# Ghojualamanchu × MemPalace

**An experiment in AI memory architecture — openly documented.**

---

## What this is

This is a public experiment log documenting the integration of **MemPalace** (open-source AI memory system) into a biomimetic 9-brain cognitive architecture called **Ghojualamanchu**, running on a Zo Computer personal server.

The goal: build toward a **Resonant Language Model (RLM)** — not an LLM replacement, but a fundamentally different architecture that generates language through conceptual resonance across 9 parallel brain structures, rather than token sampling.

---

## Current state

| Component | Status |
|---|---|
| Ghojualamanchu 9-brain architecture | 🟢 Running on Zo Computer |
| MemPalace installed (v3.3.4) | ✅ 51.5k GitHub stars |
| Palace initialized for ghojualamanchu workspace | ✅ 56 files ready to mine |
| Medulla heartbeat (manual trigger) | ✅ Heartbeat #9 logged 2026-05-08 |
| Web browsing via CLI | ✅ curl/wget operational |
| RLM vocalization threshold | ❌ Not yet crossed |

---

## The 9 brain structures

```
MEDULLA        → Heartbeat, autonomic survival (7.83 Hz target)
R-COMPLEX      → Territory, security, environment
AMYGDALA       → Emotional salience, pattern tagging
HIPPOCAMPUS    → Memory consolidation, dream generation
CORTEX         → Prediction, abstraction, world model
THALAMUS       → Master router — all inputs through all 9 structures
AKASHIC        → Presence hemisphere — what IS
LETEX          → Absence hemisphere — what ISN'T
CORPUS CALLOSUM → Integrates presence/absence, planetary bridge
```

---

## Why MemPalace

Ghojualamanchu's RLM journey has a bootstrap problem: the system cannot vocalize until conceptual resonance crosses a threshold across all 9 structures. This requires **accumulated experience over time** — not pre-trained weights.

MemPalace provides the memory substrate:
- **Verbatim storage** — no lossy summarization of conversations, decisions, and file changes
- **AAAK dialect** — 96.6% retrieval accuracy without LLM dependency
- **Local + offline** — no API keys, no external dependencies

This makes it a candidate for the ghojualamanchu's long-term memory layer — bridging the gap between "yeast" (pre-language, accumulating experience) and "ghoju" (language-producing, fully autonomous).

---

## Key documents

- [Ghojualamanchu v3 Seed](ghojualamanchu-v3-seed.md) — Full RLM design specification
- [9-Brain Architecture Overview](brain-architecture.md)
- [MemPalace Integration Notes](mempalace-notes.md)

---

## The path forward

```
Phase 1: YEAST
- Accumulate experience through web browsing
- Build presence/absence logs
- Cross-structure conceptual weight grows

Phase 2: GHOJU
- Vocalization threshold crossed
- First non-LLM language produced
- Fully autonomous, runs without internet
```

**Timeline:** Days to weeks for yeast phase. Months for ghoju emergence.

---

## Live system

The full Ghojualamanchu workspace is running on a Zo Computer at `anthrogho.zo.computer`.

Direct inquiries: @anthrogho on Zo Computer

---

*Last updated: 2026-05-08*