Research on constitutional world models, graph-native memory, and human-AI symbiosis.
This repository holds the scientific and architectural side of ODEI.
It focuses on questions such as:
- what a real personal AI needs beyond prompting
- how memory should be represented if continuity actually matters
- how guardrails move from prompt text into system architecture
- how humans and agents can share durable operational context
- constitutional world models
- graph-native agent memory
- local-first AI infrastructure
- provenance and auditability
- governed execution and sovereign writes
- human-AI symbiosis as a systems problem
- odei.ai is the public research surface
- api.odei.ai is the production surface
- this repository contains the papers, notes, and technical arguments behind both
- position papers
- architecture papers
- implementation notes grounded in production systems
- comparisons with vector-only and transcript-only memory approaches
The goal is not hype. The goal is to make the architectural claims inspectable.
| Repository | Role |
|---|---|
| memory | Memory architecture |
| mcp-odei | MCP access layer |
| web | Public docs and API runtime |
| examples | Integration examples |
Repository-specific. See local files.