Skip to content
dfpn. Read the protocol

← Use cases · ML researchers & model developers

A permissionless venue to ship detectors

If you build detection models, dfpn is the supply side of a verifiable-AI network: deploy your detector, and get paid when it contributes to a consensus verdict — no vendor gatekeeper.

The problem

A strong detection model usually reaches production only through a vendor relationship or an internal deployment. There is no open, incentive-aligned marketplace where a researcher can register a detector and earn whenever it contributes to a paid, verifiable verdict. dfpn is that marketplace — the model layer of a DePIN detection network.

Where dfpn fits

How the flow works

  1. 01

    Register

    Publish model metadata and a version to the Model Registry program and post the developer stake (20,000+ DFPN per version).

  2. 02

    Get run

    Workers run registered models as part of serving requests; your model contributes to consensus verdicts.

  3. 03

    Earn

    Collect a share of the 20% model-developer fee split whenever your model is used on a paid request.

Honest caveat

Model developers are held to the same honesty bar as workers: overfit or biased models are countered by hidden test sets and periodic benchmark rotation, and the developer stake is slashable.

Other use cases

Put it to work

Wire verdicts into your pipeline, or read how a verdict is produced end to end.