Errata

The dossier-reconciliation sweep — older surfaces aligned with the adjudicated dossiers

Date: 15 August 2026. With the two lecture dossiers signed off (Brighton A−, Sheffield A; final ruling appended to review/dossiers_audit.md, adjudication in review/dossiers_adjudication.md), the older analytical surfaces — the flagship analysis, the concordance, the essays, REPORT.md, and the v3 headnotes — were swept for claims the adjudicated findings contradict, overstate by comparison, or materially enrich. Every candidate was verified against both documents’ text and the underlying consensus sheets before any edit (rule 0). Corrections:

1. Flagship §3.10: the ε-block hedge was misplaced and over-scoped. “the same bracket immediately hedges it” and “assert, then disown the assertion’s letter” overstated both the hedge’s position and its reach. On consensus sheet 21 the ε unit runs “instrumentality—for what? God? Whole? / for spirit itself” → “X ultimately — subjectivity fulfils objectivity / by awareness & espousing & — / creative action” → “but this may be sheer myth / symbolically true”: the hedge is not adjacent to “for spirit itself”; its nearest grammatical antecedent is the subjectivity-fulfils-objectivity claim it follows, and whether it also disowns “for spirit itself” the sheet does not decide. §3.10 now quotes the intervening claim, states the antecedent limit, keeps the structural finding that survives (answer and reservation travel as one keyed insertion), and cross-links the Brighton dossier §2.4, whose adjudication set the standard (“Medium 4,” review/dossiers_adjudication.md). “Assert, then disown” is now “assert, then reserve.”
2. The concordance synthesis carried the same overstatement — “the whole is immediately hedged” — and is corrected to the block-closing hedge with the antecedent limit stated and the dossier linked.
3. Flagship §3.7 flattened the 1939→1949 paranormal comparison. “The Brighton 1949 position on the paranormal (§3.10) is already fixed here” ran two different questions together. Per the dossier’s adjudicated repair (“High 2”): 1939 grades the powers in general (“cannot see anything particularly lofty”), 1949 grades what proof of survival would add (“no new values yet emerge”), and the epistemics move from a hedged personal inclination to a reported expert consensus. §3.7 now claims only the axiological half, across the changed question, cross-linked to the dossier §3.
4. The concordance synthesis stated the bifurcation’s motive as fact. “it is audience-shaped… he preached its human half and treated the other half as a private reserve” asserted a policy the notes cannot show. Per the Sheffield dossier §4’s discipline — the notes record the omission, not the motive; audience-shaping, genre, and wartime tact are rivals the held record cannot rank — the synthesis now states the correlation as the surviving plans’ pattern, names audience-shaping as this concordance’s reading, and tiers the rivals with the dossier linked.
5. The salute essay still quoted “disbelieved God” as the 1954 book’s text. The held 1954 text reads “haunted by the ever-present ghost of a disbelieved-in God” (meditation 15); Moskowitz, quoting, drops the “in.” The flagship was corrected when the full text arrived (August 2026); the essay had kept the Moskowitz variant. MD and HTML twin both corrected.
6. REPORT.md quoted the Brighton prayer’s woven form without disclosure. The doubling phrases — “knowing not —”, “whether thou art,”, “or not.” — are diagonal margin insertions; the disclosure the flagship, the salute essay, and the dossier all carry is now stated there too.

Enrichments applied in the same pass (one sentence + cross-link each; no new corpus citations, the 677-citation invariant unchanged): flagship §3.10 records the dossier’s revision-step inference for the prayer’s diagonal stratum (offered as inference, undated strata disclosed); the flagship §4 Huxley entry notes the biography’s report of “evolutionary humanism” among Stapledon’s own platform formulas at the Exeter Present Question Conference by August 1946, with the witness-value weighed in the 1946 monograph §2.5; and the two dossier-tier v3 reading editions’ headnotes now link their dossiers. Verified findings needing no change: no older surface claims a religiously empty wartime platform (the concordance’s “almost entirely inside the humanist frame” and its own indexing of the diagnostic “pseudo-mysticism” already match the Sheffield dossier’s repaired reading); the flagship’s account of the 1946 salvation translation (§3.8) makes no novelty claim, so the monograph’s narrowed terminological-novelty finding contradicts nothing older; and the fair-copy chronology for The Opening of the Eyes was already stated correctly in the flagship §3.10 and REPORT.md.

The wording-revision pass — the two-family debate over the reworded texts

Date: 9 August 2026. The four reworded texts — the v2 performing edition, the v3 source-bounded reading edition, the 1934 blog essay (MD and HTML), and the modern talk — were put through a structured two-family wording debate: independent position papers (GPT-5.6 Sol, 56 proposals; a Fable-family critic auditing its own family’s prose, 44 proposals), crossed responses, and a merged adjudication. The full record is review/wording_debate/ (five files); the verdict table with per-site winners and the mechanism that settled each is review/wording_debate/final_dispositions.md. Outcome: 89 of 100 proposals applied in the winning formulations (many merged, belonging to neither Phase-A paper), 5 withdrawn by their proposers, and 6 disputed proposals across 5 sites deliberately left unapplied for the owner (the working-hours tableau; the coffee-bathos apology; the talk’s Superman-in-uniform line; the rowing-boat detail; the Star Maker “final form” phrase — where both sides agree the current phrase is wrong but back different replacements). Every current string was re-verified in the live file and every winning wording against its consensus sheet before application; two factual pushbacks were verified against the project record and ruled accordingly (the turquoise “Naive” third medium, analysis §6, which corrects the talk’s “blue-green” and defeats the proposal to delete the colour; and the Conway Hall witness’s true remove, §1, which the shipped apparatus wording now states everywhere it appears). The revision is chiefly epistemic — in both papers’ shared formula, keeping “no doctrine” from becoming negative doctrine and a marginal question mark from becoming cosmic purpose. The four most significant shipped changes:

1. The sheet-30 margin loses its smuggled teleology (v2 + blog). “Instrumental to what? To worship: to the apprehending and enjoying of the universe — and thus, it may be, to some contribution, some distinctive beauty, which the universe requires of us and of nothing else” → “…and thus, perhaps, to the contributing of a distinctive beauty.” The margin reads “& thus — contributing? a distinctive beauty”: the restored “To worship” is firm (strike-and-restore, double underline), the contribution is a hesitation, and the universe-requires clause had no note behind it. v3’s “and thus, contributing? a distinctive beauty” was defended by both sides and stands.
2. The “No doctrine” slide stops preaching a negative catechism (talk). “No personal God. No providence. No immortality.” → “No inference to a personal God, providence, or immortality.” — sheet 28 forbids the inference (“we may be tempted to say… but we must not”), not the propositions; sheet 31 welcomes all hypothesis. The companion teleprompter line “The experience licenses nothing.” → “The experience licenses no doctrine about its object.”
3. Sheet 11’s voluntary-sacrifice guard is no longer strained (v2 + blog). “But all might have to ‘die’ for the Ideal. Out of human unity, mankind itself might be offered up. Hold that thought; we shall need it.” → “But all might have to ‘die’ for the Ideal — out of human unity.” The notes say distributive, voluntary self-sacrifice — beside the explicit guard “none may sacrifice another against his will” — not a collective object offered up.
4. The idea’s afterlife is dated honestly (talk, slide + teleprompter). “1946 · same society: ‘agnostic piety’” / “His final position: ‘Humanism Plus. Theism Minus — but Plus also.’” → the merged 1937 / 1942 (“agnostic piety” already in print, in the closing pages of a wartime Searchlight paperback) / 1946 (the same-society formula, “not, as we’ll see, his last word”) / 1949 (Brighton: “he was still saluting — a universe he would no longer even affirm to exist: ‘whether thou art … or not’”) chronology, per analysis §§3.7b–3.10.

Also in this pass: the v2/v3 synopsis apparatus and the talk’s opening now state the Conway Hall witness at its true remove (a catalogue record carrying the delivery date and a four-sentence summary of the Feb. 1935 Monthly Record abstract — the abstract itself not yet acquired); v2’s salute paragraph ends at the manuscript’s double-underlined “which I see.” (the appended creed-coda deleted); v3’s three tinting-contract breaches are repaired and its dropped double-underlined line (“these high-flown intuitions are suspect”) restored; the blog’s flipped sheet-31 inference (“and in feeling that” → “because”), its “certainty” self-refutation in the Larger Rationalism paragraph, and its struck-insertion quotation (“more difficult ~~and more human~~ living”) are corrected; and the Otto apparatus on all surfaces gains Crossley’s 1933–34 diary-list evidence with the month left open. The v2 and v3 pages are rebuilt from their build scripts; the blog HTML twin carries every MD change verbatim (drift gate clean).

The Crossley sign-off round — adjudication of the first external audit

Date: 8 August 2026 (evening pass). A GPT-5.6 Sol adversarial audit of the Crossley integration (review/crossley_signoff_round1.md: 46 of 49 claims verified, 3 flagged, 18 candidate additions) was adjudicated item-by-item against corpus/crossley_speaking_for_the_future.txt and the raw page photographs; the full disposition record with evidence is review/crossley_signoff_fable_response.md. Every applied item was re-verified against the corpus (and, where OCR could be doubted, the page image) before application. Outcomes:

1. Doctorate metadata completed (audit flag a — corrigendum to item 2 of the entry below; dated line, history unchanged). That entry’s “All surfaces now say ‘the 1929 book’” was false as written: work/source_corpus.py (“His doctorate book”), and work/build_sources_index.py’s description feeding SOURCES.md and sources.html (“his philosophy doctorate book”), still called A Modern Theory of Ethics the doctorate. All three generator strings now read “his first philosophy book,” begun after the March 1925 degree (Crossley pp. 165–66, re-verified: “Meanwhile Olaf took Mair’s advice and began his first book”), the three shipped surfaces are regenerated, and two guard rules watch the corrected strings.
2. Nebula Maker status-timing narrowed (audit flag b). Analysis §3.6 had let two claims slip past Crossley’s evidence: that the lecturer of December 1934 “had already set that attempt aside,” and an unhedged reliance on “late in the summer of 1933.” Re-reading pp. 229–30, 243–44 and endnote 5 (p. 432): Crossley’s narrative dates Mitchison’s reading of the draft “late in the summer of 1933,” but the letter itself is “undated [but probably summer 1933]” (his own catalogue note); the “Discarded MS. of Star-Maker” label “was found forty years later” and is not dated to any moment; and the restarts are attested only from after Odd John (late 1935). §3.6 now says exactly that: the year-level dating and the draft-precedes-lecture sequence stand (they were signed off), the month is disclosed as Crossley’s synthesis over an editorial conjecture, and the draft’s status on the lecture’s evening is stated as undatable. The nebula_maker source-page provenance carries the same disclosure. A guard rule watches the old phrase.
3. The Agnes/Opening of the Eyes reconstruction de-synthesized (audit flag c — with a corrigendum to the hedge paragraph of the entry below; dated line, history unchanged). §3.10 had fused Crossley’s begun-fair-copy sentence (p. 394), his passing notebook mention (p. 396), the 1954 title page and Acknowledgement, and the naming (p. 386) into one definite scene (“part-fair-copied manuscript plus a nest of drafts and notebooks, from which she assembled the published text”). No single source states that scene. §3.10 now lists the attested pieces separately — fair copy begun in the first week of September, death on the 6th, extent unknown; notebooks attested in passing; “edited by Agnes Z. Stapledon” (title page) and the shared deciphering (Acknowledgement); the title hers (Crossley) — and marks the old scene as a synthesis. The entry below said her “documented role” was “posthumous assembly, editing, and the title”: assembly was itself part of the synthesis; the documented words are editing and the title. Two guard rules watch the old sentence’s halves.
4. Eighteen audit-proposed additions adjudicated: 15 applied as recommended, 3 applied in corrected form, 0 rejected — every quotation re-verified in the corpus first (one on the page photograph). The three corrections to the audit’s own evidence, with the details in the response document: its item M13 quoted two sentences (“There will have to be full Negro independence…”; “The Portuguese colonies are in a vile state”) that appear nowhere in Crossley — the applied passage uses the real pp. 241–42 text, and adds the held novel’s own check on Crossley’s “deleted” list (the flood-lit-Führer sentence stands in the published epilogue nearly verbatim); its item M14’s Russell material was applied as a new addition (the audit’s verification table had treated it as an existing analysis claim, which it was not), with Crossley’s “two lights of reason and love” gloss flagged against the novel’s own emblems; and its item M2 was applied with the Last Men in London passage cited from the held novel directly rather than at second hand. Analysis citation total: 459 → 516 markers (55 new Crossley, 2 new Last Men in London); the Crossley excerpt edition grows from 41 to 77 published paragraphs (ranges recomputed by work/_crossley_excerpts.py).
5. The audit’s excerpt-economy escalation is rebutted with evidence. Its three “surplus” paragraphs each carry a quotation the analysis uses: the prologue Newark paragraph supplies §5’s quoted “concise anti-autobiography” (an explicit marker now makes that linkage self-verifying); the “In the Witness Box” paragraph supplies §7’s “would have questions fired at him…” (evidence for the audit’s own verified erratum 4); and the Conway Hall paragraph supplies §1’s “Anyone happening into Conway Hall… where Olaf often spoke on politics or ethics or religion.” Nothing was removed.

The Crossley biography pass — four corrections

Date: 8 August 2026. Robert Crossley’s Olaf Stapledon: Speaking for the Future (Syracuse University Press, 1994) — the standard biography — was acquired from the owner’s page photographs, OCR’d (497 pages, zero refusals), and read in full by six dossier passes (work/crossley_dossiers/). Reading it against the project’s claims forced four corrections; each was re-verified verbatim against corpus/crossley_speaking_for_the_future.txt before application.

1. The engagement date. The analysis (§3.1) had said the Stapledon–Miller engagement “became mutual only in November 1915,” and the July 2026 errata entry below repeats that date. The full chronology is layered: Crossley documents a first acceptance just before the Bangor WEA summer school of late July 1914 — “Agnes had agreed to marry him (though by mid-August she had changed her mind)” — and a renewed acceptance in place by the time Stapledon crossed to France on 1 August 1915 (“By then Agnes had at last agreed to marry him—although the war made theirs a marathon engagement”). The letters’ own November 1915 “So now we really are engaged” marks the public engagement (Agnes’s letter proposed, in Olaf’s words, “a public engagement”), not the first acceptance. §3.1 now gives all three dates; the March 1914 “not yet his fiancée” framing stands.
2. “His doctoral thesis” ≠ A Modern Theory of Ethics. The analysis (§§2.2, 3.1, 3.2, 3.4), the blog, and the modern talk had called the 1929 book Stapledon’s doctoral thesis. Crossley separates the two: “Late in 1924 he submitted his doctoral thesis in the area of inquiry now called semiotics” (degree conferred 5 March 1925); “The one-word title of his thesis, ‘Meaning,’” — 433 pages, unpublished, deposited at the Sydney Jones Library. The Ethics was a distinct book begun after the degree (“he drafted almost half of A Modern Theory of Ethics by midsummer 1925”), published 31 January 1929. All surfaces now say “the 1929 book” (or equivalent); the doctrine-level claims are unaffected.
3. Nebula Maker is dated: 1933. The project had carried “undated; c. 1933–36 by scholarly convention” and derived all sequence claims from topical anchors. Crossley dates the fragment on letter evidence: Naomi Mitchison read “the draft of a cosmic history, from the first nebulae to the extinction of all life” in “late… the summer of 1933”; what she read was “labeled ‘Discarded MS. of Star-Maker’” and later published as Nebula Maker; his settled phrase is “the 1933 fragment known as Nebula Maker.” Analysis §3.6, SOURCES.md, the README shelf, REPORT.md, and the source-page metadata now carry 1933 (with the evidence chain disclosed). Consequence: the failed personal-God draft precedes the 1934 lecture.
4. The “only published record of Q&A” claim, narrowed. The analysis (§7), REPORT.md, the site index, the JBIS supplementary page, the source-reader page and the Interplanetary Man reconstruction page had called the 1948 JBIS discussion “the only known published record of Stapledon answering questions from a platform.” Crossley documents other published (but non-verbatim) records: his wartime Merseyside radio format “In the Witness Box,” where the weekly “witness… would have questions fired at him” (covered in the Liverpool Post, “A One-Man Brains Trust,” 11 Jan. 1945), and Stapledon’s own printed account of fielding challenges at an RAF lecture in Youth and Tomorrow (1946). All surfaces now claim only the verbatim record.

One claim was revised rather than corrected: §3.7b had described Beyond the “Isms” (1942) as “written from troop-lecturing experience.” Crossley evidences a mid-1941 Searchlight-series commission (“Secker and Warburg’s series of blue-jacketed paperbacks… designed to blueprint the post-war world”) and ties the book’s confessional turn — its preface’s “a much more personal book than I had planned” — to a crisis in Stapledon’s private life, not to the troop circuit. The claim is withdrawn. And one account was hedged at its source: the statement that Agnes transcribed The Opening of the Eyes “from his abbreviation-dense notes” rests on Moskowitz alone; Crossley shows Stapledon began a fair copy of the whole manuscript in his last weeks, with Agnes’s documented role being posthumous assembly, editing, and the title (§3.10 now says so).

The delivery date of the 1934 lecture — 30 December, not 29 December

Date: 3 August 2026. Every project document had dated the delivery of “Humanism and Superhumanism” to 29 December 1934, from the manuscript’s red inscription “South Place 29·12·34” (sheet 1; the University of Liverpool’s IIIF catalogue metadata carries the same date, “D:29 M:12 Y:1934”). A newly found Conway Hall catalogue record corrects this. The evidence chain:

1. The catalogue note. Conway Hall Ethical Society’s library catalogue holds a record for a contemporary published abstract of the lecture — Monthly Record (the journal of the South Place Ethical Society), Vol. 40 No. 2, February 1935, pp. 3–4, standard no. J1*40.2*2 — whose Notes field reads: “Lecture to the Society, December 30th 1934.”
2. The weekday. 29 December 1934 was a Saturday; 30 December was a Sunday (the Saturday date had already been flagged as an oddity in REPORT.md).
3. The convention. The South Place discourses were the Society’s famous Sunday lectures.

Conclusion: the lecture was delivered Sunday 30 December 1934; the red 29·12·34 is Stapledon’s own inscription — presumably the date he dated or finished the notes, the day before. Corrected across the analysis, README, REPORT, index, how-this-was-made, the 1934 blog essay (MD and HTML), the modern talk, the v2/v3 build scripts and their rebuilt pages, and the v1 page (header corrected, errata box extended). The consensus transcription is untouched — the manuscript genuinely reads 29·12·34, and that reading is evidence, not error. A propagation rule now guards against the stale delivery date recurring. The Monthly Record record is also integrated into the analysis (§1) and REPORT.md §9 as the lecture’s newly identified contemporary witness; acquisition of the full two-page abstract from Conway Hall is pending.

The August 2026 pass — the intellectual-context shelf

Date: 3 August 2026. Six texts the analysis had quoted at second hand were acquired and held complete (Pope’s An Essay on Man; Bentham’s The Rationale of Reward; Mill’s “Bentham”; Russell’s “A Free Man’s Worship”; Otto’s The Idea of the Holy, Harvey trans.; Julian Huxley’s Religion Without Revelation). Corrections that reading forced:

• Push-pin attribution (analysis §4): the wording “quantity of pleasure being equal, push-pin is as good as poetry” was sourced to “Bentham via Mill’s Utilitarianism”; it is from Mill’s 1838 essay “Bentham” (London and Westminster Review) — push-pin does not appear in Utilitarianism. Bentham’s own sentences (Rationale of Reward, Book III ch. 1) are now quoted and linked beside Mill’s paraphrase (§2.2).
• A silent splice disclosed (analysis §3.5/§4): Waking World’s “our own nature made perfect” and “That is just where the trouble lies” are separated by a sentence in the book; an ellipsis now marks the elision.
• The Otto identification hedged with a channel caveat (analysis §2.6(b)): the sheets’ numinous vocabulary matches Otto far beyond the two loanwords (both poles, majesty, objectivity), but Julian Huxley’s 1927 bestseller had quoted Otto’s mysterium-tremendum page verbatim, so second-hand acquisition of the loanwords in 1934 cannot be excluded (direct knowledge is certain by 1939, via the Philosophy and Living reading list).

Errata — the July 2026 revision

Date of this pass: 29–30 July 2026.

In July 2026 three independent audits of this project were run by three different model families (GPT‑5.6, Kimi K3 Max, and Fable 5; their reports are preserved unaltered in review/). The project’s owner adjudicated their findings, and the accepted corrections were applied in this pass. Every correction below was re-verified against a primary source — the consensus transcription, the 72‑megapixel archival captures, or the published corpus — before it was applied; audit claims that failed that check were not applied and are logged, with evidence, in review/FABLE5_ADJUDICATION_AND_FIXES.md (the full adjudication record).

The transcription (lectures/transcriptions_v2/consensus_v2/, Humanism 80190)

Re-reads against the 72‑megapixel archival captures (each decision logged with confidence in the adjudication record):

• Sheet 11 — the “[illegible marks]” in the margin are stitching holes with dark residue (physical damage, not writing); now described as such. The page number “14” carries the same renumbering strike as its sibling sheets (now “10 ~~14~~”). The balloon note’s “thorough-going” was kept as primary; the audit’s proposed “through-going” is preserved as an alternative reading.
• Sheet 12 — the page number “11/15” was tagged red; the capture shows plain graphite. Now rendered “11 ~~15~~”, untagged.
• Sheet 14 — the strike previously recorded on “other” in “all ~~other~~ goals are” is not present on the capture; the strike was removed (the earlier reading is preserved as an alternate). Documents quoting “all other goals are either incomplete… or positively false” are therefore correct as they stand.
• Sheet 16 — the disputed primary now reads “(using θ arguments θ 4 an used against him)” (previously “θ H can use”; both alternatives preserved); the page-number “-3” is in fact a struck “3”.
• Sheet 27 — the doctrinal margins (“conviction θ — man is transitional…”, “# but our mystery…”, “⊕ our striving toward spirituality”) and the matching inline ⊕/# marks were tagged red; the capture shows graphite throughout. Retagged. (This also removes an apparent self-contradiction in the analysis’s account of what red pencil is for.)
• Sheet 30 — the struck glyph in “awakens θ ~~·~~ spirit further” is a double-f false start of “further” (“ff”), not an “H”; the primary was flipped (old reading kept as an alternative).

Analysis (analysis/humanism_and_superhumanism_analysis.{md,html})

• The centerpiece quotation: “an unspeakable Beauty (or Majesty)” → “an unspeakable Beauty, & Majesty”, with a note that “& Majesty” is a diagonal manuscript insertion — an addition, not an alternative (and that a struck “that” precedes the sentence).
• Misquote: “to get the benefit of ourselves” → “to get the best out of ourselves” (sheet 30) — a fix the project had already made in the reconstruction but never propagated here.
• The “To worship?” narrative: the sheet‑30 margin puts the question mark on “what?”; “To worship” is written, struck through, then re-entered with an emphatic double underline. The “question mark he refuses to erase” story was replaced by the actual strike-and-restore.
• Cancelled words no longer quoted as live text: the sheet‑27 margin is quoted in its final reading (“our mystery reaches beyond the human sphere”, cancellation noted); Brighton’s “a salutory ~~emotional~~ tension” now discloses the strike.
• Brighton 1949: “for spirit itself” now carries the sheet’s own adjacent hedge (“but this may be sheer myth / symbolically true”); the closing prayer’s diagonal margin insertions are disclosed as such.
• Ages and intervals: “twenty-six” (June 1913) → twenty-seven; “Fifteen years after” → fourteen; “(now at Conway Hall)” dropped — the society had been at Conway Hall since 1929.
• Speaker and source fixes: “the living music that is man” is the President’s line in “Old Man in New World” (the Fool’s actual answer — “the Unseen makes music with us” — added); the Sirius gloss corrected to the book’s own direction (“not to the universal Tiger but to the universal Master”); “ecstatic submission” (a paraphrase in quotation marks) replaced with Darkness and the Light’s verbatim phrases; “brief music that is man” re-attributed to Last Men in London; Clarke “organized the evening” → “formally opened the discussion” (per the JBIS transcript); “Agnes publicly corrected him” → corrected in remarks printed by Moskowitz; meditation 32’s “may be but a very little nearer” → “are but a very little nearer”.
• Unsourced claims removed or hedged: “‘neither a Christian nor an atheist,’ as he often said” (unattested in the held corpus); “his coinage” of superhumanism (now “his term (possibly his coinage…)”); the false precision “in eight words”; “shd” removed from the abbreviation list (not in this manuscript).
• Media claims corrected against the captures: “Dwell on this” is a graphite boxed cue, not red; the “naive” softening is turquoise, not red; red pencil is confined to the cover’s staging (plus one stray page number).
• Hedges restored: sheet 24’s “(Wordsworth said so) but I am not sure”; the thesis’s “We seem to stand above the battle”; the 1913 star-letter’s own “whether they are mere tricks of the mind…”; Waking World’s “must somehow be the best possible” slide and its “we cannot solve it” admission added as the book’s divergences from the lecture.
• Genealogy errors: the “mechanization mitigation” is the lecture’s own, not the thesis’s (the thesis is harsher, and is now quoted: “not merely… up to the limit of his capacity, but of the very highest capacity… a scourge to society”); the salute was in print by 1929/1930 (thesis; Last and First Men), not first in 1932; the 1915 letter wakes “in my sleeping bag, alone” (the hillside is the novel’s); Agnes was his cousin, not yet his fiancée, in March 1914 (engagement: November 1915); “tunic pocket under shellfire” → “in my pocket (‘as being on the whole the safest place’)” with the letter’s own “blague… it is utter peace here” caveat; Waking World’s “Humanism and Superhumanism” heading is in its opening chapter, not the preface; the unsupported “two-thousand-year experiment” figure removed; two ellipsis-reorderings re-quoted in source order; the spliced AMTE lead-in (“the ideal is that…”) corrected to “It is demanded, namely, that…”.
• Additions: Philosophy and Living II’s admissions (“one of the most dangerous moral snares”; “a logical conflict… useless to pretend that there is not”; “complacent acquiescence in the misfortunes of others”); Star Maker’s epilogue revocation (“false in every theme and facet, yet in spirit true”) and its community-priority passage (“this, this alone, was the solid ground of existence”) set beside the “Two lights”; Youth and Tomorrow ch. 5 (previously listed as a source but never used); the late drift to fictionalist piety and the Spirit/Other identity-doubt (The Core; Death into Life); The Flames’ epistolary frame and its narrator’s “hopelessly confused and untenable”; sheet 11’s consent margin (“none may sacrifice another against his will”); sheet 29’s “lack of humour” and “taking man too seriously”; the sheet‑23 “copulation” insertion no longer silently omitted.
• Reframing: §3.9 retitled and reframed from confirmation-harvest to “the salute institutionalized, doubted, and fictionalized”; new §3.11 on hierarchy, eugenics and empire across the corpus (1908 → 1948), held together with the corpus’s own anti-fascist and anti-imperial passages; a provenance box and a “Note to the third edition” added.

Reconstructions

v3 — the diplomatic edition (lectures/reconstructed_lectures_v3/)

• The page now labels itself v3 · diplomatic in title and footer (it formerly said “(v2)”), and its “← Index” link now points at the project front page.
• The struck “originally” (wings margin) is no longer voiced; the cancellation is noted in marked apparatus.
• ~21 unmarked connective phrases (both audits’ lists, merged) are now tinted as editorial, and several over-tintings of licensed words were removed.
• Dropped note lines restored (verified against their sheets): sheet 6’s “Utilitarianism not derived from individualism but from social impulses” + “intuitive ‘oneness’” + “to passing impulse”; sheet 7’s “essentially human in man + possibility of further growth”, the boxed “[Dwell on this]” cue and the cognition/conation margin; sheet 8’s “not derived from something more basic”; sheet 22’s “As clear as I can get it” and “(True Rationalists should be Superhumanists)”; sheet 23’s “even mind itself as object” (margin now reported as the assertion it is); sheet 27’s “conviction” margin; sheet 30’s “Presented Other… a Presence; It; Reality, terrible and beautiful” cluster; sheet 16’s parenthetical, restored after the transcription re-read settled its disputed reading.
• The invented section title “…and burning houses” replaced with the sheets’ own heading words; the sheet‑28 margin note is “arrowed”, not “boxed”; “his anthology” → “an anthology”; the intro’s “every deviation… is repaired here” overclaim softened; “All goals” → “All other goals” per the corrected sheet 14.

v2 — the performing edition (lectures/reconstructed_lectures_v2/humanism_and_superhumanism.html)

• A standing foreword now labels the page an editorial performance script: its expansions are editorial and it is not quotable as Stapledon’s exact words (with a link to v3 for exactly what is his).
• False attributions removed: “as Bentham said”; “with Rudolf Otto”; “the old monks had the right words” (the term is Rudolf Otto’s, 1917 — now an italic apparatus note); “and the Ode is evidence”.
• Invented autobiography removed: “my own coinage”; “I have spent some years of my life constructing such worlds”; “I have known humanists who shame the saints”; “Every demagogue in Europe is exploiting that, this winter”.
• The sheet‑29 margin is no longer rendered twice; cancelled text is no longer voiced (“has a significance”; “originally”; the ditto misexpansion “nor slight possibility for probability” → “for certainty”); “Instrumental to what? To worship” now carries the manuscript’s emphatic restoration rather than an invented hedge; “what more can it have?” per the margin; the invented “burning houses” title fixed; “All other goals” per the corrected sheet 14.
• Where the notes end (p. 25), an apparatus note now marks the closing peroration as editorial staging.

v1 — the superseded early edition (lectures/reconstructed_lectures_v1/)

• An errata box now heads the Humanism page: superseded by v2/v3, with its verified defects listed (the invented “made an idol of the human”; the struck plan block spoken as live text; “slight possibility for impossibility”, which inverts the note’s “nor slight possibility for certainty”; the wrong archive citation). The citation itself was corrected in place (F33 → OS/F11/4); the index’s dead pre-project link was removed. The text is otherwise left as a historical record.

Other lectures (v2 pages, rebuilt from corrected caches)

• Science and Fiction: “Venerians” → “Venusians” (per the notes); the hallucinated closing section replaced with what the final sheets actually discuss (invented environments, gravity, fire); “pukka” aligned with the consensus reading.
• Corrigendum (round 2, 30 July 2026): the last clause above misstates the applied reading. The consensus primary on 80177 sheet 3 is “pucker” (with “pukka” preserved only as the alternative), and both the prose cache and the rebuilt page follow the primary: the v2 prose reads “pucker” novels. The prose was aligned with “pucker”, not “pukka.”
• Science and Literature: “Morgan upon sadism” → “Maugham”; “the why of human nature” → “the whole”.
• The Future of Britain (80143): “one woman” → the notes’ “one man?”; the struck “good” no longer voiced; invented autobiography removed. (80148): “supra-national” → “infra-“.
• Our Responsibility Towards the Future: the struck “bloody revolution” is no longer voiced.
• The template tic “Well, there it is.” reduced from five pages to a single licensed instance.
• The generator now stamps each prose cache with a hash of its consensus input and prints a staleness warning when the transcription has changed since the prose was written.

Presentations (presentations/)

• All decks rebuilt from the corrected transcription; the deck index now carries the mobile viewport tag.

Blog

blog/the_salute.{md,html}

• “fifty-year-old” → forty-eight; “At twenty-six” → twenty-seven; “Fifteen years” → fourteen; the Brighton notes ≈ nineteen months before his death; “now at Conway Hall” dropped (there since 1929).
• The salute is quoted per the manuscript (“Beauty and Majesty”, the insertion noted); “the exact centre of the manuscript” → a climactic late passage; the invented pause/duration claim removed; “He was watching Nuremberg” recast as editorial chronology; the Russell comparison softened to an editorial one; the coinage claim removed; the “To worship?” narrative corrected; “tint” → “lust of gregariousness”; “benefit” → “best out of ourselves”; the figure caption’s “circled” note is now “marginal”; “all other goals” restored per the corrected sheet 14; divider-sheet count fixed; “reconciled every disputed word” → honest (“a handful remain marked uncertain”); a provenance note added.

blog/humanism_and_superhumanism_1934.{md,html} (formerly blogpost2.html)

• File renamed blogpost2.html → humanism_and_superhumanism_1934.html (repairing the front page’s “easiest way in” card).
• The false standfirst — “The content is his, unchanged…” — replaced with an honest disclosure: the argument, structure and emphases follow the notes sheet by sheet; the prose is editorial; see the diplomatic edition for exactly what is his.
• Removed from Stapledon’s mouth: “my own coinage”; the invented citations; the Philosophy and Living (1939) formula imported into a text billed as 1934; “Every demagogue in Europe is exploiting that, this winter”; the fabricated closing peroration (the essay now ends where the notes end).
• Misquotes fixed: “say rather, a Majesty” → “Beauty and Majesty” (insertion noted); “tint” → “lust of gregariousness”; “benefit” → “best out of ourselves”; “the perceived mass” → “the perceived herd”. A provenance note added.

Modern talk (modern_talk/humanism_is_not_enough.html)

• “fifty-year-old” → forty-eight; a clear hinge added where the historical material ends (“The historical reconstruction ends here; what follows is my contemporary application”); a provenance note added.
• Unsupported claims removed or re-grounded: Woolf/Churchill cut (no source in the project); “you can tell he coached rowing” → “rowed in the college eight” (Moskowitz); “taught miners and dockers” → WEA evening classes in literature, psychology and industrial history (Moskowitz); “his notes underline the word ‘all’” → “his notes hammer the word ‘all’ three times” (no underline on the sheet); the imagined room-reaction softened.
• The salute now reads “an unspeakable Beauty and Majesty” with the insertion explained; poem placements marked as conjecture and the anthology unowned (“an anthology he never names”); “slight possibility for probability” → “for certainty” (twice); “oldest freethought congregation” standardized to the Society’s own claim (“in the world”); the “mystical Race deity” line correctly framed as the Communist indictment he reports and half-concedes; “The old monks had the right words” → “The theologian had the right words”; several smaller quotation-boundary repairs.

REPORT.md

• Faithfulness clusters “~66” → 91 (recounted from work/faithfulness_out/ by the stated method); the “remaining 14 sets” hi-res sentence now gives the non-Humanism figures (41 confirmed / 43 revised / 26 uncertain, ~630 proposals — the old sentence printed all-corpus totals); the toggle is called by its shipped label (“θ markings”); one sentence added recording the corpus-wide visual pass (222/223 sheets, 1,821 style lines); “[or Majesty]” → “Beauty and Majesty” with the insertion noted; the sheet‑30 margin narrative corrected; the Waking World “preface” mislabel fixed; Clarke “chaired” → “formally opened the discussion”.

Security hygiene (work/)

• The Google API key was masked in the two UTF‑16 run logs (98 logged request URLs redacted, encodings preserved); the project snapshot zip was rebuilt without api_keys.txt and renamed project_snapshot_2026-07-26.zip; the six Google-calling scripts now send the key via the x-goog-api-key header instead of the URL; hires_verify.py’s docstring now describes its true scope (all 15 sets). Keys were deliberately not rotated — that is the owner’s action; note that the logged key appears to be an older, since-rotated value.

Site plumbing

• v2 and v3 pages now label their own editions in title and footer (shared renderer parameterized); the v3 back-link is fixed; the presentations index gained its viewport tag; the v1 index’s dead link was removed; requirements.txt added; two vestigial style.css references removed from books/ pages; the cache-staleness warning added to the page generator.

Corrections that the audits proposed but that failed primary-source verification were rejected rather than applied — most notably the alleged strike on “other” (sheet 14, resolved instead by correcting the transcription), the “through-going” balloon reading (kept as an alternative only), and several smaller items listed with evidence in review/FABLE5_ADJUDICATION_AND_FIXES.md.

Round 2 — 30 July 2026

A GPT‑5.6 Sol re-audit of the pass above (review/gpt56_reaudit_round1.md) found that most of it held, but that twelve findings remained — chiefly six passages in the 1934 essay whose removal the record above claims but which survived in both .md and .html. All twelve were re-verified against the primary sources and repaired in this round; the full disposition record with quoted evidence is review/round2_fixes.md. Every fix below was verified against the consensus transcription or the archival captures before application; generated pages were rebuilt from their sources, never hand-edited.

The 1934 essay (blog/humanism_and_superhumanism_1934.{md,html}) — the failed round-1 lane, now landed

• Invented autobiography removed (both files): “I have spent some years of my life constructing such worlds, and I can recommend the exercise” (sheet 10 has only “Consider some hypothetical examples”); “I have known humanists who shame the saints” (sheet 29 gives only “generosity, refinement, fortitude / but: aridity?” with the margin “some humanists are more than humanists”).
• Unlicensed citations removed (both files): “Call it, with Rudolf Otto…” → “Call it a sense of the numinous…” and “Rudolf Otto had the right words: Mysterium Tremendum” → bare “Mysterium Tremendum” (no Otto anywhere in the 34 sheets; sheet 23 line “call it a sense of — ‘ð numinous’, ð ‘holy’”, sheet 24 bare “Mysterium Tremendum”); a marked editorial note now carries the Otto attribution as apparatus, matching v2’s convention. “Wordsworth said so, and the Ode is evidence; though I am not sure” → “Wordsworth said so; though I am not sure” (sheet 24: “(Wordsworth said so) / but I am not sure” — no ode).
• The “To worship” correction applied: “Instrumental to what? I do not certainly know. To worship, perhaps:” → “Instrumental to what? To worship:” — sheet 30 puts the question mark on “what?”, then strikes “To worship” and re-enters it with an emphatic double underline; the hedge “perhaps” recreated the rejected reading.

v3 diplomatic edition

• The restored sheet‑16 parenthetical, which round 1 had over-expanded into the incoherent “using the arguments the 4 an[swers] used against him”, is now neutral apparatus: the prose quotes the disputed span diplomatically — “(using θ arguments θ 4 an used against him)”, alternative “θ H can use” — and states that it is left unresolved rather than expanded. The notes column was already correct and is unchanged.

Overview documents

• Analysis §3.6: the one remaining pre-correction centerpiece quotation now reads “an unspeakable Beauty, & Majesty, which I salute though it destroy me” (sheet 28’s diagonal insertion), matching §2’s rendering. HTML rebuilt.
• README: the same centerpiece completed (with the insertion noted); the margin sequence “Instrumental to — what? To worship?” corrected to the actual strike-and-restore (the question mark belongs to “what?” only).
• REPORT: “what Stapledon coins” and “the coinage ‘superhumanism’” hedged to “his term” (priority unestablished, per the analysis); the “naive” softening’s colour corrected from “in red” to the paler turquoise (consensus tags it teal); “his anthology” → “an anthology he never names”; the uncited “at Clarke’s invitation” removed (the JBIS record documents venue and Clarke “formally opening the discussion”, not the invitation).

Blog and supplementary

• The Salute: “the oldest freethought congregation in Britain” standardized to the Society’s own claim, attributed — “by its own account the oldest freethought congregation in the world” (matching the modern talk and the analysis).
• JBIS discussion page: “at the invitation of its chairman, Arthur C. Clarke” and “chaired by the young Arthur C. Clarke” softened to what the printed record supports — Clarke “formally opened” the discussion (JBIS 7:6; corpus/interplanetary_man_discussion.txt).
• Interplanetary Man page (deferred from round 1): the header’s “at the invitation of Arthur C. Clarke” replaced with the documented facts (venue per the JBIS record; Clarke formally opening the discussion); “formally opening the debate” → “…the discussion”. Page rebuilt from build_interplanetary.py.

Changelog corrections (this file)

• “pukka” corrigendum: the round‑1 Science and Fiction entry above says “pukka” was adopted; in fact the consensus primary is “pucker” (80177 sheet 3, “pukka” preserved as the alternative) and the shipped prose says “pucker”. A corrigendum is appended at the entry; the original wording is left visible.

Tooling

• work/keys.py: fixed the parser bug that silently dropped the xAI key (its value starts with “xai” and was mistaken for a header line); all seven keys now load.
• work/qa_propagation.py added — a permanent checker holding the known-bad strings from both fix rounds (with allowed-file exceptions, e.g. books/last_and_first_men.html legitimately contains “Venerians”, and “supra-national” is a genuine reading in 80158 though corrected in 80148): scans all 284 project-authored HTML/MD files and exits nonzero on any reappearance. Current state: clean.
• work/qa_propagation.py scope fix (30 July, closing the audit loop): the GPT‑5.6 Sol round‑2 re-audit found the “supra-national” rule’s path filter (only_matching="80148") could not catch a recurrence in the combined deck presentations/the_future_of_britain.html, whose filename lacks the docid. The rule now scans every file and instead allowlists the five surfaces where the reading is genuine (the 80158 consensus sheets and the World Beyond Hitler pages); self-tested in both directions. This was the round‑2 audit’s only remaining item; it ruled the artifacts themselves converged (no corpus-checkable content error remaining).

Verification: full build chain re-run in dependency order (all exit 0; generated pages served from cache with zero API calls); qa_propagation.py clean; 0 broken internal links across the site; the 1934 essay’s .md/.html verified in agreement passage by passage. No audit report in review/ was modified; this round adds only review/round2_fixes.md.

Latter-Day Psalms re-transcription — 2 August 2026

The corpus text of Latter-Day Psalms (1914) — previously the scan PDF’s embedded text layer, riddled with running heads, page numbers and OCR garbling — was replaced by an authoritative re-transcription of all 68 page images (Gemini vision model, 300 dpi, one call per page), reconciled word-by-word against the old text with every non-trivial divergence adjudicated on the page image (work/ldp_ocr/reconciliation.md). The old text is preserved as corpus/latter_day_psalms_old_ocr.txt (marked superseded in SOURCES.md).

• Analysis quotation corrected: §3.1’s “Apollo” quotation read “Thou carest not for the happiness of men”; the 1914 printing reads “happines” (the book’s own printer’s error, confirmed on the page image at p. 32, and agreed by both the old and new transcriptions). The quotation now reproduces the printed reading with [sic]. All 8 latter_day_psalms citation locators resolved against the new text without change; the full build passes with all 212 corpus citations.
• sources/latter_day_psalms.html is now a poetry-native edition (per-source poetry mode in build_source_pages.py): the book’s own contents page linkified, psalms as centered small-caps headings with per-poem anchors, exact verse lineation with hanging indents, the 1914 pagination in the margin, and the same #pNNNN citation anchors as every other source page.

Corpus-wide faithfulness audit of the twelve generated reconstructions — 3 August 2026

The July 2026 faithfulness audit covered only the hand-written Humanism reconstruction. This pass extended it to the generated lecture pages: every prose section of every model-drafted reconstruction was audited against the verified notes by all three model families (work/faithfulness_audit_corpus.py; the same GPT / Claude / Gemini ensemble and violation taxonomy as the Humanism audit), the findings clustered, and every cluster on which two or more providers agreed was then adjudicated by a human-directed worker against the consensus sheets themselves — no repair was applied on the auditors’ authority alone. Repairs were patched into the prose caches (work/recon_sections/*.json), the patched text was re-audited in confirmation rounds until the flags converged or were adjudicated as artifacts, and the pages were rebuilt from the patched caches. The full per-lecture adjudication records — every before → after with its sheet evidence, every rejection with its counter-evidence — are in work/faithfulness_out_corpus/shard_reports/*.md.

The numbers

The original ensemble pass issued 357 section verdicts (119 sections × 3 providers; 2,739 raw findings) and clustered them into 604 agreed-by-2+ clusters and 1,341 singletons across the thirteen generated lectures. Twelve lectures were adjudicated in this pass (three shards of four); per lecture, of the ensemble-agreed clusters:

Confirmation re-audits of the patched text (2–6 rounds per lecture, 722 further API calls, converging everywhere) surfaced roughly 270 additional finer-grained repairs — mostly hedges, intensifiers and glosses the first pass had ranked below its own headline findings — which were verified against the sheets and applied the same way. Of the 1,176 singleton findings (flagged by one provider only), ~30 were repaired where they were instantly-checkable string-level misquotes or cross-provider duplicates split only by the clustering normalizer; the remaining ~1,100 are recorded but unadjudicated (many fell incidentally when an offending sentence was deleted). The thirteenth generated lecture, Is Humanism Enough? (80498; 58 agreed clusters, 165 singletons), was audited in the same ensemble pass but its adjudication is still pending; its findings are preserved in the cache.

Example repairs (before → after, with sheet evidence)

1. Misquote (Science and Literature, sheet 2): “botanising upon his mother’s grave” → “botanizing on his mother’s tomb”. The sheet quotes it: x "botanizing on his mother's tomb".
2. Meaning reversed (The Future of Britain 80148, sheet 4): “Christianity itself must learn from Buddhism and from other spiritual traditions” → “The Christians must convert—Buddhism.” The note reads ð Xxs mst c'vert / Bd'm — the conversion of Buddhism, not learning from it.
3. Negation lost (The Future of Britain 80143, sheet 4): “America has to face its own relation to fascism.” → “In America, fascism is not due.” The sheet reads — America fascism N due.
4. Cross-sheet keyed insertion restored (Personality and the Paranormal, sheets 6–7): the manuscript keys two sheet-6 lines (⊙ or better — "ad hoc re-vival" of full or partial personality; α and would explain — no positive information of ð after life) into sheet 7’s “ad hoc survival” block with //⊙ and //α marks. The prose had voiced them as an invented preview paragraph; they are now integrated at their keyed points in the Tyrrell passage.
5. Padded direct quotation (Living on Other Planets, sheet 3): “A great adventure has begun, and it must go on.” → “A great adventure must go on.” — the sheet gives the sentence in quotation marks, exactly.
6. Omission restored (Alternative Futures, sheet 4): the section now opens with the enumeration the notes open with — “In estimating the future, then, we must take three things into account: the course of economic determinism… the future of herd mentality… and the future of the will for community…” (In estimating the future, take into account: — / 1. θ course of economic determinism…), and the invented closing paragraph that had displaced it was deleted.

The commonest repair classes overall were quieter than these: deletion of invented editorial sentences (“This argument contains much truth.”), removal of unlicensed hedges and intensifiers (“almost inevitable” → “inevitable” where the note is flat; “may” → “is” and vice versa per the sheet), and un-voicing of struck-through manuscript text (e.g. Possible Futures had voiced the struck heading Future of England; Science and Literature’s peroration re-voiced a struck idea twice).

Two artifact classes found in the audit tooling itself

A substantial minority of the ensemble-agreed clusters — including 16 of the 42 in The Future of Britain (80143) and most of the 8 rejected in Our Responsibility — were artifacts of the audit’s context construction, not defects in the prose, and were rejected with sheet evidence:

1. Continuation sections audited against empty notes. Sections whose prose voices the latter half of the previous section’s sheet carry an empty sheet list in the cache. The audit script now inherits the preceding section’s sheets, but the original cached verdicts for several lectures predate that logic — one cached verdict states outright that “the source transcription for this section is empty” — so their wholesale “no basis in notes” clusters condemned prose whose licensing lines exist on the shared sheet.
2. Cross-sheet license invisible to section-scoped auditors. Stapledon keys insertions across sheets with ⊙ / # / α marks, recaps earlier sheets on summary sheets, and strikes a line on one sheet while leaving it live on another. An auditor shown only the section’s own sheets cannot see the license (e.g. Ourselves and the Future’s dinner-bell analogy is licensed by a ⊙ cross-reference to the previous sheet; Science and Fiction’s “But the focus, observe, is God and man” is licensed by a #-keyed line on the previous section’s sheet and was re-flagged in every confirmation round). Both classes are now fixed in work/faithfulness_audit_corpus.py for future runs (sheet inheritance; adjacent sheets supplied as marked context), and both are documented in the script header. The preserved caches were not regenerated — they are the record of what this pass actually adjudicated.

Two adjudicated rejections in Ourselves and the Future (“dinner-bell”, “one’s eye”) and one in Science and Fiction (“God and man”) will predictably re-flag if a future audit still scopes context to the section’s own sheets; the shard reports document the licensing lines. The Future of Britain (80148) additionally shows non-converging run-to-run variance: after four clean rounds its final confirmation round freshly flagged sentences that earlier rounds had passed; the residual items are itemized in its shard report and none identifies a misstatement of a legible note line.

Completion note — 3 August 2026 (final shard: Is Humanism Enough?, 80498)

The thirteenth generated lecture, whose adjudication the table above records as pending (its filename contains “humanism” and the three shard globs missed it), was adjudicated in a final single-lecture shard against its 35 consensus sheets. All 58 ensemble-agreed clusters adjudicated: 57 repaired, 1 rejected with evidence (the recap section’s “danger of overlooked singular points”, licensed verbatim by sheet 14, which the summary-section auditors cannot see). Confirmation re-audits (5 rounds, 135 further API calls, 0 failures, converged) added 37 finer-grained repairs — including two ⊙/#/☆-keyed relocations the round-1 pass had mis-attached (sheet 2’s #-keyed “humane social order” line moved to its key beside sheet 1’s “scientific materialism”; sheet 22’s ☆ query “or to the supremely important thing in it ?” moved to its key at sheet 23’s “attitude to θ universe”) — and confirmed 2 further recap-licensed rejects (“a wider universe of discourse”, sheets 14–15; the child’s “embryonic religion of I/you/we”, sheet 23). 22 singleton string-level repairs (~38 of its 165 singleton findings, several 3-provider duplicates split by the clustering normalizer); the rest remain recorded and unadjudicated. The analysis (§3.8) quotes this lecture’s sheets, not its reconstruction; a post-repair check found no conflict — the “My view is Humanism Plus” repair brings the prose into line with the sheet form the analysis quotes. Consolidated corpus totals therefore become: 604 clusters → 571 repaired, 33 rejected; 857 confirmation calls; ~307 confirmation-round repairs; ~50 of 1,341 singletons fixed string-level. Full record: work/faithfulness_out_corpus/shard_reports/is_humanism_enough_80498.md. Five misquote-class strings from these repairs were added to work/qa_propagation.py.

Correction pass — 5 August 2026 (three-part GPT-5.6 Sol review)

A three-part external review (GPT-5.6 Sol) of the whole project was verified claim-by-claim against the primary sources by the project coordinator; the items that survived verification were re-verified once more against the sources during application and are recorded here. Items the review proposed that failed re-verification were not applied.

Corrigendum to the completion note above (dated line, history unchanged): the consolidated totals in the 3 August completion note double-count one Atomic Power rejection that its own shard report records as “2 initially rejected (one rejection later conceded)” — the M07.2-r1 “interplanetary travel” summary trim, conceded and repaired in that shard’s round 3. The consolidated line should read 604 clusters → 572 repaired, 32 rejected with sheet evidence. The per-lecture table above already carries the parenthetical; README and REPORT now carry the corrected figures; the completion note is left as written.

The substantive corrections of this pass:

1. The manuscript’s fourth world restored (modern talk, slide + teleprompter). Sheet 10 reads “a world of many subordinates, but fulfilling their powers, & a few supermen attaining humanity,” verdict “a world of hierarchical free exercise of powers… does justice to θ ideal.” The talk had sanitized this to “all fulfilled, each to their capacity — the ideal.” Both surfaces now quote the hierarchy honestly, the teleprompter addresses it head-on (per analysis §3.11’s register), and the talk’s transhumanism section gains the documented complement: the lecture disavows the worship of supermen, not the ambition to make them — Stapledon’s directed-improvement advocacy runs 1908 (“a splendid race of men”) to 1948 (“breed new human types of men to people the planets”), both re-verified in the corpus. The same complement and a de-totalized shelf claim went into blog/the_salute.md/.html.
2. Endpoint harmonization (analysis §3.9 ↔ §3.10, and the concordance synthesis). §3.9’s frame (“fictionalizes its object, until the salute is offered to something he will not affirm to exist”) contradicted §3.10’s own finding that “the fiction passages are not where the argument came to rest.” §3.9 is retitled and reframed “institutionalized, doubted, fictionalized and restored,” keeping the three late positions distinct: the Brighton salute to an object whose existence is left unknown (“whether thou art… or not”), the practical-fiction floor of The Opening of the Eyes meditations 16–17, and that book’s restored second-person address ending in commanded action. The concordance’s “fictionalized endpoint” / “the speaking corpus thus ends where The Opening of the Eyes ends” was corrected to match.
3. Completeness language tiered (analysis §3.9, README). “A complete pass through everything else Stapledon published” was false (Saints and Revolutionaries, New Hope for Britain, Seven Pillars of Peace remain unheld — REPORT §8) and “the armature of the entire body of work” over-totalized. The claims are now tiered: documented recurrence across the held corpus; demonstrated dependence only where texts touch the lecture directly; shared armature elsewhere; nothing claimed for unheld titles. Also §3.1: “first public statement” → “first public prose statement” (Latter-Day Psalms, 1914, was published verse).
4. Conway Hall witness at its true remove (analysis §1, REPORT). The four quoted sentences are the catalogue’s summary of a contemporary published abstract (Monthly Record 40.2, Feb 1935, pp. 3–4) that has not yet been acquired; “a published two-page abstract of what was actually said” and “confirms” were corrected to the catalogue-summary framing and “already suggests.” Date evidence unchanged.
5. Otto/Freud hedging (analysis §§2.4, 2.6(b), 4). The inference that a second-hand borrower “would have no reason to reproduce the fascinans half” was invalid — Julian Huxley’s Religion Without Revelation excerpts both poles of Otto (the tremendum passage including horror “developed into something beautiful, pure, and glorious,” and the summary “both attraction and repulsion, both fascination and fear”; both now cited with self-verifying markers, 408 → 410). The verdict is restated: textual fit strong; direct knowledge certain by 1939 (P&L’s reading list carries The Idea of the Holy); for December 1934, plausible but not decidable against the Huxley channel. “Takes Otto’s phenomenology whole” keeps the content claim, drops the provenance implication. The Freud identification is restated as closest comparator, possibly the immediate source (verified: sheet 12 names only “psycho-analysis”; the Virgin example appears nowhere in The Future of an Illusion; P&L’s reading list cites The Interpretation of Dreams, Totem and Taboo and The Psychopathology of Everyday Life, not this book), with Freud’s corrigibility answer acknowledged as a real one and the closing caveat — a wish for austere submission may itself be wish-laden — kept as the live question.
6. Numbers and disclosures. 572/32 propagated (see corrigendum above). The visual A/B coverage gap is now disclosed on every methods surface: work/visual_critique/anthropic/ holds 219 of 223 files, so four sheets had two-provider review only (The Future of Britain 80143 sheet 4 and 80148 sheet 8, Science and Literature sheet 7, Our Responsibility 80340 sheet 4 — verified by listing). how_this_was_made.html’s stale “adjudication pending” paragraph replaced with the completed totals; REPORT’s “Human verification” item retitled “Human spot-check” and aligned with the methods page’s no-independent-collation register. The concordance’s counts made precise: 97 passages across the fifteen archive note-sets, indexed as fourteen lecture files (The Future of Britain survives as two sets sharing one deck).
7. The disputed-readings promise kept. All 61 span.disputed elements carried data-alts that nothing displayed. The shared notes CSS/JS now disclose the alternatives on hover and keyboard focus (tabindex, aria-label, tap-to-focus for touch), rendered from data-alts; and the θ-markings/plain toggle was reworked to expand shorthand over text nodes only — the old innerHTML regex could rewrite attributes (a data-alts="θ H can use" would have become “the H can use”). Verified by Playwright across deck, analysis embed, analysis side panel, v3 and mobile tap, both themes (25/25 checks; screenshots in work/shots/).
8. Six empty evidence panels fixed. Continuation sections with empty sheet lists (2 in Our Responsibility 80340, 4 in The Future of Britain 80143) now repeat their licensing sheet labelled “SHEET N (continued)” beside the prose.
9. v3 relabelled. “The diplomatic edition” over-claimed a documentary-editing standard; v3 is now “the source-bounded reading edition” everywhere it is described (the transcription remains diplomatic — that usage is unchanged, as are this file’s historical entries).
10. Excerpt retrofit. Four clear-cut modern editions that predated the excerpt machinery published full text (Talking Across the World, ed. Crossley 1987; the Crossley Reader’s rare pieces, 1997; Moskowitz 1979; the collected short fiction with 1976 posthumous elements). All four are now excerpt editions (13/428, 32/476, 3/20 and 8/274 paragraphs shown respectively), with honest rights notes; the front page’s rights sentence now claims only what is true. The Stapledon novels’ pages are untouched (owner decision pending).
11. Attribution added per the Liverpool IIIF manifests’ reproduction notice (“By permission of the Olaf Stapledon estate…”): deck footers, methods page and README now credit the estate and University of Liverpool Special Collections & Archives, explicitly as good-faith attribution while written permission is sought.

Fifteen new guard rules for the worst of the corrected strings were added to work/qa_propagation.py (the sanitized fourth world, the “abstract of what was actually said” framing, the fictionalized-endpoint phrasings, the old totals, “diplomatic edition,” and others).

Correction — 9 August 2026 (concordance provenance: five dating-table rows reconciled with the sheets’ own inscriptions)

The editions-expansion integration pass verified every row of the lecture concordance’s dating table against its set’s first-sheet inscriptions in the consensus transcription (lectures/transcriptions_v2/consensus_v2/). Five rows understated the on-sheet evidence; all five were corrected at the source (work/concordance/*.json, plus the year labels and chronological sort in work/assemble_concordance.py) and the concordance markdown and HTML view rebuilt. The other rows match their sheets (80446, 80340 and 80308 carry no first-sheet venue or date inscription, as their rows say).

1. Alternative Futures (81530) — the row was venue-less, dated “c. 1937–1938” on internal evidence alone. Sheet 1 carries, in teal pencil beside the title, Cambridge Univ Democratic Front / Feb 1938. The row now dates the set February 1938 from the inscription (an on-sheet inscription, uncorroborated by any external record; the internal evidence is consistent with it), and the set’s chronological sort moves after Science and Literature’s first delivery of October 1937.
2. Is Humanism Enough? (80498) — date and venue were right (November 1946, Conway Hall) but the evidence cell claimed “The sheets carry no date”. Sheet 1 carries, in red beside the title, Conway Hall 11/46. The evidence cell now quotes the inscription, with the project’s documentary dating (analysis §3.8) as the concurring independent evidence class.
3. Science and Fiction (80177) — the row said “No date on the sheets”, venue-less, “c. 1947–1948”. Sheet 1’s upper corner carries, set apart in red, Lewis's / Book Exhibition / Manchester / OCT (1947). The row now dates the set October 1947 by the inscription while keeping its stated caution (corner-set, year parenthesized, no external notice acquired; internal evidence consistent).
4. Science and Literature (80169) — the row said “No date on the sheets”, venue-less, “c. 1937–1938”. Sheet 1’s margin stacks three dated deliveries in Stapledon’s own hand: John of London Circle / 20.10.37 / WEA Lit Soc .11.38 / L'pl Univ. Lit' Club. 14.12.38. The row now lists the three occasions and their dates (20 October 1937; November 1938, day unrecorded; 14 December 1938).
5. The Future of Britain (80143 + 80148, one shared row) — the 80143 half was right (Birkenhead West / LNU Youth Groups / March 1938, quoted); but the 80148 half said only “post-Munich 1938… reused 1946” from internal headings, missing that set’s own sheet-1 inscriptions: Univ. Staff / Discussion Club in red, with three dated deliveries in teal — 27.10.38, W K Toc H Jan 1939, Beechcroft 28, Jan (no year stated for Beechcroft). The row now quotes all of them; “post-Munich” survives as the set’s own internal heading, consistent with 27 October 1938.

Three lectures’ editorial headnotes (the v3 section JSON and its blog twin, for 81530, 80177 and 80169) described the old rows; those cross-reference sentences were updated to match the corrected table, with each headnote’s own evidence-class caution kept as written. The 81530 and 80498 mismatches were reported by the content shard that prepared those editions; the sweep of the other thirteen sets found the further two.