Gap Scan — 2026-06-05

Scanned

Recent window: 4 dev-log files (2026-06-03-1510/2205, 2026-06-04-1505/2205) and ~42 pages edited in the last 48h. Rotation slice: 1 — concepts/ A–F (158 pages thinness-scanned against in-degree). Note: the 2026-06-04-2205 dev log is unprocessed by the nightly ingest (the 06-04 evening Ingest-Reflect cycle covered through 06-04-1505); its news items were left for developments-log, but its foundational-source thread (the Anthropic Institute post) was actioned here as gap type 1.

Gaps actioned (8 of ~19 found)

New pages created (live)

• UK Competition and Markets Authority (CMA) — lint-report backlog (explicitly flagged 06-04: "now recurs… likely warrants its own page") + live thread: the June 3 world-first conduct requirement making Google let publishers opt out of AI Overviews and AI fine-tuning without losing search placement. Built from the CMA's own gov.uk press release (primary) + Verge/Ars. Covers the SMS regime mechanics, the Microsoft probes, and the publisher-bargaining-power frame. confidence: medium, 3 sources.
• Institute for Progress (IFP) — lint backlog + ScreenDNA live thread: co-organizer of An Open Letter in Support of Mandatory Nucleic Acid Synthesis Screening and Recordkeeping (June 2026) with no page; entity_type: think-tank. Metascience/immigration/biosecurity profile, AI-biosecurity lane. 2 sources.
• Foundation for American Innovation (FAI) — same gap pair as IFP: ScreenDNA co-organizer, tech-right positioning (ex-Lincoln Network, Zach Graves), the notable deregulatory-think-tank-organizing-for-a-mandate angle. 2 sources.
• Frontier AI Governance — 3 inbound content links (ControlAI, Sir Stephen Fry, UK Artificial Superintelligence Security Bill (ControlAI draft, May 28, 2026)) with no page, carried on the deferred backlog since 06-02; also gap type 4 — OpenAI's Frontier Governance Framework and OpenAI Preparedness Framework V.2 had no concept anchor. Organizes the instrument spectrum (RSPs → transparency statutes → EOs → preemption bargains → prohibition proposals → pause infrastructure) from existing wiki pages. Depends on queued source: When AI builds itself (below).
• Critical Minerals and AI — carried from the 06-04 deferred backlog; the Quad-framework page and its INGEST task both link it, and the 06-04 lint cycle had to drop the same link from Pax Silica (US-Led AI Supply-Chain Initiative) because the page didn't exist. Built from existing Quad/Pax Silica coverage; no new web pulls needed.
• Meta Oversight Board — 3 inbound content links (Meta AI ×1, Paolo G. Carozza ×2) + live thread: the June 4 first-ever account-disabling case (ban upheld; systemic due-process and human-rights concerns in Meta's account governance). Primary oversightboard.com announcement cited. confidence: medium, 3 sources.

Pages expanded (live)

• AI and Privacy — slice-1 thin anchor: 342 words, sources_count: 0, in-degree 25. Expanded to ~900 words across the four axes; added the ITIF publicly-available-data report, the Frontiers agentic-AI data-leakage survey, the GDPR controller-breakdown analyses, the Foley state-patchwork overview, and the SECURE Data Act (House E&C Republican Working Group) preemption fight (June 3 hearing). 5 new (Source: URL) cites; sources_count 0 → 5; stays medium.

Link aliases fixed

• None warranted this run. Remaining high-count broken targets are intentional non-page references ([[wikilinks]], folder indexes [[standards/]]/[[industries/]]), lint-territory dangling sources (sources/anthropic-pwc-expanded-partnership, sources/trustworthy-agents-in-practice), or genuine missing pages on the deferred backlog below. The two concepts/frontier-ai-governance-adjacent fixes happened by page creation, not aliasing.

Queued — foundational sources

• When AI builds itself (Favaro & Clark, The Anthropic Institute, June 4, 2026) — gap type 1 + live thread (score 8, highest this run): the most explicit frontier-lab statement to date for building coordinated-pause verification infrastructure, plus previously unreported internal RSI-acceleration data (>80% of merged Anthropic production code Claude-authored; 8× code output/engineer). The wiki's Recursive Self-Improvement (RSI) page currently rests on secondary coverage; this is its anchor primary text. Verified: anthropic.com (primary host; corroborated by WSJ/Reuters/SiliconANGLE + Marcus response). Saved: Raw Sources/When AI Builds Itself (Anthropic Institute, 4 June 2026).md (full verbatim text). Queued: INGEST-anthropic-when-ai-builds-itself-2026.md with fold targets and contradiction-check instructions (pause framing vs. race-participation rationale; Marcus counterweight; self-reported productivity figures treated as positions).

Authenticity-verification failures

• None this run.

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

• US government equity stakes in AI companies (June 4 WSJ/NOTUS; Altman pitched Trump; Anthropic abstaining) — left for the nightly developments-log to fold (06-04-2205 file unprocessed); if it develops, a concepts/ page (state-capitalism / golden-share thread, cf. broken concepts/discretionary-state-capitalism, 2 inbound) becomes warranted. Score 4.
• CMA Google conduct-requirement decision text (gov.uk case page) — foundational primary-text ingest to upgrade the new CMA page; suggested by the 06-04 lint report. Score 4.
• entities/jack-dorsey (4) + entities/brian-armstrong (4) — perennial; score 3 each.
• entities/victor-zhenyi-wang (5 inbound) — score 3.
• entities/pim-de-witte (5 combined with pim-dewitte variant; create page, then collapse slug variants) — score 3.
• claude-code (6 inbound) — no clean home under "Where does X go?" (a product, not a company/model/concept); recommend the user decide: a concepts/ai-coding-agents alias, or sanction a product page convention. Score 3 but blocked on a schema call.
• Finance/energy entity cluster (kkr, blackstone, blackrock, apollo, calpine, eaton, sandisk, general-catalyst) — mostly inbound from one FT M&A source page; score 2.
• Slice-0 thin company anchors carried: Character.AI (597w × 31), Moonshot AI (470w × 22), CoreWeave (534w × 16), Replika (517w × 14), companies/anduril (src 1 × deg 17). Score 2–3.
• Slice-1 thin anchors not reached this run: AI Divides (Literacy / Occupational / Ethico-Philosophical) (861w, src 1, deg 32), Alignment Assemblies and Collective Constitutional AI (578w, src 1, deg 24), Agent Autonomy Spectrum (5 Levels) (556w, src 1, deg 19). Score 2.
• concepts/effective-altruism (2), concepts/universal-basic-income (3), concepts/singularity (3 — decide: distinct page vs. fold into Superintelligence). Score 2–3.
• Lint territory (route to lint, not gap-scan): sources/anthropic-pwc-expanded-partnership (6, likely supporting → fold into Anthropic); sources/trustworthy-agents-in-practice (5, logged created-then-deleted).

Dashboard rebuild (skill step 10)

Run after live page commits: python3 bin/build-dashboard.py — see audit log for result.

One-line summary

Six new pages (CMA, IFP, FAI, frontier-ai-governance, critical-minerals-ai, Meta Oversight Board) and one major expansion (ai-and-privacy) committed live; Anthropic's "When AI builds itself" verified and queued for review — the user should prioritize that ingest, as it re-anchors the recursive-self-improvement thread on a primary text.