Gap Scan — 2026-06-27

Scanned

Recent window: 3 dev-log files (2026-06-26 ×2, 2026-06-27 ×1) and ~36 content pages edited in the last 48h. Rotation slice: 9 — models/ (all) + industries/ (all).

Gaps actioned (5 of ~11 surviving candidates)

New pages created (live)

• GPT-5.6 (Sol, Terra, Luna) — recent thread and rotation-slice gap (models/). OpenAI's GPT-5.6 family (Sol/Terra/Luna) was named across the 06-25/26/27 dev-logs and discussed at length in OpenAI with no model page; the GPT-5.x slice has gpt-53-codex, gpt-54-thinking, and gpt-55 but no 5.6. Deferred 06-26 as "update-class while unreleased"; it has since crystallized — formally named three-tier limited preview (June 26), published pricing ($5/$30, $2.50/$15, $1/$6), a system card classifying all three High for cyber and bio/chem, and an independent METR evaluation. Built from the OpenAI primary announcement and system card, METR's eval, plus VentureBeat/TechCrunch/Axios. confidence: medium. Depends on queued source (METR eval) — cited inline with a provenance note.
• Artificial Intelligence Data Center Moratorium Act — recent-thread missing legislation page in a wiki core area (US AI regulation / data-center energy). The Sanders–Ocasio-Cortez Artificial Intelligence Data Center Moratorium Act (Senate S.4214; House companion introduced by AOC June 24, 2026 with 9 cosponsors) was named across the 06-26/27 dev-logs and ties into Data Center Siting / AI Power Politics (edited <48h) with no page. Built from the verified congress.gov bill record, the AOC house.gov press release (provisions), the Sanders senate.gov announcement, and DataCenterDynamics/Broadband Breakfast (the >20 MW coverage threshold). confidence: medium.

Pages expanded (live)

• GPT-5.3 Codex — slice anchor: in-degree 13, confidence: medium, and no sources_count (an uncited-in-frontmatter page on a heavily-relied frontier model). Added sources_count: 4; bumped last_updated; folded the live-thread fact that a later checkpoint, GPT-5.3-Codex-Spark, ran on engineering samples of OpenAI's Jalapeño accelerator (06-24 dev-log), linked the new GPT-5.6 (Sol, Terra, Luna) successor line, and added a typed ## Relationships section. Every prior fact and citation preserved. [1 new (Source: URL) cite]
• IsoDDE (Isomorphic Labs) — thinnest high-in-degree page in the slice (299 words, in-degree 6, sources_count: 1 understated relative to its two inline cites). Expanded the capabilities section from Isomorphic Labs' own February 10, 2026 article and technical report: the Runs N' Poses generalization result (>2× AF3 on the hardest bin), antibody-antigen accuracy (2.3× AF3, 19.8× Boltz-2 at DockQ > 0.8), binding-affinity benchmarks (FEP+/OpenFE/CASP16), and blind cereblon cryptic-pocket recapitulation. Corrected the announcement date (early 2026 → Feb 10, 2026) and added the canonical Zenodo citation (10.5281/zenodo.19699685) alongside the existing one; sources_count 1 → 3; the medium-confidence caveat (no independent benchmarking) preserved. [primary developer-host cites added]

Link / back-reference cleanups

• OpenAI — added a backlink from the existing GPT-5.6 discussion to the new GPT-5.6 (Sol, Terra, Luna) page, with the corrected Sol/Terra/Luna naming and June 26 limited-preview date.

Queued — foundational sources

• "Summary of METR's predeployment evaluation of GPT-5.6 Sol" (METR, June 26, 2026) — foundational AI-safety evaluation behind the new GPT-5.6 (Sol, Terra, Luna) page. Verified: metr.org (the issuing organization's own domain; HTTP 200, citation_date 2026/06/26, embedded bibtex confirms title/author/date; the distinctive 11.3 hr–>270 hr time-horizon range appears verbatim and matches the dev-log/TechCrunch coverage). Full text saved to Raw Sources/METR Predeployment Evaluation of GPT-5.6 Sol (METR, 26 June 2026).md; queued: INGEST-metr-gpt-56-sol-evaluation-2026-06-26.md (+ verification record in proposed-sources/metr-gpt-56-sol-evaluation-2026.md).

Authenticity-verification failures

• None this run. The one queued source verified against its canonical primary host (METR's own domain). New pages and expansions used primary/supporting web sources folded as inline (Source: <URL>): OpenAI's own domain (announcement + system card), congress.gov and house.gov/senate.gov (the moratorium bill), and isomorphiclabs.com (the IsoDDE developer host).

Deduplicated (already pending / already covered — not re-actioned)

• GPT-5.6 government-restricted release — the policy angle is already folded into OpenAI and Overview; the actioned gap is the missing model page, not the release news.
• Jalapeño (OpenAI/Broadcom inference chip) — already covered in OpenAI (06-26); hardware, not a model; folded the Codex-Spark link into GPT-5.3 Codex rather than create a chip page.
• METR's Sol eval as a development — update-class for METR (expanded 06-23); the eval is actioned here as a foundational source queue, not an entity edit.
• Leading the Future (super PAC) / super-PAC thread (06-27) — entity page already exists; the Bores-primary news is update-class for developments-log.
• Anthropic Mythos export-ban partial reversal, Micron earnings, Gary Marcus "lost its mojo," California UI/AI-exposure tool — all surfaced from the 06-26/27 dev-logs but map to existing pages (Anthropic, Export Controls (AI), AI Bubble Debate, AI Labor Disruption) or are supporting news-cycle items.

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

• industries/construction (slice; 152 words, confidence: low, sources_count: 1, in-degree 0) — the lowest-confidence page in the slice, but below the in-degree ≥6 thin-anchor threshold; a low-reliance leaf. Expand from supporting sources next pass. Score ~3.
• models/gpt-54-thinking, models/gpt-55 (slice; in-degree 9–10, medium, ~400–530 words) — mid-tier slice anchors that would benefit from benchmark/lineage depth; below this run's stronger items. Score ~3.
• companies/groq (06-26 dev-log: $650M raise, neocloud pivot, Nvidia talent raid) — recurring compute actor, no page; revisit if it recurs. Score ~3.
• companies/broadcom (Jalapeño co-developer; recurring custom-silicon partner) — still a supporting mention; deferred again. Score ~3.
• [[claude-code]] (in-degree 8, broken) — perennial; placement genuinely ambiguous (product vs. concept). Needs a deliberate products/-schema call from the user. Score ~4.
• LIFT AI Act (Literacy in Future Technologies Artificial Intelligence Act) (in-degree 11, confidence: high with sources_count: 1) — confidence overstated relative to single-source support; expand to 3 sources or downgrade. Carried since 06-25. Score ~4.
• US AI Safety Institute (NIST AISI) (in-degree 55, thin) / National Institute of Standards and Technology (NIST) (in-degree 24, thin) — the most-relied underbuilt entity pages; strong expands when their slice (5/6) rotates. Score ~5.
• entities/public-first-action (Anthropic-linked pro-regulation super PAC, 06-27 dev-log) — counterpart to the existing Leading the Future (super PAC); build as a small political-spending cluster. Score ~2.

Needs user action

• None blocking. The dashboard rebuild that had failed daily since 2026-06-06 (iCloud 2 conflict-copy folders inside the old Wiki/_meta/site/) now succeeds — the v4.x site.nosync migration removed the sync-conflict source, and bin/build-dashboard.py wrote 1,420 article pages cleanly this run. The recurring needs-review/*-dashboard-rebuild-failed.md notes can be considered resolved.

One-line summary

Created two live pages (GPT-5.6 (Sol, Terra, Luna), Artificial Intelligence Data Center Moratorium Act), expanded two slice anchors (GPT-5.3 Codex, IsoDDE (Isomorphic Labs)) and added one backlink, and queued one verified foundational source (METR's GPT-5.6 Sol evaluation) for review; the long-broken dashboard build is working again.