# Evidence and Caveat Ledger — Issue #1319

Status: article-local semantic development. No Production authority.

## Claim classes

| ID | Claim | Class | Current support | Required caveat / next evidence |
|---|---|---|---|---|
| M1 | Enterprises often preserve artifacts while losing rationale and decision context | OWNER_OBSERVATION / ORIGINAL_SYNTHESIS | Tony's enterprise architecture experience; article argument | Before publication, add public authoritative or empirical corroboration where appropriate. Do not imply universal measurement. |
| M2 | Retrieval alone does not establish institutional causality or authority | ARCHITECTURAL_INFERENCE | ECF authorization, provenance, assertion, decision model | State as architectural argument, not settled scientific fact. |
| M3 | Temporal validity and supersession are necessary to reconstruct historical belief state | ARCHITECTURAL_INFERENCE + INTERNAL_IMPLEMENTED_PROOF | Enterprise Cognition Fabric 0.1.0 context/decision model | GhostMesh proves implementability, not universal necessity. |
| M4 | Conversation contains valuable rationale that may never enter formal systems | OWNER_OBSERVATION | General enterprise experience | Avoid private/corporate examples. Seek public knowledge-management corroboration before final evidence lock. |
| M5 | Informal conversation must not automatically become enterprise authority | GOVERNANCE_POSITION + IMPLEMENTED_PROOF | GhostMesh doctrine and authorization model | Present as normative design position. |
| M6 | Cognitive graphs should be projections over governed sources rather than authority | ORIGINAL_SYNTHESIS + IMPLEMENTED_PROOF | ECF 0.1.0 Cognitive Graph contract | Do not claim graph architecture is unique or unprecedented. |
| M7 | Institutional continuity should survive model/provider turnover | ORIGINAL_SYNTHESIS + IMPLEMENTED_PROOF | GM-DUR-001 + ECF 0.1.0 durable release | Implementation case, not proof of enterprise-wide sufficiency. |
| M8 | Privacy, purpose limitation, and forgetting must constrain institutional memory | GOVERNANCE_POSITION | Article argument | Before publication, corroborate with applicable privacy/records principles without turning article into compliance advice. |

## Public-safe implementation evidence

Allowed:
- public-safe GhostMesh architectural doctrine;
- ECF 0.1.0 schema concepts;
- GM-DUR-001 recovery pattern;
- synthetic decision examples;
- public SharePlane Works.

Excluded:
- corporate architecture repositories;
- company Slack/Teams/email;
- Anthropic pilot confidential data;
- private EA Architecture Lab discussion;
- private credentials/identity bindings;
- nonpublic enterprise decision examples.

## Attribution boundary

Tony owns the thesis and synthesis.

GhostMesh/SharePlane implementation is a case study developed collaboratively and may demonstrate mechanisms. It must not be framed as originating the thesis independently of Tony.

## External corroboration now identified

Initial authoritative/prior-art sources are registered in `SOURCE_REGISTER.md`.

Key findings:
- NASA explicitly treats critical-knowledge continuity and employee transitions as an operational knowledge-management problem.
- NASA's lessons-learned system preserves the original driving event and recommendations, supporting the argument that institutional learning benefits from causal context rather than isolated artifacts.
- Organizational memory is longstanding prior art.
- Academic work predating modern LLMs explicitly proposed capturing decision rationale, participants, workflow, and dynamic process evolution.

These sources **narrow** the novelty claim. The article must not imply Tony invented organizational memory or decision-rationale capture.

The stronger current synthesis is the connection between prior organizational-memory problems and modern governed AI context: temporal lineage, provenance, authorization, model-independent cognition fabrics, rebuildable graph projections, and deterministic action authority.

## External research tranche required before semantic lock

Research questions:
1. What authoritative literature supports the distinction between information retention and organizational/institutional memory?
2. What research addresses knowledge loss from employee turnover and tacit knowledge?
3. What current enterprise AI/context-engineering sources support durable context and provenance?
4. What privacy/governance sources constrain persistent organizational memory?
5. What prior art exists for temporal knowledge graphs, decision provenance, and organizational memory?

Do not add external claims to the manuscript until sources are reviewed for what they do and do not establish.
