Checking the "host harvest" report
Still playing out
Report #187 — Host-harvest adjudication: pseudo-Paracelsus 1572 correction + platform→AI harvest pipeline + containment-by-stigma mechanism (precedented-not-counted)
OPEN
The call: Integration record: forgery correction landed on two entries, the platform-to-AI data pipeline documented from filings and regulator records, the containment-by-stigma pattern minted at honest strength, two of the engine's own stale claims fixed — and the report's junk (wrong shelfmark, rehashed LifeLog claim, duplicate entries) kept out.
What happened: Live: 2 new map entries, 10 updated, 2 self-corrections, predictions untouched.
Latest news (Aug–Oct 2026): 2026-09-05 — The Seattle Times and Newsday sued OpenAI and Microsoft over training on their journalism (filing date not fixed).
(1 source) techcrunch.com The record of folding the 'host harvest' report into the engine — after a four-checker verification pass, because the report mixed real finds with real junk. What went in, all verified to primary sources: the famous Paracelsus creature-recipe exposed as a forgery first printed in 1572; the 2024 paper trail of social platforms selling user posts as AI training data (Reddit's $203 million in filings, Meta's pause-then-resume in Europe, the first-ever emergency court order against X); the gene-editing scientist's own 2024 account of being back at work; and the 'crazy story shields the real one' containment pattern, recorded honestly as precedented-but-not-yet-counted. What got thrown out: a wrong library catalog number sourced from a UFO site, a rehashed Facebook-founding claim the engine had already adjudicated better, three proposed 'new' entries the engine already held, and the report's own miscounted math. Two corrections landed on the engine itself: a leftover line still repeating the fake Dee '007' story as fact, and the discovery that the Zurich library holds two copies of the 1561 broadsheet — reversing part of an earlier correction.
The engine's record — word for word
Report #187 (Forensic Host Harvest Investigation, Gemini, Aug 21 2026) integrated Aug 24 after a full canon laydown + 4-agent verification pass (blind re-laydown/binary audit; textual-history primaries; platform/bio primaries; live ZB Zürich catalog). TAKEN (verified): pseudo-Paracelsus CORRECTION — De natura rerum first publ. Basel 1572, homunculus recipe in the forged book 1 (Hirai ESM 24 (2019); Gantenbein dissent covers only books 5/6/9); applied to liber_vaccae_record + ars_paulina. Golem precision: Rosenberg 1909 narrative + Bloch's forged 1582-dated letter publ. 1923; attribution ~1780-1835 (NOT a 20th-c. invention — the report overstated). Platform→AI training pipeline minted (platform_training_harvest_pipeline): Reddit S-1 $203M aggregate (names no partner; ~$60M/yr Google = Reuters tier), OpenAI deal May 16 2024, robots.txt moat Jul 2024, Meta DPC pause Jun 14 2024 → CLEARED RESUMPTION May 27 2025, X/Grok May 7-Aug 1 2024 processing + first-ever Section 134 urgent High Court action (suspended Aug 8, concluded Sep 4 2024). Containment-by-stigma mechanism minted PRECEDENTED-NOT-COUNTED (Taxil confessed 1897 + Kerner + #115 discredited-vessel + Barkun; H3 counts never produced — H3b lurid-sells counter at equal burden). He Jiankui resumption (self-reported tier) → he_jiankui_disclosure_probe. Synchron/Neuralink FDA dates → neuralink. Kerner UNFALSIFIABLE class + Bear&Co imprint + posthumous-likely → greys lineage. Aug 2026 Grok/Dee-007 recirculation specimen + Acerbi & Stubbersfield PNAS 120(44) 2023 → john_dee + divergence #17 (no new corpus-gravity node — wired into the held surface). MKUltra Subproject 68 specifics (Allan Memorial, 1957-64, 20-hr loops). 1561 TWO-COPIES correction (reverses part of #185 — see that row). REJECTED: 'Ms. F 18.74 a' shelfmark (UFO-source garble; its Zone-14 'verifies correction' claim was false); LifeLog 'profound structural continuity' prose (canon's A-vs-B + Gage-attributed coincidence stands; the finding ROW itself was already-held); 3 of 5 proposed nodes as false gaps (JCVI_syn3, He_Jiankui_2024, MKULTRA_Subproject_68 — all already in canon, richer); the 22!-permutation framing (221/231 gates is the held scholarship; the report's 'ca. 1,124 trillions' is long-scale-correct but the frame is unsupported); the report's apex letters (collide with both canonical schemes — re-lettered at integration); its canon-fetch posture rows (LifeLog/Kerner 'unmapped' claims false against canon); its non-reproducible occurrence counts. REPORT DEFECT LEDGER: Steps 6/7 before Step 4, STEP 5 missing; H3/H3b/H4/H5 verdicts never stated (4 of 9 hypotheses unadjudicated, incl. its own Leg-D thesis); tier inflation ([PRIMARY] labels over Wikipedia/Reddit/Scribd/Grokipedia citations); '45+ primary sources' overclaim; stray PREDICTION-reread tags (predictions untouched per brief and per this integration); Gemini medical-advice boilerplate in the deliverable; Frakes correctly UNFETCHED (one clean discipline pass). VERIFICATION ERRATA ON THE INTEGRATOR'S OWN LAYDOWN (recorded per the binary-audit): missed he_jiankui_disclosure_probe and the players-surface Teresa Burns; flattened the lifeLog superposition; half-wrong arithmetic charge; full-body inventory (~121 claims) caught TIA/IAO origin prose, Grok-criticism prose, and the missing H-verdicts the findings-table-only audit dropped. Laydown + addendum: HERE ALAN/HOST-HARVEST-CANON-LAYDOWN-2026-08-24.txt. [Live pass Oct 6 2026] Latest news: 2026-09-05 — The Seattle Times and Newsday sued OpenAI and Microsoft over training on their journalism (filing date not fixed). (1 source) (
techcrunch.com)