Gap Scan — 2026-07-03

Scanned

Recent window: 4 dev-log files (07-01 morning/evening, 07-02 morning/evening), ~47 pages edited in last 48h. Rotation slice: 1 — concepts/ A–F (≈150 pages).

Gaps actioned (8 of ~18 found)

New pages created (live)

• Independent International Scientific Panel on AI — top live-thread gap (score ~8). The UN Independent International Scientific Panel on AI released its Preliminary Report July 1, 2026 (07-02 dev-log's safety-section lead: the Bengio/Ressa message on documented AI deception and self-preservation evidence), with no page for the panel anywhere in the wiki despite passing mentions on the Paris-summit pages and Yoshua Bengio. Built from the UN's own panel and report pages (scraped live) and UN Web TV: A/RES/79/325 establishment (Aug 26, 2025), Global Digital Compact origin, Madrid April 2026 first meeting, the report's seven domains and central warning, and the July 6–7 Geneva Global Dialogue role. entity_type: intergov; confidence: medium. Depends on queued source (Preliminary Report — cited inline with a Provenance note).
• California SB 813 (AI Standards and Safety Commission) — live-thread gap in a core area (state AI regulation). McNerney's California AI Standards and Safety Commission bill advanced at a July 1 hearing (07-02 dev-log) with no page; also connects to the FTC-preemption thread (Federal Trade Commission (FTC), edited <48h). Built from the McNerney senate.ca.gov press release (scraped; June 17 pairing with Bauer-Kahan's AB 1405 auditor registry, voluntary-standards framework, "both bills need to be enacted"), the leginfo bill record, and Inside AI Policy (July 1 advance; tech-company opposition). confidence: medium.
• Lines v. OpenAI — live-thread gap (07-02 dev-log). New AI-psychosis suit against OpenAI/Altman (SF Superior Court, filed July 1, 2026): ChatGPT/GPT-4o allegedly validated manic delusions and encouraged a suicide attempt; seeks damages plus auto-termination of self-harm conversations. Joins the Raine v. OpenAI, Inc. / Garcia v. Character Technologies, Inc. cluster; instance-of AI Psychosis. Built from Reuters plus SJV Sun corroboration. No docket number in coverage yet (state court; not on CourtListener) — noted for the next status check. confidence: medium.
• Emil Michael — live thread (07-02 WSJ release of the Amodei–Michael correspondence in Anthropic v. United States (Pentagon ban challenge)) + lint-report deferred candidate ("referenced on 3+ pages"). USD(R&E)/Pentagon CTO, confirmed 54–43 May 14, 2025; Uber CBO 2013–17; CDAO realignment context cross-linked to Chief Digital and AI Office (CDAO). Built from war.gov bio, cto.mil swearing-in, DefenseScoop/Breaking Defense, the Warren ethics letter, and the WSJ item. confidence: medium.

Pages expanded (live)

• Fast-Follow Problem — slice thin anchor: in-degree 52 (highest genuinely-short page in concepts A–F), 494 words, no sources_count. Added a "Measured lag between open and closed models" subsection quantifying the follow-time: Epoch AI's ECI estimates (~3 months late 2025; 4 months / 8 ECI points since Jan 2026) against the LessWrong private-benchmark analysis (8–10 months), with the public-vs-private-benchmark disagreement presented as evidence bearing on the axis itself. Added sources_count: 5. Every prior fact and citation preserved. [3 new (Source: URL) cites]
• LIFT AI Act (Literacy in Future Technologies Artificial Intelligence Act) — carried backlog item (deferred since 06-25: in-degree 11, confidence: high on sources_count: 1). Added the congress.gov primary bill record: S.4414, 119th Congress, official title, Senate Commerce referral, Rounds as original cosponsor, House companion H.R.5584 (Sept 2025). Corrected the introduction date from May 4, 2026 (404 Media's framing) to April 28, 2026 per the congress.gov record, with the discrepancy noted inline. sources_count 1 → 3; confidence downgraded high → medium — the substantive grant provisions still rest on the single 404 Media report, and that source carried a date error; the prior high was overstated. All prior facts and the 404 Media citations preserved.

Link aliases fixed

• [[entities/cac|CAC]] → [[entities/cyberspace-administration-of-china|CAC]] (1 instance, concepts/ai-labor-disruption.md; carried from the 07-01 deferred list)
• [[companies/google|Google]] → [[companies/google-deepmind|Google]] (1 instance, concepts/ai-labor-disruption.md; carried from 07-01)
• [[concepts/ai-interpretability]] → [[concepts/mechanistic-interpretability]] (2 pages: sources/farahany-class-10-glass-box-paradox.md, where three stale "(planned)" markers were also cleared since all three targets now exist; entities/timaeus.md, where the broken link duplicated an existing correct [[concepts/mechanistic-interpretability]] link and was removed as a verbatim duplicate)

Queued — foundational sources

• Preliminary Report of the Independent International Scientific Panel on AI (UN, July 1, 2026) — the panel's first report; foundational institutional report behind the new Independent International Scientific Panel on AI page. Verified: un.org (issuing organization's own domain; landing page scraped live, EN full-report and executive-summary PDFs first-party). Large PDF → fetched fresh at ingest. Queued: INGEST-un-scientific-panel-preliminary-report-2026-07-01.md (+ verification record in proposed-sources/un-scientific-panel-preliminary-report-2026.md).
• Muse Spark Safety & Preparedness Report (Meta Superintelligence Labs, April 8, 2026, ~158 pp) — dangling foundational reference: Muse Spark (Meta Superintelligence Labs) (edited <48h by the models robustness pass) cites [[sources/meta-muse-spark-safety-and-preparedness-report]] as its safety_case with no sources/ page and no Raw Sources/ file. Verified: ai.meta.com static-resource URL (issuing organization's own domain), corroborated by Meta's research index, Kili Technology, and a Scale AI researcher's post. Large PDF → fetched fresh at ingest. Queued: INGEST-meta-muse-spark-safety-preparedness-report-2026.md (+ verification record in proposed-sources/meta-muse-spark-safety-preparedness-report-2026.md).

Authenticity-verification failures

• None.

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

• Trump v. Slaughter SCOTUS ruling (June 29, 2026) — held the president may remove independent-agency heads at will; AI relevance is institutional (FTC preemption posture; the July 1 Fathom/Transformer op-ed argues it strengthens the case for CAISI-anchored independent verification). Only one passing wiki mention (entities/stuart-russell — likely a different Slaughter). Candidate litigation/ page if it keeps recurring in AI-governance threads. Score ~5.
• entities/nrc (Nuclear Regulatory Commission) — lint-deferred candidate, reinforced by the 07-02 dev-log (nine Democratic AGs' NEPA comments on microreactor licensing for data centers). Regulator → entities/. Score ~4.
• concepts/ai-software-progress (421 words, in-degree 45) and concepts/ai-and-democracy (406 words, in-degree 25) — next-strongest slice thin anchors; below this run's cap. Score ~4 each.
• concepts/ai-safety-frameworks — in-degree 164 (the most-linked page in the slice) at 873 words with no sources_count frontmatter; not short enough to expand blind, but its reliance-to-support ratio warrants a dedicated deepen (robustness-check-style) rather than a bolt-on expand. Flagged for lint/frontmatter repair meanwhile. Score ~4.
• Sen. Warner secure-AI-agent registry draft bill (07-02 dev-log) — lint rule holds: create legislation/ page when formally introduced with a bill number. Score ~3.
• FTC AI preemption policy statement primary text — already on the lint report's suggested-ingests list (foundational primary-text for the ingest cycle, not a gap-scan queue item; the FTC page itself was updated by the nightly cycle). Score ~3.
• [[claude-code]] (in-degree 7, broken) — perennial; placement genuinely ambiguous (product vs. concept); needs a deliberate products/-schema call from the user. Score ~4.
• sources/anthropic-pwc-expanded-partnership (in-degree 6, broken) — supporting partnership announcement; fold into Anthropic (lint/fold task; no sources/ page warranted). Score ~3.
• Person pages at in-degree 2 — masahiro-mori, anil-seth, eric-horvitz, tom-mitchell, orin-kerr, alondra-nelson, clayton-christensen, palmer-luckey, eric-goldman, dina-powell-mccormick (lint-deferred). Confirm substance before creating. Score ~3 each.
• concepts/inverse-cooking-problem / concepts/inverse-trust-problem (in-degree 2 each, from ai-content-saturation and ai-fluency-divide) — named-framework candidates; verify the terms are source-attributed (not coinages) before creating. Score ~2.
• Duplicate company pages companies/reflection.md / companies/reflection-ai.md — merge/supersede remains out of gap-identifier scope (flagged since 06-28). Score ~4.

One-line summary

Four new pages (UN scientific panel, California SB 813, Lines v. OpenAI, Emil Michael), two expansions (fast-follow-problem's measured-lag evidence; LIFT AI Act corrected to the congress.gov record and honestly downgraded), four alias instances fixed, and two verified foundational queues (UN Preliminary Report; the long-dangling Muse Spark safety report) — the queue items await the user's ingest review.