Gap Scan — 2026-07-18

Scanned

Recent window: 4 dev-log files (07-16 morning/evening, 07-17 morning/evening — the 07-17 evening file is unprocessed and left for tonight's developments-log cycle; only gap-relevant items actioned, per precedent), ~70 pages edited in the last 48h. Rotation slice: 2 — concepts/ G–O (34 pages, thinness × in-degree crossed).

Tooling notes: the firecrawl family was unavailable this session; research used the built-in web search/fetch fallback the web-research skill sanctions, with all authenticity-protocol steps run per source. The CourtListener MCP connector rejected its credentials again (second occurrence after 07-11) — user should reconnect it; court-filing verification fell back to public corroboration.

Gaps actioned (7 of ~20 found)

New pages created (live)

• CXMT (ChangXin Memory Technologies) (score ~7) — type-3 missing company page on a live thread: CXMT was named across five-plus content pages (export-controls-ai ×2, semiconductor-supply-chain ×3, chip-smuggling, overview) with no page anywhere, while the $8.6B Shanghai IPO filing (July 15), the DoD "Chinese Military Company" designation, the reported Entity List approval, and the House China committee memory-purchase-ban push (July 16) made it the compute thread's most-touched pageless actor. Built from the Reuters explainer (via Yahoo mirror of the wire), SCMP, FT, and dev-log items; Snapshot tables for revenue (Q1 +719% YoY) and implied valuation; wired to export-controls, semiconductor-supply-chain, chip-smuggling, us-china-ai-competition, BIS. confidence: medium. One field left honest: founded: unknown — no pulled source states the founding year (enrichment candidate; the Reuters explainer covers origin via Zhu Yiming but not the date).

Pages expanded (live)

• Open-Source AI / Open-Weight Models — the slice's most-relied thin anchor (in-degree 129, highest in concepts/ G–O; sources_count missing; last touched June 21 while the open-weight thread became the month's loudest). Added: Kimi K3 (largest open-source release to date, Frontend Code Arena first, AA Index third, Greenblatt 6–8-month estimate), Thinking Machines' Inkling as a US frontier-scale open-weight entrant (975B/41B MoE, Kimi K2.5 bootstrap), Z.ai's reported approach to $1B annual sales, Xi's WAIC "open source and open collaboration" keynote + 29-country WAICO, the White House open-source EO consideration, and the UK AISI 4–7-month cyber-lag finding; cross-linked the previously-unlinked Open-Weight Frontier Models companion page. 8 new (Source:) cites; sources_count set to 13.
• General-Purpose AI (GPAI) — slice thin anchor #2 (in-degree 41, sources_count missing, stale 06-06): the EU AI Act GPAI page never linked the wiki's own EU General-Purpose AI Code of Practice (Final Version, 2025) (in-degree ~95) or EU AI Office — GPAI Provider Guidelines and Enforcement Framework. Added a Compliance instruments section (Code published July 10, 2025; adequacy decisions Aug 1, 2025; 23 confirmed signatories per the April 23, 2026 EC list; xAI safety-chapter-only; enforcement powers incl. fines activate Aug 2, 2026 — mirrored from the verified EC host), plus the UK AISI open-weight lag finding and Kimi K3/Inkling in the open-source dimension. sources_count set to 8.
• Former Meta Employees v. Meta (AI-assisted layoffs) (in passing, alongside the complaint queue) — added the reported caption (Does 1–26 v. Meta Platforms, Inc.), the ADA/FMLA/PDA/Title VII claim statutes, the first-major-AI-RIF-challenge framing, and the July 22 preliminary-injunction posture from law-firm analyses; corrected filing date to July 13. sources 4→5.

Link aliases fixed

• None this run. The ≥2-inbound broken targets are _meta/ cross-namespace links, table-cell \| pipe-link artifacts (verified: [[models/glm-5-2\]] etc. resolve — the pages exist), already-queued sources, or standing carries (person/org entities at in-degree 2, claude-code placement, foundation-models disambiguation).

Queued — foundational sources

• New York Executive Order No. 62 (July 14, 2026) — type-1 dangling primary text on the window's top US thread: New York Data-Center Moratorium (2026) cites the governor.ny.gov primary inline only. Verified and captured in full from the canonical host (title, 12 GW queue recital, 50 MW definition corroborated vs. Reuters and the EO PDF). Saved: Raw Sources/New York Executive Order 62 - Data Center Moratorium.md. Queued: INGEST-ny-eo-62-2026-07-18.md.
• Rome Declaration / "Humanity at the Threshold" (Global Nobel Laureates Assembly, July 16, 2026) — type-1 dangling foundational declaration (Bletchley / We Must Act Now pattern), named as a section heading on Religious and Civil-Society Voices on AI with supporting cites only. Canonical host identified (globalnobelassembly.org/declaration — JS-rendered, queued URL-only per the Hassabis precedent); identity corroborated via Vatican News (full formal title, Capitoline Hill signing, Magnifica humanitas lineage), The Elders text mirror, and press. Queued: INGEST-rome-declaration-humanity-threshold-2026-07-18.md; trail in gap-scan/proposed-sources/.
• Does 1–26 v. Meta Platforms complaint (N.D. Cal., filed July 13, 2026) — matured 07-17 carry ("needs docket identification"): the RECAP Doc. 1 PDF is the primary (PACER case id 474171); caption/claims corroborated via CNBC, Reuters, and Butzel/Mondaq law-firm alerts. CL docket-number confirmation deferred to ingest (connector credentials rejected). Queued URL-only: INGEST-does-v-meta-ai-layoffs-complaint-2026-07-18.md.
• Regulation (EU) 2022/1925 (Digital Markets Act) primary text — matured 07-17 deferred item now that Digital Markets Act (EU) exists with a provenance note. Verified on EUR-Lex via the ELI OJ URL (CELEX 32022R1925 resolves; canonical title and dates in ELI metadata). Queued URL-only: INGEST-dma-regulation-2022-1925-2026-07-18.md.

Authenticity-verification failures

• None. Three provenance cautions recorded rather than failures: the Rome Declaration's canonical page is JS-rendered (URL-only queue; text capture at ingest); the Meta complaint's docket number awaits CourtListener (MCP credentials rejected — reconnect needed); EUR-Lex's generic legal-content TXT URL returned an empty body — the ELI OJ URL is the verified ingest URL. (Checks that did not become pulls: congress.gov returned an empty page body again for H.R. 9619 — last verified "text has not been received" 07-16, carried; Gemini 3.5 Pro ship-watch resolved negatively — Bloomberg reported the launch delayed by months on July 16, so the standing "create models/gemini-3-5 when it ships" carry is closed until an actual release.)

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

• Unprocessed 07-17 evening dev-log items — tonight's fold first, then gap-scan candidates: Meta–Anthropic ~$10B compute-rental talks, Grok 4.5 release fold-check on Grok (xAI), SpaceX–Pentagon data-center talks, Etched $20B-valuation talks (zero wiki coverage — page if references recur, Markey precedent), OpenRouter sale talks, OpenEvidence ~$20B raise (two wiki mentions), NJ A.5328 enforcement suspension, EDPB cross-regulator information-sharing call, AI Data Center Bans & Lawsuits Tracker (chatgptiseatingtheworld.com). Score ~2–4 each.
• Xi Jinping WAIC keynote transcript (mfa.gov.cn, July 17) — primary-text queue candidate once pages name the transcript itself (the keynote's content folded 07-17 onto Chinese AI Policy and World Artificial Intelligence Cooperation Organization (WAICO)). Score ~3.
• H.R. 9619 (People-First Chatbot Act) text — carried; congress.gov unfetchable this run. Score ~3.
• Slice-2 residue (two-per-run pacing): Normative Competence (high on sources_count 1 at in-deg 21 — the stochastic-parrots misrating pattern), Levels of Autonomy for AI Agents (Feng-McDonald-Zhang framework) (high/1, in-deg 19), LLM Fallacy (655w, in-deg 19), Mathematical Impossibility of Perfect Fairness (high/1), The Long Industrial Revolution (medium/1, 490w), Model Welfare (low/1), Open Global Investment (OGI) Model (medium/1). Score ~3 each.
• Substack-redirect URL upgrades flagged by the 07-17 ingest Notes (Japan AI Basic Plan revision, Australia mandatory standards, Malaysia consultation items). Score ~2.
• US AI Regulatory Approaches Compared (in-deg 36) and Labor Disruption Timelines: Who Predicts What and Why (in-deg 21) — per the 07-16 note, crystallize/query-session territory, not gap-scan lanes.
• Orphan-check sources-layer backlog (07-15): six state chatbot-act primary texts + the Model Spec blog post. Batch decision when the queue drains.
• Standing carries: claude-code/claude-cowork placement (curator-blocked), inverse-cooking/inverse-trust coined terms (needs-review 07-10), duplicate entities/cdao/government/cdao (lint), companies/reflection/reflection-ai duplicate (needs-review 07-15), companies/fairly-trained misfiling (needs-review 07-16), concepts/foundation-models disambiguation, concepts/safety-training-methodologies and concepts/a-vision-of-democratic-ai (planned umbrellas, in-deg 2), person/org pages at in-degree 2 (pam-bondi, masahiro-mori, anil-seth, eric-horvitz, orin-kerr, alondra-nelson, soufan-center, recalc-academy, et al.). CXMT founded field is an enrichment candidate (no pulled source states the year). Score ~2–3.

One-line summary

CXMT finally has a company page anchored on its IPO-week prominence across five compute pages; the slice's two most-relied thin concept anchors were expanded (open-source-ai at in-degree 129 caught up to the Kimi K3/Inkling/WAICO/EO week, general-purpose-ai now wires the GPAI Code of Practice layer it never linked); and four foundational sources were verified and queued (NY EO 62 full text, the Rome "Humanity at the Threshold" declaration, the Does v. Meta complaint, the DMA primary text) — user review needed on the four queued ingests, plus a CourtListener reconnect.