What KetQat will not tell you
Most of these are questions people arrive wanting answered. Each one is refused for a reason, and the reason is worth more than a confident number would be.
- When will quantum computers beat classical ones for my problem?
No date is produced, ever. A date requires predicting hardware progress, and this product has no basis for that which is better than the reader's own. What it computes instead is the condition: the cycle time, error rate or machine size at which the comparison would change.
How it is enforced: Structural. There is no field in the report carrying a date, so there is nothing to leave blank or fill in optimistically.
- Which quantum hardware vendor should we buy from?
No vendor ranking is produced. Device parameters you supply are used as assumptions and labelled as yours. Published roadmap figures, where shown, are labelled as targets — never mixed into a measurement, and never combined into a score.
How it is enforced: Observations, roadmap targets and user assumptions are separate kinds throughout, and no function combines them.
- Is quantum computing worth it for us?
Not without a cost model you supply. Absent one, every economic conclusion is refused by name and the report says which input was missing. It does not fall back to an industry-average price.
How it is enforced: A quantity with no value must carry the evidence class UNKNOWN, and the contract rejects any other combination — so a missing cost cannot enter the arithmetic as a zero.
- How accurate are these numbers?
They are exactly as accurate as the model and the inputs, both of which are stated. The estimates describe a machine nobody has built. Two estimators using different prefactors will disagree, and that disagreement is reported as a definitional difference rather than resolved by picking one.
How it is enforced: Every figure carries its evidence class, its assumptions and its limitations, and the sensitivity analysis reports the spread each assumption produces.
- Can you run this on real quantum hardware and tell us what happened?
Not here. This product costs workloads; it does not execute them. Nothing you upload is run, and no provider credential is stored.
How it is enforced: Nothing user-supplied executes in the web process at all. Execution lives in a separate sandboxed plane, and credentials are never persisted.
- Does a matching reproducibility hash mean the science is verified?
No. It means the bytes are unchanged. Verification recomputes the estimates and decisions from the inputs, which catches a tampered or stale conclusion — but agreement with our own model is not independent confirmation of anything.
How it is enforced: The strongest verification badge requires a human reviewer's authorization. The platform cannot award it to itself.
- Which of these two workloads is the better quantum candidate?
Not if they are not comparable. Workloads assessed under different assumptions are never ranked together, and a comparison table across assumption sets carries no aggregate row — a figure combining them would describe no machine anybody could build.
How it is enforced: Comparability is checked before any ranking is produced, and refused rather than warned about.
Why a product would publish this
A resource estimate is used to justify spending that will not pay back for years. The failure mode that matters is not an estimate that is slightly wrong — it is one that looks certain when it is not, because the reader then stops asking. Every refusal above exists to keep a question open that a number would have closed.
If you need one of these answers, that is a legitimate need and this is not the tool for it. Saying so is cheaper for you than discovering it later.
Read what is measured and what is promised, how the estimates are produced or a worked assessment, including one where the tool refuses to assess at all.