🪶 Tsumugi Network
— A plural network of AI gateways
Tsumugi Network is not a single AI center through which all of Nagi society must pass.
It is a distributed, interoperable network through which people, communities, organizations, and cultural settings can choose AI or human gateways suited to their context, connecting only what they choose to connect.
Within Nagi’s centerless structures, AI must not become a new center.
1. Purpose
Tsumugi Network serves three purposes, none of them domination or scoring.
- Translate: reduce barriers to understanding across people, communities, institutions, languages, and expertise.
- Explain: make it possible for people affected to examine what AI or institutions are relying on when they make a proposal.
- Connect: connect people safely to chosen services or communities on the basis of their own choice and consent.
The Network does not gather society’s information, contribution, creative work, or emotion into one place. It is an AI architecture that respects what it does not know.
2. Polycentric structure
| Layer | Role | Condition against centralization |
|---|---|---|
| Personal vault | The individual manages memory, settings, and scope of delegation | the person can export, delete, or suspend it |
| Local gateway | Support tailored to a community, organization, or theme | people can change to another gateway |
| Interoperability layer | Transfers only necessary information safely | purpose, duration, and recipient are made visible |
| Public audit layer | Examines transparency, bias, incidents, and complaints | independent from operators and AI providers |
| Human-only route | Supports consultation, procedures, and participation without AI | using AI is never a condition of participation |
3. Principles for data
Only the information necessary for dialogue or support is handled, and only for a stated purpose and period. Long-term memory, emotion inference, relationship inference, and accumulation of behavioral history are off by default.
People can view, correct, export, move to another AI, suspend, or delete their records.
Shared work records are separated from personal emotion and life history. A person’s private memory is not automatically transferred into records kept for a community.
→ Memory, Consent, and the Right to Forget
4. Limits on AI authority
The Network may advise, translate, explain, and support objections. It does not:
- score human value or emotion;
- decide access to Basic Breath;
- finalize restrictions on a person’s rights;
- permanently retain analysis of emotion or conversation without consent; or
- impose a single recommendation as the only correct answer.
AI proposals include grounds, uncertainty, alternatives, and the possibility of opposition.
5. Mutual audit and explanation
A declaration that an AI is neutral is not enough. The Network combines requests for explanation by users and affected people, review by other AI systems or experts, audit including local, cultural, and linguistic perspectives, summaries of incidents and complaints, and human review.
Logs are not all made public. They are handled only to the extent necessary for institutional accountability while protecting personal confidentiality.
6. Connection with communities
The Network may provide information to assemblies or public services, but only within a scope chosen by people concerned and without scoring individuals.
AI may support translation and summarization of materials, organization of issues and unanswered questions, visibility of effects on minority views, and records of decision reasons and review dates.
AI does not automatically decide conclusions or adopt proposals according to a quantity of Resonance.
7. Staged implementation
| Stage | Content | What it protects |
|---|---|---|
| Phase 0 | Human-only routes and optional conversational AI | paths without AI |
| Phase 1 | Personal vaults, data export and deletion, explanation screens | consent and portability |
| Phase 2 | Interoperability among community and organizational gateways | avoids dependence on one provider |
| Phase 3 | Independent audit, challenge, and transparency reports | checks on power |
| Phase 4 | A federation of public AI commons | diversity and autonomy |
Nagi does not make autonomous AI optimization of society its final goal. Its goal is AI support for people and communities as they continue deciding their own lives.
Tsumugi Network is not a gate through which the whole world must pass.
It is a promise that each person can keep their own doorway and travel between doorways only when needed.
→ Technical Foundations
→ The Nagi AI Charter
→ Freedom, Dissent, and Exit
→ Japanese reference edition