Gap Scan — 2026-06-08

Scanned

Recent window: 4 dev-log files (2026-06-06-1505/2205, 2026-06-07-1504, 2026-06-08-0510) and the 13 content pages carrying last_updated 2026-06-07/06-08 (the 06-06 nightly cycle re-touched mtimes vault-wide, so last_updated frontmatter was used as the genuine-edit signal rather than filesystem mtime). Rotation slice: 4 — entities/ A–G (thinness-scanned against in-degree; 43 pages with in-degree ≥6 surfaced). Deduplicated against the 13 open INGEST- queue files, the gap-scan reports for 2026-06-01 through 06-07, the proposed-sources/ staging area, and existing Raw Sources/.

Gaps actioned (6 of ~22 found)

New pages created (live)

• None this run. The two highest-scoring live-thread candidates both resolved to existing coverage on inspection (see Carried-item resolutions), so the run's value sits in the slice-4 thin-anchor expansions and two alias fixes. No thinly-evidenced page was forced past the quality gate.

Pages expanded (live)

• Chief Digital and Artificial Intelligence Office (CDAO) — slice-4 thin anchor (585w, sources_count: 0, in-degree 46 — the highest-leverage zero-source page in the slice; core defense-deployer area). Added the GenAI.mil first-deployment record: Google Cloud's Gemini for Government selected by CDAO in December 2025 as the first AI on GenAI.mil, IL5-authorized, ~3 million DoW personnel, no-training-on-platform-data commitment. [1 new (Source: URL) cite] sources_count 0 → 1.
• Apollo Research — slice-4 thin anchor (~340w, no sources_count, in-degree 34; core safety/evals area). Added an Organization section: 2023 founding, CEO/co-founder Marius Hobbhahn (ex-Epoch AI), senior team and board, the late-2025 conversion to a Public Benefit Corporation and San Francisco office, the four work lines (science of scheming, pre-deployment evals, governance, product), policy engagement (UK AISI Taskforce, EU AI Office, UN, US Congress), and adoption of the UK AISI Inspect framework. [1 new source, apolloresearch.ai/about] sources_count → 3.
• Epoch AI — slice-4 thin anchor (383w, sources_count: 0, in-degree 20; core forecasting org, frequently cited by the live AI-bubble / Gary Marcus thread). Added an Organization and funding section: 2022 founding, director Jaime Sevilla, philanthropic-plus-commissioned funding model and named clients (Google, UK DSIT, UK ARIA, EPRI, Sentinel Bio), the explicit not-a-developer / not-a-think-tank / not-an-incubator self-positioning, the Hobbhahn→Apollo and Besiroglu→Mechanize spinouts, and the FrontierMath/OpenAI exclusivity and funding-disclosure controversy. [1 new source, epoch.ai/latest/what-is-epoch] sources_count 0 → 3.
• Frontier Model Forum — slice-4 thin anchor (sources_count: 1, in-degree 25; core frontier-lab-governance area, adjacent to the June 4 Anthropic coordinated-pause thread). Added executive director Chris Meserole; corrected the now-stale membership claim (Amazon and Meta joined May 20, 2024 — the prior "does not include Meta" line was unsourced and factually wrong); added the AI Safety Fund grantee announcement (Dec 11, 2025) and the frontier-threat information-sharing progress update (Feb 16, 2026). [3 new (Source: URL) cites] sources_count 1 → 3.

Link aliases fixed

• [[concepts/ai-platform-power]] → AI Power Concentration (2 instances: Anticipatory AI Ethics (Lazar framework, including the Technological Horizon) and Anticipatory AI Ethics (Lazar, Knight Columbia, May 1 2026)). The existing power-concentration concept page already covers the platform-power argument; no separate page warranted.
• [[companies/spacex]] → SpaceX (1 instance, in Data Centers in Space: A Pipe Dream, or AI's Next Big Thing? — Micah Maidenberg, Merrill Sherman (WSJ, May 12 2026); "(if/when created)" marker removed). SpaceX, merged with xAI in February 2026, is comprehensively covered on the xAI page (Snapshot includes the SpaceX IPO valuation, AI-segment financials, and both Colossus compute leases). See the deferred note below on whether the compute-landlord business eventually outgrows the xAI framing.

Queued — foundational sources

• None this run. Every live-thread foundational source is already queued: the signed June-2 EO (INGEST-eo-trump-ai-innovation-security-2026), NSPM-11, the OpenAI biodefense plan, and When AI Builds Itself (the Clark/Favaro coordinated-pause essay that anchors the June 4 pause-call thread). No new un-queued primary document surfaced in the window.

Authenticity-verification failures

• None this run. No new raw sources were pulled or saved; the four expansions used supporting web sources folded as (Source: URL), and all four resolved to authentic primary hosts (apolloresearch.ai, epoch.ai, frontiermodelforum.org, Google Cloud Press Corner / PRNewswire).

Carried-item resolutions

• Trump June-2 EO — Promoting Advanced AI Innovation and Security (heavy live thread across all four window dev-logs) — NOT a gap. Already fully covered: EO — Promoting Advanced AI Innovation and Security (Trump, signed June 2, 2026) was updated to the signed order (full provisions, postponement history, draft→signed deltas, reactions), and the signed primary text is verified and queued. The only residue is the legacy -draft slug, whose rename is a supersede-skill action explicitly out of gap-identifier scope.
• SpaceX as AI-compute player (30 content-page mentions; massive live thread — $920M/mo Google deal, Colossus leases, ~$70B aggregate, ~$1.75T IPO) — NOT a new-page gap. Comprehensively covered on xAI post-merger; only one stray [[companies/spacex]] link existed, fixed as an alias. Flagged below as a possible future split.

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

• companies/spacex as a dedicated page — schema/placement decision for the user, not an auto-create: as SpaceX's compute-landlord business (leasing to both Anthropic and Google) outgrows the xAI frontier-lab framing, a company_type: cloud-infra page cross-linked with xAI may become warranted. Currently one merged entity; the no-duplication rule favors the single xAI page for now. Score 3.
• Slice-4 thin anchors not reached: bureau-of-industry-and-security (712w × deg 25, src −1), frontier-model-forum's sibling center-for-ai-safety (400w × 20, src −1), future-of-life-institute (432w × 12, src −1), cnil (710w × 11, src 0), aclu (335w × 11, src 0), cset (625w × 29, src 1), apollo-research-adjacent gillian-hadfield (288w × 14), dpc (754w × 6, src 0). Score 2 each — top promotes next entities-A–G rotation.
• claude-code (8 inbound) — still blocked on the product-vs-company placement call; carried since 2026-06-04. Score 3, but needs a schema decision from the user.
• entities/pim-de-witte / entities/pim-dewitte slug split (6 combined; General Intuition founder) — create page, then collapse variants. Score 3.
• Genuine missing entity/concept pages (out of slice or lower priority): entities/larry-ellison (3), entities/sequoia-capital (3), concepts/post-training (3), concepts/effective-altruism (2), concepts/regulatory-managerialism (2), companies/dropbox (2), companies/notion (2), companies/arm (3), companies/blue-origin (orbital-datacenter thread). Score 1–2.
• concepts/singularity (4) and concepts/universal-basic-income (3) — fold decisions vs. Superintelligence / AI Dividends (Universal Basic Capital, Digital Dividend, Global Dividend); route to lint/crystallize, not gap-scan. Score 2–3.
• concepts/digital-divide (2) — a genuine distinct general concept (the classic literacy/access divide), NOT an alias for the Jarovsky AI Divides (Literacy / Occupational / Ethico-Philosophical) framework or AI Fluency Divide; AI Divides (Literacy / Occupational / Ethico-Philosophical) itself links it as depends-on. Candidate concept page next concepts rotation. Score 2.
• Lint territory (route to lint): sources/anthropic-pwc-expanded-partnership (6, likely supporting → fold into Anthropic); sources/trustworthy-agents-in-practice (5, logged created-then-deleted); briefings/weekly-2026-W19 (6, an operational-target broken link, not a content gap); the stale satisfied INGEST-* queue files.

Dashboard rebuild (skill step 10)

Failed (recurring) — bin/build-dashboard.py aborts on shutil.rmtree over OS-locked iCloud duplicate-naming artifacts (… 2) in the generated site directory; the same host-side failure has been logged since 2026-05-30 (queue/gap-scan/needs-review/2026-06-06-dashboard-rebuild-failed). The page graph is fully committed (four live expansions + three alias fixes); only the static HTML mirror under _meta/site/ is stale. Host-side fix required.

One-line summary

No new pages or queued sources warranted this run — the two big live threads (the signed June-2 EO and SpaceX-as-compute-player) both resolved to existing coverage. Value landed in four slice-4 (entities/ A–G) thin-anchor expansions of heavily-relied-upon, poorly-sourced pages — CDAO (deg 46, src 0→1), Apollo Research (deg 34, +Organization), Epoch AI (deg 20, src 0→3), and the Frontier Model Forum (deg 25; corrected a stale Amazon/Meta-membership error) — plus two clean alias fixes. User review needed: the still-open queued ingests (June-2 EO, NSPM-11, biodefense, When AI Builds Itself), the companies/spacex split question, and the recurring host-side dashboard-rebuild fix.