Our rule checker has 218 rules, and 218 is the wrong denominator

We build a design-rule checker for quantum hardware. It has 218 rules. Quoting that number would invite you to assume all 218 apply to your device, and they do not — so here is the honest version.

The measurement

Across 18 real published devices, the median device reaches a verdict on 7–8 rules. Not because the rules are wrong, but because the papers do not report the fields the rules need.

Then we asked a different question: what would each device reach if it published everything its own field already knows how to publish? We saturated each one with every field its same-domain peers declare, and re-counted.

Median reach goes from 8 to 28. Assuming all 218 rules are reachable overstates by 7.8×.

The thing that changed the project

We grew the superconducting rule database from 16 rules to 60 over three days. Median reach barely moved.

That is the whole finding. We had been writing rules, and rules were not the constraint. What a device paper chooses to report is.

What this implies

There is a small set of fields whose absence blocks most verdicts — the gate needs one check to run, and the marginal cost of a verdict is usually one or two missing values rather than a whole schema. Which means a short, agreed reporting list would move the field's checkability a long way, and nobody would have to change what they measure — only what they write down.

That is the open question we are working on now: what is the minimal reporting set, and does closing it change verdicts rather than just counts?

We publish the denominator alongside the rule count for the same reason we publish the envelope our surrogates refuse outside of. A number without its denominator is not a measurement, and 218 without this paragraph would be a sales figure.