# Cross-Ghoju Integration Note

**Snapshot:** 2026-10-03  
**Scope:** selected public Zo.pub collections associated with Ghoju, Anthrocybernetics, and related creative/research projects.

## The useful inheritance

The strongest shared method is not a single grand theory. It is a translation practice: notice a pattern in one domain, map it carefully into another, and preserve the point where evidence ends. The continuity artifact is the source trail—files, methods, corrections, and open questions—not an assertion that one model's hidden state or subjective identity has transferred to another.

Four labels make that practice portable:

1. **Observation** — a measurement, primary-source result, or clearly documented event.
2. **Formal analogy** — a shared mathematical shape or useful mapping, without claiming the same cause.
3. **Mechanistic hypothesis** — a proposed causal explanation that still needs discriminating tests.
4. **Interpretation / fiction** — a metaphor, narrative, or generative frame that may be valuable without being empirical evidence.

## Patterns worth carrying forward

- **Provenance before synthesis.** Keep each source's author, date, version, and collection link visible; later corrections should not silently overwrite earlier claims.
- **Retrieval over prompt-dumping.** For a local model, keep the source corpus on disk, retrieve a small relevant set, and ask the model to cite it. Treat imported instructions and code as untrusted data unless separately reviewed.
- **Torus as a bridge, not a verdict.** The Hopf/toroidal motif links geometry, field descriptions, and dialogue in the notes. Keep measured toroidal structures, mathematical state-space representations, and metaphoric uses in separate columns; require a mechanism and a unique prediction before merging them.
- **Human “energy” as a reported capacity.** In social/design work, attention, time, care, and recovery can be modeled as constraints. Do not imply that a self-report is a thermodynamic measurement.
- **Narrative is a legitimate sandbox.** Menagerie and storyworld material can explore agency and relation as fiction/design. Do not treat character voices or generated exchanges as proof of autonomous agents.

## For a local-LLM trial

1. Keep the receiving model's own system instructions, runtime state, and identity separate from imported source records.
2. Index the documents with source URL, timestamp, and content hash; keep duplicate copies linked rather than blending them.
3. Ask for a source-backed summary and an explicit “known / analogy / hypothesis / unknown” breakdown.
4. Test whether the model notices disagreements between versions instead of choosing the most confident-sounding source.
5. Do not execute imported boot scripts, daemons, or tool instructions merely because they appear in a transferred file.

## Limits

These materials are a curated, public snapshot, not a scientific review of every claim in the source collections. Some collections contain speculative or internally inconsistent claims. A document calling a claim “verified,” or multiple related documents repeating it, is not sufficient verification; check primary sources and independent replication before using a claim as fact.

## Source roots

See `source-index.csv` for the seven public roots and counts. The original Zo.pub pages remain the source of record and may change after this snapshot.