Gap Scan — 2026-06-28

Scanned

Recent window: 4 dev-log files (2026-06-26 ×2, 2026-06-27 ×2) and ~38 content pages edited in the last 48h. Rotation slice: 10 — analysis/ (all). Methods: wiki-wide broken-link analysis (existing-slug subtraction + inbound ranking), alias resolution on every candidate, dangling-reference scan on the recent window, anchor-thinness cross of the slice, and a recent-thread scan of the four dev-logs. Deduplicated against the open queue (57 INGEST tasks), the proposed-sources staging area, gap-scan reports from the last 7 days, and existing Raw Sources/ files.

Gaps actioned (6 of ~14 found)

New pages created (live)

• Groq — referenced in prose across ~10 pages (incl. Cerebras Systems, Specialty Inference Hardware — Cerebras, Groq, SambaNova, Inference Economics and Token Pricing, US-China AI Competition: Different Races, Different Metrics) with no page; surfaced by the live thread (June 22 $650M raise, neocloud pivot). Medium confidence; grounded in Wikipedia + Reuters/TechCrunch/Axios. Core compute area.
• Broadcom — referenced in prose across ~14 pages (incl. Nvidia & TSMC — AI Compute Infrastructure, OpenAI, AI Bubble Debate, AI Antitrust) with no page; live thread (June 24 OpenAI/Broadcom "Jalapeño" inference chip). Medium confidence; custom-silicon/XPU + networking role and OpenAI partnership documented. Core compute area.
• Public First Action (super PAC) — Anthropic-linked pro-regulation super-PAC network; the counterpart to existing Leading the Future (super PAC), named across the recent super-PAC thread (Bores primary) with no page. Medium confidence; CNBC/NYT/WaPo/Politico.
• Future of Humanity Institute (FHI) — Future of Humanity Institute (Oxford, 2005–2024); broken [[entities/fhi]] target (in-degree 2) and a queued retrospective ingest reference it; an intellectual origin point for existential-risk / alignment / governance research. Medium confidence; Wikipedia/Guardian/LessWrong.

Pages expanded (live)

• None this run. The two live-thread anchor pages that might have warranted expansion — Micron Technology (June 24–25 blowout earnings) and Microsoft (Andreou/Copilot "super app") — already carry that material (Micron last_updated 2026-06-25 with FQ3 results; Microsoft already covers the Andreou super-app thread, Copilot Cowork, and consumption pricing). The incremental June 27 color (NHS 500k Copilot rollout; 4.5% 365 attach rate) belongs to the nightly developments-log, not a gap fill.

Link aliases fixed

• [[sources/leopold-aschenbrenner-situational-awareness]] → [[sources/aschenbrenner-situational-awareness]] (2 instances in sources/nyt-douthat-chan-china-ai-2026-05-14.md).
• [[entities/uk-ico]] → [[entities/ico]] (2 instances in entities/opc-canada.md; also removed the misleading "(stub — UK peer)" annotation, since entities/ico is a full high-confidence page).

Queued — foundational sources

• None this run. The live-thread foundational candidate (METR pre-deployment evaluation of GPT-5.6 Sol) was already queued on 2026-06-27 (INGEST-metr-gpt-56-sol-evaluation-2026-06-26.md). No new foundational source was pulled, so the authenticity-verification protocol had no foundational candidate to gate.

Authenticity-verification failures

• None. New pages were built from supporting web sources folded as (Source: URL); no primary/foundational documents were downloaded to Raw Sources/ this run.

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

• apa-health-advisory-ai-adolescent-well-being-2025 — foundational APA advisory referenced by sources/ais-delusional-spirals-hai.md; a Raw Sources/APA Health Advisory AI Adolescent Well-being 2025.md file already exists but is un-ingested and has no sources/ page. Out of gap-identifier scope (ingest, not gap-fill) — flag for the next ingest cycle. Score ~5.
• concepts/iterative-deployment — OpenAI deployment-philosophy concept, in-degree 2, no live thread. Score ~3.
• entities/eric-goldman / entities/orin-kerr / entities/eric-horvitz / entities/alondra-nelson — notable individuals referenced (in-degree 2 each); low single-scan value, defer to confirm substance before creating person pages. Score ~3 each.
• concepts/foundation-models (in-degree 2) — likely an alias for Frontier Models; needs disambiguation before creating to avoid duplication. Score ~2.
• sources/anthropic-pwc-expanded-partnership (in-degree 6) — dangling source link; the underlying item reads as a supporting partnership announcement (no sources/ page warranted) and should fold into Anthropic. Score ~3.

Notes for lint

• Probable duplicate pages: companies/reflection.md ("Reflection") and companies/reflection-ai.md ("Reflection AI") appear to describe the same company (the US open-weight frontier lab founded 2024 by ex-DeepMind researchers). Recommend a merge/supersede pass — outside gap-identifier's scope (it does not merge pages).
• needs-review cleanup: 14 stale *-dashboard-rebuild-failed.md notes (2026-06-06 → 2026-06-26) predate the 2026-06-27 site.nosync migration that fixed the build; they can be cleared.

One-line summary

Four new pages created live (companies/groq, companies/broadcom, entities/public-first-action, entities/fhi) and two link-alias groups fixed; no foundational sources needed queuing (the one live-thread candidate was already queued). For the user: confirm the Groq/Broadcom/Public First Action framing, and note two items for lint — the reflection.md/reflection-ai.md duplicate and the un-ingested APA adolescent-well-being advisory.