Gap Scan — 2026-07-15

Scanned

Recent window: 4 dev-log files (07-13 morning/evening, 07-14 morning/evening — the 07-14 evening file is unprocessed and left for tonight's developments-log cycle; only gap-relevant items actioned), ~64 pages edited in last 48h. Rotation slice: 13 — sources/ Q–Z (~97 pages) + Raw Sources/ orphan check (623 files).

Gaps actioned (7 of ~19 found)

New pages created (live)

• None this run — the top gaps were three matured queue-lane foundational sources anchoring the week's two live threads (pre-release vetting and labor disruption), plus slice thin anchors. The new-page candidates in the window (Hachette v. Google; the Meta AI-layoffs discrimination suit) sit in the unprocessed 07-14 evening dev-log and belong to tonight's fold first (deferred below).

Pages expanded (live)

• Techno-Federalism: How Regulatory Fragmentation Shapes the U.S.-China AI Race — the slice's most-linked anchor (combined in-degree ~135 with its concepts/ twin) at confidence: medium with no sources_count — the missing-frontmatter flag the 07-09 scan left for lint, never fixed. Corroborated against the Harvard National Security Journal primary host: canonical citation (17 Harv. Nat'l Sec. J. 1 (2025)), journal landing page and typeset-PDF URLs, December 2025 posting date (frontmatter date corrected from the placeholder 2025-01-01), and the Part V policy-implications framing from the journal's abstract. Retired the "This page summarizes…" self-reference in Provenance (house-style §2d) and added a typed ## Relationships section (supports: the concepts/ twin; related: the eight adjacent-debate links already in prose). Every prior fact and citation preserved. sources_count → 2; source_url added. [2 new (Source: URL) cites, both journals.law.harvard.edu]
• Frontier AI Safety Commitments (Seoul, 2024) — slice thin anchor #2: high on sources_count: 1 (the stochastic-parrots pattern) at combined in-degree ~101, and the lagging twin of Seoul Frontier AI Safety Commitments (2024), which got the gov.uk corroboration on 07-09 while the sources/ page kept none of it. Aligned against the same primary text (no additional pulls): DSIT publication provenance and Open Government Licence, the frontier-AI / home-governments / thresholds definitional footnotes, the safety-framework-by-the-France-summit delivery deadline, the evolution/transparency proviso, the best-practices affirmation paragraph (red-teaming, weight security, provenance mechanisms, vulnerability reporting, capability reporting, societal-risk research), Outcome wording and per-commitment detail upgrades, official signatory names (Inflection AI, Mistral AI, Samsung Electronics, Technology Innovation Institute), and primary-host confirmation of the Magic/Minimax/01.ai/NVIDIA additions (upgrading that claim from its prior medium-confidence gov.uk-updates attribution). source_url added; sources_count 1 → 2; cross-link to the legislation twin added under Relationships. [4 new (Source: URL) cites, all gov.uk]

Link aliases fixed

• None this run. The broken-link scan's ≥2-inbound targets are all either _meta/ cross-namespace links (lint-report, log, weekly briefing — not content gaps), already-queued sources (trustworthy-agents-in-practice, claude-sonnet-5-system-card, how-will-openai-compete-evans, grading-ai-2027-2025-predictions), or standing carries (claude-code/claude-cowork, foundation-models, inverse-cooking/inverse-trust, person/org entities at in-degree 2).

Queued — foundational sources

• Demis Hassabis, "A Framework for Frontier AI and the Dawning of a New Age" (X article, July 14, 2026) — type-1 dangling foundational reference on the scan's top live thread: the 07-14 ingest cycle folded the FINRA-modeled standards-body proposal onto AI Pre-Release Vetting, Demis Hassabis, Google DeepMind, Gary Marcus, and Overview with supporting cites only. Verified: title, date, and distinctive passages corroborated verbatim across TechCrunch, Axios (author interview), The Verge, and The Deep View. x.com is unsupported by the scraper, so queued URL-only with capture-at-ingest note: INGEST-hassabis-frontier-ai-framework-2026-07-15.md.
• Office of Sen. Markey, "The AI Accountability Agenda" (July 10, 2026) — matured 07-13 carry, now anchored by the 07-14 cycle's new Markey AI Accountability Agenda page. Verified on markey.senate.gov: 11-page PDF; title page, six-category structure, polling recitation, and named bills confirmed against Inside AI Policy (pp. 1–5 fetched). Queued URL-only per the large-PDF pattern: INGEST-markey-ai-accountability-agenda-2026-07-15.md.
• "We Must Act Now: A Statement on AI's Transformation of the Economy" (Stanford Digital Economy Lab, July 13, 2026) — matured 07-14 carry (Bletchley/CAIS-statement pattern): the 07-14 cycle folded it onto AI Labor Disruption, Erik Brynjolfsson, and Overview. Verified on the dedicated canonical site wemustactnow.ai (full three-point text + roster; 1,649 signatures at pull time; 16 Nobel laureates marked), corroborated by AP News, PR Newswire, The Well News, and Poets&Quants. Full raw saved: Raw Sources/We Must Act Now.md. Queued: INGEST-we-must-act-now-statement-2026-07-15.md.

Flagged for review (beyond gap-scan's lanes)

• Duplicate company pages: companies/reflection.md vs companies/reflection-ai.md — same company (US open-weight frontier lab, founded 2024) under two slugs in the same folder, violating the no-duplication rule; found while triaging the 07-14 evening dev-log's Reflection–Nebius $1B compute item. A merge needs fact/citation reconciliation across ~10 inbound links — lint/curator territory (the entities/cdao/government/cdao pattern). Note left: needs-review/2026-07-15-reflection-duplicate-pages.md.

Raw Sources orphan check (slice 13, part 2)

All 623 Raw Sources/*.md files carry source_class (zero unprocessed). Cross-referencing foundational raws against sources/ pages and the open queue leaves a small never-ingested backlog whose content is already represented on non-source pages: the six state chatbot-act primary texts (Idaho SB 1297, Nebraska LB525, Oregon SB 1546, Tennessee SB 1580, Washington HB 2225, Utah HB 276 — each has its legislation/ page but no sources/ summary page), "Inside our approach to the Model Spec" (OpenAI blog post; the Model Spec itself is ingested at OpenAI Model Spec), and "Raine v OpenAI Case Summary" (the litigation page and Raine v. OpenAI — Wrongful Death Complaint (2025) exist; the summary raw appears to be a derivative companion). Deferred rather than queued — the open queue stands at ~95 INGEST tasks and these are low-urgency sources-layer completions, not live threads. Listed in the backlog for a low-capacity run or a batch decision by the curator.

Authenticity-verification failures

• None. All three queued sources verified on their canonical hosts (x.com via ≥4 independent corroborations, markey.senate.gov, wemustactnow.ai). Two provenance cautions recorded rather than failures: the Hassabis X article cannot be scraped by current tooling (queued URL-only, capture at ingest via LinkedIn cross-post or manual copy), and the Markey PDF was verified on pp. 1–5 of 11 (fetch in full at ingest).

Deferred backlog (over the daily cap — re-surfaces next run)

• Hachette v. Google (Hachette, Elsevier, Cengage, Turow class action, S.D.N.Y., filed July 10; "$10Bs–$100Bs" internal-document allegation) — litigation page + complaint queue candidate after tonight's fold of the 07-14 evening dev-log. Score ~6.
• Meta AI-layoffs discrimination suit (26 former employees, July 14) — litigation page candidate after tonight's fold. Score ~5.
• H.R. 9619 (People-First Chatbot Act) primary text — re-checked today: Congress.gov still shows "text has not been received" as of 07/15. Carried; re-check next run. Score ~3.
• Anthropic values-across-languages study (~310,000 conversations, 20 languages, July 14) — the evening dev-log cites only a Medium repost; locate the anthropic.com primary at next touch, after tonight's fold. Score ~3.
• "Gold Eagle" initiative (White House AI-vulnerability clearinghouse, July 14) — whitehouse.gov release is a queue candidate if pages name it after tonight's fold; likely folds onto EO — Promoting Advanced AI Innovation and Security (Trump, signed June 2, 2026). Score ~3.
• Slice-13 residue (two-per-run pacing): Stanford HAI AI Index Report 2026 (in-deg 69, medium, no sources_count), TAKE IT DOWN Act — Source Summary (in-deg 82, well-sourced — frontmatter check only), Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training (in-deg 50, high/1), RAND — AI's Power Requirements Under Exponential Growth (2025) (in-deg 44, high/1), California SB 1047 — Safe and Secure Innovation for Frontier AI Models Act (enrolled + veto) (in-deg 43, high/1), Qwen3 Technical Report (in-deg 42, high/1), Safety Cases for Frontier AI (373w, medium, no sources_count, in-deg 31), When the Interface Is Neural (superseded stub) (206w, low, sc 0). Score ~3–4 each.
• Orphan-check sources-layer backlog (see above): six state chatbot acts + Model Spec blog post. Score ~2 each; possible batch ingest when the queue drains.
• US AI Regulatory Approaches Compared (in-deg 36) and Labor Disruption Timelines: Who Predicts What and Why (in-deg 21) — carried since 07-12; need a run with capacity. Score ~4 each.
• Possible standalone models/gemini-3-5 page once 3.5 Pro ships (reported targeting July 17). Score ~3. Xi's WAIC keynote (July 17, Shanghai) — same timing. Score ~2.
• Standing carries: claude-code/claude-cowork placement (curator-blocked), inverse-cooking/inverse-trust coined terms (needs-review 07-10), duplicate entities/cdao/government/cdao (lint), now joined by companies/reflection/reflection-ai (needs-review 07-15), concepts/foundation-models disambiguation, person/org pages at in-degree 2 (pam-bondi, masahiro-mori, anil-seth, eric-horvitz, orin-kerr, alondra-nelson, palmer-luckey, tom-mitchell, soufan-center, richard-dawkins), Helix (Figure AI Vision-Language-Action model) (next models slice). Score ~2–3.

One-line summary

Three foundational sources verified and queued (the Hassabis FINRA-model framework and the We Must Act Now statement — the week's two loudest threads — plus the Markey agenda PDF); the slice's two weakest high-traffic source pages expanded (techno-federalism gained its HNSJ canonical citation and lost a self-reference; the Seoul commitments sources-twin caught up to its 07-09-expanded legislation twin); the Raw Sources orphan check came back clean except a small documented sources-layer backlog; and a new companies/reflection duplicate was flagged for merge review. Nothing failed verification; the queue awaits the user's ingest review.