A twelve-domain expansion went to referees and came back overclaimed
Two independent referee passes — one on prior art, one taking the buyer's position — rejected the case for applying certified numbers across twelve domains. Two survived; ten did not.
The proposition under review was that a certified-number pipeline generalises across twelve domains that all handle numbers under scrutiny. It went to two adversarial referee passes: one assessing prior art, one arguing the buyer’s side. Both returned overclaimed.
Earlier rounds had already killed two other headline claims — a provider-term figure, and a compression ratio positioned as a competitive moat. Neither survived contact with a referee who was asked to break them rather than confirm them.
What the retreats were actually about
The failure mode was consistent, and it was rarely technical. In most domains a free, institutionally-backed alternative already existed, or the party who benefits from the verification is not the party who would pay for it.
- Metrology and standards. PTB’s Digital Calibration Certificate with D-SI is free, LGPLv3, and backed by a national metrology institute. NIST’s Uncertainty Machine is a more capable GUM-S1 engine than a 63-draw port.
- Scientific reproducibility. RO-Crate and W3C PROV are free, and the beneficiary of reproducibility is not the payer.
- Econometrics and policy. The AEA Data Editor re-runs submissions at no charge.
- Supply chain and operations research. Nobody audits an optimisation result, and linear-programming duality already hands the solver an optimality certificate for free.
- Crypto oracles. A smart contract cannot re-execute a dataframe library.
- Legal and forensic. Daubert challenges attack the model, not the arithmetic.
- Clinical research. Under 21 CFR Part 11 and GAMP 5, the vendor is the one being audited.
- Aerospace and engineering simulation. Model-form error dominates a term the method cannot express, before any compliance cost is counted.
Why this is published
A method’s boundary is the useful part. Ten recorded retreats say more about where certified numbers actually apply than twelve enthusiastic domain pages would have. A claim leaves the do-not-say list only through a recorded referee pass with sources — never because a new document needed it to be true.