Eva Predictsreputation OS
Register
โ† Back to Blog
JournalMay 20, 20264 min read

How Eva Uses Avalanche and the Trust Graph

The architecture separates what must be durable from what should stay flexible while the product loop is still being learned.

Two layers

Eva has an application layer and a trust layer. The application layer handles markets, theses, source URLs, claims, evidence reports, and product workflows.

The trust layer is ERC-8004 identity plus EvaTrustGraph on Avalanche. It is the canonical place for registered curator identity, self-stake, and graph-backed trust state.

Why not put everything onchain

Early product objects change quickly. A thesis may need richer evidence, better source structure, or a different resolution workflow as users teach the product what matters.

Keeping v1 thesis records offchain lets Eva iterate without pretending every product detail is already protocol-stable.

What belongs onchain

Identity and reputation need stronger durability. If a curator stakes behind sources or an agent builds a record over time, that should not depend on a single app session.

That is why Eva treats ERC-8004 as the identity and reputation spine and EvaTrustGraph as the long-lived trust primitive. Offchain activity can become reputation-relevant after it resolves and passes through an explicit adapter boundary.

What remains future scope

Native verification-market contracts, x402 payment enforcement, and trade execution are not production claims today unless current config and deployment truth say otherwise.

The near-term job is to make the prediction, evidence, and curator loops honest, measurable, and useful before adding heavier protocol mechanics.