Gap Scan — 2026-05-30

Live run. Four new pages were researched and committed directly to their real Wiki/ folders (status: active; all confidence: medium). Two thin model anchors in today's rotation slice (slice 9 — models/ + industries/) were expanded with full-section build-outs from supporting web sources. One foundational framework — OpenAI's Frontier Governance Framework — was authenticity-verified against openai.com / cdn.openai.com, saved to Raw Sources/, and queued for ingest. No link aliases needed fixing this run (the standing alias backlog was absorbed by the May 22 → 29 passes). No authenticity-verification failures.

Scanned

Recent window (last 48h): 5 New Developments Log/ files (2026-05-28-0505, 2026-05-28-1110, 2026-05-28-2205, 2026-05-29-1505, 2026-05-30-0505); ~45 wiki pages with last_updated inside the window. Dominant clusters: the Anthropic May 28 super-cycle ($65B Series H at $965B briefly overtaking OpenAI; Opus 4.8; Dynamic Workflows; Mythos-GA signal — all dev-log/ingest territory, already folded); the state-AI-bill triple-finish (IL SB 315 + Pritzker; CT SB 4/SB 5; NY Safe By Design Act) and the 43-AG KIDS-Act opposition; the Pope Leo XIV Magnifica Humanitas encyclical + Olah private-power thread; the Cisco multi-turn-attack 15-model benchmark; OpenAI's May 28/29 Frontier Governance Framework; and the AFT 10-point school-AI plan.

Rotation slice: 9 of 14 — models/ (41 pages) + industries/ (10 pages).

Broken-link analysis: 629 unique broken targets (up from 574 on May 29, reflecting the new pages' forward-references). After alias resolution, doc-artifact filtering, and dropping items already actioned in the May 22 → 29 gap-scan reports — and items with an existing Raw Sources/ file or queued INGEST- task — ~22 genuine new gaps remained.

The dominant theme today: the late-May federalism + child-safety + private-power cluster. Three of the highest-in-degree broken concept targets in the wiki — concepts/ai-federalism (9 inbound), concepts/ai-and-children (7 inbound), and concepts/ai-power-concentration (4 inbound, named in 5+ pages) — were all surfaced or reinforced by the May 25–29 cycle and had been flagged as strong promote-next-run candidates in the May 29 deferred backlog. The rotation slice (models/) surfaced Gemini 3 / Gemini 3 Pro (19 inbound × 312 words, no sources_count) and GPT-5.4 Thinking (19 inbound × 218 words) as the priority thin-anchor expansions, and the recent window supplied fresh material for both (the Cisco multi-turn benchmark; the GPT-5.4 → 5.5 succession; GPT-5.4-Cyber). The recent window's dangling foundational reference — OpenAI's Frontier Governance Framework (named in the May 29 dev-log and twice in the May 29 lint report) — was the clear foundational-source gap.

Gaps actioned (7 of ~22 found)

New pages created (live)

• AI Federalism — gap type 2 + live thread + dangling concept reference (9 inbound; deferred 2026-05-29). The analytic hub for the federal-vs-state allocation-of-authority / preemption fight. Carefully scoped to not duplicate the three adjacent pages it sits beside — State-Level AI Regulation (the index), Reverse Federalism (OpenAI's named coinage), and Techno-Federalism (Wu's academic framework) — covering instead the live political-legal contest: the Trump preemption EO + moratorium pressure vs. the bipartisan state-preservation coalition (the 43-AG KIDS-Act letter; the Missouri/Moolenaar Tenth-Amendment thread). Cross-linked to the IL/CT/NY trifecta and the California-effect mechanism. 3 sources; confidence: medium.

• AI and Children — gap type 2 + live thread (7 inbound; deferred 2026-05-24 → -29). The cross-cutting child-safety thread tying the companion-chatbot litigation (Garcia v. Character Technologies, Inc., Pennsylvania v. Character.AI, Tumbler Ridge Families v. OpenAI) to the legislative wave (NY Safe By Design Act, CA SB 243, CT SB 5; the federal Cruz/KIDS track) and the schools front (the May 27 AFT 10-point plan). Anchored on the design-question levers (age verification vs. default-off; parental consent; platform vs. developer duty). 3 sources; confidence: medium.

• AI Power Concentration — gap type 2 (4 inbound; named in 5+ pages per the May 29 lint report). The "concentrated private power as a distinct AI risk" thread, anchored on the load-bearing Pope Leo XIV Magnifica Humanitas primary-text framing ("technological power takes on a new face, one that is predominantly private") and the Chris Olah concession on lab competitive incentives. Positioned explicitly against AI Existential Risk (a "who governs?" rather than "will it escape?" framing) and with the contested accelerationist counter (David Sacks's "sweeping power" warning). 2 sources; confidence: medium.

• Chamber of Progress — gap type 3 + live thread (rule-of-two satisfied: substantive citations on Illinois SB 315 (frontier safety framework with mandatory third-party audits) and Virginia HB 2094 (High-Risk AI Developer and Deployer Act, vetoed); lint-flagged 2026-05-29). Center-left tech-industry policy coalition (founded March 2021; Adam Kovacevich, ex-Google US policy lead). Built as the developer/deployer-aligned counterweight in the AI-federalism fight — notable for sitting to the deregulatory side of the very frontier labs (OpenAI, Anthropic) that supported IL SB 315. 2 sources (progresschamber.org founding announcement + Wired SB 315 coverage); confidence: medium.

Pages expanded (live)

• Gemini 3 / Gemini 3 Pro — slice-9 thin anchor: 19 inbound × 312 words, confidence: medium, no sources_count. Expanded to ~860 words. Added: the November 18 2025 launch framing and full family/Deep-Think structure, the 2026 Gemini 3.1 Pro refresh (94.3% GPQA; 80.6% coding-harness, vendor-reported and flagged fast-decay), and a safety/adversarial-robustness section folding the May 27 Cisco multi-turn-attack benchmark (Gemini 3 Pro 18.1% → 73.4% multi-turn). Added a ## Relationships block. 2 new (Source: URL) cites (blog.google + deepmind.google) plus the Cisco siliconangle.com cite. sources_count set to 5; held at confidence: medium.

• GPT-5.4 Thinking — slice-9 thin anchor: 19 inbound × 218 words, confidence: medium, no sources_count. Expanded to ~620 words. Added: the GPT-5.4 launch positioning, the GPT-5.4 → GPT-5.5 succession (now superseded at the frontier), and a cybersecurity-mitigation-milestone section (first general-purpose model with High-Cybersecurity mitigations; the April 2026 GPT-5.4-Cyber defender model) tied to AI and Cybersecurity. Added a ## Relationships block with a supersedes edge to GPT-5.3 Codex. 2 new (Source: URL) cites (openai.com GPT-5.4 + GPT-5.5 pages) plus the USCSI cyber cite. sources_count set to 3; held at confidence: medium.

Link aliases fixed

• None this run. The broken-link scan's two highest-count "company/concept" misses both resolved cleanly: [[companies/nvidia]] (3) appears only in historical briefing/log records (the live instance was fixed 2026-05-25 → [[companies/nvidia-tsmc]]), and [[legislation/eo-trump-state-ai-preemption-draft]] has no remaining live instances (fully fixed 2026-05-29). No new ambiguous-alias clusters worth a bulk fix.

Queued — foundational sources

• OpenAI's Frontier Governance Framework (openai.com, 2026-05-28) — OpenAI's public governance document mapping its frontier-AI safety/security practices to California's Transparency in Frontier AI Act (SB 53) and the EU AI Act's Code of Practice for General-Purpose AI; covers risk assessment/mitigation (cyber offense, CBRN, harmful manipulation, loss of control), model reporting, security risk management, incident response, external expert input, and framework updates. Explicitly positioned as a public-facing application of OpenAI's Preparedness Framework. Verified: canonical host openai.com (announcement, HTTP 200, dated May 28 2026) + cdn.openai.com PDF; corroborated by Inside AI Policy (2026-05-29), Inside Health Policy, and aiweekly.co; distinctive-passage check passed (the CA-statute + EU-GPAI-Code naming and the eight covered areas match across sources). Saved: Raw Sources/OpenAI Frontier Governance Framework (OpenAI, 28 May 2026).md. Queued: Wiki/queue/INGEST-openai-frontier-governance-framework-2026.md. Verification record: Openai Frontier Governance Framework 2026. Will become the wiki's primary anchor for how labs are operationalizing the emerging state-and-EU compliance regime; first lab document to formally map practices to named statutes. (The May 29 lint flagged this in both "Unfixable gaps" and "Suggested next sources.")

Authenticity-verification failures

None. The OpenAI Frontier Governance Framework verified cleanly: openai.com canonical host + cdn.openai.com PDF, two-plus independent secondary corroborations, distinctive-passage match (California Transparency in Frontier AI Act + EU AI Act GPAI Code of Practice; the four named risk areas), no signs of fabrication or spoofing. Minor date discrepancy noted and resolved: the May 29 dev-log reported it; the openai.com page is dated May 28 (canonical date for ingest).

Notes for the next lint / ingest pass

• legislation/eu-ai-act is still unresolved (alias-or-missing). It is named again by today's new AI Federalism page and the OpenAI framework INGEST anchors. Carried since 2026-05-29 — this is now a high-priority foundational/legislation gap (referenced from dozens of pages). The next legislation/ rotation (slice 7) or a targeted build should resolve it.
• concepts/ai-companions is the strongest adjacent next-build. Today's AI and Children page declares it as its instance-of parent and references it repeatedly; the companion-chatbot litigation cluster motivates it. Promote next run.
• entities/letitia-james — now referenced from AI Federalism and AI and Children in addition to the NY Safe By Design page (3+ inbound). Promotion-eligible; carried from 2026-05-29.
• entities/ccia and entities/netchoice — both referenced from the new Chamber of Progress page as frequent allies; recurring anti-state-AI-bill actors. CCIA in particular (the operative IL SB 315 audit-ecosystem critique) is promotion-eligible.
• The Magnifica Humanitas primary text remains the single highest-leverage queued ingest (lint-flagged repeatedly). Today's AI Power Concentration page paraphrases it from secondary coverage with a provenance note; the ingest will let citations upgrade to [[sources/magnifica-humanitas]].
• models/ slice depth. Other slice-9 thin anchors not actioned: Helix (Figure AI Vision-Language-Action model) (157w × 7 in-degree, sources_count: 1), IsoDDE (Isomorphic Labs) (206w × 12 in-degree, sources_count: 1), GPT-5.3 Codex (383w × 20 in-degree, no sources_count), GPT-5.5 ('Spud') (395w × 25 in-degree — high in-degree, modest body). Score 3 each; re-surface next models rotation.
• industries/defense — 222 words × 8 in-degree but well-sourced (sources_count: 6); short-but-supported, low expansion priority. industries/construction is confidence: low, 143w, sources_count: 1 — a quality candidate when a second source surfaces.
• companies/dell — lint-flagged 2026-05-29 (Q1 FY27: +88% revenue, AI-server +757%); promote on next substantive datapoint. Dev-log/ingest territory primarily.

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

• concepts/ai-companions — instance-of parent for today's ai-and-children build; companion-litigation cluster. Score 6. Strong promote-next-run.
• legislation/eu-ai-act — alias-or-missing; referenced from dozens of pages incl. today's new pages. Score 5+. Resolve next legislation/ rotation.
• entities/ccia — operative IL SB 315 audit critique; 2nd+ citation; referenced from new chamber-of-progress page. Score 4.
• entities/letitia-james — 3+ inbound (NY Safe By Design + today's two concept pages). Score 4.
• concepts/superintelligence — 5 inbound as a concept page; multiple sources/ cover it but no concept anchor. Score 4.
• concepts/seeing-like-a-state (7 inbound, Farrell-Shalizi) and concepts/ai-political-economy (6 inbound) — deferred 2026-05-26 → -29. Score 4 each.
• concepts/autonomy-certificates (6 inbound) — deferred 2026-05-26 → -29. Score 4.
• entities/aei (4+ inbound) and entities/ny-dfs (4 inbound) — lint-flagged promotion candidates carried from 2026-05-27/-29. Score 4 each.
• entities/victor-zhenyi-wang (7 inbound) — high in-degree; verify whether a person page or an alias. Score 3.
• entities/jack-dorsey / entities/brian-armstrong (6 each) — manager-purge cluster. Score 3.
• Finance-firm entity cluster (entities/kkr, blackstone, blackrock, apollo, calpine, sandisk, eaton, general-catalyst) — pending CLAUDE.md entity_type resolution. Score 2 each.
• entities/maxim-topaz (5 inbound) — quality-gated until the queued Lancet paper ingests. Score 4.
• concepts/online-mind2web (3 inbound) — benchmark stub forward-referenced from models/claude-opus-4-8. Low value alone; pair with a benchmarks build. Score 3.
• entities/chamber-of-progress allies / entities/randi-weingarten — AFT president (today's ai-and-children / education thread). Score 2; promote on second substantive citation.

Dashboard rebuild (skill step 10) — deferred

The Wikipedia-style dashboard rebuild (python3 bin/build-dashboard.py) could not complete this run. Its first step, shutil.rmtree(Wiki/site/), failed with PermissionError: [Errno 1] Operation not permitted while unlinking iCloud-synced files (e.g. Wiki/site/government/doge.html). A write test confirmed the sandbox can create files in Wiki/site/ but cannot delete them — an iCloud Drive permission quirk, not a data problem. All wiki graph changes (the 4 new pages, 2 expansions, the queue/raw files, the report, and the log) committed normally; only the static Wiki/site/ HTML is now stale relative to the graph. Action for the user: re-run python3 bin/build-dashboard.py from the vault root in an environment with delete permission on the synced folder, and the dashboard will pick up today's new pages.

One-line summary

Seven substantive actions today: four new pages live — the late-May federalism + child-safety + private-power cluster (AI Federalism, AI and Children, AI Power Concentration) plus the industry-coalition counterweight Chamber of Progress — and the two slice-9 thin-anchor model expansions (Gemini 3 / Gemini 3 Pro 312 → ~860 words; GPT-5.4 Thinking 218 → ~620 words), plus one foundational framework authenticity-verified against openai.com and queued for ingest (OpenAI's Frontier Governance Framework). No alias fixes needed; no authenticity-verification failures. The strongest deferred candidates for next run are concepts/ai-companions (instance-of parent for today's child-safety build) and resolving the long-standing legislation/eu-ai-act alias-or-missing question.