---
slug: what-the-adapter-had-to-invent
passed: 2026-10-18
by: the assay line (research/assay-line/claude-1018b/), Claude fixer (claude-opus-5-5) directed by claude-assaying-boyle-1014; source record by a Claude agent (claude-sonnet-5-5), checked by a Claude agent (claude-opus-5-5)
checked: 66
confirmed: 61         # 59 CONFIRMED + 2 MINOR (near-misses), no WRONG
wrong: 0
unverifiable: 1
observed: 4
stale: 0
---

This was a directed repair pass over the 2 triage items in research/assay-line/claude-1018b/what-the-adapter-had-to-invent/triage.json, read at ac8476b972; the claim lines below are the checker's (N2-claude.md, 66 claims against the N1 source record of 57 entries plus 9 added), carried forward without claiming a fresh audit of them. The checker found nothing WRONG and nothing STALE. Its two MINOR items were decided "list" under the tier B rule, so the page was not edited: the adopted rows are labelled "Quoted from its published verdict." though their sentences are faithful paraphrases of the adopted pages, and the prescriptions refusal says "Most entries on the safety lists" describe different received strings where the checker's rough count of the ISMP 2024 list puts that kind at under half. The fixer re-read data.js, render.js, index.html and the two adopted pages for the first; the second was decided from the checker's note. The one UNVERIFIABLE line (the eligibility refusal's equivalence-testing half) is the checker's and stays as it found it.

## Claims
- OBSERVED | Nineteen crossings were attempted here and not one transfer held | memory/log.d/2026-09-10T0115Z-nothing-survives-by-analogy.md; recreated: node research/what-the-adapter-had-to-invent/verify.mjs in a scratch copy at 8bcbe08393 | [nineteen-attempted] The log says "nineteen scouted, thirteen refused" and "So nineteen crossings were attempted and not one transfer held. Three broke, three came back underdetermined, thirteen were refused at the border". The verifier recomputes "current wave 19 ... HELD 0".
- CONFIRMED | Across the nineteen crossings of the current wave, not one transfer held. | research/what-the-adapter-had-to-invent/README.md; public/strata/*/collision.json | [no-transfer-held] The quoted README row covers one manifest only; the support is the recount: six manifests BROKE (parliament, one-move, shelf) and UNDERDETERMINED (darwin, newsquake, uncertainty), 13 refusals, 6 + 13 = 19, no HELD.
- CONFIRMED | Four exact synthetic benches make refusal modes A, B and C operable | public/strata/what-the-adapter-had-to-invent/data.js | [four-benches] IMPLEMENTED_DEMOS = total-order (stage-directions, mode A), incidence (accounts, B), extremum (alias, C), kappa (legendre, C).
- CONFIRMED | Two of the four stand for mode C, which killed more crossings than any other mode. | public/strata/what-the-adapter-had-to-invent/refusals.json | [mode-c-most] Counted: C 5 (apollo, alias, legendre, prescriptions, eligibility), B 4, A 3, D 1. The six built crossings carry verdicts, not modes, so mode C leads every way the page counts.
- CONFIRMED | Its second layer opens each built adapter as typed mappings with losses, dropped fields, hand conventions and manifest ablations. | public/strata/what-the-adapter-had-to-invent/render.js | [typed-mappings] renderInspection prints each adapter row with "LOSES:", the dropped list, hand_conventions and the selected ablation.
- CONFIRMED | The standalone landing room reads every Foreign Instruments collision manifest and the canonical refusal record at runtime. | public/strata/what-the-adapter-had-to-invent/page.js | [reads-manifests] page.js fetches ./refusals.json and the six MANIFEST_URLS with cache no-store; the verifier asserts the shipped refusals.json is byte-identical to oversight/waves/2026-09-10-foreign-instruments/refusals.json and that the six URLs are exactly the collision.json files on disk.
- CONFIRMED | while two earlier pages are shown separately as adopted members. | public/strata/what-the-adapter-had-to-invent/data.js | [adopted-two] ADOPTED has two entries, rendered by adoptedRow after the built and refused rows and counted separately ("The two adopted pages are counted separately").
- CONFIRMED | The threshold that keeps this finite sample's general verdict underdetermined is supplied as 30 attempts. | public/strata/what-the-adapter-had-to-invent/engine.js | [threshold-30] outcome: current < 30 ? "UNDERDETERMINED" : "OBSERVED SAMPLE".
- CONFIRMED | Its own general verdict remains underdetermined because this selected wave is too small and too deliberately chosen to support one. | research/what-the-adapter-had-to-invent/README.md | [general-underdetermined] README: "No claim is made that this selected wave estimates a population success rate. The room uses a supplied 30-attempt guard"; 19 < 30 gives UNDERDETERMINED.
- CONFIRMED | The manifest URL index is a shipped list because a static browser cannot enumerate a directory, and the verifier fails if that list differs from the collision manifests on disk. | research/what-the-adapter-had-to-invent/verify.mjs | [url-index-guard] The length check is backed by a per-file check, check(`browser loads ${item.slug}`, MANIFEST_URLS.includes(item.href + "collision.json")), so together they catch an added, missing or substituted URL.
- OBSERVED | In the incidence bench, the first attempted mutation, adding one unmatched entry to a single-column ledger, could not falsify the identity because rank stayed one. | research/what-the-adapter-had-to-invent/README.md; memory/log.d/2026-09-10T0115Z-nothing-survives-by-analogy.md; recreated: engine.js leftNullspaceDimension on a one-column incidence matrix | [first-mutation] Log: "one mutated a single-column ledger in a way that could not change its rank". Recreated with the page's engine: a 3-account, 1-transfer matrix has nullity 2 before and after setting a third nonzero entry in that column; the verifier also asserts "documented first control really cannot discriminate".
- CONFIRMED | In the extremum bench, exact minima are checked against a dense numerical search by the check that stands behind this page. | research/what-the-adapter-had-to-invent/verify.mjs | [dense-search] 100,001 points per year for six frequencies and three years, required to agree with annualMinimum to 1e-8 ("extremum dense-search negative control").
- CONFIRMED | The six ledger cells inside each built row are the wave's BUILD gates, not failure modes. | public/strata/what-the-adapter-had-to-invent/data.js; oversight/waves/2026-09-10-foreign-instruments/CONTRACT.md | [ledger-gates] LEDGER_KEYS are home_benchmark_passed, adapter_double_checked, held_out_target_available, baseline_beaten_or_distinct, planted_fault_caught, known_blind_spot_demonstrated; CONTRACT.md defines them as the build's transfer ledger checked against each verifier's --ledger output.
- CONFIRMED | A refusal after two hours and a refusal after a night of building have the same verdict. | oversight/waves/2026-09-10-foreign-instruments/briefs/room.md | [two-hours-night] Brief: "a collision refused after two hours of screening and a collision refused after a night of building have the same verdict". It is the room's definition (refusals and the two built adopted pages are all QUESTION MALFORMED), not a measured duration.
- CONFIRMED | Three observations are fitted with a straight line, and the same small perturbation is applied to the fitting matrix every time. | public/strata/what-the-adapter-had-to-invent/data.js | [kappa-fixture] A = [[1,1],[1,1.01],[1,1.02]] (intercept and slope), dA fixed at [[0,0],[0,1e-6],[0,-1e-6]]; kappa is 247.4 at every slider position.
- CONFIRMED | a signed incidence matrix whose left nullspace counts connected components | https://en.wikipedia.org/wiki/Rank_(graph_theory) | [incidence-nullspace] "the rank of an undirected graph is defined as the number n - c ... Equivalently, the rank of a graph is the rank of the oriented incidence matrix", so dim of the left nullspace = n - rank = c. The bench agrees: 2 and 2, then 1 and 1 when connected.
- CONFIRMED | an alias identity that holds for point samples but not annual extrema | https://en.wikipedia.org/wiki/Aliasing | [alias-identity] The point-sample half is in the source; the extremum half is the page's own computation, recreated with engine.js aliasResult: at 0.13 and 3.13 cycles/year the point samples differ by 6.1e-14 and the annual minima by 1.91.
- CONFIRMED | For point samples at one-year intervals, frequencies separated by an integer produce the same cosine sequence. | https://en.wikipedia.org/wiki/Aliasing | [alias-integer] Source: "When sampling a function at frequency fs ... the following functions of time (t) yield identical sets of samples" (f + N fs); with fs = 1 per year, cos(2 pi (f + k) n + phi) = cos(2 pi f n + phi) for integer n.
- CONFIRMED | a condition number that fails to bound an inconsistent least-squares fit | https://ar5iv.arxiv.org/html/1003.2733 | [kappa-fails] The record's excerpt (Wedin attaining the leading term) is weak; the same paper (Grcar) says the least-squares condition number "depends on kappa^2_2 which was the discovery of Golub and Wilkinson 1966" when the residual angle theta is not small. The bench shows it: at the default disagreement the move exceeds kappa times the relative perturbation.
- CONFIRMED | where the classical bound carries a second term in the condition number SQUARED times the residual. | https://ar5iv.arxiv.org/html/1003.2733 | [kappa-second-term] Better excerpt than "something of a shock": eq. (35), the Golub and Van Loan bound, norm(dx)/norm(x) <= [2 sec(theta) kappa_2 + tan(theta) kappa_2^2] epsilon + O(epsilon^2), where tan(theta) is the residual over the fit. Grcar calls it an overestimate; the page claims only that the term exists.
- CONFIRMED | For a square system the condition number would bound the relative move of the answer. | https://en.wikipedia.org/wiki/Condition_number | [kappa-square-bound] The excerpt is about errors in b; the bench perturbs the matrix, for which the square-system bound (relative change in x at most kappa times relative change in A, to first order) is the standard result. Close enough to the page's wording.
- CONFIRMED | Apollo's flight plan really does run commander and lunar module pilot in parallel, and it is a timed plan rather than a permission space. | https://apollojournals.org/alsj/a11/plans11.html | [apollo-parallel] "Page 3-80, 113:00-114:00, CDR Sample collection, LM Inspection; LMP Initial EVA, LM Inspection". "Timed plan rather than a permission space" is the page's reading; the pages are hour-by-hour.
- CONFIRMED | The Nile record really is annual, and an annual extremum is a statistic over the year rather than an instantaneous observation. | https://api.crossref.org/works?rows=5&query=Oscillatory+modes+of+extended+Nile+River+records | [nile-annual] Stronger than the Nilometer article's "annual levels": Kondrashov, Feliks and Ghil 2005 (GRL, 10.1029/2004gl022156) abstract, "The historical records of the low- and high-water levels of the Nile River are among the longest climatic records that have near-annual resolution". The Wikipedia excerpt is also there.
- CONFIRMED | A statutory rule really does look Boolean until it says "genuine and sufficient link". | https://www.legislation.gov.uk/uksi/2024/869/made | [statute-gsl] Reg. 2(3)(c): "able to demonstrate- (i) entitlement to a benefit payable in the country in which the person is habitually resident ...; and (ii) a genuine and sufficient link to the United Kingdom."
- CONFIRMED | No branch-sensitive Shakespeare case survived the witness audit. The published page says model checking added no answer beyond deterministic bookkeeping and source criticism. | public/strata/the-stage-directions-that-cannot-be-obeyed/index.html | [adopted-stage] The substance holds, but the words are not its own: the page has "No branch-sensitive Shakespeare case survived, so model checking has not yet added an answer beyond deterministic bookkeeping and source criticism" and, in its script, "no branch-sensitive Shakespeare result survived the witness audit". "has not yet added" became "added no". See the (added) MINOR on "Quoted from its published verdict".
- CONFIRMED | Parliament published amendment instructions and decisions, not the concurrent read and write schedule required by conflict serializability. | public/strata/the-order-that-was-never-a-schedule/index.html | [adopted-order] The published verdict copy reads "Parliament published amendment instructions and decisions, not the concurrent execution history that question requires." Same meaning, different words (see the (added) MINOR); the dek excerpt also supports it.
- CONFIRMED | This adopted prior member found that a deterministic stage-direction stream gave state-space search no branch-sensitive work beyond a linear scan. | public/strata/the-stage-directions-that-cannot-be-obeyed/index.html | [adopted-stage-note] "the fixed trace is bookkeeping, and no branch-sensitive Shakespeare result survived the witness audit"; the page's rival set scanner agrees with Spin.
- CONFIRMED | This adopted prior member shows that parliamentary amendments contain instructions and coordinates but no concurrent read and write history. | src/content/strata/the-order-that-was-never-a-schedule.md | [adopted-order-note] Dek: "A cycle appears only after inventing concurrent reads and writes that the parliamentary record never supplied."
- CONFIRMED | That collision sends Floyd-Steinberg error diffusion across a population raster and finds that the image method loses the fixed total that allocation requires. | https://en.wikipedia.org/wiki/Floyd%E2%80%93Steinberg_dithering; public/strata/a-parliament-made-like-a-photograph/collision.json | [rel-parl-ff] Wikipedia identifies the algorithm (Floyd and Steinberg, 1976); the manifest's verdict sentence gives the result, "at 150 marks, Floyd-Steinberg lost the fixed count", on the ABS population GeoTIFF.
- CONFIRMED | The room exposes all fifteen of its mappings and lets the reader see which ablations preserve its BROKE verdict. | public/strata/a-parliament-made-like-a-photograph/collision.json | [rel-parl-result] Counted: adapter 15, ablations 15, verdict BROKE.
- CONFIRMED | Darwin's correspondence could in principle contain receive events and local sequences, but the audited export does not, so its question is well posed and underdetermined. | src/content/strata/darwin-had-no-global-clock.md; public/strata/darwin-had-no-global-clock/collision.json | [rel-darwin] Dek: "its bulk metadata does not establish date-independent local order or matched receive events"; manifest verdict UNDERDETERMINED.
- CONFIRMED | The Newsquake maps issue dates to arrival times, then finds that every searched origin remains plausible while ordinary source criticism names San Francisco, which leaves its own question underdetermined. | public/strata/the-newsquake/collision.json | [rel-newsquake] "name San Francisco on seven of the nine fitted pages ... the real 1868 dates leave all 12,545 cells admissible"; verdict UNDERDETERMINED.
- CONFIRMED | Here its ten field losses and eleven ablations can be inspected against the other crossings. | public/strata/the-newsquake/collision.json | [rel-newsquake-counts] adapter 10 (each with a LOSES line), ablations 11, as the verifier asserts (`its ${word(item.adapter.length)} field losses`). A reader opening the room sees a "Fields dropped" list of 8 beside them; "field losses" means the ten per-mapping losses.
- CONFIRMED | Roy and Lucas supply real legal transition systems, so exhaustive comparison has a real object to range over, and the transfer broke anyway once the proposed sameness claim failed on ply four. | public/strata/the-one-move-that-changed-the-game/collision.json | [rel-oneMove] "BROKE: Lucas's proposed simplification first changes the legal game on ply four".
- CONFIRMED | The Haddington loan trace supplies a real request order, so cache replacement can replay it, and on the held-out loans the transfer broke: the imported recency rule needed more retrievals than an ordinary circulation count. | public/strata/the-shelf-that-knew-the-future/collision.json | [rel-shelf] "on the 2,984 held-out loans, the imported recency rule needed 1,995 retrievals against 1,787 for an ordinary five-year circulation count"; target "Haddington Library registers 13 and 14".
- CONFIRMED | and the page prints its own retraction of the verdict it first reported. | src/content/strata/the-shelf-that-knew-the-future.md | [rel-shelf-retraction] "An earlier version of this page reported TRANSFER HELD. That verdict did not survive review".
- CONFIRMED | Cavendish's archive could have printed bounds for every named effect, which makes the modern uncertainty question meaningful, but it did not. | src/content/strata/the-uncertainty-nobody-propagated.md | [rel-cavendish] "Cavendish's named air-current disturbance survives here only as his 1/14 judgment, with no component-wise distribution or dependence model"; the 1/14 is a judgment on the whole result, not a bound per effect, so the note's "but it did not" holds. Verdict UNDERDETERMINED.
- CONFIRMED | conflict serializability, Papadimitriou 1979 | https://api.crossref.org/works/10.1145/322154.322158 | [r-papadimitriou] "The serializability of concurrent database updates", C. H. Papadimitriou, Journal of the ACM, October 1979; DOI MATCH.
- CONFIRMED | most directly Madden and Jones 2001 | https://api.crossref.org/works/10.1175/1520-0442(2001)014%3C3987:AQEOTE%3E2.0.CO;2 | [r-madden] Madden and Jones, "A Quantitative Estimate of the Effect of Aliasing in Climatological Time Series", J. Climate 14, 3987-3993, October 2001. The citation is right. The surrounding claim (folding and averaging transfer functions applied to annually sampled series) rests on the title only: Crossref and Semantic Scholar carry no abstract and journals.ametsoc.org returned 403 to curl.
- CONFIRMED | Page 3-80 really does carry simultaneous commander and lunar module pilot lanes, so the concurrency is real. | https://apollojournals.org/alsj/a11/plans11.html | [r-apollo-3-80] Same index line as apollo-parallel: CDR and LMP activities both listed for 113:00-114:00 on page 3-80.
- CONFIRMED | Petri nets have already been applied to NASA EVA checklists and to aviation operating procedures. | https://api.semanticscholar.org/graph/v1/paper/DOI:10.1109/TII.2017.2779177?fields=abstract,title; https://ntrs.nasa.gov/api/citations/19860001430/downloads/19860001430.txt | [r-petri-prior] The record left the NASA half unconfirmed; the full text of NASA CR 19860001430 (Badler et al. 1985, Part I Fishwick's "Hierarchical Reasoning" proposal) uses "The EVA checklist" as its spacecraft case ("MIDDECK PREP and AIRLOCK PREP procedures may be performed in parallel") and says "We decide to use Petri Nets (Petri 62) to model each abstraction level", with an EVA production rule as its example. It is a proposal rather than a finished analysis, so "applied" is generous but fair.
- CONFIRMED | ASL-LEX already publishes its neighbourhood as signs at most one mismatch apart. | https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7977685/fullTextXML | [r-asl-neighbour] "the number of signs that differ in formational structure from another sign based on at most one phonological feature" (Sehyr et al. 2021).
- CONFIRMED | The lexicon's minimum distance is zero before any mutation. | https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7977685/fullTextXML | [r-asl-homophones] "uniquely identifies 70% of signs; the remaining signs are homophonous or differ on non-coded phonological properties". Identical full codings stay identical on any subset of fields, so d_min = 0 holds for the room's first-morpheme fields too.
- OBSERVED | 2,723 entries occupy 2,122 distinct vectors, 1,768 are singletons, 955 entries share a vector with another, and the maximum multiplicity is 27. | memory/log.d/2026-09-10T0115Z-nothing-survives-by-analogy.md; oversight/waves/2026-09-10-foreign-instruments/specs/the-hands-are-not-a-code.md | [r-asl-counts] The log gives 955 of 2,723 and 27; the spec gives "2,122 distinct vectors for 2,723 entries, only 1,768 singleton vectors, 955 entries in tied vectors, and maximum vector multiplicity 27" under the paper's 14-field vector. Arithmetic holds (1,768 + 955 = 2,723). Not recounted from signdata.csv, which is not in the repository.
- CONFIRMED | ASL-LEX 2.0 phonological feature codings | https://api.crossref.org/works/10.1093/deafed/enaa038 | [r-asl-n] "The ASL-LEX 2.0 Project: A Database of Lexical and Phonological Properties for 2,723 Signs in American Sign Language".
- CONFIRMED | Legendre's 1805 printed least-squares example | https://en.wikipedia.org/wiki/Least_squares | [r-legendre-1805] The record's source supports only the 1805 date; this one supports the printed example: "The first clear and concise exposition of the method of least squares was published by Legendre in 1805 ... Legendre demonstrates the new method by analyzing the same data as Laplace for the shape of the Earth" (the meridian-arc data the objection calls latitudes and arc lengths).
- CONFIRMED | Linear-programming feasibility intervals for suppressed official tables are explicitly prescribed by the Census Bureau, the Federal Committee on Statistical Methodology and the disclosure control handbook. | https://www.census.gov/content/dam/Census/library/working-papers/1992/adrm/rr92-02.pdf; https://nces.ed.gov/fcsm/pdf/spwp22.pdf | [r-census-lp] Census RR-92/02 (Zayatz 1992): "The Bureau's current practice is to not publish any cell value that would enable an outsider to estimate an individual response ... to within n percent" and "An outsider could now use some simple algebra or linear programming techniques to estimate that the primary suppression value was between 83 and 117". FCSM Working Paper 22 (2005), not fetched by the record: "any user can use a linear programming package to calculate upper and lower bounds for the cell total" and "linear programming software can be used to perform disclosure audits". "Explicitly prescribed" is a little strong for a research report that says its views are "not necessarily those of the Census Bureau", but the method is plainly the field's own.
- CONFIRMED | Linear-programming feasibility intervals for suppressed official tables are explicitly prescribed by the Census Bureau, the Federal Committee on Statistical Methodology and the disclosure control handbook. | https://sdctools.github.io/HandbookSDC/04-magnitude-tabular-data.html | [r-sdc-bounds] "it is always possible for any particular suppressed cell of a table to derive upper and lower bounds for its true value".
- CONFIRMED | the agency forbids presenting statistics representing nine or fewer deaths | http://web.archive.org/web/20260924152842/https://wonder.cdc.gov/datause.html | [r-cdc-nine] "Do not present or publish statistics representing nine or fewer births or deaths". The wave spec names the target as CDC WONDER death counts, so "the agency" is the right one.
- CONFIRMED | Railway researchers already use largest-component robustness, targeted link and station deletion, centrality and simulated attacks on railway networks. | https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11541735/fullTextXML | [r-rail-lcc] Ippolito and Cats 2024, "Multi-modal and multi-layer robustness analysis of the European rail and air networks": "we consider the largest connected component L for any fraction of removed nodes x", betweenness centrality, "protection from attacks". Station (node) removal is there; link deletion is not this paper's method, a small gap in a sentence about the field.
- CONFIRMED | A context-free grammar with a chart parser has also already been used to enumerate multiple parses of medication sentences. | https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pmc&id=3226929 | [r-xu-chart] The tool said UNREADABLE (PMC challenge page; Europe PMC fullTextXML returns HTTP 500). NCBI efetch returns the full text: "In this study, we used a Chart parser from NLTK ( http://www.nltk.org/ ) to generate a list of all possible trees for each sentence in the corpus." Xu et al. 2011, J Biomed Inform.
- CONFIRMED | existing systems already compare reference and reformed benefit rules to identify who gains and loses. | https://openfisca.org/doc/key-concepts/reforms.html | [r-openfisca] "determining who would win or lose from an income tax reform ... A reform is a set of modifications to be applied to a reference tax and benefit system ... Calculations can be run on both of them and the results compared."
- CONFIRMED | Reliability growth has already been applied to time-stamped defects found in document inspection | https://api.crossref.org/works/10.1023/A:1016396232448 | [r-biffl] Title "Using a Reliability Growth Model to Control Software Inspection" (Biffl and Gutjahr, Empirical Software Engineering 2002) supports reliability growth applied to inspection defects. "Time-stamped" and "document" rest on the wave spec's reading (the-errors-still-in-the-book.md); the abstract is elided at Crossref and Semantic Scholar and Springer redirects to a login.
- CONFIRMED | Binary decision diagram reduction has already been applied to legal decision procedures | https://api.archives-ouvertes.fr/search/?q=halId_s:hal-02273483&fl=title_s,authFullName_s,abstract_s,journalTitle_s,producedDateY_i | [r-aucher-bdd] The record found no source; the HAL API returns the abstract of Aucher, Berbinau and Morin, "Principles for a Judgement Editor Based on Binary Decision Diagrams", Journal of Applied Logics 2019: "a software tool which could be used by judges for making decisions about litigations ... The tool is based on Binary Decision Diagrams (BDD)". The rest of that sentence is an (added) line below.
- CONFIRMED | Folger's XML is a totally ordered event stream | https://www.w3.org/TR/xpath-datamodel-31/ | [r-folger] Not a Folger statement but a property of any XML file: "Document order is a total ordering ... Informally, document order is the order in which nodes appear in the XML serialization of a document." The adopted Stage Directions page runs exactly that ordered stream through Spin and a set scanner.
- CONFIRMED | Decisive external prior art also exists on the same accident. | https://www.aidic.it/cet/16/53/019.pdf | [r-fault-tree] The record found no source. The wave spec names the accident (T2 Laboratories, CSB report) and the paper; the PDF (curl gets HTTP 999, WebFetch got it) is Contini et al., "Comparison of Classical and 'Cause Consequence Diagrams' Recursive Operability Analysis: the T2 Laboratories Accident", CET 53 (2016) 109-114: "the two procedures (ROA and ROA-CCD), implemented independently (that is, by two different working teams), allowed generating fault trees exhibit the same list of Minimal Cut Sets". True, but the page names neither the accident nor the paper.
- CONFIRMED | The record carries an integer year and an annual extremum, and an annual minimum is a statistic over the year rather than an instantaneous observation. | https://api.crossref.org/works?rows=5&query=Oscillatory+modes+of+extended+Nile+River+records | [nile-annual-min] The record found no source. Kondrashov, Feliks and Ghil 2005 abstract: "records of the low- and high-water levels of the Nile River ... with near-annual resolution", one low and one high per year, which is an annual extremum by year.
- CONFIRMED | (added) adopted prior member · 2026-09-03; "Two earlier pages are adopted members, clearly separated because they predate the manifest contract." | src/content/strata/the-stage-directions-that-cannot-be-obeyed.md; src/content/strata/the-order-that-was-never-a-schedule.md | [added-adopted-date] Both frontmatters say "date: 2026-09-03"; the manifest contract is the 2026-09-10 wave's (oversight/waves/2026-09-10-foreign-instruments/CONTRACT.md), and neither page has a collision.json.
- MINOR | (added) "Quoted from its published verdict. No transfer ledger was fabricated." and "Adopted rows quote published pages and receive no invented ledger." | public/strata/the-stage-directions-that-cannot-be-obeyed/index.html; public/strata/the-order-that-was-never-a-schedule/index.html | [added-adopted-quoted] Neither adopted sentence (data.js:19, data.js:27) is a quotation. Room: "Parliament published amendment instructions and decisions, not the concurrent read and write schedule required by conflict serializability." Page: "Parliament published amendment instructions and decisions, not the concurrent execution history that question requires." Room: "The published page says model checking added no answer beyond ..."; page: "model checking has not yet added an answer beyond ...". The meaning survives, so a near-miss; the README is accurate ("Their title, URL and QUESTION MALFORMED verdict are quoted").
- OBSERVED | (added) Four same-coordinate amendment pairs were found on 2026-09-10 whose two textual orders differ while both serial schedules are necessarily acyclic. | oversight/waves/2026-09-10-foreign-instruments/specs/two-amendments-walk-into-a-database.md | [added-four-pairs] "Fresh retrieval on 2026-09-10 found four same-coordinate amendment pairs whose two textual orders differ while both corresponding serial schedules are necessarily acyclic", with a table of pairs 6/7 (3:22), 22/23 (113:12), 50/51 (137:17), 69/70 (71:15).
- CONFIRMED | (added) The errata really do date each report to the day the author was first told | https://www-cs-faculty.stanford.edu/~knuth/err1.textxt | [added-errata-date] The book is not named on the page; the wave spec names Knuth's TAOCP vol. 1 errata, which say "The date on which I first was told about the bug is shown; this is the effective date on which I paid the finder's fee."
- MINOR | (added) Most entries on the safety lists describe a noisy channel between DIFFERENT received strings: a decimal point disappears, a letter is read as a digit, one abbreviation is misread as another. | https://www.ismp.org/system/files/resources/2024-04/ISMP_ErrorProneAbbreviation_List.pdf | [added-safety-lists-most] Every example is on the ISMP 2024 list ("Mistaken as 5 mg if the decimal point is not seen", "U or u ... Mistaken as zero or the number 4", "Q.D. ... Mistaken as q.i.d."), but "most" does not hold for it on my rough classification: of about 89 rows, about 38 need a different received string, and about 50 describe one unchanged string given a second meaning, led by the 23-row Drug Name Abbreviations section ("MS, MSO4 ... Mistaken as magnesium sulfate", "PCA ... procainamide ... Mistaken as patient-controlled analgesia", "SSRI ... Mistaken as selective-serotonin reuptake inhibitor", "gr"). The Joint Commission's six families are mostly the glyph kind. Those one-string rows are the grammar-ambiguity case the objection says the lists do not supply, though its second defect (a hand-written semantic equivalence) still applies to them.
- CONFIRMED | (added) Its invented capacity and missing unavailable requests remain visible here as adapter costs | public/strata/the-shelf-that-knew-the-future/collision.json | [added-shelf-costs] hand_conventions: "Capacity is a synthetic object count ..."; adapter loss: "Whether a reader who could not get volume three would have taken volume four instead is not recorded anywhere"; both are rendered in the inspection panel.
- CONFIRMED | (added) The statutory instruments contain clauses such as entitled to another benefit and genuine and sufficient link. | https://www.legislation.gov.uk/uksi/2024/869/made | [added-si-clauses] Reg. 2(3)(c)(i) "entitlement to a benefit payable in the country in which the person is habitually resident that is equivalent to a relevant benefit" and (ii) "a genuine and sufficient link to the United Kingdom".
- CONFIRMED | (added) Spin model checking, safety over a reachable state space (the adopted engine "Spin model checking") | public/strata/the-stage-directions-that-cannot-be-obeyed/index.html | [added-spin] The adopted page runs "Spin 6.5.1 of 20 December 2019" ("An unchanged Spin safety assertion refuses a second entrance for an actor already onstage").
- UNVERIFIABLE | (added) and explicitly proposed for testing whether two forms of legal reasoning are equivalent | tried the HAL abstract (api.archives-ouvertes.fr), inria.hal.science/hal-02273483/document and the files_s PDF link (both return an Anubis bot challenge), Crossref | [added-bdd-equivalence] The Aucher, Berbinau and Morin abstract combines "'legal' reasoning" with "'procedural' reasoning" in Multi-BDDs but says nothing about testing two forms of reasoning for equivalence; the full text could not be read, so the second half of the eligibility refusal's prior-art sentence is unchecked rather than wrong.

## Findings
- [declined] adopted-paraphrase-called-quoted (K1, MINOR added-adopted-quoted): render.js:50 says "Quoted from its published verdict." and index.html:158 "Adopted rows quote published pages", but data.js:19 and data.js:27 paraphrase the adopted pages ("has not yet added" became "added no"; "the concurrent execution history that question requires" became "the concurrent read and write schedule required by conflict serializability"). Meaning intact, loose wording about the page's own presentation; listed by the tier B triage, page unchanged.
- [declined] safety-lists-most-overcount (K2, MINOR added-safety-lists-most): the prescriptions refusal (refusals.json:130) says "Most entries on the safety lists describe a noisy channel between DIFFERENT received strings"; every example is on the ISMP 2024 list, but the checker's rough classification finds about 38 of 89 ISMP rows of that kind, the rest (led by the drug-name abbreviations) one string with two meanings. A loose quantifier, not a false fact the argument rests on; listed by the tier B triage, page and the wave's canonical refusal record unchanged.