QAISS is building a neutral quantum-AI coordination layer. Follow the public roadmap and verify every claim.

Docs · Specification

Benchmark methodology

Protocol v0 (draft): workload, timing, required metadata and comparability rules.

PublishedUpdated 2026-10-14

Workload

Brickwork circuit: a Hadamard layer, then four layers of RY(θ) on every qubit followed by CZ on alternating neighbour pairs. Angles come from a seeded PRNG (seed 2026), so every device runs the identical circuit.

Timing

  • Measure state evolution only (sampling excluded).
  • Three repetitions; report the median.
  • Small sizes are batched until at least 4 ms elapse so timer resolution does not dominate.

Required metadata

  • Protocol id and version
  • Hardware (cores, reported memory, GPU model where relevant)
  • Software and driver versions
  • Precision
  • Date and time

Comparability rules

  • Only compare records with the same protocol id.
  • Never publish a benchmark without its metadata.
  • Always include a classical baseline when comparing quantum or hybrid methods.
  • Treat browser numbers as illustrative: background load and power settings matter.

Release notes

Engineering updates,
not hype.

Research drafts, releases, and every change to a public claim. No price talk.

Subscriptions open once the project publishes its privacy policy and mailing endpoint.