address, 191 of the 191 rows in the index the build of 2026-10-10 carried there → the note says the box is on /developers/ and on /search/, which carry one index and one client, and every result row links the address its target page names as canonical none: no part of the rule holds for this entry’s fields
route recorded: /developers/
/corrections/ correction: a surface correction on /corrections/ that names many routes linked only the first. Each entry now links every object it names and carries a category placed by a published rule. each correction as its note and its before-and-after, with no category; a surface correction naming many routes linked only the first of them → each correction opening on its note's first sentence, with a category placed by a rule the page publishes and a link to the row it names and, kept apart, to every route it records; each row page a correction names by its instrument states how many name it, the last date, and links its history presentation
61 routes recorded
/developers/ correction: /developers/ said its Model Context Protocol server is listed in no registry, an absence no search measured; it now says this record has submitted it to none. the Model Context Protocol section of /developers/ said the server is listed in no registry, a statement about every registry, none of which this record searched → the section says this record has submitted the server to no registry, which is a statement about this record's own acts none: no part of the rule holds for this entry’s fields
route recorded: /developers/
/packet/anvisa-ai/ correction: from 2026-09-08 the structured trail of each packet page named its row at an address with the site's origin written twice. It now names the row's own address. the structured trail of each packet page naming the row it belongs to at an address with the site's origin written twice → the structured trail of each packet page naming the row at its own address on the site presentation
132 routes recorded
/observatory/ correction: 19 built files said no run had been authorised to write the signal store. Since 2026-10-21 the nightly is authorised to write two of its sources; none has been written yet. an absent meta file recorded that no run had been authorised to write the signal store → an absent meta file records that no run has written the store; the nightly is authorised to write ecfr and federalregister since 2026-10-21, and the write preconditions each meets are stated beside it none: no part of the rule holds for this entry’s fields
19 routes recorded
/how/ correction: /how/ printed 0 corrections on every build since 2026-09-05, while /developers/ printed the whole count as published corrections. /how/: 0 correction(s) recorded so far; /developers/: 272 published corrections; the 120 entries this record no longer publishes to readers → one count read by both pages, with three labelled subsets: document claims, site machinery and figures; the 120 bound to 2026-09-06, the day it was measured figure
routes recorded: /how/ · /developers/
/ correction: from 2026-10-11 the front page said this record held 0 confirmed revisions while it counted revision entries written; 7 confirmed revisions stood by 2026-10-12. The diff collection holds 0 entries, and this record holds 0 confirmed revisions. Both are stores nothing has been promoted into: the block beside this one counts what detection has staged, and no verdict has moved any of it here. → The diff collection holds 0 entries, and 0 revision entries have been written. Both are stores nothing has been promoted into: the block beside this one counts what detection has staged, and no verdict has moved any of it here. A confirmed revision is counted by the act that confirmed it, whether or not a revision entry has been written for it: 7, the confirmed revisions in the block headed What this record holds. figure
route recorded: /
/r/eu-gmp-annex-17/ correction: on 856 pages a trail began with Board, a section or Home, a structured trail named the front page Tracker, or a link used a retired label or /methodology; each is re-based. trails beginning with four words for their root, a structured trail naming the front page Tracker, links naming /how/ Methodology and the board Findings or tracker, and links to /methodology → every trail from the declared root, every structured trail naming what a visible trail names, no link to a navigation destination carrying a retired word, and none to /methodology wording
856 routes recorded
/r/anvisa-ai/ correction: the absence section stated what this record's readings hold or did not find; its heading now says so. Row pages now open on four facts and five actions. each row page opening on its title, then the evidence coverage card with its twelve rungs open, then the bands; a section headed "What cannot be established" inside the Evidence band → each row page opening on four facts and five actions, then the coverage card with its rungs folded, then "What this record does not establish", then the bands (on the worked example, its walkthrough before the bands) wording
132 routes recorded
/ correction: some cells printed 0 for a count that was never taken. On 35 entry pages a window with no observation or digest read 0, and every signal cell read 0 of 0; those now read not measured. a zero printed as a bare 0 whether or not anything was measured, and some zeros printed as a dash, "none" or "not reached" → every zero standing as a figure of its own rendered by one component with its state: a measured zero as 0, and a count that was not taken, including a count over an empty population, as "not measured" presentation
405 routes recorded
/changes.xml correction: /changes.xml left out seven corrections about documents that /corrections/ lists. Its derivation now reads the page's own predicate. /changes.xml carried no item for seven corrections about documents that /corrections/ lists, dated 2026-09-07 to 2026-09-18 → /changes.xml and the Atom changes and corrections feeds read the predicate /corrections/ reads, so every correction the page lists is an item of the derivation behind them; /changes.xml carries those of the seven within its newest items none: no part of the rule holds for this entry’s fields
route recorded: /changes.xml
/ correction: the 2026-09-26 entry says compact rows set two lines of each board row smaller. From 2026-10-04 that did not hold of the reason line. compact rows set the board's jurisdiction and reason lines smaller, as the entry dated 2026-09-26 states → compact rows change row spacing only and set no type size; both lines are drawn at one size in both settings none: no part of the rule holds for this entry’s fields
route recorded: /
/api/v1/probe.jsonl correction: the probe pointer and partition headers named an index address no build wrote; they now name /api/v1/probe.json. the pointer record and every partition header named the partition index at /api/v1/probe/index.json, an address no build has written → both name /api/v1/probe.json, where the index is served identity/URL
13 routes recorded
/ correction: the structured data typed this record as a natural person, and gave a key's day as the day a row page was modified. the front page's Dataset typed its creator as a Person; each row page's Article gave the row's last_verified day, where one was recorded, as the day the page was modified → the creator is typed as the Organization the row pages already name as publisher; each row page's dateModified is the record's day, as the front page's is identity/URL
59 routes recorded
/api/v1/schema/gxp-obs-v2.json correction: GXP-OBS version 2 said every version 1 field statement was carried unchanged, while two had been replaced by the signal's, and its floor refused every evidence state. content_hash and instrument_id in version 2 carried the signal's statement only; the floor was version 1's required list alone → each of the two names carries every family's statement as one branch of anyOf, version 1's first and unchanged; the floor is stated per family none: no part of the rule holds for this entry’s fields
routes recorded: /api/v1/schema/gxp-obs-v2.json · /developers
/developers/ correction: 49 of 97 redirect rules named one spelling; modelled against the build, the other reached no page, 7 of them published URLs. The map now names both; the live site was not read. 49 of 97 redirect rules named one spelling of a route; under this record's model of the edge, the other spelling reached no built page → under that model, each of the 142 literal sources and each of their 128 other spellings reaches a page this build serves; no request to the live site was made presentation
17 routes recorded
/verify/ correction: from 2026-09-05 /verify linked the key rotation procedure to a section of /methodology that the site no longer served, so the link led to /how, which carries no such section. how a key would be rotated, linked to /methodology#keys → how a key would be rotated, linked to #keys on /verify, where the procedure now stands presentation
route recorded: /verify/
/how/ correction: three sentences on /how and /developers said only a person writes a status, while two published statuses rest on recorded machine reads. a status is written by a person pressing a key, and no machine writes a status here; every published status written by hand; statuses change only when a person confirms one against the primary source → every confirmation states the act that made it none: no part of the rule holds for this entry’s fields
routes recorded: /how/ · /developers/
/changes/ correction: from 2026-10-11 /changes said no candidate had been confirmed by analysis, while the record held derived acts confirming revisions of staged candidates. No candidate has been confirmed by analysis. → each candidate a recorded act of analysis confirms is listed under confirmed revisions, with the act's basis and day identity/URL
route recorded: /changes/
/changes/ correction: /changes counted a candidate a person had recorded as a change of the witness among those with no receipt, so from 2026-09-15 the count ran one above the receipts. 40 change candidates pending review → 39 pending review, and 1 was recorded by a person as a change of the witness, the publisher’s page about the document none: no part of the rule holds for this entry’s fields
route recorded: /changes/
/archive/ correction: an address minted by a web archive was labelled on /archive in words that read as a copy this record keeps, and five record pages cited an address this site does not serve. a web archive's capture address was labelled in this record's own archive words: a column headed “Archive” with a link reading “archived”, a field named “Archive copy”, “not yet archived” where no address was held, a sentence saying an independent archive holds a capture, and a note naming “the archive link” as the remaining route for five dead links of which one holds an address; the opening paragraph gave every record a SHA-256 of the file as retrieved, where 5 of 18 carry none, and the head description called every record a single retrieval with one SHA-256; five record pages cited an address with spaces in it that this site does not serve → /archive defines observation, retained snapshot and third-party capture apart, names the column and the field “Third-party capture”, labels the link as an address at the archive's host, says this record holds the address only and has not read what it serves, states “none held” where no address is held, gives the SHA-256 only where a file was retrieved, in the opening paragraph and the head description alike, counts capture addresses in their own section, and each record page cites its own address identity/URL · wording
20 routes recorded
/ correction: the tracked total was printed without its tiers, and the two observation totals each appeared without the other. Every page now states both partitions. the tracked total was printed on 238 of 998 pages; only the home page named both tiers' figures, and no page said how many core values the schema supplies. The raw and reader-facing observation totals were printed apart, each without the other → every page states the tracked total by tier, with how many core rows write core and how many write no tier and are filed core by the schema, and both observation totals with their split presentation
256 routes recorded
/candidates/ correction: the state of a change candidate with no decision on record named a person as what the candidate was short of. It now reads pending review, as Program 7 2.5 orders. a change candidate with no receipt recording a decision on it was in the state “pending human review”, and /changes counted such candidates as awaiting “a person’s reading” → the same candidates are “change candidates pending review”, each in the state “pending review”; the words state that no receipt records a decision, and nothing else wording
42 routes recorded
/graph/ correction: the line over citation edges, and the definition behind it, named them as associations this record drew. A citation edge is a publisher's own text naming another document. the line over citation edges, on the citation graph and in the citations section of each entry page that has one, named them “contextual mapping”, the words this record uses for an association it drew; /independence#mapping counted a citation edge among those associations → the line over citation edges reads “publisher-stated citation relationships — not a regulatory applicability determination”; /independence#mapping no longer counts a citation edge as an association this record drew, and the edge's class is stated at /independence#citations wording
39 routes recorded
/explore/ correction: two pages used a word for a surveyed field. /explore now calls what this record reads its coverage, as Program 7 2.5 orders; /crawlability states its limit without the word. /explore opened by naming the publishers, jurisdictions and documents this record reads as a “landscape”, and /crawlability closed a paragraph with the same word → /explore opens “This record’s coverage:” before the same figures, and /crawlability says that two agents cannot describe how access varies across agents in general wording
routes recorded: /explore/ · /crawlability/
/ correction: from 2026-09-06 the front page headed its verification split "What this record can vouch for"; on 2026-10-10, 76 of 132 tracked entries under it sat in states naming no check by a person. What this record can vouch for → Evidence coverage wording
route recorded: /
/observatory/ correction: a window or series in which no digest differed was worded as if nothing had changed. What this record measured is that it detected no difference; the words now say so. where no observation in a window or series differed from the one before it, entry pages, authority pages, the observatory, the coverage contract and the observation packets worded the result as “no change” → the same results read “no detected difference”, or say that no difference was detected, on every surface that renders them wording
279 routes recorded
/r/anvisa-ai/ correction: 132 of 132 row pages put this record's status under the publisher's label; 125 had no read, among them 57 of 58 with a status other than unknown. In place from 2026-09-07 (300b4d2bd). Document status, what the publisher asserts: in force as of 20 August 1997 → Publisher lifecycle field: none captured · Recorded status: in force as of 20 August 1997 (basis: the act named under Established by) wording
132 routes recorded
/r/eu-digital-omnibus-ai/ correction: 5 row pages drew Effective date as a dash while the calendar entry each lists under Dates dates one. In place from each entry's admission, 2026-07-29 to 2026-08-04. Effective date: — → Effective date: 20 August 1997 (basis: a calendar entry citing its source, date confidence confirmed; listed under Dates) presentation
5 routes recorded
/corrections/ correction: 6 nightly deploy receipts from 2026-09-25 to 2026-10-03 named a build no upload carried, and two corrections cited a receipt as a delivery. a deploy receipt's build id computed from the commit at the deploy gate, and cited in two corrections as the build a deploy served → the receipt reads its build id from the dist/ the upload sends, and a deploy is called deployed only when the wire serves that id none: no part of the rule holds for this entry’s fields
route recorded: /corrections/
/ correction: the board's 90-day bar worded its “!” mark as the document changing, on 38 of 132 rows whose own tail read no confirmed change. The mark records that a digest differed. the “!” mark read “the document changed” on the board, and “content changed” on entry and observation pages; on entry pages each mark’s name gave only a date and a sequence, and the count of marked records read as changes to the retrieved text → the “!” mark reads “digest differed” on every surface that draws, defines or counts it, each mark’s tooltip and accessible name included wording
routes recorded: / · /r/us-21-cfr-part-11/ · /observations/us-21-cfr-part-11/
/how/ correction: /how/ said differing digests meant differing text; 105 readings differ because the thing hashed moved, and 11 more were compared across a tag-reader change. two days with different digests were said not to have seen the same text, with no condition on the rules or the thing hashed → holds only for digests under the same rules over the same thing; a move of the thing hashed is marked from 2026-09-08, a move of the rules from the first stamped nightly none: no part of the rule holds for this entry’s fields
route recorded: /how/
/packet/anvisa-ai/ correction: from 2026-09-08 each packet page said every byte of its archive is served at its own address, which the JSON never was, and on 6 of 132 pages named a leaf slice the archive lacks. each packet page: the archive holds this row's slice of the published leaf file and the published root, every byte of it is served at its own address as well, and it originates nothing → each packet page lists its own archive's files from the list the archive is written from, and says what each file is; a row with no leaf line has no leaf file listed none: no part of the rule holds for this entry’s fields
132 routes recorded
/r/iec-62304/ correction: from 2026-09-15 the observed cadence on 28 row pages counted days withheld under a host's filed terms as days the address was reached; on 18 it named an interval ending on one. intervals measured between every day holding an observation, a day holding only records withheld under a host's filed terms counted as a day this address was reached → intervals measured between the days on which a request was made, and the withheld days stated as their own class none: no part of the rule holds for this entry’s fields
28 routes recorded
/ correction: from 2026-09-15 the board's 90-day bar drew days withheld under a host's filed terms as could not be retrieved: 480 days on 28 rows by 2026-10-03. No fetch was made on any of them. a day on which no fetch was made, the host's filed terms withholding it, drawn as a day that could not be retrieved → that day drawn with its own mark, / withheld by the publisher’s terms — no fetch made presentation
route recorded: /
/api/v1/timemap/iec-62304.json correction: from 2026-09-15 the TimeMaps of 28 rows named their original as the address that was asked, from a newest record withheld under a host's filed terms, on which nothing was asked. original_basis: url, the address that was asked — this record holds no answered address for the observation it was taken from → original_basis: url, the address the observation names — that observation was withheld by the publisher’s terms — no fetch made, so this record asked no address and none answered wording
28 routes recorded
/observations/iec-62304/ correction: from 2026-09-15 the observations and row pages of 28 rows said no response arrived of records on which no request was sent, each withheld under a host's filed terms. a record written in place of a fetch cited as a request to which no response arrived → that record cited as withheld by the publisher’s terms — no fetch made, with no request sent presentation
56 routes recorded
/ correction: from 2026-09-15 records withheld under a host's filed terms were counted and dated as fetches on the front door and board, 28 row pages, 9 authority pages and their index. records written in place of a fetch counted among the fetches that answered and the entries fetched, and dated as the day a document was read, checked or accessed → those records counted apart, as withheld by the publisher’s terms — no fetch made, and every such date taken from a record on which a request was made figure · presentation
40 routes recorded
/r/iec-62304/ correction: from 2026-09-15 the observation strip on 28 row pages drew withheld records as unchanged: 480 marks by 2026-10-03. No fetch was made on any of them, so nothing was compared. a record written in place of a fetch, drawn in the observation strip as · unchanged → that record drawn as / withheld by the publisher’s terms — no fetch made presentation
28 routes recorded
/ correction: the board's 90-day bar said no check ran on 36 days over 9 rows, 2026-09-14 to 09-17, when the step fetched each document and every record it tried to write was refused. the “?” mark of the board’s 90-day bar read “no check ran that day” → the “?” mark reads “no request for the document recorded that day” wording
route recorded: /
/ correction: from 2026-09-15 the board's 90-day bar drew 126 days on 6 rows as days no check ran, though the store holds a request for the document on each; 12 of the 126 from 2026-09-23. the board’s 90-day bar drew “no check ran that day” on any day holding no record classed as the document, including days on which a request for the document was sent → a day on which the store holds a request for the document and no digest of it is drawn “could not be retrieved — nothing hashed”, except a day whose request was answered from an address not declared as the document’s and hashed there, which is drawn with a mark of its own, “answered from an address not declared as the document’s — that answer hashed”; the “?” mark falls only on a day on which the store holds no request for the document, and its words say so none: no part of the rule holds for this entry’s fields
route recorded: /
/papers/ correction: from 2026-07-30 /papers/ described a deposit and a DOI as owed to its preprint and dataset. No deposit service is used; the release is the citable object. /papers/: two sentences described a deposit with an outside service, and a DOI from one, as owed to the preprint and to the dataset → /papers/: both removed; no deposit service is used, and the citable object is this record's own release none: no part of the rule holds for this entry’s fields
route recorded: /papers/
/r/us-21-cfr-part-11/ correction: from 2026-09-25 the front door set one entry apart under “Start with a document”, and no page said why. The entry's page now says. the front door set one entry apart under “Start with a document”, and neither page said why → the entry's page says why it was chosen, what each reason rests on and where one stops, and that being chosen is not a statement of weight presentation
routes recorded: /r/us-21-cfr-part-11/ · /
/bundles/ correction: /bundles/ published five named bundles from 2026-09-08. Each began as a name, and a rule was fitted to it: a grouping this record authored. Retired; the addresses answer as dated pages. /bundles/: 5 named bundles, each a phrase from the filed order of 2026-09-07 with a rule read out of it; 3 had pages: 2 listed the entries their rule selected, 1 stated that its rule selected none, and 2 linked their rule as a board lens → /bundles/: the lens query documented; the 3 bundle addresses answer as dated retirement pages carrying each rule and count, and no retired rule is offered as a lens; the 2 names that never became a rule, AI in biopharma and OT cybersecurity, are named there with the words of each that no facet carried none: no part of the rule holds for this entry’s fields
5 routes recorded
/bundles/ correction: /bundles/ said a second, uncited classification field existed and was not read there. From 2026-09-18 that field held only cited publisher subjects. /bundles/: the corpus records "a second, looser set of classification values" that "carry no quoted span and no citation", and whether they may be published is an open question → withdrawn: from 2026-09-18 that field, topics, holds only each entry's own subject_as_stated values, every one carrying the publisher's quote and address none: no part of the rule holds for this entry’s fields
route recorded: /bundles/
/ correction: from 2026-09-06 the board's compact-rows control hid two published lines in its rows: each row's jurisdiction, and the reason given on each row marked not re-checked. Both now stay. compact rows on the board hid the jurisdiction line of every row and the reason line of every row marked not re-checked, and drew the 90-day bar shorter → compact rows set those two lines smaller and closer, hide nothing, and keep the 90-day bar at its height presentation
route recorded: /
/topic/ correction: topic identifiers were described as a scheme this record keeps; they are record identifiers for publisher-stated values. Two scheme words also leave this record's own voice. topic identifiers named "controlled identifiers" on 132 entry pages and five other pages, "controlled concept identifiers" on /crosswalk, "the topic vocabulary" on /conventions, /observatory and 77 retirement pages, and "a second vocabulary" on /activity; "ontology" or "taxonomy" in this record's own voice on six surfaces and in a comment in the deployed redirect rules → record identifiers for publisher-stated values, in words read from the binding count on each build; neither word in this record's own voice anywhere in the built site wording
221 routes recorded
/how/ correction: /about, written without a trailing slash, answered 404 on 2026-09-24 while /about/ redirected to /how/. Both forms now redirect there. /about answered 404; only /about/ redirected to /how/ → /about and /about/ both redirect to /how/ presentation
routes recorded: /how/ · /about (redirected, 301, to /how/)
/concordance/ correction: four concordance headings were not exact words, letter case kept, in every listed text and are retired; five stay; the claim that listed entries address one subject is withdrawn 9 headings, each grouping a singular with its plural and ignoring capitals, published as sections of different instruments that address the same subject → 5 runs of exact words, letter case kept, held by the retained text of all 15 listings under them in the store read to 2026-09-23; 4 retired on 2026-09-24 with dated notices at their own addresses figure · wording
35 routes recorded
/developers/ correction: /developers said the audit surfaces it lists were not indexed or linked from the reader's navigation. All four were in the header navigation, and two were indexed. they are not indexed or linked from the reader's navigation → They are live and permanent, and each keeps its address. none: no part of the rule holds for this entry’s fields
route recorded: /developers/
/how/ correction: /how named a list of regulations that no control in the build writes, and omitted the two keys the Developers shelf writes; the sentence is now composed from the key list. Two controls on the board → every control that saves a value is named with its page and key, the board's new orientation dismissal included; /developers says the fourth condition is carried none: no part of the rule holds for this entry’s fields
routes recorded: /how/ · /developers/ · /
/integrity/ correction: the RECONCILIATION family on /integrity gains invariants whose findings can name a page as well as an entry; it reports now and withholds the deploy from its first measured zero. Each finding names the entry it is about → Each finding names the entry or the page it is about and links to it none: no part of the rule holds for this entry’s fields
route recorded: /integrity/
/llms.txt correction: /llms.txt claimed a date on every entry, one value and one display word for every status not verified, one header set on every address, and surfaces past their own pages. Every page shows a last-updated date; every instrument shows when it was last verified → the count of instruments the board marks verified, each with the date of that check none: no part of the rule holds for this entry’s fields
route recorded: /llms.txt
/ correction: no section row marked the page a reader was on — 0 of the 19 pages carrying one, measured at 8e623e1f. The comparison set a slashed link against an unslashed path. a section row that marked no page as the one being read → each row marks the page a reader is on presentation
route recorded: /
/ correction: the front page's heading, title, descriptions and share cards described this record as broader than its positioning sentence: no selection stated, and standards bodies called regulators. a dated record of what regulators published → a dated record of what selected regulatory and standards sources published none: no part of the rule holds for this entry’s fields
routes recorded: / · /llms.txt · /og/gxplex.png
/verify/ correction: the description of /verify, shown in search results and link previews, said every command on the page runs on every build. One does not. The description was replaced whole. Recompute the archive's daily integrity root from the published data. Every command on this page is run against the live artifacts on every build. → This record's integrity root with its date and manifest, whether a third party has timestamped it, what this proves and what it does not, and the shell commands that reproduce it. none: no part of the rule holds for this entry’s fields
route recorded: /verify/
/verify/ correction: /verify's list of published files said every command on the page reads root.txt. The recompute does not; it reads leaves.txt, and its output is compared with root.txt. every command below → the timestamp check and the OpenTimestamps check; the recompute’s output is compared with it none: no part of the rule holds for this entry’s fields
route recorded: /verify/
/verify/ correction: /verify addressed its reader in the second person in seven places, and four of its sentences told the reader what to do, one in its description. The new text addresses no one. (i) Checks you can run yourself → (i) Checks a reader can run alone wording
route recorded: /verify/
/verify/ correction: /verify said its timestamp check prints the time the token carries. It prints a verdict and no time. The time stays inside the token, and the page shows its date. the command below prints the time it carries → prints a verdict and no time; the instant stays inside the token none: no part of the rule holds for this entry’s fields
route recorded: /verify/
/developers correction: two pages derived, correctly, that this build served no TimeMap address. A narrowed one is now served, and both sentences change with the route set they are read from. none: no part of the rule holds for this entry’s fields
routes recorded: /developers · /api/v1/timemap.json
/api/v1/series.jsonl correction: 117 published lines restated a publisher's reservation of rights in this record's own voice. The published field now carries a state; the observation store keeps what it recorded. terms_withheld carried a prose cause restating a publisher's terms → terms_withheld carries the host and the filed state; the prose stays in the store none: no part of the rule holds for this entry’s fields
route recorded: /api/v1/series.jsonl
/crawlability/ correction: an attribution of a public-domain dedication to EUR-Lex content was withdrawn before filing. The named page carries no such sentence; publications.europa.eu is filed unclear. a public-domain-dedication sentence, expected on the Publications Office copyright notice → no such sentence is on that page; the sole token match is inside a CSS colour none: no part of the rule holds for this entry’s fields
route recorded: /crawlability/
/crawlability/ correction: 189 cells published "no" where no request was made, which a reader takes as a refusal by the publisher. Both columns now name which of four a non-request is. a row with no recorded status rendered as "no" in the accessible columns → each such row names which kind of non-request it is none: no part of the rule holds for this entry’s fields
route recorded: /crawlability/
/crawlability/ correction: the crawlability census now asks each host's permission basis before it opens a socket. 34 rows are withheld. The figures on this page changed meaning; no publisher changed anything. 132 rows probed and reported crawlable or not crawlable → 97 probed; 34 withheld on a recorded permission basis and reported unmeasured none: no part of the rule holds for this entry’s fields
route recorded: /crawlability/
/crawlability/ correction: iris.who.int publishes a restriction on automated use. Its 2 row(s) are withheld from collection as of 2026-09-19. The retained bytes are kept and not displayed. 2 row(s) on iris.who.int fetched on this record's own initiative → withheld — the publisher's terms restrict this act none: no part of the rule holds for this entry’s fields
route recorded: /crawlability/
/crawlability/ correction: ispe.org publishes a restriction on automated use. Its 4 row(s) are withheld from collection as of 2026-09-19. The retained bytes are kept and not displayed. 4 row(s) on ispe.org fetched on this record's own initiative → withheld — the publisher's terms restrict this act none: no part of the rule holds for this entry’s fields
route recorded: /crawlability/
/crawlability/ correction: store.astm.org publishes a restriction on automated use. Its 2 row(s) are withheld from collection as of 2026-09-19. The retained bytes are kept and not displayed. 2 row(s) on store.astm.org fetched on this record's own initiative → withheld — the publisher's terms restrict this act none: no part of the rule holds for this entry’s fields
route recorded: /crawlability/
/crawlability/ correction: webstore.iec.ch publishes a restriction on automated use. Its 1 row(s) are withheld from collection as of 2026-09-19. The retained bytes are kept and not displayed. 1 row(s) on webstore.iec.ch fetched on this record's own initiative → withheld — the publisher's terms restrict this act none: no part of the rule holds for this entry’s fields
route recorded: /crawlability/
/crawlability/ correction: www.isa.org publishes a restriction on automated use. Its 1 row(s) are withheld from collection as of 2026-09-19. The retained bytes are kept and not displayed. 1 row(s) on www.isa.org fetched on this record's own initiative → withheld — the publisher's terms restrict this act none: no part of the rule holds for this entry’s fields
route recorded: /crawlability/
/crawlability/ correction: www.pmda.go.jp publishes a restriction on automated use. Its 1 row(s) are withheld from collection as of 2026-09-19. The retained bytes are kept and not displayed. 1 row(s) on www.pmda.go.jp fetched on this record's own initiative → withheld — the publisher's terms restrict this act none: no part of the rule holds for this entry’s fields
route recorded: /crawlability/

How each correction is placed in a category

Each entry above is placed by the fields it records — its change type, its stated class, and the values it records before and after — and never by reading its note. It stands under every category whose rule holds, so one entry can stand under two; where no rule holds, the entry says so and stands under none.

status/date
change_type is status_change or date_change; or from or to is, whole, a calendar day
identity/URL
correction_class names referent; or from and to are both, whole, web addresses
figure
correction_class names figure
wording
correction_class names wording or account
provenance
— no field an entry records states this, so no entry is placed under it
presentation
correction_class names rendering

Stated classes that no rule above reads: derivation, self_description. An entry stating only these stands under a category only where another rule holds for it.

Days on which no request for a document was recorded

Each line is a run of days, inside an entry’s window, on which the store holds no request for the document that entry names. The days are counted from the store, from each entry’s first observation classed as its document to , the last day the store holds, over the 132 entries that hold such an observation. The reason for each run is stated beside it, with what the daily log records on the same dates: where each nightly run stopped, and what it recorded for the step the reason is about.

Runs of days with no request for the document, by first day
DaysEntries What the store holdsWhy What the daily log records
to , and continuing: the cause stands on the last day the store holds nist-csf 94 entry-days with no request for the document recorded; on each, another address of the entry was read The nightly reads this entry's source address, the publisher's page about the framework. The address of the document the entry declares was read on 2026-07-31 only, when the entry was added, and no nightly step reads it. Which of the two addresses the entry declares is an open question filed for decision, and until it is answered the document's own address has no nightly reader. From 2026-08-01 to 2026-09-28, read as of the night of 2026-09-28: 129 runs on 59 of the 59 dates; 74 completed, 22 stopped at tests, 11 stopped at build, 10 exited 130, 5 exited 1, 5 stopped at gatecheck, 2 stopped at lint. The step “observe — documents” is recorded as ok in 107 runs, not run in 17 runs, degraded in 5 runs.
to eu-gmp-annex-11, fda-ai-guidance, fda-guidance-agenda, ich-q12, ich-q13, ich-q14, ich-q2r2, ich-q9r1, nist-ai-rmf 45 entry-days with no request for the document recorded; on each, another address of the entry was read The step that reads the address of the document each of these entries declares had been run once, by hand, on 2026-09-07, and the nightly run did not call it. On each of these nights the nightly read another address of the entry and not the document's. The step joined the nightly on 2026-09-12 and first ran there on 2026-09-13. From 2026-09-08 to 2026-09-12, read as of the night of 2026-09-28: 5 runs on 5 of the 5 dates; 4 completed, 1 stopped at tests. The step “continue document series” is recorded as not run in 5 runs.
iec-62304, iso-13485, iso-14971, iso-81001-1, iso-iec-22989, iso-iec-23894, iso-iec-24028, iso-iec-27001, iso-iec-27002, iso-iec-27017, iso-iec-27017-2026, iso-iec-27018, iso-iec-27018-2025, iso-iec-38507, iso-iec-42001, iso-iec-42005, iso-iec-5259, iso-iec-sc42-programme 18 entry-days with no request for the document recorded; on none of them does the store hold any record of the entry These entries' host has filed terms that withhold automated collection, and on this night the collector was to write a record in place of a fetch for each. It built each record and returned before writing it, so the store holds no record of any of these entries for the day. The collection step is recorded as degraded that night. The branch was repaired the same day, and from 2026-09-15 each night holds the record. From 2026-09-14 to 2026-09-14, read as of the night of 2026-09-28: 1 run on 1 of the 1 date; 1 stopped at build. The step “observe — documents” is recorded as degraded in 1 run.
to eu-gmp-annex-11, fda-ai-guidance, fda-guidance-agenda, ich-q12, ich-q13, ich-q14, ich-q2r2, ich-q9r1, nist-ai-rmf 36 entry-days with no request for the document recorded; on each, another address of the entry was read The step ran on each of these nights and fetched each document, and every record it tried to write was refused. From 2026-09-13 the store accepted a write only from a caller marked as the collection pass, and this step had not been given the mark. The daily log records the step as ok on every one of these nights, because a refused write did not fail the step. The step was given the mark on 2026-09-18, when an attended run wrote that day's reading of all nine. From 2026-09-14 to 2026-09-17, read as of the night of 2026-09-28: 4 runs on 4 of the 4 dates; 2 stopped at build, 1 completed, 1 stopped at tests. The step “continue document series” is recorded as ok in 4 runs.

The inferences this record declines to draw — each with the gate that refuses it and the correction on this page that dated it — are listed at what this record refuses to infer.

Reporting an error

corrections@gxplex.com — please include the page and the primary source that contradicts it. No form, no tracking, no third party.

Last updated: