Precision ladder · computed live in your browser
The Virtual Collider
Numbers nobody fitted, against measurements the precision community made.
Three particle-scale quantities — the proton-to-electron mass ratio, the gravitational constant, and the neutron–proton mass difference — each predicted by a closed-form, parameter-free expression from G's Ø Predictions (Parts I–III), and each confronted here with the value the precision-measurement community actually produced (CODATA 2018/2022, vintage labelled per value). The single measured dimensionless input is α, the fine-structure constant — Part II additionally uses CODATA me (ħ and c are exact by SI definition). Every verdict on this page is computed by your browser from the formulas below — predictions, residuals and PASS/DEAD are never typed in; the CODATA comparison values themselves are pinned constants shown in the source, and the Python proof runs the identical arithmetic.
What this is — and what it is not
This page does not simulate particle collisions, and it has no affiliation with CERN or any accelerator programme. Nothing here models a beam, a detector, or an event.
What it is: the particle-physics face of the five predictions. The "measured" column is the distilled output of decades of real accelerator and precision experiments — the CODATA constants — confronted with closed forms that never saw them. Arithmetic vs measurement; the reader judges whether that is structure or luck.
1The precision ladder
One log axis, eleven decades. Each bar runs out to that Part's |relative residual| — the fractional gap between prediction and measurement. Further left is finer agreement. The red tick is G's own published tolerance: a bar ending past its red tick means that Part is dead. Every position below is computed from the three formulas at page-load; nothing is drawn by hand.
2The three predictions, in full
Formulas verbatim from G's published verify.py (Appendix B).
The predicted and residual cells are computed by this page's JS from those
formulas — the same arithmetic as the self-verifying Python proof, to the displayed digit.
The PASS marker is not typed in; it is the live result of
|residual| ≤ tolerance.
Part I — proton-to-electron mass ratio
| Quantity | Value |
|---|---|
| Predicted | — |
| Measured (CODATA 2022) | — |
| Residual | — |
| G's tolerance (kill switch) | 0.017 ppb |
| Verdict (computed live) | — |
Part II — gravitational constant
| Quantity | Value |
|---|---|
| Predicted | — |
| Measured (CODATA 2018/2022) | — |
| Residual | — |
| G's tolerance (kill switch) | 1.0 % |
| Verdict (computed live) | — |
At 0.69 % of a 1.0 % tolerance, Part II is the coarsest of the three and the first place a sceptic should push. It is shown at the same scale as the others precisely so that difference is visible, not smoothed over.
Part III — neutron–proton mass difference
| Quantity | Value |
|---|---|
| Predicted | — |
| Measured (CODATA) | — |
| Residual | — |
| G's tolerance (kill switch) | 5.0 ppm |
| Verdict (computed live) | — |
3 · How each one dies
The framework indexes its own refutation — the corpus publishes 549 falsification conditions, and each Part above carries its own. Outside tolerance = dead, with no appeal:
- Part I is dead if mp/me sits more than 0.017 ppb from the closed form.
- Part II is dead if G sits more than 1.0 % from α²¹·(1+1/π)·ħc/me².
- Part III is dead if (mn−mp)/me sits more than 5.0 ppm from the closed form.
The Python proof asserts exactly these three gates and exits non-zero if any fails — the PASS on this page and the PASS in CI are the same arithmetic.
4Run the same check offline
Nothing here asks for trust. The proof is Python standard library only, runs in under a second, needs no network, and exits non-zero if any residual falls outside G's stated tolerance.
cd the420code/proofs/particle-bench
python3 demo.py # Parts I–III, verbatim formulas, self-verifying gate
5Honest limits
- Three quantities, not a theory of QCD. No hadron spectra, no cross-sections, no running couplings. "Where's QCD?" is a fair question — the falsification registry holds the kill switches for that line of attack.
- Reproducing arithmetic ≠ nature agreeing. This page establishes that G's published numbers follow from G's published formulas. Whether those closed forms are physics or numerology that happens to land is the open question, and it belongs to specialists.
- The measured column is CODATA, not raw events. Decades of accelerator and precision work, distilled — that is what the predictions are confronted with here.