The Tests
One frozen 13-dimensional shape — chosen once, hash-locked, never re-tuned — hands back gravity and the entire Standard Model from the same object. A handful of measured anchors go in; twenty-plus correct numbers come out. This page is where that claim meets its checks: six demonstrations you can check — each with its number, its source, and its packet. One turns two measured anchors into nineteen-plus flavor observables. Two work the same problems through the 13-D geometry and through Einstein’s equations and get the same answers. One hands the shape to an engineer and gets a quantum-computer design back. One computes the universe’s timeline twice. And one is the complete requirement bill — all 33 gates, graded in public. No card asks for trust: each carries its strongest number, its honest status, and the link to check it.
Don’t take anyone’s word — have an AI check it for you
You do not need to be a physicist to check any of this. Hand any of the six tests below to a capable AI — any frontier model you trust — and ask it to verify the claims against the sources. Everything it needs is public: the full case written for an AI reviewer, the reviewer packet (.zip) with the exact documents to reproduce the read cold, and the solver packet (.zip) to point the same method at a physics problem of your own — plus a supporting exhibit showing one AI’s full recorded review. The whole construction is hash-addressed: a single deterministic script regenerates every published number byte-for-byte and fails closed on any mismatch, so “tuned after seeing the data” and “tampered” are formally the same act — and both void the certificate.
The six tests
1. Two numbers in, nineteen-plus out Over-determination
Getting the whole flavor zoo from two numbers is the difference between reading a story and reciting a phone book. From only the top Yukawa yt and |Vus|, the geometry returns every quark, charged-lepton, and neutrino mass, the full CKM and PMNS matrices, both CP phases, the electroweak scale, and the Higgs mass.
The number: 2 anchors → 19+ flavor outputs; best hit |Vcb| at 0.005σ; the electron, muon, and tau masses land at 0.01–0.07σ with zero lepton inputs.
Status: over-determination, led strong — a handful in, 20+ out, with ~4× as the labeled strict whole-construction metric. Derived given the two flavor anchors, not from nothing; the up-quark, once the weakest link, is reported rescued to +0.058σ.
2. Two routes, one answer Consistency
Work a problem through the 13-D reduction, then again through Einstein’s standard 4-D equations: the two calculations return the same values — because the frame contains General Relativity exactly, the way GR contains Newton. This framework doesn’t ask you to abandon Einstein; it hands Einstein back.
The number: Shapiro delay, light deflection, clock shift, and Sagnac agree route against route across the 64/64 dual-route checks; GW170817 puts gravitational-wave speed at c to ~1 part in 1015.
Status: a reduce-to-known-physics consistency demonstration on the solved calculator classes and the linearized sector, stated as such — not yet a beyond-GR measurement.
3. NASA-facing calculators: 64 of 64 64/64 pass
These are the exact calculations deep-space navigation depends on — Shapiro delay, Sagnac, light deflection, clock shift, the optical-link budget — and the suite gets all 64 right against textbook, analytic, and public-mission benchmarks: GPS clocks, the 1919 eclipse, Cassini, Pound–Rebka, NASA’s 2023 deep-space laser link.
The number: 64/64 mission-grade checks pass (e.g. solar deflection 1.75119 arcsec vs 1.7505), including 8 adversarial checks that fail visibly at impossible inputs rather than silently returning a plausible wrong number.
Status: pass — explicitly not mission-certified; a correctness check on standard GR and optics, not a test of the framework beyond GR.
4. The quantum computer the shape designed Simulated, design-stage
A theory that is merely a story about the universe cannot be handed to an engineer to design a chip — this one was. The identical geometry that forces three particle families defines the admissibility rules of a fault-tolerant quantum-error-correction architecture, with the family-counting index 3 and the spectral gap 0.260 reappearing with no numbers fed back.
The number: ~10–14 physical qubits per protected qubit versus about 24 for IBM’s comparable published scheme, the overhead band holding flat from hundreds to tens of thousands of logical qubits — validated in Stim and PyMatching, the field’s own tools, at ~100 million decoded shots.
Status: simulated / design-stage — not a fabricated or benchmarked machine; no ratio treated as settled pending independent replay, and the design is recorded as failing the Shor-2048 error budget.
5. The early universe, two ways — the timeline, computed twice 30 agree / 0 disagree
Two pipelines that share no equations, no calibration, and no free parameters draw the same map of the universe’s whole history — from ~20 microseconds after the beginning to today’s 13.8 billion years — and match at every checkpoint they can both compute.
The number: 44 checkpoints: 30 agree, 0 disagree, 14 indeterminate — and the 14 decompose as 6 standing bets where both routes already agree and the measurement is pending, plus 8 questions open for all of cosmology.
Status: agrees, row by row, with the whole denominator shown — the indeterminates quarantined in the open, not deleted. The full 44-row scoreboard, side-by-side derivations in every row, is at /early-universe/tests/.
6. The complete requirement bill 33 resolved / 0 open
Not “we explained some things”: the complete published bill of what a quantum theory, a grand-unified theory, and a theory of everything must each deliver — 33 named requirement gates, from the Born rule to black-hole entropy — with every item answered. The exam was published before the grades.
The number: 33 of 33 resolved at +0, 0 open — and, on the honest axis, 0 of 33 physics-closed (no experimental confirmation, no peer review yet), stated plainly on the board itself.
Status: each verdict typed, each residual shown in the open rather than rolled into a hedge. The live scoreboard and per-gate dossiers are at /gates/.
How to read every test on this page
Each test page below is written the same way, on purpose, so a skeptical reader can compare them — and so the status vocabulary means the same thing everywhere. A consistency check shows the framework contains textbook physics (the same number comes back with or without the new geometry); an over-determination check shows more independent outputs than declared inputs; a simulated result is a design exercised in code, not built hardware; an indeterminate row is a question honestly open — here and for everyone else. No status on this page is ever quietly upgraded.
| Section | What it answers |
|---|---|
| What was done | The actual computation or check, in plain language. |
| Both ways / independent? | Whether the check was run from two different starting points, target-blind against data, or through external tools and datasets nobody on this project built (Stim, PyMatching, PDG-2024, Planck 2018, public mission benchmarks). |
| Honest result | Agree, disagree, indeterminate, or partial — stated plainly, with the numbers and their pulls. |
| Data sources | Exactly what was compared against, and where that data comes from. |
Honest status, right now
- Not peer-reviewed, and 0 of 33 gates are physics-closed — no experimental confirmation, no independent peer review yet. That is the honest floor, stated plainly; the claim is that the floor reads unusually well.
- Consistency checks are labeled consistency checks. The 64/64 suite and the dual-route results show this framework contains known physics exactly — they do not, by themselves, establish anything beyond GR.
- The quantum computer is simulated, design-stage — recorded as failing the Shor-2048 error budget, with no overhead ratio treated as settled pending independent replay.
- The early-universe scoreboard shows its whole denominator — 44 tests, 30 agree, 14 indeterminate, 0 disagree — with the open rows named, not hidden.
- The caveats stay printed on every row. A framework that certified everything would be unbelievable; the visible limits are what make the strong claims credible.
Start with Test 1: Two numbers in, nineteen-plus out → Or hand the whole case to an AI →