AI Policy Wiki
Dashboard

Gap Scan — 2026-06-02

Daily gap hunt — rotation slice 12 (sources/ I–P). Live run: 5 new pages created, 1 foundational source verified and queued (CDT Dark Patterns taxonomy). Cleared the long-deferred Farrell–Shalizi / private-credit / China-committee / e-acc backlog. No clean alias fixes remained (prior runs cleared them).

Live run. Five new pages were researched and committed directly to their real Wiki/ folders (status: active; all confidence: medium). One foundational institutional report — the CDT Dark Patterns in AI Chatbots taxonomy — was authenticity-verified against cdt.org and queued for ingest. No link-alias fixes were warranted this run: the highest-count remaining broken targets are intentional folder-index references ([[standards/]], [[industries/]]) or distinct deferred concepts, not alias errors. No authenticity-verification failures.

Scanned

Recent window (last 48h): 5 New Developments Log/ files (2026-05-30-2205, 2026-05-31-1510, 2026-06-01-0510, 2026-06-01-1505, 2026-06-02-0510) plus ~30 wiki pages edited in the window. Dominant clusters: Anthropic's confidential S-1 / IPO race with OpenAI; Alphabet's $80B equity raise (incl. $10B Berkshire); HPE forecast; OpenAI moving to weaken Nvidia's CUDA moat; Harvey's Legal Agent Benchmark (LAB) ("<10% partner-task pass"); the Meta AI support-bot Instagram breach; Florida v. OpenAI (AG Uthmeier suit naming Altman personally); SoftBank's €75bn France build; SpaceX Space Force AMTI; Palo Alto Networks × Claude Mythos vuln-hunting; Anthropic's Conway agent platform; Mecka AI robotics-data raise; evaluation-awareness research; the CDT dark-patterns taxonomy; 2026 state-legislature close-outs. Most are dev-log/ingest territory already folded by the nightly cycle.

Rotation slice: 12 of 14 — sources/ I–P (88 source pages). Confirmed slice 12 per (date.today()-date(2026,1,1)).days % 14. As on the slice-11 sweep, the slice's thin pages are deliberately-compact source summaries, none with high enough in-degree to qualify as thin anchors; the slice surfaced no orphaned-source or missing-anchor gaps beyond the recent-window/rotation ledger.

Broken-link analysis: 2,491 distinct [[wikilink]] targets vs. the existing-slug set. After excluding historical references in briefings/, log.md, log-archive/, lint-report.md, queue/, and the index files — which inflate raw broken counts with no live-graph cost — and running alias resolution, the genuine live-content actionable set is far smaller than the raw 631-broken count. This filtering corrected several apparent top gaps (companies/nvidia, eo-trump-state-ai-preemption-draft, sb-53-supplementary, maxim-topaz) that exist only in briefing/log/queue records and were already resolved or are alias errors in dead text.

Gaps actioned (6 of ~20 found)

New pages created (live)

Pages expanded (live)

  • None this run. Prior runs (through 2026-06-01) cleared the live alias backlog (chip-export-controls, anthropic-economic-index-q1-2026, the Mackintosh column aliases, eo-trump-state-ai-preemption-draft, nvidia). The remaining high-count broken targets are intentional folder-index references ([[standards/]], [[industries/]] — explicitly used as "folder index" / "each profession has its own page" pointers) or distinct deferred concepts (concepts/singularity is linked alongside superintelligence as a separate "related" item, not an alias). No unambiguous rewrite warranted.

Queued — foundational sources

  • CDT — Dark Patterns in AI Chatbots: A Taxonomy to Inform Better Design (CDT Research; lead Michal Luria, w/ Adinawa Adjagbodjou, CMU; May 27 2026). Gap type 1 + live thread (named across the 2026-06-01 dev-log). A 37-pattern / 5-category taxonomy the Companion Chatbot Harms — Cross-Cutting Analysis and AI Companions threads should cite by name. Verified: cdt.org (publisher domain; report + taxonomy PDFs both publisher-hosted), corroborated by independent dev-log coverage. Saved: Raw Sources/cdt-dark-patterns-ai-chatbots-2026.md (landing-page summary stub; PDFs to pull fresh at ingest). Queued: INGEST-cdt-dark-patterns-ai-chatbots-2026.md (+ verification trail in gap-scan/proposed-sources/).

Authenticity-verification failures

  • None — the single foundational source pulled (CDT) verified cleanly against its publisher domain.

Carried-item resolutions

  • Farrell–Shalizi concept cluster — partially cleared: Seeing Like a State (applied to AI) created live. concepts/ai-political-economy (3 inbound) remains the central sibling anchor — promote next run (it is the more reusable node and now has two pages linking to it expecting it).
  • concepts/effective-accelerationism — RESOLVED (created). Its paired sibling concepts/effective-altruism (2 inbound) remains deferred.
  • entities/letitia-james, entities/jaan-tallinn, concepts/ai-companions, concepts/superintelligence — confirmed created on the 2026-06-01 run; no longer gaps.

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

  • concepts/ai-political-economy (3) — central Farrell–Shalizi sibling; now expected by two live pages. Top promote next run. Score 5.
  • concepts/effective-altruism (2) — pair with the new e/acc page. Score 3.
  • entities/jack-dorsey (4) + entities/brian-armstrong (4) — the manager-purge / founder-mode cluster (anchors Intelligence Replaces Hierarchy, Guardian source). Score 4 each.
  • concepts/autonomy-certificates (4) — Knight Columbia / Kevin Feng levels-of-autonomy proposal; deferred since 2026-05-26. Score 4.
  • entities/center-for-democracy-and-technology — entity gap surfaced by the CDT ingest; create alongside or just after the CDT source ingest. Score 3.
  • entities/mike-gallagher (3) — now expected by the new House-China-committee page (links to it). Promote next run. Score 4.
  • concepts/frontier-ai-governance (3), concepts/universal-basic-income (3), concepts/singularity (3, decide: distinct page vs. fold into Superintelligence). Score 3 each.
  • Entity finance/energy cluster: entities/kkr (6†), entities/blackstone (6†), entities/blackrock (5†), entities/apollo (5†) — †raw counts inflated by briefings; live in-degree lower, but the new private-credit page motivates pulling at least KKR/Blackstone/Apollo into entities/. Score 3.
  • entities/victor-zhenyi-wang (5) — quality-gate deferral persists (sparse public record; Mulligan-Marda-Wang co-author). Defer.
  • claude-code (6) — still quality-gate-ambiguous (concept page vs. companies/anthropic subsection). Defer pending scoping decision.
  • Lint territory (route to lint, not ingest): sources/anthropic-pwc-expanded-partnership (likely supporting → fold into companies/anthropic); sources/trustworthy-agents-in-practice (logged created-then-deleted).
  • Thin analysis/ anchors: Federal AI Adoption — Patterns and Tensions (655w × 11), Companion Chatbot Harms — Cross-Cutting Analysis (615w × 9) — expand the latter when the CDT taxonomy is ingested.

Dashboard rebuild (skill step 10)

This run committed live page changes (6 new pages), so a rebuild is warranted. Rebuild attempted, failedbin/build-dashboard.py aborts at its initial shutil.rmtree(Wiki/site) step with PermissionError: Operation not permitted: 'doge.html'. This is the known iCloud-sync quirk documented since 2026-05-30: Wiki/site/doge.html is a dataless iCloud placeholder the sandbox cannot unlink (a chmod -R u+w retry did not clear it). The wiki page graph is fully committed regardless; only the regenerated static Wiki/site/ HTML is stale. Needs a rebuild from an environment with full iCloud-materialized write access (or evicting/redownloading the Wiki/site/ placeholders first).

One-line summary

Slice 12 (sources/ I–P). 6 gaps actioned: five new pages live — Private Credit & AI Infrastructure and Constellation Energy (the live AI-buildout-financing thread), House Select Committee on the Strategic Competition between the United States and the Chinese Communist Party (export-controls anchor), Effective Accelerationism (e/acc) and Chinese AI Policy (core-area backlog), and Seeing Like a State (applied to AI) (Farrell–Shalizi cluster) — plus one verified foundational source queued (CDT Dark Patterns taxonomy). No alias fixes warranted; no verification failures. Next run: promote concepts/ai-political-economy, entities/mike-gallagher, and the CDT entity/source ingest.