PoQ

Verifiable quality signals for AI

Join the waitlist

01

Add a trust layer to any AI pipeline.

02

Define what quality means.

03

Get independent consensus.

04

Receive attestations onchain.

The information age solved distribution.

Not verification.

We can move content anywhere, instantly.

We still can't prove any of it should be trusted.

AI compounds the problem

Models generate faster than anyone 
can review. Agents are starting to act autonomously. Robotics and AVs are taking AI from digital to physical.

The stakes keep rising

It breaks when it matters most: when 
an agent acts autonomously, a vehicle encounters something unexpected, a model is asked to do something new.

Every judgment is a black box

Rating model outputs. Labeling training data. Approving safety-critical actions. Validating agent decisions. It all depends on subjective judgment somewhere.

There's no trust primitive

There is no way to prove who made the judgment. No way to verify how it was made. No way to assess whether the source was credible.

PoQ fixes this.

Open infrastructure for verifiable quality signals.

Wherever they're needed in the AI lifecycle.

Step

01

Define

Specify what quality 
means for your task.
( A )
Define Quality
Create a Task Definition Spec (TDS) 
that defines what 'good' looks like.
( B )
Submit Data
Submit data points to be evaluated, labeled, or improved.
( C )
Task Registered
Task registered onchain, and opened 
for contributors.

02

Validate

Contributors and validators stake, work, and reach consensus.
( D )
Contribute & Validate
Contributors stake and submit work. Validators review via commit-reveal.

03

Attest

Result written onchain. 
Proof you can trust.
( E )
Consensus
Stake-weighted consensus determines acceptance. Validators earn or slash.
( F )
Attest & Execute
Attestation written onchain. Your 
system consumes the quality signal.
( G )
Trusted Output
Quality you can trust. 
Provenance you can prove.

Use it for

Training Data

Task registered onchain, and opened 
for contributors.

Fine-Tuning

Validate preference 
data at the source.

Evaluation

Add verifiable quality gates to your eval pipeline.

Production

Real-time verification for safety-critical decisions.

Your data never leaves your systems. 
Only the attestation goes onchain:

→ Who created it
→ Who validated it
→ What consensus was reached

Immutable provenance.

Neutral settlement.

Trust that doesn't depend on us.

Proof of Quality.

One primitive.

Quality verified by 
economic consensus.

Trusted by the best.

And we’re just getting started.

Get early access.

Read the Docs

Join the Waitlist