Gap Scan — 2026-05-23

Live run. The 7 new pages below were researched and committed directly to their real Wiki/ folders (status: active, confidence: medium). One foundational source is verified and queued at Wiki/queue/INGEST-topol-marcus-llm-patient-outcomes.md, awaiting a reviewable ingest.

Scanned

Recent window (last 48h): 8 New Developments Log/ files (2026-05-20-2205 through 2026-05-23-0511, all saved within the last 48h); ~60 wiki pages edited 2026-05-21 to -23. Rotation slice: 2 of 14 — concepts/ G–O. Broken-link analysis: 23,371 wikilinks across 1,212 pages → 537 unique broken targets / 792 instances. After alias resolution, doc-artifact filtering (wikilinks, wiki/index.md, etc.), and dropping supporting-source slugs that intentionally get no page, ~24 are genuine, actionable missing-page or dangling-source gaps. The 8 highest-scored were actioned (the 6–10 daily cap).

Gaps actioned (8 of ~24 found)

New pages created (live)

• Illinois SB 317 — Consumer Artificial Intelligence Notice Act — Consumer Artificial Intelligence Notice Act. Live thread (2026-05-23 dev-log): passed the Illinois Senate 57-1 on 2026-05-21. Bill text fetched from the Illinois General Assembly and the page built from the engrossed statutory text.
• California AB 2013 — Generative AI Training Data Transparency — Generative AI Training Data Transparency. 6 broken inbound links; a source summary (AB 2013 — Training Data Documentation (California)) existed but the schema requires a legislation/ page for a tracked state AI law. Built from the source page + the codified Civil Code § 3111 provisions.
• Sapient Intelligence — Singapore AI lab behind the Hierarchical Reasoning Model and, on 2026-05-21/22, the open 1.15B HRM-Text model. Live thread (May 22–23 dev-logs); not created by the nightly ingest.
• Daniela Amodei — Anthropic co-founder and President. 5 broken inbound links; Wiki/log.md and index.md both record the page as "created" two cycles ago, but the file was never actually written — a silent gap the broken-link scan caught.
• Anthropic Interpretability Team — Anthropic's mechanistic-interpretability team (Chris Olah). 4 broken inbound links, each explicitly annotated (stub) by the linking page — the wiki was waiting for this page.
• Kadrey v. Meta — the Chhabria training-data fair-use ruling; one of the wiki's two anchor AI-copyright decisions, referenced from Training AI is Not Fair Use? — Ben Thompson (Stratechery, July 2025) and Hachette et al. v. Meta (and Mark Zuckerberg) with no case-tracking page.
• NAIRR — National Artificial Intelligence Research Resource — National AI Research Resource. 3 broken inbound links; the wiki's anchor for US public AI research infrastructure, grouped with the AI Safety Institutes and Frontier Model Forum in policy-infrastructure discussions.

Pages expanded (live)

None this run — the recent-window and rotation-slice anchor pages scanned were already adequately built (several were strengthened in yesterday's run).

Link aliases fixed

None this run. Two candidates were investigated and rejected as alias fixes by the alias-resolution discipline:

• sb-53-supplementary (appeared with 6 hits) — resolved to a deleted former source page; the only live references are in a historical lint report and the archived log. No live wiki page links to it → not a gap, no fix needed.
• entities/anu (4 instances) — not an alias for the existing Anu Bradford. Context confirms it refers to the Australian National University (Seth Lazar's institution), a genuine missing page → deferred, not auto-fixed.

Queued — foundational sources

• "Have LLMs improved patient outcomes?" — Gary Marcus, Marcus on AI, 2026-05-03. Anchors 8 broken [[sources/topol-marcus-llm-patient-outcomes]] links (incl. industries/healthcare, where it is called "the binding empirical constraint" on healthcare-AI value) and is flagged by the lint report. Authenticity verified — PASS for the anchor essay (direct firecrawl scrape of the author's own Substack — the canonical host; full text obtained; corroborated). Saved: Raw Sources/Marcus - Have LLMs Improved Patient Outcomes 2026.md. Queued: Wiki/queue/INGEST-topol-marcus-llm-patient-outcomes.md. Verification trail: Wiki/queue/gap-scan/proposed-sources/topol-marcus-llm-patient-outcomes.md.

Authenticity-verification notes

One source verified — PASS (Marcus essay). No sources rejected; nothing fabricated was saved.

Provenance caveat surfaced and handled transparently: the wiki's slug topol-marcus-llm-patient-outcomes implies a co-authored Topol-Marcus paper. No such paper exists. The named argument is synthesized from three real, separate documents — the Marcus essay (verified, saved), Eric Topol's companion review "The Paradox of Medical AI Implementation," and a Nature Medicine editorial. The INGEST task keeps the existing slug so the 8 inbound links resolve, but instructs the ingest to build the source page honestly as a three-document cluster and to verify the Topol review and the Nature Medicine editorial (which has a resolvable DOI) before folding in their claims.

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

• concepts/agi — 4+ broken links, marked (planned) by Embodied AI vs AGI (China's Race Redirect); distinct from AGI Timelines (general AGI concept / discourse vs. forecasting). Score 3. Sapient Intelligence's page adds 2 more inbound links this run.
• legislation/iowa-sf-2417 — 4 broken links; missing state AI law page. Score 2 — verify the bill is genuine before drafting.
• entities/anu (Australian National University) — 4 broken links; academic-institution page (Seth Lazar's institution). Score 1.
• concepts/long-industrial-revolution — 4 broken links; AI-and-economic-history concept. Score 1.
• companies/civitai — 4 broken links (mis-filed under entities/civitai); media-gen company. Score 1.
• government/nsf — National Science Foundation; needed for the X-Labs $1.5B initiative (May 23 dev-log) and as NAIRR's lead agency. Score 2. (Deliberately not linked from the new NAIRR page to avoid creating a fresh broken target.)
• Foundational sources flagged by broken links: trustworthy-agents-in-practice (9 hits — verify it is a real paper/report), anthropic-pwc-expanded-partnership (7 — likely supporting, belongs folded into Anthropic, a lint cleanup not a sources/ page), the OpenAI Erdős unit-distance proof artifact (May 22–23 dev-log).
• Missing model/feature pages from recent dev-logs: Gemini 3.5 family (Flash/Pro), Antigravity 2.0 — expected to be folded by the nightly developments-log run; flagged for next-run confirmation.
• Missing entity pages (3–6 broken links each): david-mcdonald, amy-zhang, rakshit-trivedi, nik-marda, victor-zhenyi-wang, herbert-simon, brian-armstrong, jack-dorsey.
• Carried items not yet pursued: WISeR (AI Medicare prior-authorization model), Ex-Im Bank "ExportAI" initiative, NSF X-Labs program page.

One-line summary

Seven new pages created live — three from live news threads (Illinois SB 317, Sapient Intelligence, and the corrected Daniela Amodei page that was logged as created but never written), four closing dangling-reference and stub gaps (AB 2013, the Anthropic interpretability team, Kadrey v. Meta, NAIRR) — plus one foundational source verified, saved, and queued, with a slug-conflation issue on the "Topol-Marcus" reference surfaced and handled honestly.