Scanned
Recent window: 4 dev-log files (07-07 morning/evening, 07-08 morning/evening), ~52 pages edited in last 48h. Rotation slice: 7 — legislation/ (all, ~115 pages). Note: the 07-08 evening dev-log (2026-07-08-2205) had not yet been processed by the nightly developments-log cycle at scan time; its non-gap folds (Grok 4.5 release, Amazon bond sale, Northslope acquisition, Meta Alberta data center, Epoch EBR-bench/CVE data, Illinois SB 315 signing detail) are left to that cycle.
Gaps actioned (9 of ~19 found)
New pages created (live)
- GPT-Live (GPT-Live-1 / GPT-Live-1 mini) — live-thread gap: OpenAI launched GPT-Live-1 / GPT-Live-1 mini on July 8 (07-08 evening dev-log), replacing Advanced Voice Mode for 150M+ weekly ChatGPT voice users; no page existed and the nightly cycle had not yet run. Built from the openai.com launch post and the system card on OpenAI's deployment-safety domain. confidence: medium. [Depends on queued source: GPT-Live system card — cited inline with a provenance note.]
- European Data Protection Board (EDPB) — live-thread gap: the European Data Protection Board adopted anonymisation and web-scraping-for-generative-AI guidelines at its July 8 plenary (07-08 evening dev-log), and the board — already named on AI and Privacy, CNIL (Commission nationale de l'informatique et des libertés), and Irish Data Protection Commission (DPC) — had no page despite being the EU's central GDPR-consistency regulator for AI training data.
entity_type: regulator, confidence: medium. [Depends on queued source: Guidelines 03/2026 on web scraping — cited inline with a provenance note.] - MiniMax — live-thread + rotation-adjacent gap: named on 8+ pages (us-china-ai-competition, distillation, synthetic-data, the Seoul commitments signatory list) with no page; the 07-08 evening dev-log added a 2.7-trillion-parameter model plan (The Information exclusive, flagged single-source). Built from Reuters IPO coverage (Jan 5 and Jan 9, 2026), WSJ first post-IPO results, and the gov.uk Seoul endorser list.
company_type: frontier-lab, confidence: medium.
Pages expanded (live)
- Seoul Frontier AI Safety Commitments (2024) — top slice thin anchor: in-degree 97 on
sources_count: 1. Corroborated against the gov.uk primary text and added the Paris-summit safety-framework delivery deadline, the best-practices affirmation paragraph (red-teaming, weight security, provenance mechanisms, vulnerability reporting), the evolution/transparency proviso, and the definitional footnotes (home-government = HQ country; threshold-definition latitude). sources_count 1 → 2. [4 new (Source: URL) cites, all gov.uk.] - Executive Order 14110 (RESCINDED) — slice thin anchor #2: in-degree 82 on
sources_count: 1. Added the Federal Register primary citation (88 FR 75191, doc. 2023-24283) with URL, the order's stated principles and the White House AI Council establishment per the FR abstract, and the verified EO 14148 rescission citation (FR doc. 2025-01901). sources_count 1 → 3. [3 new (Source: URL) cites.]
Link aliases fixed
[[sources/anthropic-pwc-expanded-partnership]]→(Source: <https://www.anthropic.com/news/pwc-expanded-partnership>)(6 instances across Healthcare — AI Deployment, Anthropic Claude Healthcare 2026, Anthropic Claude For Small Business 2026). Not a missing page: the raw file exists (Raw Sources/PwC is deploying Claude….md) and is classifiedsource_class: supporting, so nosources/page will ever exist — the wikilinks were citation-format errors. This clears a backlog item deferred in every gap-scan since 2026-07-02.
Queued — foundational sources
- Bender & Koller 2020, "Climbing towards NLU" — deferred in the 07-05 and 07-06 scans; the form-vs-meaning/octopus-test predecessor to Stochastic Parrots Bender 2021. Verified: ACL Anthology (DOI 10.18653/v1/2020.acl-main.463). Queued:
INGEST-bender-koller-climbing-nlu-2020.md(URL-only convention). - EDPB Guidelines 03/2026 on web scraping in the context of generative AI — live-thread primary text (adopted July 8; consultation to Oct 30). Verified: edpb.europa.eu PDF linked from the board's own plenary press release, corroborated by MLex. Queued:
INGEST-edpb-web-scraping-genai-guidelines-2026-07-08.md. - GPT-Live system card — system card for the July 8 launch. Verified: deploymentsafety.openai.com, linked directly from the openai.com launch post; eval tables and SAG Preparedness determination confirmed. Queued:
INGEST-gpt-live-system-card-2026-07-08.md.
Authenticity-verification failures
None. All three queued sources verified on primary hosts; the EO 14148 Federal Register URL added to Executive Order 14110 (RESCINDED) was also fetch-verified before citing.
Needs review
[[claude-code]]/[[claude-cowork]]placement — in-degree 7 broken target, deferred seven consecutive runs, now carrying a heavy live thread (MIIT "backdoor" alert July 8, hidden-tracker exposure, Alibaba employee ban). Alias resolution confirms no existing page; the block is schema-level (no folder for products), so per the quality gate a note with three options was left atWiki/_meta/queue/gap-scan/needs-review/2026-07-09-claude-code-product-page-placement.mdinstead of guessing a folder.
Deferred backlog (over the daily cap — re-surfaces next run)
- Kokotajlo retrospective "Grading AI 2027's 2025 predictions" — broken link (2 inbound from Situational Awareness: A One-Year Retrospective — Nathan Delisle (LessWrong, June 2025)); foundational-essay queue candidate (score ~4).
- California AI safety commission bill (advanced July 6 per Inside AI Policy) — legislation-slice candidate; needs bill-number identification before a page can be built (score ~3).
- EDPB anonymisation guidelines (Guidelines 02/2026) — companion primary text to the queued web-scraping guidelines; cited on European Data Protection Board (EDPB); follow-on ingest candidate (score ~3).
- Mistral Robostral Navigate (first robotics model, July 8) and MiniMax 2.7T model — model-page candidates once the nightly cycle folds them and second sources appear (score ~3).
- Slice thin anchors below the top two: BIS Framework for AI Diffusion — Interim Final Rule (RESCINDED) (in-deg 64, src 1), Paris AI Action Summit Declaration (2025) (60, src 1), China — Interim Measures for the Management of Generative AI Services (59, src 1), G7 Hiroshima Code of Conduct for Advanced AI (2023) (55, src 1), California SB 1047 (VETOED) (43, src 1) (score ~4 each; two-per-run pacing).
- Argentina "automated company" bill (Milei/Duprat) — deferred 4th time; non-core, no new coverage (score ~2).
- Duplicate pages
entities/cdao.md/government/cdao.md(in-deg 57) andcompanies/reflection.md/companies/reflection-ai.md— lint territory, not gap-identifier. concepts/foundation-models(in-deg 2, broken) and person/org pages at in-degree 2 — standing low-priority backlog.- APA health advisory (AI and adolescent well-being) — dropped from queue candidates:
Raw Sources/APA Health Advisory AI Adolescent Well-being 2025.mdalready exists (dedup rule 3); its broken link resolves at ingest.
One-line summary
Three live pages created (GPT-Live, EDPB, MiniMax), the legislation slice's two biggest single-source anchors reinforced from primary hosts, a seven-run-old PwC citation-format backlog cleared, three primary sources verified and queued for review, and the claude-code placement question escalated to the curator.