When fairness must be provable, the source matters.

Software randomness is excellent. Modern cryptographic PRNGs — /dev/urandom, ChaCha20, AES-CTR — pass every statistical test and are the right choice for the vast majority of applications: session tokens, simulations, game mechanics, A/B tests. Quey picks up where they stop — when the question isn't whether your result is random, but whether you can prove it.

But there are moments where knowing that your result is random isn't enough — you need to prove where it came from. When a giveaway platform faces accusations of rigged draws, when a gaming site must demonstrate fairness to its players, when an auditor asks for the provenance of your entropy — the answer "we used Math.random()" doesn't hold up. That's where the source matters.

Quey exists for those moments. We extract entropy from a measurable physical process — photonic shot noise — and serve it through an API that produces cryptographically signed, independently verifiable results. Not better randomness. Provable randomness.

Our entropy doesn't come from an algorithm. It comes from a single piece of hardware: a NoIR optical sensor sealed in a light-tight enclosure, connected to a Raspberry Pi 5.

With no incoming light, the sensor's output is dominated by photonic shot noise and dark current — stochastic physical processes that no algorithm can predict or reproduce. This raw signal is the foundation of everything Quey serves.

To condition that signal without introducing bias, we run a multi-stage extraction pipeline: least-significant-bit extraction, Von Neumann debiasing, and SHA3-256 cryptographic conditioning (NIST FIPS 202). The source has been validated to NIST SP 800-90B with a min-entropy of 0.6236 bits/bit, and the conditioned output passes 188 out of 188 SP 800-22 tests. The full validation certificate — including acknowledged limitations — is available here.