Hub
Docs

Changelog

60 entriesSince Jun 6, 2026

Everything that has shipped on 44B, newest first — new sections, data ingests, API surfaces, the Art. 44-B filing workflow, and the fixes worth telling you about.

July 2026

  1. Data

    +84 papers landed in the Library from the daily feeds

  2. Data

    +381 papers landed in the Library from the daily feeds

  3. Data

    Overnight harvest: +64 incidents · +300 papers

  4. Data

    +12,156 papers landed in the Library from the daily feeds

  5. Data

    Overnight harvest: +3,263,876 papers

  6. Data

    +3,252,378 papers landed in the Library from the daily feeds

  7. Data

    Overnight harvest: +578 papers

  8. Data

    +147,076 papers landed in the Library from the daily feeds

  9. Data

    Overnight harvest: +186,388 papers

  10. Enhancement

    Answers can look past what 44B holds when the corpus is thin — outside pages cite the same way as anything else

  11. Enhancement

    Search across the corpus now matches meaning as well as words, in Ask, in Chat, and over MCP

  12. Data

    One corpus — the 34,443-paper research index merges into the Library, now 73,958 documents

  13. Enhancement

    Related papers on every Library document, drawn from the whole corpus instead of a third of it

  14. API

    Search drops the `research` corpus — it is accepted as an alias for `library`, which now holds those papers

  15. New section

    Evaluations — 26 leaderboards, 20 running on live 44B data, each with a plain explanation

  16. Enhancement

    GPT-5.2 and GPT-5 mini join the 44B agent — choose your model in Chat

  17. Data

    The corpus today: 41,530 documents, 34,443 accountability papers, 2,883 incidents from 46 feeds — all queryable via the API and MCP

  18. New section

    A glossary of 57 working terms, A to Z — the vocabulary behind every number on the site

  19. New section

    Sources — the five agencies and archives behind 44B, with live ingest counts

  20. New section

    Help and a real feedback intake — every note gets a reference number

  21. Data

    +34,389 papers landed in the Library from the daily feeds

  22. New section

    Providers — 72 hosts on one roster, with pricing across 85 models

  23. Enhancement

    Ask 44B streams its answer live, naming each retrieval stage as it works

  24. API

    The 44B MCP server — plug the whole registry into your own model at /api/mcp

  25. New section

    Chat — ask the record, not a model: nine live registry tools, every claim linked

  26. API

    Cited answers via POST /api/v1/answer — and shareable /ask permalinks

  27. Data

    Lab publications now refresh nightly — frameworks, model cards, and posts as they publish

  28. Data

    The models table becomes a leaderboard — 17 Artificial Analysis evaluations as sortable columns

  29. New section

    Docs — the public API reference, with a served OpenAPI file

  30. Data

    +639 papers landed in the Library from the daily feeds

  31. New section

    Pricing — Pro and Power plans for programmatic access

  32. New section

    Account — your plan, your API keys, and your usage in one place

  33. API

    One blended search over five corpora at GET /api/v1/search — keyed and metered

  34. Data

    +266 papers landed in the Library from the daily feeds

  35. Data

    +237 papers landed in the Library from the daily feeds

  36. Data

    +268 papers landed in the Library from the daily feeds

  37. Data

    +478 papers landed in the Library from the daily feeds

  38. Data

    +191 papers landed in the Library from the daily feeds

  39. Data

    +298 papers landed in the Library from the daily feeds

June 2026

  1. Registry

    The Slack bridge — every filing gets its own channel, mirrored into the gated portal

  2. Registry

    The Art. 44-B filing workflow goes live — five filing types, a review desk, a public docket

  3. Registry

    Public incident reporting under § 1422(1) — anyone can put a report on the record

  4. Enhancement

    Search rebuilt — one relevance-ranked list across five corpora, with a filter panel

  5. Registry

    Registry accounts with roles — developers file, the Office reviews

  6. New section

    Research — 32,193 AI-accountability papers at launch, searchable by topic

  7. Data

    Semantically related papers on every research page, computed from SPECTER2 embeddings

  8. New section

    Dashboard — the whole registry through eight switchable lenses

  9. Design

    The home page becomes the AI Almanac broadsheet masthead

  10. Enhancement

    The whole site answers in under a second — every index serves one page at a time

  11. Data

    Incidents rebuilt from the public record — severity, status, and source on every row

  12. Data

    The Library becomes a research corpus — 28,600 documents at launch

  13. Data

    New papers land daily — the arXiv feed refreshes the Library every morning

  14. Data

    Semantic Scholar citations and abstracts across the benchmarks and the Library

  15. New section

    Benchmarks — a 531-evaluation catalog, each with its paper and leaderboard

  16. Data

    LLM Stats ingested — licenses, openness, and governance on every lab and model

  17. Data

    Artificial Analysis ingested — every lab and every model, with the full metric set

  18. Design

    The Library reader — hosted PDFs embedded, everything else rendered from parsed structure

  19. New section

    The Commons — a discussion drawer that follows you across the site

  20. New section

    Labs — 37 organizations on the record at launch, with public profiles

  21. New section

    44B ships — the first public build

Dated the day each change landed. Sourced from this repository’s own history.