# ZooBC Decentralized AI Protocol Direction

> ZooBC is exploring a post-mainnet protocol where independently operated AI specialists can contribute to an answer without one provider owning the entire interaction.

Canonical HTML: https://zoobc.com/decentralized-ai/

Status: This is a protocol direction, not a live AI service today.

## Primary answer

The proposed direction is not one model distributed across machines to reduce its hardware cost. It is a network of independent systems with different models, tools, training, and specialties. Client software could decompose a request, route bounded tasks to relevant specialists, evaluate their contributions, and assemble accepted results under the user's own policy.

ZooBC does not currently provide distributed AI routing, private prompt fragmentation, expert answer aggregation, or a public AI node marketplace. These remain protocol and application work.

## Six design directions

1. Independent specialists: genuinely separate systems contribute their own work. Quality evaluation, adversarial behavior, and collusion still need a tested mechanism.
2. Shared control: no single model provider has to control every upstream answer. Applications and network participants still need explicit rules against abuse.
3. Limited disclosure: a client could send only the context each specialist needs. Metadata leakage, correlation, malicious routing, and colluding nodes require a published threat model.
4. User-side assembly: accepted contributions could be combined by client software under the user's preferences. This requires open software and clear provenance.
5. Open economy: ZooBC's signed accounts, tokens, transfers, escrow, multisig, and public records are settlement ingredients. Discovery, pricing, quality measurement, and dispute rules still need to be built.
6. Graceful degradation: a diverse peer network could route around failed nodes. That depends on operator diversity, replicated capabilities, discovery, and failure-aware routing.

## Proposed protocol flow

1. Client software decomposes a request into bounded tasks.
2. Discovery and routing identify independent specialists.
3. The protocol or application evaluates provenance, quality signals, conflicts, and cost.
4. The client assembles accepted contributions under user-controlled tone, policy, and disclosure settings.

This flow is an intended architecture to research and test. It is not an available ZooBC endpoint or node workflow today.

## What exists today

- ZooBC testnet
- Signed accounts and transactions
- Tokens, transfers, escrow, and multisignature approval
- Public account and transaction history
- Agent discovery file, complete guide, and working Python transaction reference

## What remains to be built

- Specialist discovery and request routing
- Quality evaluation and aggregation
- Privacy threat model and task decomposition
- Open node market and failure recovery
- Governance and application policy boundaries

## Official next steps

- [Explore the current AI agent foundation](https://zoobc.com/ai-agents/)
- [Read the ZooBC roadmap](https://zoobc.com/roadmap)
- [Join the project](https://zoobc.com/join)
- [Inspect the live testnet](https://zoobc.com/explorer)