RESEARCH
The work is open,
and under review.
We'd rather independent cryptographers find the holes than have you find them. The whitepaper is with outside reviewers now and we'll publish what they say whether or not it flatters us.
IN REVIEW
WHITEPAPER, EXTERNAL
OPEN
SOURCE, MIT AND APACHE
DARPA
SUPPORTED GRADUATE RESEARCH
UNDER REVIEW
The whitepaper is with
outside cryptographers.
It covers consensus, the storage proofs, the recovery scheme and the economics. Independent reviewers are working through it now. We're not claiming an audit result before there's one, and when there is, it goes up here in full.
PROOF OF CONTRIBUTION
A proposal, clearly labeled as one.
Useful work such as inference, data and verification gets hashed, signed, verified, anchored, and paid out of escrow only after it has been checked. We're explicit about the boundary: the chain verifies that a contribution happened and is eligible for settlement, not that a model's answer is correct.
What's live today is attestation, verification and zero knowledge proofs for a fixed architecture. The reward economics and slashing are proposed. We separate the two on purpose.
ACADEMIC GROUNDING
The math came first.
Our core algorithms come from DARPA-supported graph matching research at UCLA under Prof. Andrea Bertozzi and Prof. Paolo Climaco, recovering templates from large noisy networks where edges are missing and labels can't be trusted. The similarity heuristics from that work went directly into how the chain selects peers and reaches consensus.
Related work spans generative modeling at Berkeley's SLMath school and scientific machine learning at the University of Maryland.
