Compared · as we understand their public materials · corrections welcome

MIZAN and AuditZK: what each one proves.

We name competitors and neighbours, because hiding them is what weak positions do. This page states what AuditZK proves, what MIZAN proves, and where the two sit relative to each other.

What AuditZK proves

AuditZK attests to track records using trusted execution environments: the computation runs inside protected hardware and the hardware vouches for it.

What MIZAN proves

MIZAN uses a STARK generated in the RISC Zero zkVM. The verifier trusts mathematics, not a hardware vendor's key. A 225 KB receipt verifies offline in about 81 ms with no assumption about Intel, AMD, or anyone's enclave.

Where they sit

Same goal, different trust model. A trusted-hardware attestation is only as strong as the hardware's attestation chain; a STARK has no such dependency.

What nobody on this list does

To our knowledge, no one else proves the anti-overfitting framework inside the proof itself: the deflated Sharpe ratio with the trial count as a committed Merkle leaf count, recomputed in-circuit. That is the unclaimed square, and it is the one the standard occupies. If that changes, this page will say so.

Other comparisons: vBase · Proven · Knidos · Darwinex · Numerai · QuantConnect.