How to Use This Collection
Historical and canonical texts. The collection distinguishes the original publications, preserved in the Historical Lexicon Archive, from their corrected canonical reference texts in Parts I–VI: the versions corrected, reviewed and approved for reference use. Most original entries are represented in both forms. Texts excluded from the canonical reference collection remain preserved and identified in the historical archive. The archive gives each original as a line-numbered plain-text transcription of the published post (A-001, A-002, …), not as a copy of the original web page; each section's header states how markup was normalized and how the text has been checked. The online edition gives every post in full; the PDF edition gives four of them (Constitutional AI, Mixture of Experts, Lexicon (3) and Circuit Sparsity) as selected transcriptions. Corrections never alter the archive.
Finding a term. Use the Alphabetical Table of Contents (concepts A–Z, with their part), the Thematic Table of Contents (by part), or the Index of Terms, which also lists variant names, abbreviations and related terms. Online, every record has its own anchor and the edition is searchable.
Concepts and records. A concept may occur in several 2025 texts — for example a full-length entry and a lexicon installment. The collection gives each concept one heading, with an Internal Concept ID (AIKC-######), and under it each record (one occurrence in one text) with its own Public Entry ID (AIK-2025-T-O-KEY[-NN]; T is the type — F full-length, E expanded, S short, L installment — and O the origin, P published in 2025 or N new).
Marks.
- † — a provisional concept identity: the concept's distinctness or attestation is not verified. A provisional record whose headword is unattested is presented as historical text under its provenance note, not as an endorsed definition (AES v1.0 §10).
- ◆ New for the 2025 Collection — a full-length entry commissioned for this collection (origin
N); it was not published in 2025. - [Cnn], [E01], [ST01] — in a corrected text: a correction, an editorial addition adopted by the editor, and a structural rearrangement into the AES section order. Unmarked text is the original's wording. In lexicon records, a corrected record line gives the canonical wording wherever a correction applies.
- Excluded (disposition X) — a 2025 text kept in the archive but not carried into the canonical text: the whole of New AI Lexicon 2025 (6), and the Launch Edition's "Instruction Tuning 2.0".
Dates. Each record gives a Date of Appearance — the earliest verified public appearance documented by the available evidence, never inferred from publicity, and not a claim of absolute priority; "Not established" where no date is supported — separately from the date the 2025 text was published (AES v1.0 §3). Times of day in this volume are Central Time (America/Chicago) unless stated otherwise.
Checking the record. Corrections and editorial notes are listed by record; the References list the works cited; Appendices A–G give the standard, the identifier rules, the provenance policy, the digest of editorial rulings, the conflicts and recusals, related editorial material, and the provenance register with each record's dated history.