Skip to content
Read • Question • ConnectAnthony’s notes & researchReturn with a new question
Anthony’s NotesA personal field guide
From the notebook

Citation Audit

🔍 Citation Audit — 30 Rhetoric

Audited 2026-08-01, after stripping 3,112 utm_source=chatgpt.com tags vault-wide. Every external URL in 30 Rhetoric was checked for reachability; a prioritized subset was checked for whether the source actually says what the note claims.

Reachability ≠ verification

236 of 290 URLs returned HTTP 200. That proves a page exists at that address. It does not prove the page supports the sentence it’s attached to. Only the rows marked ✅ below were opened and matched against the claim. Everything else is ⚠️ by default.


📊 Scorecard

NoteCitesUniqueWikipediaDeadWeak sourceBot-blocked
American Exceptionalism (5 merged + cites repaired 2026-08-02)24812349 → 3 ✅018
Christian Nationalism634239 → 0 ✅1 → 0 ✅03
HOW TO create the Rhetoric46418094
Crime41240003
Loss of Connection37201 → 0 ✅09 → 0 ✅3
Boy Crisis30150102
Morality29220006
Gun Control1260001
Foreign Policy (audited 2026-08-02)000000
Conservatism (audited 2026-08-02)000000
Transgenderism110010

These returned 404, 401, or no response. A dead date-slugged URL from a news outlet is the fabrication signature — the model knew the story and invented a plausible address.

Confirmed fabricated

Note citesReality
who.int/…/24-04-2024-vaccines-have-saved-more-than-150-million-lives-over-the-past-50-years → 404The story is real, same date, wrong slug and wrong number. Correct: Global immunization efforts have saved at least 154 million lives. Fix the URL and the figure.

Reuters — correction to the original finding (2026-08-02)

My earlier characterization was wrong

The first pass grouped these with 404s and called a dead date-slugged news URL “the fabrication signature.” A control test settles it: Reuters returns HTTP 401 for every request, including its own homepage and deliberately nonsense paths. Its 401s carry zero information about whether an article exists. These four are unverified, not dead, and nothing here supports a fabrication claim.

By contrast, WHO does discriminate — a fake path returns 404, the real one returns 200 — which is why the WHO finding above stands as genuine fabrication.

The humanitaian typo remains odd, but a typo is not proof. Open these in a browser:

  • reuters.com/…/global-child-mortality-rates-dropped-2022-progress-slow-un-says-2024-03-13/
  • reuters.com/…/moderna-covid-19-vaccine-patent-dispute-headed-court-us-nih-head-says-2021-11-10/
  • reuters.com/world/africa/us-provide-another-424-million-sudan-aid-2024-09-25/
  • reuters.com/world/africa/us-give-414-mln-humanitaian-aid-congo-2024-08-07/ ← note the typo “humanitaian” in the slug

Also dead:

  • britsoc.co.uk/media/1wnfhhmq/bsa41_national_identity.pdf and the matching BSA-41 press release
  • disappearingpod.com/episodes/the-real-father-of-modern-medicine-penicillin/
  • computerhistory.org/blog/jack-kilby-and-robert-noyce/
  • fred.stlouisfed.org/series/LNS11300001 — mistyped or invented ❌ My verdict was wrong (corrected 2026-08-02). FRED returns nothing to automated requests for every series including CIVPART, which certainly exists. Unverified, not dead. Second host-blocking false positive after Reuters — the lesson is to control-test a host with a known-real and known-fake path before calling any link dead
  • en.wikipedia.org/wiki/Disestablishment_of_state_churches_in_the_United_States — in Christian Nationalism

✅ Christian Nationalism — REPAIRED 2026-08-01

Zero Wikipedia citations remain. The note now cites Stanford Encyclopedia of Philosophy (9), Yale Avalon Project primary treaty texts (5), Justia (3), Founders Online (2), Encyclopedia Virginia, Library of Congress, State Department, and the World Happiness Report.

Case law and constitutional text now carry official reporter citations — Torcaso v. Watkins, 367 U.S. 488 (1961), U.S. Const. art. VI, cl. 3 — instead of URLs. A reporter citation cannot rot and cannot be fabricated, which makes it strictly better than any link.

22 ⚠️ needs source flags remain across 15 topics (Thirty Years’ War, the Inquisitions, Salem, St. Bartholomew’s Day, Vesalius, Index Librorum Prohibitorum, City upon a Hill, Christian views on slavery, Civil Rights Movement, religious tests, secularism in India, Taliban, East–West Schism, denominations list, Massachusetts disestablishment). Britannica was the natural substitute for most, but it blocks automated verification — and inserting an unverified URL is the exact failure this audit exists to prevent. Topic labels were preserved, so nothing was lost; each flag marks a source still to be supplied by hand.


🚫 The original finding (historical)

39 of 42 unique citations (93%) were Wikipedia. This is the note covering your most active debate project, and it violates the sourcing standard in your own method note.

The claims themselves are mostly about Supreme Court cases, which have authoritative free sources. Swap directly:

ClaimReplace Wikipedia with
Everson v. Board of Education (1947) — incorporation of the Establishment ClauseJustia 330 U.S. 1 (already cited in the note — just drop the Wikipedia duplicate)
Torcaso v. Watkins (1961) — no religious test for public officeJustia 367 U.S. 488 (already cited)
Article VI no-religious-test clauseconstitution.congress.gov

You already have the good citation sitting next to the bad one in several places. This is mostly deletion, not research.


✅ Loss of Connection — REPAIRED 2026-08-02

Zero weak sources remain. All 9 replaced: YouTube, Facebook (×3), Business Insider, Amazon (×3), Academy of Ideas, Wikipedia.

Three of these were substantive errors, not just bad links:

  1. Wrong institution. The “Friendship Recession” data is the American Perspectives Survey / Survey Center on American Life, not a Harvard report. Saying “Harvard” in a debate invites a correction that costs you the point.
  2. Stale figure. 3% (1990) → 12% (2021) → 17% (2024), per SCAL’s 2024 update. Network size has not recovered post-pandemic. The note said “tripled or quadrupled”; it’s now roughly quintupled — your claim was understated, which is the rarer and better kind of error to find.
  3. Wrong author entirely. The Holt-Lunstad 50%-survival meta-analysis was cited to an Amazon product page for Sherry Turkle’s Alone Together — in three places. Now cites Holt-Lunstad et al., PLoS Medicine 2010, the actual 148-study meta-analysis, which is also the source Morality already cites correctly.

My audit undercounted this note

The original sweep flagged 3 weak sources here. The real number was 9 — my filter listed YouTube, Facebook, and Business Insider as soft domains but not amazon.com or academyofideas.com. Domain-list filtering catches known-bad hosts; it does not catch a plausible-looking host attached to the wrong claim. Only reading the note found those.


⚠️ The original finding (historical)

Three authoritative-sounding claims are attributed to sources that aren’t the source. These are the ones that get taken apart in a live exchange.

Claim in noteCurrently cited toWhat it should cite
”American Perspectives Survey (2021) found Americans report fewer close friendships…“a YouTube videoThe Survey Center on American Life APS 2021 report directly
”A synthesis from Harvard’s Friendship Recession report… no close friends quadrupled since 1990 (to ~12%)“a Facebook group post⚠️ Attribution is muddled. The data is American Perspectives Survey / Survey Center on American Life; Harvard Kennedy School’s Leadership & Happiness Lab published a piece titled “The Friendship Recession” (Feb 2025) that reports it. Say “American Perspectives Survey,” not “Harvard report.” Also the figure has moved to ~17% (2024) — 12% is the 2021 number.
”in 1990, Gallup found only about 2–7% with zero or one friend”Business InsiderTrace to the original Gallup item or drop the precision

✅ Verified this session

Opened and matched against the claim:

ClaimSourceResult
”8% of Americans say they have no close friends, only 38% say five or more” — Loss of ConnectionPew, Oct 2023✅ exact match, both figures
2022: >48,000 firearm deaths, >50% suicide, >40% homicide — Gun ControlCDC Fast Facts✅ figures confirmed; page blocks automated fetch, confirm in browser before quoting verbatim
WHO immunization lives savedWHO, 24 Apr 2024⚠️ story real, URL fabricated, number wrong (154M not “150M+”) — see above
RAND “moderate evidence” for background checks — Gun ControlRAND Science of Gun Policy⚠️ could not verify — RAND blocks automated access. Their ratings are graded (inconclusive → limited → moderate → supportive); confirm the tier before saying “moderate” aloud

🟢 Clean notes

Crime and Morality came through with zero Wikipedia cites, zero dead links, and zero soft sources. Crime cites Brennan Center, NBER working papers, the National Research Council synthesis, and ProPublica — and states the opposition case in its own strongest form before rebutting. Morality cites Moffitt’s Dunedin work in PNAS, Holt-Lunstad’s 148-study meta-analysis, Curry’s Oxford morality-as-cooperation study. These two are the model for what the others should look like.


🔁 Structural finding

The four Five American Exceptionalism versions held 622 of the 883 citation instances in all of 30 Rhetoric — the same material duplicated four times, with the dead links propagated into each copy. Fixing a citation there means fixing it four times, or fixing it once and leaving three stale copies to mislead you later.

Merge to one canonical note before doing any citation repair on it. Per vault convention, move the superseded versions to .trash/ rather than deleting.


📋 Repair order

  1. Christian Nationalism — delete the 39 Wikipedia cites. ✅ Done 2026-08-01. Remaining hand-work: supply sources for the 22 ⚠️ needs source flags.
  2. Loss of Connection — re-source the social-media citations; update 12% → 17%. ✅ Done 2026-08-02. All 9 weak sources replaced; figure updated to 17% (2024).
  3. American Exceptionalism — merge → one, then repair citations. ✅ Both done 2026-08-02 (it was five versions, not four).
    • Fabricated WHO URL fixed; figure corrected to 154M.
    • Measles figure corrected: note said “nearly 59M, 2000–2024”; WHO gives ~60M, 2000–2023.
    • Dead Computer History link → Nobel Prize (Physics 2000, Kilby).
    • All 12 previously undefined footnotes (25–37) now sourced and verified — these carried the note’s most quotable numbers (154M immunization, 14.4M/19.8M COVID, NIH 99%/$187B, smallpox, ORT, Borlaug, CT, integrated circuit).
    • Remaining: 3 ⚠️ needs source (penicillin, ORT’s 50M figure, containerization/UPC), 2 dead BritSoc links, 1 dead podcast link, 4 unverifiable Reuters links — all annotated in place.
  4. Gun Control — open RAND, confirm the evidence tier. See Position Card.
  5. Boy Crisis — fix or drop the FRED series ID. ✅ Corrected 2026-08-02 — the link was never dead; my check was. Position card built: Boy Crisis - Position Card.
  6. Foreign Policy and Conservatism — never audited. ✅ Audited 2026-08-02, and the two concrete defects (shipping figure, malformed heading + typo) are fixed. Write real steelmen ✅ done 2026-08-02. Outstanding: source or downgrade the Niebuhr quote, and decide on the Ismay paraphrase.

⚠️ Foreign Policy & Conservatism — audited 2026-08-02

Both notes contain zero external URLs. That is the inverse of the problem everywhere else in this folder, and it is not equally forgivable in the two cases.

Foreign Policy — riskiest note in the folder

It is built almost entirely on quotable lines attributed to named authorities, with nothing to check them against. Verified this session:

ClaimVerdict
J.Q. Adams, 1821 — “goes not abroad in search of monsters to destroy”✅ wording and date correct
Kissinger — “The test of policy is how it ends, not how it begins”✅ genuine, but from a 2014 WaPo op-ed on Ukraine, not Diplomacy
NATO Article 5 invoked once, after 9/11✅ correct
”80–90% of world trade by volume moves by sea”✅ Fixed 2026-08-02 → ~80%, UNCTAD cited inline
NATO/Ismay paraphrase⚠️ real quote is “Soviet Union out, Americans in, Germans down”; the note’s softened version reads as evasion
Niebuhr — “humble and limited in scope”⚠️ in quotation marks, unverified
Kennan — “how many wars it can avoid”✅ already labeled a paraphrase. The only quote handled correctly, and the model for the rest

Conservatism — defensible sourcing, broken structure

Zero URLs is largely fine here: it argues a position in your own prose and paraphrases rather than quotes, so there is no fabrication risk. Its problems are different:

  • ✅ Malformed heading and “Personal Responsbility” typo — both fixed 2026-08-02. All eight ## headings are now well-formed and anchor-linkable.
  • ⚠️ Frontmatter has no Genre/tags/Category, so the note is invisible to every Base and query in the vault.
  • ⚠️ Framing vulnerability: the core claim invokes “Scruton, Locke, and Mill” — two canonical liberals. The strongest objection isn’t “conservatism defends privilege” (the only one the note states); it’s “you’ve described classical liberalism and labeled it conservative.” Burke appears once, in passing. Missing: Hayek’s “Why I Am Not a Conservative,” the postliberal charge that fusionism already conceded, and the selection-effect objection to tradition.

✅ Steelmen written 2026-08-02. Both notes now carry opposition cases stated at full strength plus reply matrices with pressure questions.

  • Foreign Policy — the Restrainer (your criteria never bind; Iraq passed all five on paper) and the Interventionist (you’ve never priced the cost of watching: Rwanda, Srebrenica, Syria).
  • Conservatism — Hayek (conservatism is a tempo, not a direction), the selection effect (tradition preserves what survived, not what was good), and the postliberal charge (fusionism caused the dissolution it complains about).