A physics engine, and a verification tool enforced on top of it. PhysWall inverts a closed, published, non-linear physical law at a single measurement point — and refuses when the inverse is not unique. Seven laws, one engine, and the same refusal in all of them.
A physical-logic engine that decomposes a measurement gap along four axes — the machine, the law, the object, and which definition — and runs every domain through one structure.
RF matching, PCB loss, orbital conversion, photonic scattering, stream gauges, ocean waves, basketball. Same engine, same refusal: when the measurement cannot carry a conclusion, it says so instead of answering.
Three questions, and what would settle each one.
Each of these has an answer. None of them has been found here — and for one of them the answer turns out to be that the measurement cannot decide, which on this site counts as a result. The first correct solution to each gets published with the name of whoever sent it. The ones already settled are on the other page.
BIPM publishes the difference between the practical temperature scale ITS-90 and thermodynamic temperature. At the triple point of water, 273.16 K, the fitted polynomial gives −0.07 mK.
Is that a real physical offset, or is it noise?
⚠ This said “that difference is −0.07 mK” until an outside reviewer read the source. The data point entering the fit is (T − T90) at the triple point = 0.0(1) mK — zero, with an uncertainty of 0.1 mK inherited from the Boltzmann constant when the kelvin was redefined in 2019. The 12th-order polynomial is deliberately not forced through zero there, so it returns its own value nearby. Those are different objects and the page was calling one by the other's name.
the document "Updated Estimates of the Differences between
Thermodynamic Temperature T and the ITS-90
Temperature T90", BIPM 2022, Table 1
Everything you need is in one row of that table. The answer is not a number.
⚠ And the reviewer got further than the puzzle asks. The stated uncertainty on that point is 0.1 mK and the value is 0.07. The figure is smaller than the uncertainty on it, so the measurement cannot tell an offset from noise — which is the answer, and it sits in the uncertainty column rather than the value column. The puzzle stays up because working that out yourself is the point, and because we had the right question for the wrong reason.
Vega's spectrum peaks at 322 nm. Wien's law gives a surface temperature. What is it?
Wien λ_peak · T = 2.8977719e-3 [m·K] Vega A0V, and it rotates at roughly 274 km/s spectrum CALSPEC alpha_lyr_stis_011.fits (STScI)
⚠ And read that file's header before answering. It is not one measurement — it is a composite: a Kurucz model below 1152 Å, IUE data to 1675 Å, STIS fluxes above that, and a model again past 10200 Å. Which piece 322 nm falls in is part of the question.
The arithmetic is one line. The question is whether the arithmetic is the question.
A field goal is a closed law with a tolerance window, which is exactly what this site runs. Declare it, and it comes apart. Find out where.
geometry uprights 18.5 ft apart · 10 ft crossbar
snap and hold add 17 ft behind the line
model the same one used everywhere here:
tolerance window / angular scatter, through erf
NFL make rate, 2024 season, by distance band
20-29 yd 93.8%
30-39 yd 92.9%
40-49 yd 83.8%
50+ yd 76.7%
⚠ Bands, not single yards, because a per-yard table is not published anywhere. And the band matters: kickers in 2024 are about forty percentage points better from 55-59 yards than they were in 1999-2003, so a figure without a season attached is a figure without a meaning.
Fit one angular scatter across all four bands and it lands remarkably close — 1.25 degrees, within about two points everywhere:
predicted actual miss 20-29 yd 98.0% 93.8% +4.2 30-39 yd 92.1% 92.9% -0.8 40-49 yd 84.0% 83.8% +0.2 50-59 yd 75.9% 76.7% -0.8
Three of the four sit inside a point. And the fourth is the short one — where the model is too optimistic, by four points, on the easiest kicks on the board.
The question is why the model misses where the kick is easiest. Geometry says a twenty-yard attempt is nearly a certainty. Something removes four percent of them, and it is not the angle.
⚠ The answer is not in the trajectory at all. Ask what can go wrong with a field goal that has nothing to do with how it was struck — and note that whatever it is, it should be roughly constant across distance, which is exactly the shape of a residual that only shows up where everything else is easy.
⚠ And this one is different from the other four: the answer is a limit of this tool, not of the data. If you find it, you have understood the engine better than any explanation page here would teach you.
Defect density on a chip line doubled this week. What changed?
Published work on patterning at the smallest nodes reports that a 0.4% change in exposure dose doubles the defect density. So a doubling is exactly what a fraction of a percent of drift produces, and the measurement cannot separate that from a process that actually changed.
Better defect counting does not fix it — the ambiguity is in the relationship, not the count. The full working is here, including what the same relationship does settle.
Send the number and the closed form, or the reasoning where there is no number. A figure on its own can be guessed, and fifty guesses would arrive within the hour — so it is the form we record, not the value.
POST /api/puzzle
{"puzzle": "t90", "value": ..., "form": "...", "name": "..."}
Submissions are timestamped by the server, not by your browser. Nothing else is stored — no IP, no cookie, no email.
24 hours, or until both are solved. Then the answers go up, with the solutions that got there first.
Neither is a puzzle. They are the reasoning the puzzles come out of, written out.
An unchanged 36% shooter can look like 18% or 58% over twenty attempts. On when a number carries a conclusion — and what causes the gap when it does.
On measuring quantities that cannot be measured, and on the cases where the gap between a measurement and a law is not an error but the result.
They are not brain teasers. Each one is a place where the obvious move gives a confident wrong answer, and the four tools on this site exist because that kept happening to us.
Question three has no numeric answer. Question four's answer is that the question is malformed. If that annoys you, that is the point — a tool that always returns a number is the thing we are arguing against.
PhysWall was developed and architected by Gadi Zion.
Built on PhysWall — the same engine reads antenna bandwidth, conductor loss, bit erasure and heat limits. It answers what the measurement implies, and refuses when the measurement cannot say.⚠ Check this instead of believing it. Every number here reproduces from a source that is named, and the claims that turned out wrong are still printed next to what replaced them. The same engine runs all of these — it asks how much a measurement allows you to conclude, and refuses the same way in every field. The same engine runs all of these — it asks how much a measurement allows you to conclude, and refuses the same way in every field. How to check each one →