Appendix A. AES v1.0 — Aikipedia Annual Collection Editorial and Format Standard

AES v1.0 as frozen November 14, 2026, in its header revision of the same day, which names the editor W.H.L. and changes no other wording (SHA-256 aeda2699…2092). The frozen file is preserved unchanged (SHA-256 bae6930d…8a81).

Aikipedia Annual Collection — Editorial and Format Standard (AES v1.0)

Standard AES — Aikipedia Annual Collection Editorial and Format Standard
Version v1.0 (frozen November 14, 2026; header revision November 14, 2026)
Status Frozen November 14, 2026, by the editor's decision (W.H.L., 00:51 CT), from candidate v0.2 (November 14, 2026). AES and AES-ID v1.0 in force from the freeze; v0.9 superseded and preserved
Proposed by GPT-6 (as "AES v1.0," November 4, 2026)
Amended by Muse Spark 1.3 (review of November 4, 2026; amendments incorporated below and marked)
Adopted by W.H.L. (editor), November 4, 2026
Amendments adopted by W.H.L. (editor), November 13, 2026 (eleven points; see §11 and the document history)
Consolidated by Muse Spark 1.3, November 13, 2026, from the editorial decision package of November 11, 2026 (draft of record) and Claude Opus 5.5's mapping of November 12, 2026
Pilot specimen AIK-2025-F-N-RLVR — RLVR (accepted November 4, 2026; the reference specimen for this standard)
Identifiers AES-ID v1.0, documented authoritatively in the Aikipedia Master Register (~/workspace/aikipedia-master-register.md, §1); summarized in §10 below

Freeze notice. This document is AES v1.0, frozen November 14, 2026, by the editor's decision (W.H.L., 00:51 CT), from candidate v0.2. GPT-6's changes-only check of v0.2 passed on all six checks, and the freeze was recommended. AES and AES-ID v1.0 are in force from the freeze; v0.9 is superseded and preserved unchanged, as are candidates v0.1 and v0.2. The §10 provisional-identity presentation sentence was confirmed by GPT-6 in its pre-freeze review of v0.1 (item 10). Amendments are adopted verbatim from the decision package except for the two cross-reference adaptations noted in the document history; the two review rulings (§2, §3) are entered as ruled on November 14, 2026.

Governing principle. Aikipedia Annual Collections shall preserve the complete publication record of each year, present verified and transparently corrected reference texts, and maintain consistent editorial structure, source traceability, and searchable digital access across all volumes.

1. Structure of each annual volume

Order Component Requirement
1 Title and copyright pages Required
2 Editor's Preface Required
3 Annual Introduction Required
4 Editorial Method and AI Disclosure Required
5 Alphabetical Table of Contents Required
6 Thematic Table of Contents Required
7 Thematic Parts I–VI Standard framework; adaptable
8 Historical Lexicon Archive All original annual installments
9 Corrections and Editorial Notes Required
10 References and Provenance Register Required
11 Alphabetical Index of Terms Required

The six thematic parts serve as a default classification system, not an immutable taxonomy; new subject areas may justify additional categories in future volumes.

Publication policy. Every original Aikipedia text published within the collection year must be represented. Newly commissioned entries are added separately, with clear identification (see AES-ID origin codes).

Archive and canon (Amendment A). The collection carries two textual objects, and the standard distinguishes them explicitly: the Historical Lexicon Archive preserves original texts as published, traceably; the canonical collection text is the corrected, reviewed text, and it is the citable text of the collection. Where an original fails the evidence rules of §6 (for example, entries whose references do not support their claims), its disposition — retain with annotation, correct under a correction ID, or exclude from the canonical text with a stated note — is decided per record during the source audit and recorded in the Master Register. No original is silently altered, and no corrected text is presented as the original. Exclusion is only ever from the canonical reference text, never from the historical record: an original excluded from the canon remains represented in the Historical Lexicon Archive — or, where even archival reproduction is inappropriate, in a documented exclusion register — with its disposition and the reason recorded. Historical completeness and canonical reliability are protected by different instruments and are never traded against each other.

2. Full-length entry structure

Eight core sections, plus standardized metadata and reference elements. Optional sections may be added where the subject requires them.

Order Section Requirement
— Title and Entry Metadata Required
1 Definition Required
2 Origin and Historical Development Required
3 Conceptual or Technical Framework Required
4 Applications and Implementations Required when applicable
5 Evidence and Evaluation Required
6 Significance and Implications Required
7 Limitations, Debates, and Open Questions Required
8 Related Concepts and Distinctions Required
— See Also Required where relevant
— References and Sources Required
— Editorial Notes Required for substantive corrections
— Entry Provenance Required

Variants: for historical or conceptual entries without practical implementations, §4 is titled Examples and Illustrations. For software systems, §3 may cover architecture, components, interfaces, or operating principles. For technical methods, §5 reports actual empirical evidence, including evaluation conditions and limitations; adoption or diffusion material included in §5 is labeled as report-level evidence and distinguished from measured results. Evidence (§5) and Significance (§6) are kept separate as a mandatory rule: empirical observations and their interpretation are not mixed within a section. Significance must be stated explicitly in §6, however briefly; it may not be left merely distributed across other sections. For a corrected original, §6 may be assembled from the original's own wording drawn from any published text state of the same record, provided each passage is traceable in the correction register's concordance and any passage restored from a non-base version is registered as a restoration (see §7). Assembled wording must be the original's, apart from section labels and connective headings; a new significance claim requires an editorial addition (§2, addition class).

For a corrected original, an empirical claim retained in its documented original location need not be relocated solely to populate §5; its original placement is identified in the concordance and, where necessary, an Editorial Note.

Where an original publication contains no evidence for §5 (or no material for another required section), the corrected text either (a) carries an editorial addition, numbered E01, E02, … within the record, labeled inline and in the correction register, drawn only from sources available by the original publication date, and approved by the editor separately from the corrections; or (b) marks the section 'Not in original.' The choice is the editor's, per record. An editorial addition takes no correction ID and is recorded in the Master Register as an editorial addition, not a correction. Editorial additions are used sparingly; absence, plainly marked, is preferred where no addition is needed to prevent misreading.

The complete eight-section structure is required for a corrected standalone original (type F, origin P): every section appears, and a section the original does not support is marked 'Not in original' (an editorial addition may fill it instead, per the addition class). A contained occurrence record (type S or E, origin P) is presented in the occurrence's own published form — a concise corrected record carrying the AES metadata, the marked corrections, and the provenance annotations — and the eight-section structure is not required. An expanded (E) occurrence may carry section headings where the original has them, but is not restructured into the eight-section order.

3. Entry metadata

Every entry opens with a compact, consistently formatted metadata block (title; thematic classification; and a table carrying Date of Appearance, Date of Original Publication, Date of Collection Publication, Collection, Entry Status, Original Contributor, Review Status, Format Standard, and Entry ID — as in the pilot specimen).

The three dates are never conflated: Date of Appearance (defined below), Date of Original Publication (first publication of this text), and Date of Collection Publication. Date of Appearance: the earliest verified public appearance documented by the available evidence (never inferred from publicity), distinguishing that milestone from any claim of absolute historical priority. Where no sufficiently supported date is established, the field states 'Not established.' Contributors and reviewers are identified by model name and release number. The Entry ID follows AES-ID (§10) and is shown as "To be assigned" until the record is accepted into the Master Register.

Every corrected original carries a Base version field in its metadata, stating the published text state the canonical text is corrected from and the basis for the choice (the post's own declaration, the editor's choice, or sole published version). For an installment, the base version is stated per occurrence where occurrences differ. Each published text state records its contributor as credited (see the release-numbers paragraph in this section), so a record whose states have different contributors names each.

Where an original publication does not state a contributor's or reviewer's release number, the record reads 'as credited; release not stated,' and no release is ever inferred or assigned. Where different credits within one post conflict or differ, all are recorded as credited and are not reconciled by inference. Where the editor states a release on the record, both facts are recorded: the credit as given in the report or post, and the editor's statement, each attributed to its source. This paragraph governs historical credits only; new contributions, audits, and reviews must state model name and release number at opening, and a report without them is held, not published on disclosure or routing alone.

4. Length and level of detail

Entry category Suggested length Treatment
Short lexicon entry 150–500 words Concise definition and context
Expanded entry 500–1,500 words Definition, origin, technical explanation, references
Full-length entry 2,000–3,500 words Complete standard structure
Extended scholarly entry 3,500–5,000+ words Complex subjects requiring substantial evidence and debate

The 1,500–2,000-word range is handled flexibly. Length is never a reason to add filler or omit important evidence. All originally published shorter texts remain included under the historical-completeness policy.

For originally published records (origin P), record type follows the form of publication: a standalone publication is type F (ruling X12); a contained entry is type S or E under ruling X11. The length bands in this section apply as guidance to newly commissioned entries (origin N) only. A corrected original is never padded to reach a band, and length alone never changes a record's type.

5. Formatting and typography

A restrained academic style, suitable for both 6 × 9-inch print and GitHub Pages. Exact typography is finalized after testing the 6 × 9-inch page layout.

Element Standard
Body typeface Source Serif 4 or equivalent readable serif
Headings Source Sans 3 or equivalent sans-serif
Body text 10.5–11 pt for print
Line spacing Approximately 1.2–1.3
Entry title 17–20 pt, bold
Major section heading 13–14 pt, bold
Subsection heading 11–12 pt, semibold
Paragraphs Consistent spacing; no manual blank-line padding
Tables Numbered when substantive; clear headers; source notes
Figures Numbered captions, original or appropriately licensed
References Consistent author/title/year/URL or DOI format
Hyperlinks Clickable online; complete source addresses in reference records

Production rules established in the pilot: (i) tables must remain readable — a substantive table begins on a fresh page where needed, header rows repeat if it must split, and no row is ever broken across pages; if a table cannot fit at body size, reduce its type size or set it landscape (RLVR pilot, item R5); (ii) mathematical notation, subscripts, Greek characters, and symbols are checked in both the print PDF and the GitHub HTML; (iii) figures and tables carry descriptive captions understandable without the surrounding paragraphs; (iv) print builds may use a compact form of the §3 metadata block, but the canonical master record always carries the complete metadata.

6. References and evidence rules

References are kept in two categories. Primary Sources [P]: original research papers, official specifications, original project documentation, first-party announcements. Additional References [R]: independent analyses, reviews, contextual scholarship, and other supporting material.

Every important numerical, historical, or technical claim is traceable to a specific source, not merely to a bibliography. Further:

For an originally published text, the sourcing rule covers: (a) numerical and quantitative claims; (b) historical, dating, and priority claims; (c) attribution claims (authorship, provenance, and who said or did what); and (d) empirical and evaluative claims. A claim in these classes that cannot be verified at a source available by the original publication date is qualified, corrected, or removed. Definitional wording, descriptive practice statements, and interpretation clearly framed as such are retained unless shown to be wrong; where a retained statement rests on the original alone and its status matters, an Editorial Note records that. For newly commissioned entries, every important claim remains traceable to a specific source as stated above.

Material added to a corrected original — a correction, an editorial addition, or a restored citation — must rest on a source available by the original publication date. Later sources may be used in the audit, the correction register, and the entry's provenance to establish what was knowable at publication or to demonstrate an error, and in labeled retrospective Editorial Notes; they must not silently supply content to the historical body of the entry. Where a correction's evidence postdates the original publication, the correction register states that fact and the editor rules on the correction individually.

Eligibility (Amendment B). A term is eligible for an annual collection only if it first appeared in the collection year or earlier. First appearance is the gate; prominence during the collection year is a ranking preference among eligible terms, not the eligibility test. (Rule set by the editor, November 4, 2026.)

Cross-reference integrity (Amendment D). Every See Also and Related Concepts target must exist in the collection, or be explicitly marked as an external or standalone entry outside it. Broken internal references are defects of the same class as broken source links.

7. Provenance and correction record

The AP-2025-01 provenance policy is a permanent component of this standard. Every entry carries a concise Entry Editorial Record: original contributor (model + release), reviewers (model + release), first draft date, historical coverage, publication dates, version history, and relevant correction notes, linking to the draft and review history — and, for corrected original publications, to the original published text and itemized correction notes. Substantive corrections are traceable to a permanent correction ID.

A correction to a canonical text carries a permanent correction ID of the form <Public Entry ID>/Cnn, where nn is sequential within the record, assigned when the corrections are approved. Correction IDs are recorded in the Master Register under the record's details (the register carries no separate correction-ID field). A correction number once assigned and then withdrawn is retired: it is recorded as withdrawn, is never reused, and no other correction takes its number. Audit finding IDs are a separate series and are not correction IDs; the correction register maps each correction to the finding(s) it resolves. Editorial additions (§2) take no correction ID and are recorded as editorial additions, not corrections.

A restructuring of a corrected original into the AES section order carries a structural correction ID of the form <Public Entry ID>/STnn, separate from the Cnn sequence, with a concordance filed in the correction register showing every passage's disposition. The restructure is normally ST01; a further structural change to the same record takes ST02, ST03, and so on. A record's ST numbers, like its C numbers, are permanent; a withdrawn ST number is retired and never reused (see the correction-ID paragraph above). Grandfathered exception: Chain-of-Thought Prompting (AIK-2025-F-P-COTPROMPT) keeps /C14 (class ST, as ruled under O7); it is not renumbered to ST01.

Material restored from another published text state of the same record takes a correction ID (class: restoration), states 'restored from' its source state in the correction register, and is recorded in the Master Register with the restoration noted. Fragments, off-post texts, elided drafts, and embedded reviews are never restoration sources (rulings X9/X15).

Fragments, off-post texts, and embedded peer reviews are provenance annotations, not entries or versions; a review annotation records whether the review was applied (Review Applied?).

Records are literal (Amendment C). A review, verdict, or status is recorded when it occurs, on the version it concerns, by the person or model who performed it. Templates and specimen records carry neutral statuses ("In review"); no template pre-records an outcome. The approval authority for every entry is the editor of record. Peer-review outcome, editorial approval, and publication status are three separate facts and are recorded separately, in the entry's record and in the Master Register; a reviewer's acceptance is a recommendation to the editor, never editorial approval itself.

Both editions are generated from the same approved master text (approved Markdown entry + structured metadata): a print edition (PDF; book pagination; alphabetical and thematic contents) and a GitHub edition (HTML; year / A–Z / theme browsing; full-text search), sharing indexes and the provenance register, with permanent IDs, cross-references, versions, and corrections held in common. Original historical texts remain separately preserved; the shared master is the approved corrected collection edition.

9. Mandatory and flexible elements

Mandatory Flexible
Definition and origin Number of technical subsections
Historical dating and eligibility (§6) Additional examples
Source-supported factual claims Diagrams and illustrations
Limitations and open questions Specialized comparison tables
Related concepts and references Entry length beyond minimum guidance
Provenance and corrections Topic-specific headings
Peer review and editorial approval Secondary thematic classifications

The standard enforces editorial rigor, not identical article shapes.

10. Identifiers (AES-ID v1.0)

Entry, installment, and container records carry permanent public identifiers under AES-ID v1.0, grammar AIK-YYYY-T-O-KEY[-NN], assigned on acceptance into the Aikipedia Master Register, which is the authoritative documentation of the scheme (grammar, permanence rules, the two identity levels — internal concept ID vs. public entry ID — and the assignment log). The register, not this document, is where identifiers are assigned and reserved. A public ID identifies a record within a single collection; a concept treated again in a later collection receives that collection's own ID, linked through the internal concept ID, and earlier collections' records stand unchanged.

The November 9, 2026 rulings, adopted under AES-ID v0.9 and carried into this text, govern the scheme: contained entries are type S (short) or E (substantively expanded) under ruling X11; the -NN suffix is allowed within one parent under ruling X1; registration order is original publication date, then published sequence, the earliest occurrence keeping the bare key, under ruling X13; a standalone publication has one F-P record serving as both parent and occurrence under ruling X12; fragments, off-post texts, and embedded peer reviews are provenance annotations, not entries or versions, under rulings X9/X15. Keys CONSTAI and CIRCUITSP are confirmed.

A provisional concept identity — assigned where a variant's distinctness or a term's attestation is unverified — carries a normal Internal Concept ID; provisionality is a classification state recorded in the record details, not a separate ID series, and never appears in the Public Entry ID. A provisional record whose headword is unattested is presented in the collection visibly as historical text under its provenance note, not as an endorsed definition. If the concept is later merged into another concept by the editor's ruling, its Internal Concept ID is marked Retired — merged into the surviving ID, is never reused, its entry records are re-pointed to the surviving concept ID with the re-pointing verified, and the merger is recorded in the assignment log. If it is withdrawn without merger, its ID is Retired under the never-reuse rule. If its distinctness or attestation is confirmed, the provisional marking is removed and the ID continues unchanged.

11. Adoption and pilot

AES v0.9 completed its pilot: newly commissioned entries, corrected standalone originals, and registered installments were produced under it, and AES-ID was tested against the complete 2025 inventory. This v1.0 standard consolidates the eleven amendments adopted by the editor on November 13, 2026, the November 9, 2026 AES-ID rulings, and Register §1, rule 10. Revision v0.2, incorporating the editor's rulings on GPT-6's pre-freeze review of v0.1, passed GPT-6's changes-only check, and the editor froze v1.0 on November 14, 2026. Production continues under v1.0; packages registered under v0.9 stand, and no identifier assigned under v0.9 changed at the freeze.


Document history: Consolidated November 4, 2026, by Muse Spark 1.3 from GPT-6's AES proposal of the same date, Muse's adopted amendments (A: archive/canon disposition, §1; B: eligibility clause, §6; C: literal provenance records, §7; D: cross-reference integrity, §6; E: corrected original in the pilot, §11), GPT-6's review refinements (mandatory evidence/significance separation, §2; print metadata compaction as a production option, §5), and the RLVR pilot's production lessons (§5).

Clarifications, November 4, 2026 (GPT-6's review of the two documents; accepted and entered the same day): the completeness rule — exclusion is from the canonical text only, never from the historical record (§1); the three-status distinction (§7); the collection scope of public IDs (§10; Register §1, rule 9). One factual note on that review: its suggestion that RLVR's editorial approval might still be pending does not apply — the editor's approval was given separately on November 4, 2026, and is recorded as such in the Register; the three-status presentation is adopted because the distinction is sound, not because the approval was missing.

Consolidation to v1.0 (candidate), November 13, 2026 (Muse Spark 1.3): this candidate applies the eleven amendments adopted by the editor on November 13, 2026 (rulings 8/9/13, recorded by Claude Opus 5.5), from the editorial decision package of November 11, 2026 (draft of record) as mapped by Claude Opus 5.5 on November 12, 2026 (point 7 amended by the per-state contributor sentence); carries the November 9, 2026 AES-ID rulings (X11, X1, X13, X12, X9/X15; keys CONSTAI, CIRCUITSP) and Register §1, rule 10, into the standard text; and submits the result for GPT-6's pre-freeze review. Point 3's sentence is placed in the §2 Variants paragraph (the §6 discussion), the §6 table row carrying no prose. Two cross-references in the adopted wordings were adapted to resolve within this document (point 7's sentence: 'AES §3, point 8' → the release-numbers paragraph; point 10: 'see point 4' → the correction-ID paragraph); no adopted wording was otherwise changed. The point-9 presentation sentence is flagged pending confirmation (item 10). Companion update at freeze: the correction-ID recording sentence enters Master Register §2. No GPT-6 draft of the v1.0 points has been filed with the consolidator.

Revision to v0.2 (candidate), November 14, 2026 (Muse Spark 1.3): this revision incorporates the editor's rulings of November 14, 2026 (recorded by Claude Opus 5.5) on GPT-6's pre-freeze review of v0.1 (reviewer identified as 'GPT-6 (OpenAI)'; no release number stated; verdict Accept with minor revisions): ruling 1 enters GPT-6's §2 sentence as proposed; ruling 2 enters the §3 Date of Appearance definition with Claude's amendment keeping v0.9's 'never inferred from publicity'; the §10 provisional-identity presentation sentence's confirmation flag is removed, GPT-6 having confirmed the sentence without change and applied it to the 'Constitutional AI 2.0' case (item 10). GPT-6 also confirmed the §3 per-state contributor sentence (point 7) and the §7 three-status distinction, and disclosed its role as the original proposer of AES v0.9. Two scope limits are recorded: GPT-6 reviewed a PDF export (SHA-256 532bee0743c28f7ae9063ef7fa6a9a4dd729178a8c3a39c2ac58160660f34241), a different representation from the versioned document (SHA-256 fc585318f418e92837de1ccfeb8c434d2bcec130be08089271143a9b0eb9c438), not a version mismatch; and it did not have the decision package for a literal amendment-by-amendment diff — Claude's check compared the candidate against the package. At the freeze, the AES-ID heading and version references become v1.0, and the register checks recorded in the rulings file are performed.

Freeze to v1.0, November 14, 2026 (Muse Spark 1.3): the editor froze AES and AES-ID together at v1.0 (Woo, 00:51 CT; decision recorded by Claude Opus 5.5), from candidate v0.2, following GPT-6's changes-only check of v0.2 (pass on all six checks; checksum independently matched; freeze recommended). The frozen text is v0.2's text with the version references updated to v1.0 (title, Version, Status, and Identifiers lines; §10 heading and scheme reference; §11 adoption paragraph) and this decision recorded; no other wording changes. AES v0.9 is superseded and preserved unchanged, as are candidates v0.1 and v0.2. Freeze-time register actions (Master Register §1 heading and version references with a check against §10; §2 Date of Appearance definition and correction-ID recording sentence; rule 10 merger confirmation; Base version fields; assignment-log entry) are recorded in the register.

Header revision, November 14, 2026 (Muse Spark 1.3, consolidator): under the editor's rulings of November 14, 2026 (reviews of collection draft v0.1, item 1 — the editor's name reads W.H.L. in every record that names the editor), the header of this document names the editor W.H.L. (Status, Adopted by, Amendments adopted by, Freeze notice), and the Version line identifies this revision. No other wording changes: the frozen v1.0 body is identical to the frozen file (SHA-256 bae6930dae70d6b191f6f5a0a1cfc13176e804b5017432493d4230c81ff68a81), which is preserved unchanged. The freeze record above keeps its as-recorded wording; 'Woo' there refers to W.H.L.