Skip to main content

Conformance & quantum verification

The proofs that ship with a result tell you that computation is sound. Two further surfaces let you check bigger claims: that a whole node's physics is unmodified, and that a quantum result is what it claims to be — each with the same doctrine, never say "trust me."

Conformance — prove a node hasn't been altered​

Every node exposes a public conformance surface. Ask for it and the node re-runs every golden vector of a fixed corpus through its canonical compute path — no database, no clock, no modulation, just the physics — and returns the wire-form results:

curl -s https://zeqsdk.com/api/zeq/conformance

You then compare those results bit-for-bit against the corpus's expected values. The corpus itself is served statically and is content-addressed — identified by a SHA-256 over its canonical form — so you can confirm you're checking against the real, unmodified set of vectors, not one the node handed you.

The power is in the failure mode: a single mutated constant anywhere in a node's physics fails at least one vector. You don't have to trust that a node runs the true kernel — you can prove it, yourself, from outside. Because the node bundle is byte-identical across the lattice, every node conforms to the same corpus, and a node that didn't would be caught the moment anyone checked. A node can also publish a conformance badge so its status is visible at a glance.

Quantum verification — honest about its own limits​

The framework can verify outputs from quantum hardware, and the interesting part is where it refuses to. It verifies a quantum result only where a checkable ground truth exists, and it never implies more:

Verifiable — exact ground truth

Bell / CHSH (classical bound S ≤ 2, Tsirelson S ≤ 2√2), Deutsch–Jozsa, Bernstein–Vazirani, and Grover — classes whose correct answer is mathematically determined, so the check can't itself be wrong.

Not verifiable — and it says so

Random-sampling / "supremacy" circuits have no efficient classical ground truth by design. These return not_verifiable with the reason — never a fabricated "confirmed."

That distinction is the whole point. A dishonest verifier would stamp everything "verified." This one publishes a coverage map of exactly what it can and cannot check, and holds the line: where truth is knowable, it proves it; where it isn't, it tells you plainly rather than inventing confidence.

# what can and cannot be verified, stated up front
curl -s https://zeqsdk.com/api/quantum-verify/coverage