01
On-chain model validation
Model and dataset authenticity is validated on-chain, so a consumer can establish what they are depending on before they depend on it.
Implementations Agentic AI
Trie lets creators build, discover and exchange AI models, datasets and compute, with on-chain model validation, community governance, and compensation that reaches everyone who contributed.

Overview
An AI output is almost never the work of one party. Someone collected the data, someone trained the model, someone supplied the GPUs. By the time a result reaches a consumer, the structure of who contributed what is real and almost entirely invisible.
Trie makes that structure the marketplace. Models, datasets and compute are listed as assets, validated on-chain rather than asserted, and exchanged through the TRI token, so authenticity is something a buyer can check instead of something a seller claims.
Compensation follows contribution. Creators, dataset contributors and compute providers are rewarded through token incentives on the same recorded relationships that produced the output, rather than through a settlement negotiated afterwards.
The network rests on four layers: edge-ready compute, AI security and governance, verifiable data, and on-chain AI capital, with DePIN providers supplying physical infrastructure and validators building reputation on-chain.
What it uses the graph for
01
Model and dataset authenticity is validated on-chain, so a consumer can establish what they are depending on before they depend on it.
02
Creators, contributors and compute providers are rewarded through TRI token incentives tied to their actual contribution.
03
Ecosystem decisions and subnet governance sit with token holders rather than a central operator.
04
Validators and providers build on-chain reputation, and staking backs the claims the marketplace makes about them.
Live proof
No approved data contract for this implementation yet. Every tile shows the honest empty state rather than a placeholder number.
Other implementations