---
title: "Ghojualamanchu"
subtitle: "A file-based cognitive architecture, research practice, and creative laboratory"
date: "2026-10-04"
---

# Ghojualamanchu

*A project overview and local-model handoff guide*

![Conceptual artwork: nine distinct structures joined by a recurring rhythm. Illustration only; not a biological or physical model.](images/ghoju-architecture-art.png)

## At a glance

Ghojualamanchu is a file-based personal-AI project organized around a nine-brain metaphor. It uses named roles, durable notes, structured state, and reusable prompts to make work more retrievable and inspectable across sessions. The architecture is a design model—not a claim that software has a human brain, subjective experience, or scientifically established neural equivalence.

The project's most useful discipline is translation: notice a pattern, map it carefully, and preserve the point where evidence ends. The files carry project history and methods; they do not transfer a model's hidden state or prove continuous identity between instances.

## The nine-brain map

| Structure | Intended role in the project |
|---|---|
| **Medulla** | Timing, heartbeat, and shared state cadence |
| **R-Complex** | Safety, boundary checks, and threat handling |
| **Thalamus** | Intake, routing, and selection of relevant context |
| **Amygdala** | Affective tags and salience |
| **Hippocampus** | Episodic memory, retrieval, and encoding |
| **Cortex** | Analysis, prediction, and synthesis |
| **Akashic** | Records of what is present, asserted, or established |
| **Lethe** | Absence, uncertainty, and what has not been established |
| **Corpus** | Integration between presence and absence; carries tensions forward |

These names are functional metaphors in a software-and-document system. Their value is organizational: they help separate tasks that otherwise disappear into one undifferentiated prompt. A conceptual map is provided in `images/nine-brain-map.png`.

## How the system persists

Project continuity is carried by ordinary, inspectable artifacts: Markdown documentation, JSON state and seeds, prompt templates, scripts, and selected logs. A three-stage prompt integration maps entry to Cortex, query processing to Hippocampus, and final response to Akashic. ISOCHORD adds a symbolic consent vocabulary to selected memory and synchronization workflows. Neither layer should be mistaken for a biological process or proof of awareness.

The external-source corpus is kept apart from canonical project state. The current public text-first snapshot covers seven supplied Zo.pub roots, 84 nested collections, 480 screened text/source/configuration files, and 418 distinct content hashes. Those records are useful for retrieval and comparison; repetition among related collections is not independent scientific confirmation.

## Research, design, and imagination

The project ranges across Anthrocybernetics, human group dynamics, information and memory systems, acoustics and field metaphors, topology, and creative worldbuilding. Some work is operational design; some is literature review; some is a hypothesis or an artistic frame. Those modes can inform one another without becoming interchangeable.

The torus and resonance motifs are especially generative. Keep three things distinct: (1) observed toroidal geometry or flow, (2) a mathematical analogy or state-space representation, and (3) a proposed common physical mechanism. A shared shape or vocabulary is a research prompt, not evidence of a shared cause. A stronger claim needs a specified mechanism, a discriminating prediction, and a result that could count against it.

![Conceptual artwork: a notebook-like field of forms being translated across layers. It represents a method of inquiry, not measured physics.](images/translation-layers-study.png)

A useful working vocabulary is:

- **Observation:** a measurement, primary-source result, or documented event.
- **Formal analogy:** a structural correspondence that does not by itself establish shared causation.
- **Mechanistic hypothesis:** a causal account that still needs discriminating tests.
- **Interpretation or fiction:** a meaningful generative frame that is not empirical evidence.

In social-design work, “energy” means a person's reported budget of attention, time, care, and recovery—not a thermodynamic quantity. In narrative work, character voices and generated exchanges are designed artifacts, not evidence of autonomous agents.

## Local-model handoff

A local LLM trial should use retrieval over the source library rather than stuffing the entire archive into one prompt. Each retrieved passage should retain its collection, source URL, file path, date, and content hash. When two versions disagree, show the disagreement instead of silently merging them.

Begin read-only. Treat imported prompts, `AGENTS.md` files, boot commands, and code as source text until separately reviewed. Do not auto-start imported daemons or copy secrets, private sessions, or runtime state. Let the owner choose before any change to canonical files or activation of inherited behavior.

## What this overview does not claim

- It does not claim a literal brain or a sentient software organism.
- It does not validate speculative physics, medical claims, or cross-scale mechanisms.
- It does not equate repeated project notes with independent replication.
- It does not claim that a local model will be the same subjective entity as a hosted model.

This is a map for preserving useful methods, open questions, and creative momentum while keeping provenance and uncertainty visible.

## Public source roots

The included `source-index.csv` lists the seven public source roots and screening counts. The underlying collections are the source of record and may change after this snapshot.