Gap Scan — 2026-07-22

Scanned

Recent window: 4 dev-log files (2026-07-20 0805/2206, 2026-07-21 0805/2205), ~75 pages edited in last 48h. The 07-21 evening file (31 new items) is unprocessed — its pageless candidates defer to tonight's ingest cycle per run-order precedent. Rotation slice: 6 — entities/ O–Z (~95 pages; in-degree × thinness scan surfaced ~22 thin-anchor flags).

Gaps actioned (7 of ~30 found)

New pages created (live)

• State of Tennessee v. Meta Platforms (Instagram design trial) (score ~7) — the 07-21 backlog's top matured candidate ("litigation page candidate, score ~6 once folded"): the 07-21 ingest cycle folded jury selection in State of Tennessee v. Meta Platforms onto Meta AI with no case page behind it; alias check confirmed no page covers it under any slug. Built from JURIST (fetched in full: October 2023 filing by AG Skrmetti, Chancery Court for Davidson County, TCPA claims, autoplay/Reels/notifications/disappearing-content feature allegations, Zuckerberg warnings allegation, bifurcated seven-week trial, $1,000-per-violation penalties, Section 230 defense and Meta spokesperson quote, the March 2026 New Mexico $375M verdict, and the August 18 N.D. Cal. trial detail — 29 states' federal children's-privacy claims plus CA/CO/KY/NJ consumer-protection claims) with the tn.gov complaint PDF cited as the primary text and the Daily Record and Fox Business cites relocated from Meta AI. Backlink added on Meta AI; entry added to Index. confidence: medium, sources 4.

Pages expanded (live)

• PauseAI — the slice's top thin anchor: in-degree 16 at 385 words, sources_count missing, stale since 06-06 while the protest thread (AI Backlash, edited in-window) stayed live. Added from the organization's own site (canonical host for org facts): founding (Utrecht, May 2023, Joep Meindertsma; first action at Microsoft's Brussels lobbying office), organizational structure (national chapters incl. PauseAI US under Holly Elmore; CEO Maxime Fournes; board chaired by Meindertsma incl. Otto Barten; Stichting PauseAI legal entity), the protest history (Bletchley Nov 2023 → OpenAI SF Feb 2024 → 13 countries May 2024 → French summit Feb 2025 → largest-to-date at Google DeepMind London June 2025 + first PauseCon), the April 2026 treaty proposal (IAEA-modeled AI Safety Agency, GPU-tracking verification, US–China participation, superhuman-AI deployment vetoes, copyright-training ban and creator liability), PauseCon London September 2026, and the current campaign opposing federal preemption of state AI regulation. Every prior fact and citation preserved. sources_count → 4; confidence stays medium (org facts largely self-published). [2 new (Source: URL) cites]
• Ted Cruz — slice thin anchor #2 and a core-area actor (in-degree 7 at 142 words, medium/1, stale since 06-06) sitting on the live federal-preemption thread. Added his Senate Commerce chairmanship, the September 10, 2025 five-pillar AI policy framework and SANDBOX Act (S. 2750 — OSTP-administered federal regulatory sandbox for AI waiver applications; verified on commerce.senate.gov, the primary host, with TechPolicy.Press corroboration; Abundance Institute / U.S. Chamber / ITI support from the committee's WTAS release), and a cross-reference section for the June 2026 JAWBONE Act (cite carried from that page). The unprocessed 07-21 evening item (Cruz's objection to the Blackburn preemption package) was deliberately not folded — it belongs to tonight's cycle. Every prior fact and citation preserved. sources_count 1 → 5; confidence stays medium. [4 new (Source: URL) cites]

Link aliases fixed

• None. All ≥2-inbound broken targets in the window are standing carries (curator-blocked placements, queued sources, planned umbrellas, in-degree-2 person pages) or _meta cross-links; no new unambiguous aliases surfaced.

Queued — foundational sources

• Nathan Lambert, "Kimi K3: The open-weights escalation" (Interconnects, July 20, 2026) — the 07-21 backlog's "Kimi K3 essay cluster," matured: the 07-21 cycle folded the "natural buffer" argument onto Open-Weight Frontier Models and the K3 assessment onto Kimi K3 with supporting cites only. Verified: author's own publication, full text fetched; "natural buffer" passage and eight-lab ranking confirmed verbatim. Saved: Raw Sources/Kimi K3 The Open-Weights Escalation.md. Queued: INGEST-kimi-k3-open-weights-escalation-2026-07-22.md.
• Zvi Mowshowitz, "On Kimi K3: Its Capabilities And Related Discontents" (July 20, 2026) — cluster item #2, named on Kimi K3 ("Zvi 4–6-month lag") with a supporting cite only. Verified: author's own publication, full ~86K-character text retrieved; the "at least four and my median guess is six" passage confirmed verbatim. Queued URL-only per the large-web-report pattern: INGEST-on-kimi-k3-zvi-2026-07-22.md.
• Gary Marcus, "China has all but caught up. The US is not going to 'win' the AI war." (July 20, 2026) — cluster item #3, folded onto US-China AI Competition: Different Races, Different Metrics and Open-Weight Frontier Models ("Marcus CERN-for-AI") with supporting cites only. Verified: author's own publication, full text fetched; seven-options structure and CERN-for-AI proposal confirmed. Saved: Raw Sources/China Has All But Caught Up - Marcus.md. Queued: INGEST-china-caught-up-marcus-2026-07-22.md.
• Alex Turner, "I tried to stop Google DeepMind's Pentagon deal. Then I quit." (Transformer, July 21, 2026) — type-1 dangling foundational reference on the Pentagon-contract thread: the 07-21 cycle folded its claims onto Google DeepMind with a supporting cite only. First-person primary-source essay for an episode the wiki tracks across three pages. Verified: commissioning publication's canonical host, full text fetched; all folded claims ("all lawful use," "supply chain risk," OpenAI loopholes, fewer restrictions) confirmed verbatim with resolvable outbound identifiers. Saved: Raw Sources/I Tried to Stop Google DeepMinds Pentagon Deal - Turner.md. Queued: INGEST-turner-deepmind-pentagon-2026-07-22.md.

Authenticity-verification failures

• None. (H.R. 9619 was not retried this run — the pull budget was exhausted at the cap after 10 pulls; congress.gov has returned an empty body on four prior attempts. Carried.)

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

• Unprocessed 07-21 evening dev-log items — tonight's fold first, then gap-scan candidates: OpenAI's Hugging Face incident disclosure + the HF security-incident post (source-queue candidates once pages name them, score ~6 — the sandbox-escape/ExploitGym episode); UK AISI "cheating behaviour in frontier model evaluations" blog (source-queue candidate ~5); University of Tennessee Research Foundation v. Anthropic (litigation page candidate ~5 — first university patent suit against a major AI developer); Sony Music's second Udio suit (fold onto UMG Recordings v. Suno (AI music training data)-adjacent coverage or new page ~4); Illinois SB 315 third-party-audit law (legislation page candidate ~5); New Jersey FAIR Act rent-algorithm law (legislation page candidate ~4); Gemini 3.6 Flash / 3.5 Flash-Lite / Flash Cyber releases (models fold; the standing models/gemini-3-5 candidate matures only when 3.5 Pro ships); SemiAnalysis Meta-infrastructure critique (source-queue candidate ~4); the Blackburn–Cruz Oval Office preemption episode (folds onto Ted Cruz / David Sacks / AI Federalism); Q2 lobbying disclosures; OpenAI board additions; White House $200B research-funding redirection. Score ~2–6 each once folded.
• Steve Newman, "Anecdotes Everywhere, Evidence Almost Nowhere" (Second Thoughts, July 20) — essay queue candidate; the 07-21 cycle folded its 0.97pp-GDP and 16%-early-career figures onto AI and Productivity with a supporting cite. Score ~4.
• Slice-6 residue (two-per-run pacing): Susie Wiles (330w, in-deg 10, sources_count missing), Sean Cairncross (380w, in-deg 9, sources_count missing), Sen. Elizabeth Warren (high/2 at 319w), Sanjog Misra (270w, medium/1), Seth Lazar (212w, medium/1, in-deg 9), Robert Chesney (137w, in-deg 9), Special Competitive Studies Project (SCSP) (medium/1 at in-deg 8 — misrating pattern), Project Liberty Institute (291w, medium/1), Steven Adler (low/1, 126w), Tom Brown (sources_count missing), Rakshit Trivedi, Stanford Cyber Policy Center (high/1), Global Dialogue on AI Governance (200w, in-deg 7), Ted Lieu (134w), Radha Plumb (low/1), Rishi Bommasani (high/1), Packy McCormick (high/1), Stewart Baker, Pim de Witte, Office of the Privacy Commissioner of Canada (OPC) (sources_count: 0 regulator page — the garante pattern). Score ~3 each.
• H.R. 9619 (People-First Chatbot Act) primary text — carried (congress.gov unfetchable ×4; not retried this run against the pull budget). Score ~3.
• Substack-redirect URL upgrades (Japan AI Basic Plan revision, Australia mandatory standards, Malaysia consultation — 07-17 ingest Notes). Score ~2.
• Orphan-check sources-layer backlog (07-15): six state chatbot-act primary texts + the Model Spec blog post. Batch decision when the queue drains.
• Standing carries: claude-code/claude-cowork placement (curator-blocked), inverse-cooking/inverse-trust coined terms (needs-review 07-10), companies/reflection/reflection-ai duplicate (needs-review 07-15), companies/fairly-trained misfiling (needs-review 07-16), entities/cdao/government/cdao duplicate (lint), concepts/foundation-models disambiguation, concepts/safety-training-methodologies and concepts/a-vision-of-democratic-ai planned umbrellas, person/org pages at in-degree 2 (pam-bondi, alondra-nelson, eric-horvitz, orin-kerr, palmer-luckey, soufan-center, masahiro-mori, anil-seth, recalc-academy, et al.). Score ~2–3.

One-line summary

Created the Tennessee v. Meta litigation page the 07-21 backlog called for, expanded the slice's two thinnest high-traffic entity anchors (PauseAI, Ted Cruz), and queued the four matured foundational essays on the open-weight escalation and the Pentagon-deal episode (Lambert, Zvi, Marcus, Turner) — all verified on canonical hosts; the 07-21 evening file's candidates await tonight's fold.