PhysWall inverts a closed, non-linear physical law at a single measurement point — and refuses when the inverse is not unique.
On measuring quantities that cannot be measured.
Every measurement has a moment where the number that came out is not the number expected. That moment is almost always called an error — in the instrument, in the model, or in the assumption between them. You look for the source, correct it, and measure again.
This is about the cases where the gap is not an error. It is the result.
A physical law connects quantities. Measure one side and compute the other — that is the ordinary direction, and it is what a calculator does.
Run it the other way and something different happens. You measure what you can reach, and what falls out is a quantity nothing measures.
Nobody measures how steady a basketball player's hand is from shot to shot. Nobody measures the roughness of copper sealed inside a circuit board. Nobody puts a sensor under three thousand metres of ice.
These are not hard measurements. They are measurements that cannot be made at all — the thing is inaccessible, or destroyed by the act of reaching it, or simply has no sensor that responds to it.
But a closed law connects each of them to something that can be measured. And read from the other end, it hands you the figure nobody took.
A calculator returns a number. Always. Including when the measurement cannot support one.
Which is the useful failure to think about. If a shooting percentage comes from twenty attempts, inverting it gives a release consistency — and that number is meaningless, because the percentage itself is not stable at twenty attempts. The arithmetic is correct and the answer is worthless.
So the useful tool is one that refuses.
When the derived number does not match what a law predicts, the difference sits in one of four places. They move independently:
the instrument what the measurement itself could not resolve the formula what the law leaves out the object what this particular sample did the definition which version of the quantity was meant
The fourth is the one people skip, and it is often the whole answer. Two laboratories can measure the same thing correctly, disagree, and both be right, because they were measuring quantities that share a name and not a definition.
Then the honest output is no verdict — not a large error bar. An error bar says the answer is uncertain. No verdict says the question, as asked, does not have one.
Not by the number agreeing with an expectation, because expectations can be adjusted. By two independent routes to the same quantity agreeing.
Carbon dating against tree rings is one: a decay curve and a count of growth rings, which share no physics at all. The kilogram is another — weighing against an electrical force, and counting atoms in a silicon sphere, which the literature itself calls two redundant and independent methods.
One measurement can be wrong in a way that survives forever. Two routes that do not share an assumption cannot.
Not that the mathematics is new. Bias separation in Kalman filtering, Type A against Type B in the GUM, and the BIPM key comparisons all got there first, and we checked fifteen times.
Not that any of it has been measured against hardware: zero measurements of our own, in every domain.
And not that a shooting percentage of 85% is achievable. Inverting it returns a release consistency no human hand can hold — which is the bound refuting itself, and exactly what a bound is for.
PhysWall was developed and architected by Gadi Zion.