Research integrity · The Advay evidence standard

Every technical claim carries the evidence that produced it.

Quantum hardware is a field where confident language runs ahead of measured reality more often than it should. Advay publishes to one standard instead: a statement is only as strong as the evidence class attached to it, and the class is always stated.

The single rule

A simulation is not a measurement, and a target is not an achieved result.

Everything else on this page is a consequence of that sentence.

Evidence classes

Six labels, from architecture intent to independent review.

These labels appear in public copy, technical notes and platform reports. They are the vocabulary a reader needs to know how much weight a given statement can carry.

Designed

Architecture or engineering intent

The processor programme coordinates qubit, coupling, resonator, readout and connectivity decisions while keeping proprietary geometry and process parameters controlled.

Simulated

Computed under a declared model

A result produced by specified software, assumptions, parameters and numerical methods.

Projected

Model-based extrapolation

An estimate derived from stated assumptions, reference profiles or scaling models.

Fabricated

Physical artefact produced

A device, test structure, package or related component exists, without implying successful operation.

Measured

Observed on physical hardware

Data was collected from a physical setup with configuration, method, uncertainty and limitations recorded.

Independently Reviewed

External technical review

A qualified external party reviewed the defined artefact, method or evidence within an explicit scope.

Claim lifecycle

How a statement earns the right to be published.

A claim does not move up an evidence class because confidence has grown. It moves because a specific artefact — a simulation record, a fabrication record, a cooldown log, a measurement report — now exists to support it.

01

Question

State the engineering question, the decision it informs and the evidence class that would be sufficient to answer it.

02

Design Basis

Record the architecture assumptions, constraints and acceptance criteria the claim will be tested against.

03

Simulation

Compute under a declared model, with software version, parameters, numerical method and known model limits recorded.

04

Projection

Extrapolate only where the assumptions and reference profile are stated, and label the result as a projection.

05

Fabrication

Note that a physical artefact exists — without implying that it has been operated successfully.

06

Measurement

Collect data on identified hardware with configuration, method, repeat count, uncertainty and limitations recorded.

07

Reconciliation

Compare measured behaviour with the prediction, record the delta, and update the model before the claim is reused.

08

Publication

Publish with the evidence class attached. A claim without its class and configuration is not published at all.

Disclosure standard

What every published Advay result states.

If a result cannot be accompanied by the items below, it is not ready to be published — internally or externally.

✓The evidence class: Designed, Simulated, Projected, Fabricated, Measured or Independently Reviewed.
✓The system configuration the result belongs to, including the reference profile or hardware identifier.
✓The method, software version and parameters sufficient for a specialist to interpret the result.
✓The assumptions the result depends on, and the conditions under which it would not hold.
✓The sample or repeat count, and the uncertainty associated with the reported value.
✓The known limitations, unresolved risks and what the result explicitly does not establish.
✓The date of the result and the date the underlying model or configuration last changed.

Claim boundary

What Advay does not say, regardless of internal confidence.

A public claims register governs the wording used on this website, in reports and in technical communication. These are the boundaries it enforces.Achieved device performance No coherence, fidelity, gate-speed, error-rate or yield figure is published for SQPU256 or any Advay device until it has been measured on identified hardware and reported with its configuration and uncertainty.

Achieved device performance

A defined design basis or implementation plan that has not yet been physically realised.

Qubits that exist

SQPU256 is described as a 256-qubit target architecture. It is never described as 256 working, high-fidelity or logical qubits.

Physical execution

A Tvarit result is never described as having run on SQPU256 or another Advay QPU unless the report identifies a measured hardware backend and records the test configuration.

Manufacturing status

Layout or mask data is described as prepared for review or under review. Terms such as foundry-approved or tape-out ready are used only after formal written sign-off.

Partnership and endorsement

Organisations are named, and logos used, only with written permission and a defined scope. Advay does not claim approval, endorsement, certification or selection it cannot document.

Simulation as measurement

A simulated or projected result is never presented, summarised or visualised in a way that implies it came from a physical measurement.

Controlled disclosure

Detail beyond the public level is shared under agreement.

Public pages are deliberately non-proprietary: they describe scope, method and interfaces rather than geometry, process parameters or partner-confidential data. Reviewers, foundries, laboratories and institutional partners who need the underlying evidence receive it through a scoped technical review under an appropriate confidentiality arrangement.