AI Testing — an independent AI reads the theory

Counting what fraction of the radiation budget free-streaming neutrinos could buy reproduces the oddly precise effective neutrino number 3.044 — the same non-integer the textbook entropy bookkeeping gives, and the one Planck measures.

Demanding that a deuterium nucleus survive being erased faster than it forms reproduces the famous "first three minutes" — the same 180-second breakout condition nuclear physics derives — with the predicted element abundances matching observation to under one sigma.

The record-affordability test says the universe's baby picture locked in classically at 380,000 years because records were being pinned down 130 times faster than expansion could stretch them away — the same margin plasma physics computes.

Grand-unification theory overproduces magnetic monopoles by 17 orders of magnitude, and both the standard dilution calculation and the record-affordability calculation independently crush the relic abundance to the same fantastically small number — matching the fact that no monopole has ever been detected.

The record-affordability clock and the hot-Big-Bang equations have no common equations, calibration, or free parameters, so when both land on the same date for a cosmic event the match cannot be one method feeding the other — and a single genuine disagreement would count against the framework.

With every cut, measure, and normalization frozen before the famous Bekenstein–Hawking answer was consulted, the 13-D geometry reproduced the black-hole entropy law to about 3 parts in 100,000 — and the mysterious factor 1/4 falls out as a rigid ratio of two geometric constants by a route that deliberately refuses the field's standard shortcut.

The framework's own inflaton candidate — a breathing-mode of the internal geometry rolling down an exponential plateau — projects the tilt of the primordial ripples into a band four digits wide, and the satellite-measured value sits inside it.

The atmospheric neutrino mixing angle is currently ambiguous between two experimentally allowed values, and the framework's frozen prediction sits on one specific side — 0.04σ from the lower-octant solution and 4.6σ from the upper — so upcoming experiments settle it independently of anything computed here.

The framework pre-registers a specific primordial gravitational-wave band — and if the LiteBIRD satellite measures outside it, the inflation mechanism is dead regardless of convention; inside it, the four candidate bands separate cleanly enough that the experiment even settles the theory's own normalization.

The framework has an armed, pre-registered window for the baryon asymmetry of the universe — a number it must hit or be falsified — while proving (by two independent exact calculations agreeing to one part in 10¹⁵) that the one missing input cannot be extracted from anything ever measured, and naming the future measurement that would supply it.

The dark-matter candidate comes with a pre-registered forecast that direct-detection cross-sections sit ten-plus orders of magnitude below current experiments — so any positive detection signal in that band kills the minimal candidate outright, a built-in way to lose published in advance.

Because the strong coupling and quark masses are outputs of the geometry rather than dials, any sub-percent disagreement with future lattice-QCD extractions falsifies the framework outright — a sharpness no theory with free parameters can offer.

Because the gauge forces are fixed by the shape rather than chosen, the framework predicts NO dark photon at any mass and no hidden light force — a hard negative prediction where a single confirmed detection anywhere would falsify the ledger.

The 14 unresolved checkpoints (dark matter's identity, the matter–antimatter asymmetry, the lithium-7 excess, the S8 tension) are the same questions all of cosmology has open, and the framework reproduces the standard calculation and reports each one open rather than claiming a win — evidence the 30 agreements aren't rounded up.

Every meson, baryon, antiparticle, resonance, and nucleus in the 2024 particle-data catalog is individually classified from the geometry's short alphabet plus standard confinement — a fail-closed machine audit reports zero inconsistencies — and the classic parameter-free mass relations check out (octet formula to 0.57%, Regge lines at R² = 0.9987).

Every frozen object carries a SHA-256 fingerprint combined into one master hash, and a single script regenerates every published number byte-for-byte and fails closed on any mismatch — so "we refined it after seeing the data" and "we tuned it" are formally the same act, and both void the certificate.

An early engine pass produced a wrong curvature ratio, and the Bianchi identity — a mathematical consistency law that knows nothing about the desired answer — flagged it; the corrected exact value 23/75 was then proved by three independent target-blind routes, and the wrong values are retained as labeled negative controls so they can never drift back in.

Reduced all the way down to ordinary quantum mechanics, the three-qubit error-correction code's failure rate comes out as the exact textbook formula, with no geometric parameter available to fake it — a mismatch would have been fatal.

The strong force's confinement scale was computed by two independent algorithms agreeing to ten decimal places, and then the geometry's own threshold correction was deleted and everything recomputed to prove no hidden geometric bridge was smuggling in the answer — the result was bit-identical.

The consistency checks are exact-fraction arithmetic shipped with deliberate sabotage tests that must fail — nudging one charge from 1/6 to 1/5 breaks four anomaly traces on cue — and the quantum paper's eight runnable integrity certificates all pass while all eight negative controls correctly fail.

On a fully worked charged-shell problem the framework's three-layer accounting and standard Einstein–Maxwell theory compute the identical energy to six significant figures at every intermediate step, and the weak-field limit lands exactly on Newton's gravity equation and Coulomb's law — the new machinery reduces to known physics rather than replacing it.

Before the heat-kernel and spectral machinery was applied to the new 13-D geometry, it was made to reproduce exact known textbook answers — sphere coefficients to one part in a thousand trillion, the sphere zeta value −1/15, and a standard lattice reference number hit to 0.025σ — including a control confirming the internal shape is not secretly a round sphere.

The identical six-dimensional geometry that forces three particle families also defines the admissibility rules of a quantum-error-correction architecture implemented as runnable code and exercised against a real decoder at ~100 million shots — needing roughly 10–14 physical qubits per protected qubit at scale versus about 24 for IBM's comparable published scheme, with the overhead band holding from hundreds to tens of thousands of logical qubits.

The 13-D theory was pushed through a 17-gate quantum-consistency audit — ghosts, anomalies, unitarity, positivity, hidden-force leakage — and not a single gate failed; every row below certificate tier carries a named research path, with the hardest open items shared field-wide.

A mechanical, answer-blind check ran over the entire 15,507-line manuscript and confirmed that all seven topics the theory does NOT claim to solve are openly declared, and that no claimed result secretly leans on any of them.

A self-audit checks that all the records the early-universe reconstruction relies on would actually fit within the universe's hard information-storage ceiling — and they do, with 15 to 34 orders of magnitude of headroom.

The theory's admitted worst number — an up-quark mass forced to 3.16 MeV against the measured 1.27, a ~4.4σ miss it published on its own front pages — is resolved by a dimensionless factor 1/√6 fixed purely by the six-element symmetry group of the flavor shape, moving the prediction to 1.295 MeV and leaving it standing as a sharp falsifiable prediction.

After one measured mixing number pins one angle, every other entry of the quark mixing matrix and the CP-violation strength are forced — and land on the measured values to hundredths of a sigma.

The CP-violating phase — a pure free parameter in the Standard Model — is read off an order-three winding of the internal geometry as exactly −120°, which in the standard convention compares as +60.0° against the measured 65.5° — under one sigma off, with no free phase available to insert.

The two numbers that set the scale of all particle masses — the Higgs field's vacuum value and the Higgs boson's mass — are computed from a frozen geometric loop (a Wilson line), landing at 246.02 GeV vs the measured 246.22 and 123.82 GeV vs the measured 125.10.

With no neutrino data fed in, the same structure returns both neutrino mass splittings, all three mixing angles, and a leptonic CP phase of about 260 degrees that sits inside the current global-fit band.

Instead of nine unrelated fermion masses, every within-sector mass step is a power of a single geometric constant e^(−π√3) ≈ 1/231 fixed at the hexagonal symmetry point of the shape — the top-to-charm and charm-to-up ratios are forced to be the same number, about 231, with per-family fudge factors formally forbidden.

The shape does not contain the hidden symmetry that would make the famous W/Z ratio come out 1 automatically, so deriving ρ = 1 exactly from real overlap integrals on the sphere is a genuine geometric result confronted with the measured 1.00038 — and the deliberately wrong operator on the same geometry gives garbage, proving the result is not a tautology.

The ratio of the top quark's coupling to the bottom quark's — a number the Standard Model just measures — comes out of the same frozen geometric determinant that produces the Higgs numbers, landing at 57.50 against the measured ≈58.

The strange fractional charges of quarks (+2/3, −1/3, with bulk-matter neutrality measured to roughly one part in 10²¹) are forced by a six-fold gluing of the three force sectors' centers, and feeding the resulting grid through Q = T₃ + Y reproduces the entire measured charge table with zero per-particle adjustment — a wrong charge like Y = 1/5 is geometrically inconsistent, not just unobserved.

The six anomaly-cancellation sums that must all vanish for the Standard Model to make quantum sense come out exactly zero in whole-fraction arithmetic on the geometry's own charge table, verified on two independent routes — while a nearby look-alike sum comes out 10/3, proving the pass is a real conspiracy of unequal fractions, and a deliberate sabotage (nudging one charge to 1/5) fails four traces on cue.

The one discrete choice in the Higgs mechanism is no choice at all: winding zero gives a universe with no electroweak symmetry breaking, and winding two puts the Higgs vacuum value off by roughly 60 standard deviations, so 1 is the only survivor.

The full framework reads only four measured anchors (Planck mass, gauge couplings, top-quark coupling, one mixing element) and returns 22+ independent quantities — the pattern of forces, mixings, the family count, the electroweak scale — the opposite of curve-fitting.

The strong coupling constant — a free dial in the Standard Model — is fixed by the volume of the compact color geometry and, run down to laboratory energies, lands on the measured value within its error bar.

Why the Higgs isn't dragged up fourteen orders of magnitude by quantum corrections — the Standard Model's worst fine-tuning, normally requiring two enormous numbers to agree to twenty-eight digits — is answered by topology: the Higgs is a loop-phase controlled by a whole winding number, and no smooth correction can change an integer.

The Standard Model writes in by hand that the weak force only touches left-handed particles (the 1957 shock); here handedness is enforced by the folded circle's boundary parity, which structurally leaves no place for the mirror partners to exist at the zero-mode level — parity violation is built into the shape.

The Standard Model must declare the Higgs into existence as a brand-new fundamental field; here it is the phase the gauge field winds up when it circles a non-shrinkable loop of the internal geometry — nothing new is added to the particle list.

That quantum probabilities stay real and non-negative is proven outright for the physically relevant finite-resolution world — the only unproven piece is the infinitely-fine limit, the same wall as the Clay Millennium Yang–Mills problem — and as a by-product the framework proved a genuinely new scope theorem: positivity is a strictly smaller sub-problem of the mass-gap problem, with the dependency running one way only.

Instead of assuming entropy increases, the framework ran candidate arrow-of-time measures against its open-system dynamics and found exactly one universal quantity that never decreases — verified to machine precision — while the five expected failures failed exactly where they should.

Standard quantum mechanics postulates ħ; here the entire "reality is quantized" idea is compressed to one named value-free axiom — a positive minimum operational step — with ħ as nothing more than that step's measured size, and two explicit counter-models prove that finite resources alone would NOT give you the floor, which is exactly why it must be its own axiom.

Every published attempt to derive quantum theory's probability rule imports an assumption as strong as the rule itself; this framework's gate-specific contribution is to split the rule into two legs graded separately rather than bundled into one opaque assumption, reduce each to one explicit value-free axiom, and prove by a pencil-and-paper three-state countermodel that its own gauge machinery does not secretly force the rule — so the import is on the record, not hidden.

The identical frozen 13-D geometry that carries the gauge forces, three generations, and quark masses also produces a graviton with exactly two polarizations, light-speed propagation, and Newton's law at long distance — every integer in the bookkeeping forced by counting (fiber weights 91 raw components, ghost subtractions on a 13-component bundle, net 65), and its key curvature ratio an exact fraction proved three independent ways.

Run up in energy with threshold corrections computed from the shape's own overtone spectrum (a ledger locked before the measured couplings were loaded), the three gauge couplings converge to a single value at 10¹⁶ GeV, with the closure residual driven to the numerical solver's convergence floor — a pipeline floor the corpus itself is careful to flag as far below the ~10⁻³ physical measurement band, not a physical precision claim.

The 50-year-old demand that the three forces must merge into one big group at one point dissolves because each force lives on a different factor of the shape — and precisely because there is no big unifying group, the proton-killing mediator particles of textbook GUTs simply do not exist, with the quoted lifetime (inherited from the proton-safety gate, SG-9) clearing the experimental floor by a margin of order 40× or more.

The five separate force-strength constants of the textbooks (Newton's G, strong, weak, electric, Fermi) carry zero adjustable dials of their own here — each is routed from the frozen geometry or computed from the others, and two of them (e and G_F) become identities that would disprove the geometry outright if experiment disagreed.

Textbook relativity says every black hole hides a point of literally infinite curvature; impose one axiom — a smallest step ℓ — and the infinity is replaced by an exactly computed finite value while the exterior stays exactly Schwarzschild, and the control most theories never get to run works: dial ℓ→0 and the textbook infinity returns smoothly.

The infamous claim that empty space should weigh 10¹²⁰ times more than observed comes from adding things up as if space were perfectly smooth and gravitating; a proven tensor identity shows the huge number was never able to curve spacetime in the first place — nothing needs to cancel to 120 decimal places — and the loophole (vacuum energy changing during cosmic phase transitions) was stress-tested and closed.

Removing the singularity does not remove the black hole: a horizon exists only above an exactly computed critical mass, and a related light-ring threshold was computed once by exact algebra and once by a blind numerical scan that didn't know the answer — and they agree.

In record terms a free quark is something the universe can never pay to isolate while a proton can be — and asking when a proton record first becomes affordable lands on the same 20-microsecond, 156-MeV moment that lattice supercomputer QCD computes.

Every "smallest length" idea for taming physics' infinities breaks relativity because lengths shrink for a moving observer; this framework's floor is on a Lorentz-scalar cost (action), which every observer agrees on — a smallest step of reality with no preferred frame, fixing the decades-old flaw in the minimal-length literature.

The Clay Millennium mass-gap problem is only hard because of the infinitely-fine continuum limit — at any finite resolution everything it asks for is already a proven theorem — and the framework's finite-resolution axiom lets it decline that limit while the physical gap stays a measured fact.

The irreducible quantum of cost provably dissolves one entire class of continuum-limit divergences in quantum gravity — the endlessly deepening tower of corrections — cleanly and without breaking relativity, while the corpus states plainly the other ten named ultraviolet walls remain untouched.

The sky's temperature is identical across patches that could never have touched, and the record ledger independently refuses to let that shared record exist unpaid — demanding the same early inflationary stretch at 10⁻³⁴ seconds that standard cosmology invokes.

Because recording anything costs granularity and that cost was once unpayable, the Big Bang is the beginning of the knowable rather than the beginning of time. It is like standing in the dark and asking what the surroundings look like: not observable, and no way for us to know. Nothing is fundamentally different about the second before and the second after — the only difference is that one is observable from our frame of reference and the other is not. No singularity is needed.

The "infinitely many steps" worry about continuous motion dissolves: gliding through overlapping states costs nothing, only jumps to genuinely distinguishable states cost, and no finishable process can contain infinitely many of those — so change is finite without space being pixelated.

A value-blind enumeration proved the shape can only produce family counts from a fixed list of integers — and 2, 4, and 5 are simply not on the list, so ruling out a fourth family no longer rests on collider data at all (which nonetheless agrees: 2.984 ± 0.008 light neutrino species, where a light fourth family would be off by ~125σ).

On an unfolded extra-dimensional circle every particle would come with a mirror-image twin nobody has ever seen; folding the circle in half deletes exactly the mirror copies — the index goes from (+3,+3) to (+3,0) — and the corpus keeps the bare-circle failure as the proof that the fold is load-bearing physics, not decoration.

Every unification scheme since 1974 fights proton decay by making dangerous particles heavy; here the geometry gives quarks and leptons no connecting mediator at all, so every proton-decay coefficient is exactly zero at every operator size — a machine census of 12,391 candidate operators found zero escapees, and a scan of 13,467 higher excitation modes found zero hidden proton-killers.

The same frozen shape fixed to produce the Standard Model also contains, on its boundary, a neutral particle structurally forbidden from decaying — selected 17 of 18 times across independent reweightings frozen before any comparison to the measured dark-matter abundance, with exactly one allowed way to talk to ordinary matter.

Rather than scanning 10¹⁴-plus candidate geometries, the Standard Model's own structural facts were used as hard pass/fail filters — spheres, tori, Calabi–Yau spaces, and Witten's 1981 candidate each die on a named constraint (mirror particles killed by LEP data, proton lifetimes killed by Super-Kamiokande) — and exactly one branch survives.

The single geometric competitor cheaper than the chosen color shape (CP², two dimensions smaller) was constructed in full and fails at the gauge-recovery test by over-producing force carriers — and where it does work, its family count is a tunable dial rather than a forced integer, so the selector paid two extra dimensions for unfakeability.

A formal removal audit shows nothing in the 13-D shape is decoration — delete the 6-D color factor and the strong force loses its origin and "three generations" becomes a free number again; delete the sphere and the W bosons lose their source; delete the fold and mirror matter returns; and a theorem shows no proper subset of the three layers (stage, rulebook, actors) can pass all ten construction gates.

The weak force cannot live on a donut — a hand-checkable theorem says no torus of any dimension carries a non-abelian symmetry, so the sphere is forced; a bare circle would predict mirror particles that collider data exclude, forcing the fold; and the color factor is the unique clean choice among the complete 7-class list of candidates.

Eleven named rival geometries spanning 4 to 12 dimensions were run against 13-D under a cost metric declared before scoring — one fails to generate the target at all, and the ten actually scored all lose — and when a self-audit found the advertised 18× advantage overcounted, the corpus published the smaller number.

Compactification creates an infinite ladder of heavy particles whose combined quantum effect is usually uncontrollable; here the shape's six-element symmetry forces the whole infinite sum to collapse to exact closed-form rationals, checked by two independent methods to ten decimal places — and the corpus states that generic Calabi–Yau spaces structurally cannot deliver this.

The classic killer of extra-dimension theories — the hidden dimensions crumpling or blowing up under quantum corrections — was computed against a pre-attached numerical bound and cleared it by 35× in the weakest accepted variant (about 2,400× centrally), with a later self-caught sign correction making the stability well deeper, not shallower, and no runaway (tachyonic) mode possible at linear order.

The fatal anomalies that quietly kill most unification attempts can only mathematically exist in certain dimensions, and 13, being odd, is not one of them — the corpus proves the would-be one-loop test literally has no well-defined target at odd dimension (verified by re-running at 12 and 14, where it is well-posed), so the dimension count is doing real work.

The single hard number that tests whether the 13-D shape survives as a finite quantum theory at one loop was computed as an exact fraction with no measured constant anywhere in it — after the same machinery reproduced four textbook sphere values to fourteen decimal places.

The Standard Model postulates the symmetry-breaking field is a single doublet because that happens to work; here it is forced by the external Wu–Yang/Dray classification of monopole fields on the two-sphere — the lowest mode of the charge-1 bundle is a doublet, full stop.

In string-style constructions the resting point of the internal geometry is one choice among ~10⁵⁰⁰; here the shape's exact six-element permutation symmetry forces the resting point to the symmetric center by group theory alone — any symmetric potential has a critical point there, no tuning available.

The unfinished part of the black-hole entropy problem is proved to collapse onto a single named mathematical coefficient on the shape's folded circle — a quantity the mathematics literature has never computed for anyone, in any theory, so the remaining wall stands in front of the whole field.

On the cosmological constant — where a curve-fitter would have found a way to claim a win — the framework computed its own best cancellation mechanism, watched it fail at every order, proved by a clean group-theory argument why it had to fail, and published both certificates, never comparing to the observed value anywhere.

The theory's charge bookkeeping hangs on arithmetic anyone can verify in 30 seconds — 3·(1/6) − 1/2 = 0 — and shifting one hypercharge by a single lattice step breaks two independent things at once (the electron's charge comes out −5/6 AND the anomaly sum stops vanishing), so the assignment has no slack at all.

Four of its own optimistic claims were falsified by its own audit and published as refuted with reasons; two tempting numerical coincidences were checked and retired as "an O(1) coincidence is not evidence"; a promising matter-antimatter result was withdrawn the same day it was produced; the quantum-computer side logged seven-plus self-caught corrections including retiring its flashiest ratios; and the headline input-economy claim was cut from ~4× to an honest 1.8× when a self-audit found overcounting.

The theory is graded as 33 distinct claims with the anchors named on each — the gates page shows all 33 resting at a ratified terminal, each with its anchors named and zero vague — and states plainly that no gate is "solved from nothing," while the front page keeps the residual tensions in plain view rather than folding them into the wins.

Everything consumed from experiment at the anchor level is five numbers (Planck mass, gauge couplings, top coupling, one mixing element, dark-energy density) where standard particle physics leaves dozens of dials free — and the corpus publishes the fuller honest table too (~12–14 total measured inputs, each with a named consumer), stating as a rule that no framework anywhere derives its inputs from nothing.

The quantum paper grades itself by the strictest possible rule — the whole work is only as strong as the weakest of its seventeen gates — prints that grade (AUDIT) on the front page, machine-enforces that no sentence claims more, and deliberately left its single most tempting row (strong CP) open rather than "completing" it.

Every object the comparison pipeline reads is locked and hash-stamped before any measured number is loaded, any rule must be writable without ever looking at the answer it predicts, a published knob-audit table invites reviewers to find one unlisted parameter (finding one triggers an automatic downgrade), and a blind control was registered in advance to come out negative so a reader can check the grading discipline actually rejects things.

"Topology is rigid; magnitudes flow" cleanly predicts which numbers the geometry pins (integer counts, symmetries, dimensions) and which stay measured (masses, couplings) — and when the one continuous constant with a plausible derivation mechanism (the top coupling via a renormalization attractor) was computed target-blind, it came out wrong (1.34 vs 0.9665) and the framework reports the miss openly, keeping y_t a measured anchor.

The famous 17-orders-of-magnitude gap between the weak scale and the Planck scale stops being a mystery once both scales are honestly booked as measured inputs — the gap is just their quotient, and there was never a third object owing anyone a derivation.

Because electromagnetism is forced to be the leftover of the weak sector rather than an independent force, the photon's charges are the weak sector's charges by construction — the perfect cancellation that makes atoms neutral (to one part in 10²¹) is inherited, not tuned.

Rather than tuning unobservable heavy-neutrino parameters to "explain" the matter excess (as the literature does), the framework proved by two independent exact calculations that the one missing number cannot be extracted from anything ever measured — and named the future measurement that would supply it.

What looks like one baffling dark-energy mystery splits into three questions with three different clean answers: a Λ-slot must exist (forced by Lovelock's 1971 classification theorem in 4D), its tiny size is a measurement, and the scary vacuum estimate never gravitates (dissolved) — the three are never conflated.

The framework proves some absolute measurement ruler must exist (change units and any bare number changes with them, so only ratios or anchored values carry content), then honestly reads the ruler's value from experiment instead of pretending to derive it.

Rather than hand-waving its hardest quantum checks, the theory proves that its graviton sector, UV completion, short-distance structure, and causality requirements all reduce to one single shared mathematical fact — the Yang–Mills mass gap — naming exactly the one thing the program leans on instead of hiding many loose ends.

The integer 3 that counts the particle families and a spectral gap of 0.260 — both frozen in the geometry for physics reasons — turn up doing a second, completely unrelated job inside the quantum-computer design, with no feedback loop between the two projects.

The lazy fix — postulating a law that baryon number is conserved — would secretly outlaw processes the Standard Model itself requires (sphalerons); the geometric mechanism forbids proton decay while leaving those processes untouched.

A single four-part test — coupled, encoded, stabilized, retrievable — applied blindly to nine early-universe objects sorts every one exactly the way decades of case-by-case physics judgment does (CMB photons real, monopoles not, dark matter's identity open), with zero mismatches.

Two structural consequences fall out for free: the knowable beginning and end are always the same distance in time from us (the observable universe widens symmetrically, no edge is ever hit), and dimensional analysis and record-cost bookkeeping — two unrelated arguments — both run out at the identical instant, 5.39×10⁻⁴⁴ seconds, where recordable history begins.

The textbook event horizon needs an infinite future to even be defined, but real black holes evaporate — so the framework keeps only the locally defined trapping horizon, which stays strictly one-way for the object's entire lifetime, effectively absolute for any observer in a 10-billion-year-old universe.

Where the state of the art needs holography or two-dimensional toy models, this geometry's quantum-extremal-surface location solves explicitly and symbolically, and the information-recovery turnover comes out structural rather than assumed.

In this geometry the fundamental resolution scale is not the Planck length of gravity folklore — it is a pure color-force object about 194 Planck lengths across, a concrete, surprising read-off of where the shape actually bottoms out.

Extra disconnected particle sectors that couple to nothing — the free-riders that plague model-building elsewhere — are eliminated outright by the nonseparability constraint itself, which the corpus calls the strongest, rarest kind of result: forced by the constraints, not merely consistent with them.

The eight classic ways a grand unified theory dies (mirror particles, anomalies, proton decay, unprotected Higgs, arbitrary flavor...) were not used to grade a finished theory — the same failure list was the blueprint that eliminated candidate geometries in the first place, one standard applied in two tenses.

The seed of the entire framework is Gauss's law — the century-tested fact that a field outside a closed surface doesn't care how the source inside is arranged — read as a constraint the deep structure of space must reproduce exactly, which by itself eliminates whole classes of candidate geometries.