Gap Scan — 2026-06-19

Scanned

Recent window: 5 dev-log files (2026-06-17 ×1 of the trailing pair, 06-17 22:05, 06-18 15:11, 06-18 22:05) and ~30 wiki pages edited in the last 48h (companies/openai, companies/anthropic, companies/zhipu-ai, models/glm-5-2, several concepts/ and entities/, legislation/creator-act, government/gsa, etc.). Rotation slice: 1 — concepts/ A–F.

Gaps actioned (5 of ~14 found)

New pages created (live)

• Midjourney — live thread (06-18 dev-log: Midjourney's move into healthcare with a full-body ultrasound scanner). Widely-referenced generative-image lab (prose mentions across 3 wiki pages) with no company page. Built from Bloomberg, MarketWatch, Yahoo Finance, BBC, and Variety; covers the self-funded image business, the Disney/NBCUniversal copyright suit (June 2025), and the June 2026 medical-imaging hardware pivot. confidence: medium.
• Mark Lemley — dangling entity gap: [[entities/mark-lemley]] (in-degree 2, from Fair Learning — Mark A. Lemley & Bryan Casey (Texas Law Review, 2021)) with no page; a much-cited IP/copyright scholar directly relevant to the live copyright threads (CREATOR Act, NO FAKES Act). Built from the Stanford "How Generative AI Turns Copyright Upside Down" PDF and Reg Review, anchored to the existing Fair Learning source page. confidence: medium.
• Automated Alignment Research — recurring backlog gap (deferred 06-16 and 06-17; prose-referenced in Sequent (Sequent Research) and Jan Leike) and in today's rotation slice (A–F). Built from OpenAI Superalignment / Jan Leike's own statement of the "third approach" and Anthropic's April 2026 "Automated Alignment Researchers" weak-to-strong study (PGR 0.97 vs 0.23 human baseline; reward hacking; "alien science"). confidence: medium.

Pages expanded (live)

• AI Existential Risk — thin anchor (in-degree 44, confidence:low, sources_count:0 — heavily relied upon, poorly sourced). Added a "Notable statements and expert surveys" section (CAIS 2023 "Statement on AI Risk"; Bengio-led International AI Safety Report) and strengthened the debates section with the skeptic position and a cross-link to Automated Alignment Research. Every prior fact, table row, and citation preserved. last_updated 06-06 → 06-19, confidence low → medium, sources_count 0 → 3. [2 new (Source: URL) cites]

Link aliases fixed

• None this run. The high-count broken targets were either _meta operational (lint-report, log, briefings/…, queue/… — out of scope), already-queued/has-raw foundational sources (deduped), or ambiguous ([[models/gpt-5]] could resolve to gpt-53-codex / gpt-54-thinking / gpt-55 — deferred for human judgment rather than risk a wrong merge).

Queued — foundational sources

• "How will OpenAI compete?" (Benedict Evans, 19 Feb 2026) — dangling foundational reference (gap type 1): named by Aggregation Theory and Benedict Evans via the broken [[sources/how-will-openai-compete-evans]] (in-degree 2), with no page or raw file. Verified: ben-evans.com (author's own site; title, 2026-02-19 publish date, and five distinctive passages all confirmed). Saved: Raw Sources/How will OpenAI compete (Benedict Evans, 19 Feb 2026).md. Queued: INGEST-how-will-openai-compete-evans-2026-06-19.md; verification record in gap-scan/proposed-sources/how-will-openai-compete-evans-2026.md.

Authenticity-verification failures

• None this run. The one queued source passed verification against its canonical host.

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

• "From AGI to ASI" (Google DeepMind) — already queued by the 06-18 scan (INGEST-agi-to-asi-deepmind-2026-06-18.md); not re-actioned. Deduped.
• "Trustworthy agents in practice" (Anthropic) (in-degree 5) — raw file exists and INGEST-trustworthy-agents-in-practice-anthropic.md is already queued. Deduped.
• sources/anthropic-pwc-expanded-partnership (in-degree 6) — supporting-class partnership mis-linked as a sources/ page; raw file exists. Fold to inline (Source:); hand to lint. Recurring (06-16, 06-17). Score 2.
• entities/fhi / Future of Humanity Institute (in-degree 2) — an FHI retrospective source is already staged (INGEST-fhi-retrospective-asterisk-2026-06-09.md); revisit the entity page once that ingest decides the landing page. Score 3.
• concepts/ai-persuasion — live thread (06-18 Oxford persuasion study; plus the Farahany "Persuasion Exchange" raw source), but 0 wiki in-degree (prose/topical hole, not a dangling link). Candidate once it accrues inbound links or a foundational source is verified. Score 3.
• Security thread: SearchLeak / "parameter-to-prompt injection" (06-17) — folds into Prompt Injection via developments-log rather than a new page. Score 2.
• Central-bank entity stubs (entities/federal-reserve, entities/ecb, entities/bank-of-england, in-degree 2 each) — better created as a cluster. Score 2–3.
• In-degree-2 entity stubs named by ingested sources: entities/eric-horvitz, entities/masahiro-mori, entities/kevin-klyman, entities/palmer-luckey, entities/pam-bondi. Score 1–2.
• concepts/foundation-models — genuinely distinct from Frontier Models but low in-degree (1). Score 1.

Dashboard rebuild — blocked (environment, 9th consecutive run)

python3 bin/build-dashboard.py again failed at shutil.rmtree(SITE) with PermissionError: [Errno 1] Operation not permitted: 'analysis 2'. The empty iCloud sync-conflict directories under Wiki/_meta/site/ (analysis 2, companies 2, concepts 2, entities 2, industries 2, models 2, primers 2, sources 2, standards 2) still cannot be removed from the sandbox. The three new pages and one expansion are live and correct in the graph; only the generated HTML site is stale. Action for the user: delete the stray "<folder> 2" conflict directories under Wiki/_meta/site/ from Finder, then re-run python3 bin/build-dashboard.py. A needs-review note is filed (gap-scan/needs-review/2026-06-19-dashboard-rebuild-failed.md).

One-line summary

Created three pages live (Midjourney, Mark Lemley, Automated Alignment Research) and expanded the heavily-relied-upon AI Existential Risk from low to medium confidence; verified and queued one foundational source (Benedict Evans, "How will OpenAI compete?") for review; dashboard rebuild remains blocked on iCloud sync-conflict directories needing a one-time Finder cleanup.