Skip to content

HANDOFF.md

Latest commit

 

History

History
429 lines (356 loc) · 52.8 KB

HANDOFF.md

File metadata and controls

429 lines (356 loc) · 52.8 KB
Raw
audience both
doc_kind state
cursor fa2d584 (LOG.md 0020)
ancestor_cursors
last_updated 2026-06-04
sealed_by
status working
supersedes
title Session Handoff
type readme
version not versioned

Document kind (debate 009 A3): this is the framework unit's State doc (snapshot) under the Paperwork Standard — regenerated, paired with the append-only LOG.md. cursor: is the high-water mark into LOG.md (see that file's cursor convention note). A re-prime walking LOG since the cursor will pick up events 0007–0009 (this session's paperwork-blog-series work) and re-fold them into the next regeneration.

Known dogfood gap: this snapshot is still in its end-of-session-4 layout (Status table + sections), not the two-part ## Mechanical / ## Narrative structure described in blog #8 and distribution/handoff-mechanics.md. The full regeneration into that layout is a follow-up; for now the surgical update preserves the existing Status table while updating the cursor + new entries.

Session Handoff — 2026-05-23 (after paperwork-blog-series build)

Snapshot of Dead Light Framework state. Read this first to resume work without re-deriving context. Supersedes the 2026-05-19 handoff (end-of-debate-012-seal).

What this project is

Dead Light Framework — a software development governance methodology for human + AI agent collaboration. It composes on top of existing methodologies (Agile, Scrum, RUP, …) rather than replacing them. The distinctive idea: a frozen source of authority (the Astronomican) that no participant — human or agent — can rewrite at will.

Working tree (project-owner workstation): C:\Works\_Researchs\dead-light-framework (alternate workstation: D:\Works\source\dead-light-framework). Reference / case-study codebase (not yet integrated): C:\Works\_Researchs\lore-weave.

Status

Artifact Status
README — Manifesto + Map Drafted; AI-collaboration literature anchors added in session 3 (F-28 remediation)
Glossary — for debate Working draft (final assembled bottom-up later)
Phase 0 — The Reckoning Sealed (all calibration questions resolved in debate 003)
Phase 1 — The Astronomican Partial — debates 001 and 004 resolved; 6 known questions still open (see § Open questions)
Phase 2 — Codex per Chapter First Chapter (Adeptus Administratum) sealed via debate 005 on 2026-05-11; v1.1 amendment 2026-05-13 (script-invocation references per debate 007 §G; commit 8ebc4da); v1.2 amendment 2026-05-13 (deployment-protocol references per debate 008 — §1 override mechanism, §8 bootstrap + deployment-matrix references). Remaining Chapters pending real-project trigger. Codex amendment-procedure precedent established at §10 — implementation amendments don't require new debate; design amendments do. v2.0 design-amendment 2026-05-19 via debate 009 — §8 re-priming replaced with the sealed Paperwork Standard §14 (legacy fallback retained for un-migrated projects).
Documentation architecture & distribution Sealed via debate 006 on 2026-05-13. Migration executed across commits 1a857c2 (folder restructure) → c4a574e (frontmatter backfill, 25 files) → 9e9afb8 (distribution/ populated with sealed framework + 4 templates + lore-weave snapshot). Distribution v0.6.0 ready at distribution/ (separable from repo; clone target for adopters).
Scripting infrastructure Sealed via debate 007 on 2026-05-13. All 3 tiers implemented (12 scripts at scripts/). Tier 1 (validate_frontmatter, check_links, sync_distribution — commit 1246b25). Tier 2 (frontmatter_set, bump_version, snapshot_case_study, update_handoff_tree — commit 34eec65). Tier 3 (release.py 8-step orchestrator, new_debate, new_chapter_codex, new_case_study, ivp_phase5 — commit 60cd4fa). End-to-end pipeline tested via release.py minor --apply (v0.6.0 → v0.7.0). Codex v1.1 via debate 007 §G; Codex v1.2 via debate 008. Python 3.10+ with python-frontmatter + typer + markdown-it-py.
Deployment protocol Sealed via debate 008 on 2026-05-13. Standardises HOW AI agents engage with the framework at runtime — parallel meta-standard to debate 006 (which standardised paperwork). 10 sub-decisions covering bootstrap pattern (CLAUDE.md auto-loaded for Claude Code A-tier; .cursorrules for B-tier; system-prompt paste for raw API), provider tier model (A Claude Code / B GPT-4+/Gemini 2+/Cursor / C local LLMs), session protocol (hybrid full + lightweight re-priming), authorization mechanism (D4 inline ack + audit trail), instance identity recording (audit trail + commit trailer), inter-instance handoff (HANDOFF dual purpose), IDE/tooling matrix (in distribution/deployment-matrix.md), multi-project Codex overrides (.aa-codex-overrides/ per Imperial+Sector pattern), adopter onboarding (5-step quickstart), failure modes catalogue (F-J1..F-J7). Phase 1 implementation: CLAUDE.md + distribution/deployment-matrix.md + distribution/templates/aa-codex-overrides-example/ + Codex v1.2 patch. Phase 2/3 tooling scripts deferred.
Phase 3 — Heresy detection Inquisitor-Detect half opened 2026-06-04 via debate 014 (proposal). Splits the bundled Inquisitor → an early advisory detect function (pull forward from Phase 0; the IVP is its embryo, non-blocking, eventually-consistent) + a late enforce half. Grounded in research brief (Censorate / Inspector General precedent). Enforce-half + standalone chapter still pending Phase 1/2.
Phase 4 — Re-consecration Not started
LoreWeave case study application Not started — recommended next step
IVP audit infrastructure Spec v0.3 + slash command + Phases 1, 2 (full coverage), 3, 4, 5 complete; all 38 findings (F-01–F-39, with F-26 retracted) remediated; Phases 6, 7 queued
Framework-wide voice policy + cleanup pass Policy #3 sealed (commit 344bbb2, session 4). Voice cleanup pass applied across README, phase-0/1 working drafts, calibration docs, glossary, case study files, debates 002/003/004/006, audit findings executive verdicts (commits 1d81e0d → 68daccb). Codex v1.0 skipped (sealed — voice cleanup deferred to future Codex Re-consecration debate).
Multi-Actor Coordination Standard — sister to Paperwork Standard DECIDED 2026-06-04 via debate 013 (opened 2026-05-28 as proposal). Design locked: project owner accepted all 10 sub-decisions (C/E revised to event-relative C4′/E2′ — wall-clock→event currency + dormant backstop + evict-once-audited); 3 review concerns resolved (C1→authority-routing defers to debate 015; C2→MA-08 wired to debate 014-D; C3→non-LOC tier trigger tracked as a Paperwork Standard §2 re-consecration). "Decided" = design locked, NOT validated — MA-01..MA-10 benchmark with a control remains the seal-as-proven gate. Spawned debate 014 (Inquisitor detect/enforce split) + debate 015 (authority hierarchy; Codex HS-4/5 carve-out). (Original proposal context retained below.) Triggered by LoreWeave Pass-1-deeper-audit F-005 + §1.6.5 boundary flag (project-owner reflection: parallel branches × parallel agents × shared governance files → uncoordinated writes). Scope (project-owner-decided 2026-05-28): C1 convention + C2 file-based-lease BUILD; C3 runtime tier SPEC only (free-context-hub Phase 15 implements); compose human patterns; branches in scope; benchmark scenarios pre-registered FIRST. 10 sub-decisions A..J with recommended answers awaiting project-owner walk-through. Cross-references: tier-agnostic Principles 1-6 defined; benchmark scenarios MA-01..MA-10 cross-mapped; industry-standards anchors (TBD, GitHub flow, Chromium OWNERS, Conventional Commits, §8.3, AMAW prefix, Blog #4, CRDTs/OT/distributed locks). Next step: project-owner walks A..J in chat; debate seals (debate-013 status proposal → decided); benchmark scenarios pre-registered as framework/audit/multi-actor-benchmark-scenarios-*.md; framework/multi-actor-standard-for-debate.md v0.1 drafted; AMAW → IVP → seal v1.0. Estimated effort: ~1-2 weeks of bottom-up build (same path as Paperwork Standard).
Paperwork blog series — Tier 1 (#2–#10) CONTENT-COMPLETE 2026-05-23. 10 posts drafted value-first (Why → How → Results) with companion practitioner cards in distribution/. Covers every HIGH-value adopter-applicable section of the Paperwork Standard: §3 State/Log in #2, §1.2 CAP + §2 M-tiers + §4.2 layout in #3, §5.1 OPORD + §6.1 DTN + §10.4 basic refer-back in #4, §14 re-priming in #6, §6.3 events + §10 lifecycle in #7, §7.1–7.5 snapshot + eviction + staleness in #8, §12 sectoring + debate-001 4 conditions in #9, §11 cross-unit + §10.4 refer-back full in #10, §15a residuals threaded throughout; #5 covers the self-skepticism workflows (AMAW / plateau / IVP). #1 is the series introduction; format-aligned with #2–#10. Multi-agent review pass (mechanical / technical-accuracy / cold-start-practitioner) cleared the 10 posts for publish — 0 HIGH technical-accuracy (no v1.0-era overclaim leaked back); MEDIUM fixes applied; LOW polish skipped per agents' "would not mislead" classification. Distribution: 7 practitioner cards + 2 templates surfaced via INDEX/README/CHANGELOG. #2 published on dev.to 2026-05-22; #3–#10 ready to publish at 2–5 day cadence (~20–50 day buffer). Hashnode API gate-blocker (Pro plan now required for all queries+mutations) — adopter publishes manually. LinkedIn API has no public personal-article publish — same. Tier 2 optionals (federation, edges/internals omnibus, benchmark catalog) explicitly deferred to Chapters series / on-demand.
Paperwork Standard — document organization SEALED v1.0 (2026-05-19, project owner; framework/paperwork-standard.md). Built bottom-up across this session: v0.1 (State/Log + M-tier), v0.2 (3 RC fixes), v0.3 (4 RC fixes via distributed systems + Mission Command + DTN), v0.4 (cleanup of Run-3 determinate gaps: §1 framing fix, repo-root = Imperial-tier-hosting with no separate root unit, cross-repo graduation precision + binding-context manifest, commit-per-state-change cadence, session_id orphan attribution, two-threshold working-set bound, periodic snapshot rebase, periodic deep drift verification), v0.5 (restore tier nesting), v0.6 (authority-layer events + Case-7 cleanup), v0.7 (Domain-4: decision.referred_back + activates_when), v0.8 (coherence polish), v0.9 (§8.4 content integrity via git blob SHA — last "to be wired" closed). Benchmark history (7 independent cold-start sub-agents/run): Run 1 (v0.1) 0/3/4 → Run 2 (v0.2) 1/6/0 → Run 3 (v0.3) 1/6/0 → Run 4 (v0.4) 2/5/0 → Run 5 (v0.5) 2/5/0 (plateau confirmed; 3 consecutive whack-a-mole swaps at the CAP-AP boundary). Run 6 was a different test — a live authority-hierarchy simulation (agents as human commanders vs subordinate Execution-tier agents): authority model held structurally (no tier breached bounds; escalation routed to nearest common superior, validating the v0.5 nesting under load; binding decision propagated) = PARTIAL, with clean additive authority-layer gaps. v0.6 folds those in (decision.overrode_frozen_intent + mandatory rationale §10.3; dispute.resolved §9.3; `kind: request

Framework-wide policies (never violate in future drafts)

  1. 40k vocabulary is naming and shared metaphor only. Justification, structural arguments, and assessments of effectiveness must rest on real-world organizational systems with observable track records (constitutional federalism, religious institutions, military doctrine, established corporate practice, open-source governance, established software methodologies, …). When the 40k name and the real-world precedent diverge in implication, the real-world precedent governs the design.

  2. Industry standards over framework-invented formulas. Calibration questions (thresholds, time budgets, team sizes, audit scopes, failure measurement) anchor to documented external practice — COCOMO II, CMMI v3.0, ITIL 4 / 5, ISA 320, DORA, McCabe, Miller, Brooks, Dunbar, Cohen's κ, Bommasani et al. 2021 Foundation Models, etc. PM and Council retain authority to select, calibrate, or override, but the starting point is always documented external practice, never framework arithmetic. Operationalised by the IVP audit (see § IVP). Argument-warrant tier-floor (added IVP v0.3): load-bearing arguments — not just citations — must rest on external anchors.

  3. Practitioner voice, not authority pronouncement. The framework author is a working developer with approximately ten years of cross-project experience, not an academic, industry methodologist, or recognised authority. All framework writing — outward-facing (README, blog posts) and internal (debates, phase drafts, Codices, case studies, IVP findings) — frames hypotheses as personal exploration; qualifies load-bearing claims with first-person or tentative language; refuses to position the framework as correcting or superior to established methodologies (e.g. "Agile assumptions break" → "Agile was not designed with agent participants in scope"); and presents external standards (CMMI, COCOMO, ITIL, Toyota, Bommasani, …) as anchors the framework defers to, not credentials the framework owns. Outward-facing pages carry an explicit "Who is writing this" framing near the top. Universal-sounding pronouncements are rewritten as framework-internal commitments. Operationalised by an author-voice pass to be added to IVP (likely Phase 5b in spec v0.4).

Decisions locked (all sessions)

Debate Decision (one-line)
001 — Laws Count Cap and Multi-Repo Scaling Hierarchical Imperial + Sector Astronomicans with inherit-and-add rule. Default = single Astronomican; split only when ≥ 2 repos + dedicated teams + cross-team contracts + genuinely local decisions are all true. Anti-survivorship-bias selection-criteria note added in session 3 (F-30 remediation).
002 — Retrofit vs Greenfield Introduce Phase 0: Reckoning as prerequisite to Phase 1 for retrofit (greenfield runs lightweight or skips). Reckoning-first effective-date model with classification: Keep / Fix-now / Fix-by-date / Reconsider-Law. Mechanism-based warrant for retrofit precedents (TRC's "explicit past-violations machinery" is the durability anchor) added in session 3 (F-31 remediation).
003 — Phase 0 Calibration Significance heuristic (6 categorical bullets), Reckoning Team composition rule (≥1 active-IC mandatory + ≥1 tenure-spanning + ≥1 outside-scope), soft time budget with 80%/100% escalation, lightweight greenfield Phase 0 (Assumption Surface + Commitment Audit + Stakeholder Map).
004 — Cap Revision: Miller Citation Correction Cap on Immutable Laws and Guiding Principles raised from ≤ 5 to 5–9 with target ~7 (anchored on Miller 1956's actual 7 ± 2). Same caps apply to Imperial and Sector. Driven by IVP finding F-01 (Miller misquote).
005 — First Chapter: PM / High Lord Aide First concrete Chapter sealed: Adeptus Administratum (Codex v1.0). Multi-role (PM + High Lord aide); Draft + Notify authority (5 trigger types); D4 lifecycle (task-scoped instance + persistent role + artifact continuity via 7-step re-priming protocol); E2 multiplicity (singleton per Astronomican; concurrent instances allowed). Driven by project-owner observation during LoreWeave case-study kick-off that PM threshold-setting on a 358-KLOC project is a bottleneck.
006 — Documentation Architecture and Distribution Template Full doc architecture sealed. A4 folder topology (root framework/ + case-studies/ + distribution/); SemVer versioning (initial v0.6.0); document-type-aware organization (phase = arc42; debate = ADR-extended; codex = 10-section; audit = IVP-spec; case-study = Diátaxis how-to; reference = Diátaxis reference); 500-line hard cap + 300 soft + 100 per-section soft; YAML frontmatter (9 fields) + human summary block; role-based reading guides (for-pms / for-ics / for-ai-aides / for-adopters); strictly one-way framework/ → distribution/ sync with no adopter audit (standard OSS fork-and-modify per Linux/React/arc42/ISO); I0 — no customization protocol (folder convention + status frontmatter classify implicitly); dual-audience design rules (7 rules incl. Mermaid diagram convention + tool stack recommendation).
007 — Scripting Infrastructure Authorise Python tooling under scripts/ to handle mechanical work. All 3 tiers implemented (12 scripts). Library stack: python-frontmatter + typer + markdown-it-py. Dry-run-by-default for writes; strict transactions. Adds LLM token-cost as third design axis (alongside human-time + correctness). ~99% reduction on mechanical work.
008 — Adeptus Administratum Deployment Protocol Standardises HOW AI agents engage with framework at runtime. 10 sub-decisions covering bootstrap (A5 provider-tiered with CLAUDE.md + .cursorrules + start_aa_session.py fallback), provider abstraction (B3 tiered Claude/GPT/Gemini/local), session protocol (C4 hybrid full + lightweight re-priming), authorization (D4 inline ack + audit trail), identity recording (E4 combined), handoff (F4 HANDOFF dual purpose), tooling matrix in distribution/deployment-matrix.md, multi-project Codex overrides via .aa-codex-overrides/ — Imperial+Sector pattern re-applied at Codex level, 5-step adopter quickstart, 7 failure-mode catalogue F-J1..F-J7. Codex v1.2 implementation-amendment applied same-day per v1.1 precedent.
Council composition multi-role PM is a Council member. Council requires ≥ 3 distinct functional perspectives (minimum-diversity rule). Council size 3–7 (anchored on Dunbar / Brooks for group dynamics — not affected by debate 004's Miller-anchored cap revision). Small-team accommodation: AI-assistant Chapters as aides; first Chapter (Adeptus Administratum) sealed via debate 005.
Six embedded answers (debate 002) No Phase 0 sealing; smaller Reckoning Team produces, Council reviews; full attribution; no fixed grandfather cap; no fixed sunset horizon; PM-defined re-reckoning cadence.

IVP — Independent Verification Pass

Audit workflow specified, executed once (rodage 2026-05-08), then run end-to-end across Phases 2 (extension), 3, and 4 in session 3 (2026-05-09 → 2026-05-10) with remediation between phases. All 30 surfaced findings remediated (F-01–F-31; F-26 retracted via erratum after re-reading). The framework now has a substantially complete external-reviewable evidentiary check through Phase 4.

  • Spec (v0.3, 2026-05-09): framework/audit/independent-verification-pass-for-debate.md. 7 phases; pre-registered rubric; industry-pragmatic source-tier hierarchy; argument-warrant tier-floor (new in v0.3); cluster-level Phase 4 option (new in v0.3); multi-phase-run pattern with audit-erratum convention (new in v0.3); changelog at § 11.
  • Slash command: .claude/commands/ivp.md — /ivp [scope] to re-run.
  • Inventory (2026-05-08): framework/audit/inventory.md — 44 load-bearing claims, 76 distinct citations, 40 defined terms, 11 analogy invocations. Note: inventory paraphrases capture Phase-1-state at audit time and may lag remediation; per IVP v0.3 § 5 Phase 3 pre-step, refresh against current source files before re-running Phase 3.
  • Findings:
    • findings-2026-05-08.md — Phase 2 rodage on 30 citations. 17 V / 11 P / 1 C; 3 HIGH (F-01–F-03) + 4 MEDIUM (F-04–F-07) + 6 LOW (F-08–F-13). All remediated.
    • findings-2026-05-09.md — Phase 2 extension over the queued 46. 33 V / 12 P / 1 C; 3 HIGH (F-18 Schein/Argyris, F-19 RUP/XP spike, F-20 Lean Startup scope-as-variable) + 2 MEDIUM (F-14, F-21) + 6 LOW (F-15–F-17, F-22–F-24). All remediated. 76/76 citations now have a Phase 2 verdict.
    • findings-phase3-2026-05-09.md — Phase 3 Citation Appropriateness on 73 verified citations. 70 APPROPRIATE / 3 STRETCHED. 1 MEDIUM (F-25, reduced to LOW via erratum) + 2 LOW (F-26 retracted via erratum; F-27 Goodhart). All practically-effective findings remediated. Headline: policy-1 check on R-001 (40k metaphor) PASSES.
    • findings-phase4-2026-05-09.md — Phase 4 Argument Analysis on 41 LB claims (clustered into 7) + 4 debates. 35 SOUND / 4 WEAK-WARRANT / 0 FALLACIOUS / 2 UNFALSIFIABLE-but-conventional. 2 MEDIUM (F-28 central-premise WEAK-WARRANT, F-31 retrofit survivorship) + 2 LOW (F-29 stress-test threshold, F-30 multi-tier survivorship). All remediated.
    • findings-phase5-2026-05-11.md — Phase 5 Internal Consistency across 40 terms, 4 decided debates, ~20 numeric thresholds, both framework-wide policies, and README link integrity. 0 CRITICAL / 0 HIGH / 4 MEDIUM (F-32 phase-1:299 stale "5 Laws max", F-33 calibration-standards Brooks anchor, F-34 Dunbar layer-1 overshoot, F-35 phase-0:34 Time budget) + 4 LOW (F-36 phase-1:309 stale open question, F-37 HANDOFF Two-tier sharpness reflection, F-38 debate-001 metadata annotation, F-39 three small framework-internal heuristics). All 8 remediated. Headlines: policy-1 (40k naming-only) PASSES across all 11 inventoried analogies; README link integrity PASSES; design decisions of debates 001–004 reflected at all prescriptive sites — drift was concentrated in meta-commentary sections.
  • Phases 6, 7: not yet executed. Phase 6 = External Benchmarking (places framework in adjacent literature on AI-agent governance, software governance, frozen-spec patterns). Phase 7 = Synthesis Report (integrates all per-phase findings files into a single artifact for outside reviewers).

Document tree

Tree post-debate-006-migration (2026-05-13). Restructured per debate 006 §3 sub-decision A4 (hybrid): root folders framework/ + case-studies/ + distribution/. Working drafts in framework/; outward-facing sealed snapshot in distribution/; project applications in case-studies/.

dead-light-framework/
├── .claude/
│   └── commands/
│       └── ivp.md
├── README.md
├── HANDOFF.md
├── LICENSE
├── blogs/
│   └── 01-the-emperor-is-dead-the-light-remains.md
├── case-studies/
│   └── lore-weave/
│       ├── README.md  [working]
│       ├── methodology-notes.md  [working]
│       ├── pm-threshold-decisions.md  [draft]
│       ├── reckoning-record.md  [working]
│       └── reckoning-team-record.md  [working]
├── distribution/
│   ├── examples/
│   │   └── lore-weave-snapshot/
│   │       ├── README.md  [working]
│   │       ├── methodology-notes.md  [working]
│   │       ├── pm-threshold-decisions.md  [draft]
│   │       ├── reckoning-record.md  [working]
│   │       └── reckoning-team-record.md  [working]
│   ├── framework/
│   │   ├── audit/
│   │   │   ├── findings-2026-05-08.md  [working]
│   │   │   ├── findings-2026-05-09.md  [working]
│   │   │   ├── findings-phase3-2026-05-09.md  [working]
│   │   │   ├── findings-phase4-2026-05-09.md  [working]
│   │   │   ├── findings-phase5-2026-05-11.md  [working]
│   │   │   ├── independent-verification-pass-for-debate.md  [working]
│   │   │   └── inventory.md  [working]
│   │   ├── chapters/
│   │   │   └── adeptus-administratum/
│   │   │       └── codex.md  [SEALED]
│   │   ├── debates/
│   │   │   ├── 001-laws-count-and-multirepo-scaling.md  [DECIDED]
│   │   │   ├── 002-retrofit-vs-greenfield.md  [DECIDED]
│   │   │   ├── 003-phase-0-calibration.md  [DECIDED]
│   │   │   ├── 004-cap-revision-miller-correction.md  [DECIDED]
│   │   │   ├── 005-first-chapter-pm-high-lord-aide.md  [DECIDED]
│   │   │   ├── 006-documentation-architecture-and-distribution.md  [DECIDED]
│   │   │   ├── 007-scripting-infrastructure.md  [DECIDED]
│   │   │   ├── 008-adeptus-administratum-deployment-protocol.md  [DECIDED]
│   │   │   └── README.md  [working]
│   │   ├── phases/
│   │   │   └── phase-0.md  [SEALED]
│   │   ├── calibration-standards.md  [SEALED]
│   │   └── pm-calibration-guide.md  [SEALED]
│   ├── templates/
│   │   ├── aa-codex-overrides-example/
│   │   │   ├── README.md  [fillable]
│   │   │   ├── additional-hard-stops.md  [fillable]
│   │   │   ├── additional-operational-bounds.md  [fillable]
│   │   │   └── notify-trigger-extensions.md  [fillable]
│   │   ├── astronomican-template.md  [fillable]
│   │   ├── pm-threshold-decisions-template.md  [fillable]
│   │   ├── reckoning-record-template.md  [fillable]
│   │   └── reckoning-team-record-template.md  [fillable]
│   ├── CHANGELOG.md  [working]
│   ├── INDEX.md  [working]
│   ├── README.md  [working]
│   ├── VERSION
│   ├── deployment-matrix.md  [working]
│   ├── for-adopters.md  [working]
│   ├── for-ai-aides.md  [working]
│   ├── for-ics.md  [working]
│   └── for-pms.md  [working]
├── framework/
│   ├── audit/
│   │   ├── findings-2026-05-08.md  [working]
│   │   ├── findings-2026-05-09.md  [working]
│   │   ├── findings-phase3-2026-05-09.md  [working]
│   │   ├── findings-phase4-2026-05-09.md  [working]
│   │   ├── findings-phase5-2026-05-11.md  [working]
│   │   ├── independent-verification-pass-for-debate.md  [working]
│   │   └── inventory.md  [working]
│   ├── chapters/
│   │   └── adeptus-administratum/
│   │       └── codex.md  [SEALED]
│   ├── debates/
│   │   ├── 001-laws-count-and-multirepo-scaling.md  [DECIDED]
│   │   ├── 002-retrofit-vs-greenfield.md  [DECIDED]
│   │   ├── 003-phase-0-calibration.md  [DECIDED]
│   │   ├── 004-cap-revision-miller-correction.md  [DECIDED]
│   │   ├── 005-first-chapter-pm-high-lord-aide.md  [DECIDED]
│   │   ├── 006-documentation-architecture-and-distribution.md  [DECIDED]
│   │   ├── 007-scripting-infrastructure.md  [DECIDED]
│   │   ├── 008-adeptus-administratum-deployment-protocol.md  [DECIDED]
│   │   └── README.md  [working]
│   ├── phases/
│   │   ├── phase-0-for-debate.md  [SEALED]
│   │   └── phase-1-for-debate.md  [partial]
│   ├── calibration-standards.md  [SEALED]
│   ├── glossary-for-debate.md  [working]
│   └── pm-calibration-guide.md  [SEALED]
└── scripts/
    ├── README.md  [working]
    ├── bump_version.py
    ├── check_links.py
    ├── frontmatter_set.py
    ├── ivp_phase5.py
    ├── new_case_study.py
    ├── new_chapter_codex.py
    ├── new_debate.py
    ├── release.py
    ├── requirements.txt
    ├── snapshot_case_study.py
    ├── sync_distribution.py
    ├── update_handoff_tree.py
    └── validate_frontmatter.py

Note (post-debate-007): The tree block above is between <!-- DOC-TREE-START --> and <!-- DOC-TREE-END --> markers and can be regenerated by python scripts/update_handoff_tree.py --apply. Annotation comments (←) are stripped on regeneration; if you want them back, re-edit the tree manually after regeneration or extend the script's annotation feature.

Open questions still on the table

Phase 1 (from phase-1-for-debate.md Note on Method)

Severity HIGH:

  • Two-tier sharpness — when is something an Immutable Law vs a Guiding Principle? Content question affecting every Astronomican.
  • Failure of sealing — what happens when stress-test divergence > 20% on sealing day, or Council deadlocks? Without an exit ramp, real teams stall.

Severity MEDIUM:

  • Council size lower bound — is 3 truly enough?
  • Storage of seal mechanics — signed git tag vs multi-sig vs notarized hash.

Severity LOW:

  • Pre-work questions — 5 sufficient or a 6th on agent boundaries?
  • Session format — co-located vs distributed timezone.

(The cap-count question and the "First-time vs retrofit" question were both previously listed and have since been closed — the former by debate 004, the latter by debate 002 + phase-1 §10.)

IVP methodology (carried open in spec v0.3 § 9)

  • Should the IVP spec itself be subject to a "meta-IVP" pass periodically? v0.3 status: spec now references multiple phase-specific findings files; meta-IVP is increasingly relevant.
  • For multi-reviewer runs, what inter-rater-reliability metric is reported (Cohen's κ on classification, agreement rate on verdicts)? v0.3 status: F-29 remediation introduced Cohen's κ as anchor for the framework's stress-test threshold; using the same metric for IVP's own inter-rater reliability would be coherent.
  • Recursion-risk for governance-citing-governance — when Dead Light cites PRINCE2/ITIL/CMMI, does it inherit the cited framework's evidentiary issues? When and how flagged in Phase 6?
  • Should Phase 4 fallacy checklist be expanded based on which fallacies actually surface in real runs? v0.3 status: survivorship bias surfaced multiple times (F-30, F-31); already in v0.2 checklist; no expansion needed yet.
  • (New v0.3) Formal definition of "argument cluster" for Phase 4 cluster-level decomposition.
  • (New v0.3) Should remediation commits include a Findings-addressed: trailer for traceability?

Glossary

Final glossary deferred — to be assembled bottom-up as later phases force real definitions. The 25 candidates and 11 needs-debate items are catalogued in framework/glossary-for-debate.md.

Multi-actor / Inquisitor / Authority (opened 2026-06-04)

  • debate 013 DECIDED (design-locked, not validated) — pending: author standard v0.1; run MA-01..MA-10 benchmark with a control; open Paperwork Standard §2 re-consecration (non-LOC tier trigger).
  • debate 014 (Inquisitor detect/enforce split) — proposal; 10 sub-decisions A..J pending walk-through. Splits glossary term-11. Mechanically grounded by 4 research briefs in framework/research/: oversight-detection-precedent, contradiction-detection-formal (Sub-decision H — conflict = ≥2 live same-subject decisions, mutually-exclusive, no supersession edge; subject = Bobbio 4-scope tuple), consequence-tracing-impact-analysis (Sub-decision I — impact radius via derived_from + agent auto-capture), conflict-resolution-and-policy-detection (Sub-decision J Compiler-vs-Court + the resolution schema).
  • Inquisitor open threads: Gap B (constraint-intersection at instantiation) evidence-thin — dedicated pass owed; severity-metadata uncovered; the CAP "post-hoc forced by no-real-time" claim is an unproven conjecture; DMN/firewall unverified (cite Cedar/Margrave only).
  • debate 015 (Authority Hierarchy & Delegated Span) — proposal; 6 sub-decisions A..F pending. Amends Codex HS-4/HS-5 (bounded agent-over-agent authority) → a future Codex re-consecration. Span = info-processing-capacity; reserved-risk line = existential/ethics/law/security → human; F = conflict-resolution schema (partial-order: authority-tier override + LegalRuleML Override edges + configurable specificity/recency tie-breaks + ambiguity→human — NOT a hardcoded total order).
  • Candidate cross-standard Principle (owner to confirm): governance thresholds — staleness, tier, span — are measured in decision/event-density, not human-pace units (calendar, LOC, headcount); wall-clock is a backstop only. Surfaced 3× (debate 013 E/C; input #2; debate 015 A).

Recommended next step (updated 2026-05-28 after Pass 1 extension)

LoreWeave Phase 0 Pass 1 has materially advanced; project-owner gating items remain. Pass 1 status:

  • ✅ PM Threshold Decisions signed off 2026-05-14 (not pending; HANDOFF was stale on this).
  • ✅ §1.1 covered: foundation/governance/contracts (2026-05-14) + knowledge-service + frontend substructures (2026-05-23).
  • ✅ §1.1 LW-04 planned-services reconciliation: 5 of 6 (orchestrator / rag-index / story-wiki / qa-extraction / workflow-job) confirmed absorbed in knowledge-service; continuation unclear.
  • ✅ §1.3 discrepancies: D-R-001..D-R-007 logged (added 005 README staleness, 006 a11y CI gap, 007 UI copy leak).
  • ✅ §2 + §3 AI candidates: D-001..D-020 + AR-001..AR-005 + F-001..F-004 surfaced.
  • ⏳ Project-owner gating: confirm §2 candidates (especially D-009/D-010 missing rationale + the D-015 SHA-to-CULT_001 mapping which needs re-verification); confirm §3 candidates; write §4 independent contribution.
  • ⏳ Pass 2 not started: 14 remaining services + contracts/api/ deep + agentic-workflow/ + crates/ + design-drafts/.

Two viable paths from here, the publishing buffer still removes time-pressure:

  1. Continue LoreWeave Pass 1 → Pass 2. Once project owner confirms the AI candidates and writes §4, run Pass 2 (remaining services + contracts). After Pass 2 the Reckoning Record is ready for Phase 1.
  2. Phase Chapters series (blog #11+) — Codex pattern, Adeptus Administratum chapter intro, free-context-hub (the runtime tier teased in #3). Per the project owner's plan, slower (debate + benchmark per Chapter); deferred until ready.

Pre-Pass-1 scan summary (still accurate, retained for context): ~358 KLOC across 1717 files, 1314 commits in 6 months, mmo-rpg/design-resume branch active, frontend v1→v2 rename complete, license + infra pivots in last month.

The framework is now equipped with:

  • IVP through Phase 5 — within-document evidentiary check substantially complete.
  • Adeptus Administratum Codex v1.0 — first Chapter sealed, available for PM-aide and Council-aide invocations during Phase 0 and Phase 1.
  • Three spec departures (D-1 anonymized attribution; D-2 AI-aide-first Implicit Principles; D-3 AI-aide-drafts-PM-thresholds) logged and re-evaluated as compliant with the sealed Codex.

How to resume:

  1. Review case-studies/lore-weave/pm-threshold-decisions.md. Three items flagged for owner attention: (a) §1 #2 "lowered 5 person-days → 1 day for solo pace"; (b) §3 "design-drafts/ and poc/ — committed decisions or exploration?"; (c) §4 "~6 person-week initial commitment" — confirm or scope down.
  2. Add adjustments in the "Adjustments by project owner" section of pm-threshold-decisions.md if any.
  3. Check three sign-off checkboxes at the bottom and sign.
  4. Open a new chat session — a new Adeptus Administratum instance re-primes per Codex §8 protocol and begins Pass 1 Current State Audit of LoreWeave (docs/01_foundation/ + docs/02_governance/ + docs/03_planning/ + frontend/ + services/knowledge-service/).
  5. The Pass 1 work fills reckoning-record.md sections 1, 2 (partial), 3 (partial) for the in-scope areas.

Alternative next steps

If LoreWeave mapping is not the right next move:

  • IVP Phase 6 (External Benchmarking) — places the framework in adjacent literature (AI-agent governance, software governance, frozen-spec patterns). Heaviest IVP phase; produces audit/benchmark.md.
  • IVP Phase 7 (Synthesis Report) — integrates all per-phase findings files. Lightweight now that Phases 1–5 are done; one of the cleanest "wrap up the audit story" moves available.
  • Phase 1 Failure of sealing — short close-out debate; ceremony exit-ramp question.
  • Phase 1 Two-tier sharpness — deeper internal debate; Law vs Principle distinction.
  • Phase 2 draft (Codex per Chapter) — start defining Codex spec, including AI-assistant aide Chapters referenced from Phase 0 / Phase 1 anti-patterns.

Recommended ordering if doing further audit work over LoreWeave: LoreWeave first (it informs Phase 6 benchmarking with real empirical data), then Phase 6 + 7. Phase 7 alone (without Phase 6) is also viable as a synthesis of the now-complete Phases 1–5.

Conventions to preserve

  • All documents in framework/, case-studies/, and distribution/ use English; conversation language with the project owner is Vietnamese.
  • Working drafts use the *-for-debate.md suffix; final docs drop the suffix.
  • Specific arguments live in framework/debates/NNN-topic.md with status (open / recommended / decided / superseded).
  • Each debate has a Decision section to be filled when the project owner decides — never pre-fill it.
  • The README's "planned" links indicate forthcoming docs; they should not be made into broken-link claims.
  • IVP separation-of-concerns: audit-output files (framework/audit/) and remediation edits go in separate commits; audit run never modifies framework documents in the same pass. Per session-3 lesson: for this small docs-only repo with single-owner review, keep the separation at the commit level — do not default to separate branches per audit/remediation per phase. The 2026-05-09 multi-phase run created 7 branches + 7 stacked PRs and the project owner explicitly called this "chaos". For future runs prefer one branch per IVP run with multiple commits, OR one branch per phase max — not separate branches for each audit-vs-remediation pair.
  • IVP spec changes go between runs, never during a run; rubric tables are pre-registered per pass. (Note: v0.3 clarified this is "between runs", not "between phases of the same run".)

How to resume

  1. Read this file end-to-end.
  2. Read README.md for the framework's elevator pitch (now includes AI-collab literature anchors).
  3. Skim framework/phases/phase-0-for-debate.md and framework/phases/phase-1-for-debate.md — the two main artifacts.
  4. Skim framework/debates/README.md for decision history (6 debates decided).
  5. If touching evidentiary claims or calibration anchors: skim IVP spec v0.3 § 4 rubric, § 5 phase procedures (especially Phase 3 inventory-paraphrase pre-step and Phase 4 cluster option), and § 11 changelog before editing.
  6. Confirm with project owner which path to take from "Recommended next step" or "Alternative next steps" before starting work.
  7. For PR strategy on multi-phase IVP runs: do not propose 7 stacked PRs; propose either a single branch with multiple commits or at most one PR per phase. See "Conventions to preserve" above.

Repo state at end of 2026-06-04 governance session (debate 013 decided; 014/015 opened)

Multi-actor / Inquisitor / authority arc, all on main:

  • 4cd30aa — case-study lore-weave F-006: manual contradiction audit of LoreWeave's five docs/sessions/ files (11,838 lines): two parallel current-states, same-session BLOCKED-vs-SHIPPED (K17.9), 11 legacy duplicate rows. The empirical anchor §1.6.5 asked for. (reckoning-record §1.6.6 + F-006 + AR-007; methodology §3.3/audit-trail.)
  • 78c00b9 — framework/research/oversight-detection-precedent.md: 2-pass deep-research brief (47/50 claims verified) — Censorate, Inspector General, tax-audit deterrence, Stasi, regulatory capture. New framework/research/ dir.
  • a878f61 — debates: 013 DECIDED (accept A..J; C/E→event-relative C4′/E2′; 3 concerns resolved); 014 opened (Inquisitor detect/enforce split, pull detect forward); 015 opened (authority hierarchy, Codex HS-4/5 carve-out); README index updated.
  • this commit — LOG 0014–0017 appended + this HANDOFF refresh; cursor → a878f61 (LOG 0017).

Resulting actions pending: author framework/multi-actor-coordination-standard.md v0.1; pre-register MA benchmark with control; open Paperwork Standard §2 re-consecration; walk debates 014/015 sub-decisions; draft the Codex HS-4/5 carve-out.

Repo state at end of 2026-06-04 blog-entry-cost + debate-013-inputs session

  • Branch: main. One commit covers: (1) blog entry-cost/clarity pass; (2) debate-013 candidate inputs; (3) LOG 0013.
  • Blog entry-cost pass (013-independent, working drafts): added a per-post "New here?" 30-second recap to #3–#10; converted the series nav from an inline · list to a numbered list with a "← you are here" marker across all 10 posts; reworked #3's lead ("governance tier"→"how much structure") + subtitle and #8's subtitle to lead with pain/payoff rather than mechanism. Triggered by two independent external reviews flagging #3's high entry cost for readers arriving cold.
  • Two technical refinements DEFERRED to debate 013: (a) concurrency-control-vs-CAP is topology-dependent and a distinct wall from the temporal eventual-consistency one; (b) a non-LOC tier trigger (decision-density / domain-count). Captured as candidate inputs in debate 013 § "Inputs surfaced post-proposal" — not applied to blog #3 (a downstream artifact must not get ahead of the unsealed standard; #3's coarse framing stays consistent with the sealed Paperwork Standard §1.2/§15a). The LOC-trigger input may ultimately be a Paperwork Standard §2 amendment, not C-tier scope.
  • Standing cursor: in this HANDOFF: 4bf19fc (LOG.md 0012) — a re-prime walks LOG since that SHA and re-folds event 0013 (this session).

Repo state at end of 2026-05-28 LoreWeave Pass 1 deeper-audit session

  • Branch: main, clean, latest commit pending = "Pass 1 deeper doc-governance audit".
  • This session's work (2 commits):
    • eadca45 — checkpoint: Pass 1 services audit (knowledge-service + frontend + docs/03_planning structural scan) + AMAW fabrication incident logged.
    • (this commit) — Pass 1 deeper analysis: reckoning-record §1.6 added (doc governance maturity assessment with volume forensics + scorecard + migration sketch); F-004 corrected (Living Worlds UI exists in frontend-game/, doc-drift not implementation gap); F-005 added (canonical case-study finding); AR-006 added; AR-005 superseded; Principle 10 added; §1.6.5 boundary flag for next Standard (multi-actor concurrent work).
  • Headline finding (F-005): LoreWeave has 615 markdown files / 205,739 LOC / 55 state-like surfaces, with 0 of 7 paperwork-standard mechanisms in place (AMAW partially adopted). Project owner reports doc-fatigue empirically. This is the canonical case the Paperwork Standard exists to address; Phase 1 prescribed action is M1-paperwork migration per §1.6.4.
  • Next-phase boundary: multi-actor / multi-branch concurrent work — punted to runtime tier by Paperwork Standard §15a; LoreWeave's parallel agents × parallel branches × no division-of-labor convention is the case study demanding a concrete sister standard. Next conversation thread.
  • Standing cursor: in this HANDOFF: eadca45 (LOG.md 0010) — re-prime walks LOG since that SHA and re-folds event 0011 (this session's deeper-audit event).

Repo state at end of 2026-05-23 paperwork-blog-series session

  • Branch: main, clean, pushed to origin/main (no diverged work).
  • Recent commits (top of log): 
    29b42d9  distribution: surface the practitioner cards in INDEX, README, CHANGELOG
5f2c219  blogs #3-#10: review pass — apply findings from multi-perspective audit
8174fce  blog #10: cross-unit protocol + full refer-back — closing Tier 1 paperwork
b6f5c40  blog #9: sectoring trigger — the four conjunctive conditions for M1 → M2 split
73f9e0a  blog #8: HANDOFF mechanics — two-part structure, eviction rule, 90/180 staleness triage
76595a4  blog #7: LOG style guide — six event kinds and four lifecycle tags
1a9990b  blog #6: cursor-based re-priming protocol
cba2b5a  blog #5: draft "Try to Break Your Own Framework"
e4f9dd8  blog #1: align format with #2-#4
4619de9  blog #4: rewrite value-first — three patterns with before/after
effda24  blog #3: rewrite value-first — a 3-question tier-decision test
(earlier #2 reworks 2026-05-22 …)
  • Paperwork-series buffer: 10 posts ready; #2 already published on dev.to (2026-05-22); the rest publish at adopter cadence.
  • Distribution 0.8.0 plus [Unreleased] CHANGELOG entry documenting the 7 practitioner cards + 2 templates added since 2026-05-22. No VERSION bump yet (additive surfaces).
  • No external dependencies, no build, no tests — this is still a documentation repository.

Earlier repo state at end of Session 4 (preserved)

  • Branch: main (clean as of end of session 4; ahead of origin/main by 15 commits — not yet pushed pending project-owner authorization).
  • Recent commits on main (top of log, end of session 4): 
    (commit pending) HANDOFF refresh after debate 007 seal + Tier 1 scripts
1246b25  seal debate 007 + implement Tier 1 scripts (validate / check_links / sync_distribution)
39dd8f7  open debate 007 — Scripting Infrastructure (post-debate-006 automation)
fc5ee16  Phase 3 — HANDOFF refresh + final sweep (debate 006 migration complete)
9e9afb8  Phase 2 — populate distribution/ with framework + templates + lore-weave snapshot
c4a574e  Phase 1b — YAML frontmatter backfill across framework/ + case-studies/
1a857c2  Phase 1a — folder restructure + cross-reference updates per debate 006
7f58439  seal debate 006 — Documentation Architecture and Distribution Template
68daccb  apply practitioner voice to audit findings executive verdicts
d18472d  apply practitioner voice to debates 002 / 003 / 004 / 006
52702fe  apply practitioner voice to operational + case-study docs
f60a606  apply practitioner voice to phase-0 and phase-1 working drafts
1d81e0d  apply practitioner voice to README + add "Who is writing this" section
344bbb2  add framework-wide policy #3 (practitioner voice, not authority pronouncement)
5077769  open debate 006 — Documentation Architecture and Distribution Template
e35dcb6  propagate Adeptus Administratum Codex seal across framework + case study
78d4bb4  seal debate 005 + create Adeptus Administratum Codex v1.0
f4c7ba5  consolidate HANDOFF for end of Session 3  ← end of session 3
  • All 38 surfaced findings (F-01 through F-39, with F-26 retracted via erratum) are remediated.
  • IVP infrastructure: Phases 1, 2 (full coverage of all 76 citations), 3, 4, 5 complete.
  • Phase 2 (Codex per Chapter): first Chapter (Adeptus Administratum) sealed via debate 005 on 2026-05-11; Codex v1.0 at framework/chapters/adeptus-administratum/codex.md; additional Chapters deferred to real-project trigger.
  • LoreWeave case study: Phase 0 scaffolding complete; pre-Pass-1 work (folder, PM threshold proposal, scan, spec departures D-1/D-2/D-3) committed; pending PM sign-off of pm-threshold-decisions.md before Pass 1 begins.
  • IVP spec at v0.3. Phase 5b (author-voice pass) referenced in framework-wide policy #3 but not yet drafted into spec v0.4.
  • IVP Phase 5 complete; Phases 6, 7 queued.
  • Session 4 work summary: Framework-wide policy #3 (Practitioner voice) sealed and applied across 17 files (commits 344bbb2 → 68daccb). Debate 006 opened (5077769), refined through decide-debate dialogue (sub-decision D Lifecycle revised from D3 → D4 task-scoped instance; sub-decision I revised from I4 markers → I0 no-protocol; sub-decision J extended with J.6 Mermaid + J.7 tool stack), then sealed (7f58439). Three-phase migration executed: Phase 1a folder restructure (1a857c2 — docs/ → framework/, case-studies/ at root, distribution/ skeleton); Phase 1b YAML frontmatter backfill across 25 files (c4a574e); Phase 2 populate distribution/ with sealed framework copy + 4 fillable templates + lore-weave-snapshot (9e9afb8); Phase 3 HANDOFF refresh + final sweep (fc5ee16). Then debate 007 (Scripting Infrastructure) opened (39dd8f7) and sealed same-day (1246b25) with Tier 1 scripts implemented (validate_frontmatter / check_links / sync_distribution under scripts/). First blog post drafted earlier in session at blogs/01-the-emperor-is-dead-the-light-remains.md. Layer count surfaced: framework now has four sealed-concern layers (project governance / operational aide / document architecture / tooling infrastructure) — each fractal of frozen-authority thesis per debate 007 §17. LLM token-cost added as a third framework-design axis (alongside human-time + correctness).
  • No external dependencies, no build, no tests — this is a documentation repository.