Gap Scan — 2026-07-05

Scanned

Recent window: 4 dev-log files (07-03 morning/evening, 07-04 morning/evening), ~60 pages edited in last 48h. Rotation slice: 3 — concepts/ P–Z (74 pages).

Gaps actioned (7 of ~18 found)

New pages created (live)

• Meituan / LongCat Team — top live-thread gap (score ~6: +3 live, +2 core frontier-models/China, missing-company type 3). LongCat-2.0 (Meituan) (created 07-04 by the nightly cycle) had a developer: Meituan with no company page anywhere; Meituan is one of China's most prolific open-weight shippers (LongCat-Flash 560B late 2025, Flash-Thinking, Flash-Omni, LongCat-2.0 1.6T June 30). Built from VentureBeat's detailed launch feature (scraped; Owl Alpha stealth identity, 50k+ domestic-ASIC training claim, benchmarks, pricing, company history), SCMP, and Interconnects. company_type: frontier-lab. confidence: medium. Back-linked from the longcat-2 lead.
• Akrites — backlog carry (deferred 07-04, score ~3) reinforced by the 07-03 evening dev-log's full launch detail (score ~6 with live thread + cyber core area). Linux Foundation coalition (SIRT + standardized CVD; AWS/Anthropic/Google/Microsoft-GitHub/Nvidia/OpenAI/Citi/JPMorganChase among founders; Alpha-Omega seed funding) referenced on AI and Cybersecurity with no page. Built from the LF's own June 25 press release (scraped live — primary host; note: the ai-cybersecurity mention said June 26, corrected to June 25 per the release itself) plus the JADEPUFFER context. entity_type: coalition. confidence: medium.
• Central Intelligence Agency (CIA) — live-thread gap (score ~5: +3 live, +2 natsec core). Ratcliffe's June 30 AWS Summit keynote (fast-track AI/quantum, "digital nuclear weapons," cyber sword-and-shield reorg, AWS $1B intelligence-credit program) appeared in the 07-04 evening dev-log; the CIA was named across 10+ pages (nsm-ai-national-security, ai-national-security, claude-mythos-preview, palantir) with no deployer page. Built from Bloomberg, AFP/France24 (scraped), and Washington Examiner, cross-anchored to the $9B spy-agency AI funding item already sourced on NSPM-11 — Artificial Intelligence in the National Security Enterprise (Trump, June 5, 2026). confidence: medium.

Pages expanded (live)

• Stochastic Parrots (Bender, Gebru, McMillan-Major, Mitchell, 2021) and the Octopus Test — slice thin-frontmatter anchor: confidence: high on sources_count: 1 at in-degree 13. Added a "Publication and reception" section from the canonical ACM DL record (FAccT '21, March 2021, open access, DOI, the as-printed "Shmargaret Shmitchell" author credit) and the American Dialect Society's 2023 AI-related Word of the Year selection (primary press-release PDF). sources_count 1 → 3 — the high rating now has the 3-source support it previously lacked. Every prior fact and citation preserved. [2 new (Source: URL) cites] Foundational paper itself queued (below), per the precedence rule.

Link aliases fixed

• [[litigation/nyt-v-openai]] → [[litigation/nyt-v-openai-microsoft]] (1 instance, litigation/bartz-v-anthropic.md; stale "(stub)" annotation removed — the target is a full page).

Queued — foundational sources

• "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜" (Bender, Gebru, McMillan-Major, Shmitchell — FAccT 2021) — type-1 dangling foundational reference: Stochastic Parrots (Bender, Gebru, McMillan-Major, Mitchell, 2021) and the Octopus Test tracks it by name as the canonical LLM critique with no sources/ page or raw file. Verified: dl.acm.org publisher DOI landing page (open access; DOI 10.1145/3442188.3445922), corroborated by the ADS 2023 WOTY release. Open-access PDF → fetched fresh at ingest. Queued: INGEST-stochastic-parrots-bender-2021.md (+ verification record in proposed-sources/).
• "The TESCREAL bundle: Eugenics and the promise of utopia through artificial general intelligence" (Gebru & Torres, First Monday 29(4), April 2024) — type-1 dangling foundational reference: TESCREAL Framework (Gebru and Torres) (in-degree 13, sources_count: 1) attributes the framework to the authors but cites only the MIT TR feature; the peer-reviewed paper is un-ingested. Verified: firstmonday.org journal OJS landing page (DOI 10.5210/fm.v29i4.13636; full text free, CC BY-NC-SA). Queued: INGEST-tescreal-bundle-gebru-torres-2024.md (+ verification record in proposed-sources/).

Authenticity-verification failures

• None. Both queued papers verified on their canonical publisher hosts; all new/expanded pages were otherwise built from supporting web sources folded as (Source: URL) (linuxfoundation.org, france24.com/AFP, venturebeat.com, scmp.com, bloomberg.com, washingtonexaminer.com, dl.acm.org, americandialect.org).

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

• Single-essay framework pages in the P–Z slice flagged as thin anchors — Radical Optionality (in-deg 40), Three-Lane Standards-Based AI Governance (17), Specification Hazard (19), Three Privacy Problems AI Creates (18), Standards as Litigation Evidence (12), Three Theories of Victory (US / EU / China AI Governance) (10), PEAT — Proactive Elite Alignment Theory (17), Two Conditions and Three Weapons of AI Manipulation (12). Inspection shows these are complete, well-structured single-essay framework pages whose sources_count: 1 is accurate (the framework has one source by nature); blind expansion would add padding, not support. Per the 07-04 precedent, the right treatment is a reception-section pass where independent commentary exists — deferred as a batch candidate. Score ~4 each.
• Sociotechnical AI Risk Governance (in-deg 30) and Scaling Laws (in-deg 98, missing sources_count), Techno-Federalism (135, missing sources_count), Recursive Self-Improvement (RSI) (63, missing sources_count) — frontmatter repair/confidence review for lint, not blind expansion. Score ~4.
• Brain2Qwerty Nature Neuroscience paper (Meta, June 29–30) — named on Meta AI; peer-reviewed and foundational by default, but the full text is paywalled on nature.com, so an ingest could not capture the primary text; currently well-covered as supporting cites. Revisit if the wiki's neural-interface thread deepens. Score ~5.
• Argentina "automated company" bill (Milei/Duprat) — recurred in the 07-03 and 07-04 dev-logs with bill-text detail; still a proposal without passage. Lint rule (page on formal advancement) holds; folds into AI Personhood meanwhile. Score ~4.
• entities/federal-reserve — broken target (in-deg 2) plus the 07-04 Warsh institutional-review AI panel item. Borderline; a page is defensible if the Fed's AI-productivity work recurs. Score ~3.
• Google FARO proposal — folded into Google DeepMind/AI Pre-Release Vetting by the nightly cycle (07-04); page only if it institutionalizes. Score ~3.
• [[claude-code]] (in-deg 7, broken) — perennial; placement genuinely ambiguous (product vs. concept); needs a deliberate products/-schema call from the user. Score ~4.
• sources/anthropic-pwc-expanded-partnership (in-deg 6, broken) — supporting partnership announcement; fold into Anthropic (lint/fold task; no sources/ page warranted). Score ~3.
• Bender & Koller 2020 ("Climbing towards NLU," ACL) — the octopus-test companion paper, also named on Stochastic Parrots (Bender, Gebru, McMillan-Major, Mitchell, 2021) and the Octopus Test; queue candidate next run if today's parrots ingest doesn't cover the need. Score ~3.
• Person pages at in-degree 2 — pam-bondi, masahiro-mori, anil-seth, eric-horvitz, orin-kerr, alondra-nelson, clayton-christensen, palmer-luckey, eric-goldman, tom-mitchell. Confirm substance before creating. Score ~2–3 each.
• Duplicate company pages companies/reflection.md / companies/reflection-ai.md — merge/supersede remains out of gap-identifier scope (flagged since 06-28). Score ~4.
• litigation/authors-guild-v-openai (broken, 1 instance) — genuine missing litigation page (separate class action from NYT v. Microsoft, OpenAI et al.); low in-degree, defer. Score ~3.

One-line summary

Three new pages (Meituan, Akrites, CIA) close the largest live-thread and backlog gaps; the stochastic-parrots anchor now has the sourcing its high confidence claimed; two canonical critique papers (Stochastic Parrots, TESCREAL bundle) verified and queued for the user's ingest review.