Prometheus Spec
Censorship-resistant trustless state channels for machine learning
...This work addresses the problem of computational integrity proofs (correctness of the actual computations) made by some non-trusted (anonymous) third party without repeating the whole volume of the actual computing. The issue is usually addressed by use of probabilistically checkable proofs (PCP), however, such simple approach leave untouched the question of coupling the payments for the consumed computational resources with the actual proof. The purpose of our protocol is to scale distributed calculations by obtaining results of nearly perfect (measurable) quality, without the need to repeat calculations or check the computational integrity proofs by the client.