Nicolas Cantu
f3b7cf27ab
**Motivations:** - Align v0/livre.md with corrective chapters requirements (auditability, stable lexicon, explicit layers, memory disambiguation). **Root causes:** - "mémoire / héritage / past-dependence / non-Markov" vocabulary was used without systematic state/projection/Markov/type declarations. - Audit artifacts (glossary, hypothesis IDs, historical correspondence) were not centralized in v0/livre.md. **Correctifs:** - Add M0 rule + standard declarations, and local M0 declarations where memory/heritage/non-Markov are structural. - Add M1 rule: formulate past-dependent arguments on a minimal closed extended state, then discuss projections. - Add historical NCI → canonical-lexicon correspondence table (explicitly labeled as historical). - Extend the minimal normative glossary (incl. "mémoire transmissible" and "variable cachée") and the hypothesis library identifiers. **Evolutions:** - Document M0/M1 and audit artifacts changes in fixKnowledge/. - v1: replace split sources with consolidated livre.md. **Pages affectées:** - v0/livre.md - fixKnowledge/2026-02-17-v0-livre-m0-memory-declaration.md - fixKnowledge/2026-02-17-v0-livre-normative-glossary-hypotheses-correspondence.md - v1/livre.md (added), v1/* (sources deleted) Co-authored-by: Cursor <cursoragent@cursor.com>
2.7 KiB
2.7 KiB
| author | date | scope | type | related_corrective_chapters | ||
|---|---|---|---|---|---|---|
| 4NK Team | 2026-02-17 | v0/livre.md (global) | fix |
|
Fix: Add normative glossary, hypothesis library, and historical NCI correspondence in v0/livre.md
Problem
Corrective chapters require the manuscript to be auditable:
- a stable canonical lexicon (one term per concept),
- explicit layer markers (E/M/P/D) at the level of definitions/results,
- a library of hypothesis identifiers to avoid implicit “status” drift in interpretative sections,
- and a way to keep NCI vocabulary out of the core while still providing a historical reading aid.
Impacts
- Without a normative glossary + hypothesis IDs, audits (interdits/synonyms/layer drift) are hard to perform mechanically.
- Without an explicit historical correspondence table, readers may re-import NCI meanings by inference, or “forbidden terms” may reappear without a controlled location and label.
Cause
The book content accumulated chapter-by-chapter, but the editorial/audit artifacts were not centralized as normative tables.
Root cause
No single, book-level “contract” existed to:
- define canonical terms + rejected synonyms,
- define hypothesis IDs for conditional statements,
- isolate NCI terms into an explicitly labeled historical appendix/table.
Fix applied
Applied the correction only in v0/livre.md:
- Added a normative glossary structure and a minimal table (canonical term, definition, layer, dependencies, internal references, rejected synonyms).
- Added a hypothesis table (atomic hypotheses and reusable packages), enabling consistent references in conditional/interpretative passages.
- Added an explicitly labeled historical correspondence table “NCI → canonical lexicon”, placed outside the core narrative and introduced as a reading aid (not a justification).
Affected pages
v0/livre.md
Deployment / regeneration considerations
- If
v0/livre.mdis regenerated fromv0/chapitre*.md, these artifacts can be overwritten. The generating pipeline (or the source chapters) should carry the same glossary/hypothesis/correspondence blocks to keep audits stable across regenerations.
Analysis / verification steps
- Search
v0/livre.mdfor NCI terms (e.g.NCI,Néon,vortex,bit utile) and verify they appear only in the explicitly labeled historical correspondence table (and not in the core). - Search for rejected synonyms (e.g.
cône de futur,espace des futurs,futur possible) and verify they appear only in “rejected synonyms” lists / historical tables. - Verify that conditional “implication” passages reference hypothesis identifiers (H‑F, H‑Det, H‑Cpt, …) rather than implicit status claims.