What it does
Rhizome Brain is a runnable prototype of a living, tree-structured AI memory. It keeps competing hypotheses separate, continuously revises their valences, preserves weakened branches, and can reuse a preserved mechanism to understand a new situation, search for solutions, or construct an adapted metaphor.
The demo includes deterministic scenarios that run locally without an API key. An optional server adapter supports OpenAI calls while keeping separate logical-agent branches.
Why it matters
Conventional summaries often merge incompatible interpretations and erase minority hypotheses too early. Rhizome Brain keeps provenance, relations, revision history, and uncertainty visible. A hypothesis can weaken without disappearing and later regain a precise causal role.
What Codex accelerated
The creator supplied and corrected the conceptual mechanism. Codex helped reconstruct the architecture, turn it into a runnable interface, implement automatic graph creation and dynamic valences, correct canonical node identity during revisions, add safe JSON export, write tests, and prepare jury documentation.
Privacy and testing
The included demonstrations are fictional. No historical conversation or personal case is included. Raw event text is omitted from export by default, and nothing is persisted automatically.
Run locally with:
python3 server.py
Then open http://localhost:8000/demo/.
Built With
- codex
- gpt-5.6
- javascript
- openai-api
- python
