What the geometry forces, and what it measures
Sort every number this framework could be asked to explain into two piles — "the geometry forces this" and "this is measured" — and a sharp line appears. It runs exactly between the discrete and the continuous.
This is not a slogan we chose; it is a pattern that fell out of auditing the whole list of "just-is" numbers one by one. It is worth stating plainly because it is simultaneously one of the framework's strongest claims and one of its most honest admissions of limits.
The forced side — discrete and topological
Some quantities the geometry genuinely pins, and it pins them as a consequence of shape, not as a fitted value:
- The number of matter generations, N = 3. This is the topological index \(|\chi(K_6, E)| = 3\) of the internal shape — an integer, computed two independent ways with the same answer. You cannot smoothly dial an integer; it survives any continuous deformation of the geometry. Given the frozen shape and the observed matter content, three families is forced, not chosen. (Reported honestly as
DERIVED-GIVEN-E; provenance: the SG-3 dossier.) - The gauge group SU(3)×SU(2)×U(1). The forces are the isometries (symmetries) of the internal factors — color from \(K_6=SU(3)/T^2\), the weak force from \(S^2\), hypercharge from the circle: 12 generators, nothing extra, nothing missing (provenance: the SG-2 dossier).
- The dimension count, 13 = 4 + 6 + 2 + 1, and how it splits.
- Simple structural ratios — e.g. the Born exponent \(p=2\), and the curvature/kinetic ratio \(8/3\) that governs the inflaton's steepness.
Why these are forced: a topological invariant is rigid. An index, a symmetry, a dimension — none of them can change under a smooth deformation. Geometry is exactly the right kind of thing to pin them.
The measured side — continuous magnitudes
Other quantities the theory does not derive, and it says so:
- the Planck scale \(M_{\rm Pl}\),
- the gauge couplings \(\alpha_i\),
- the fermion masses / Yukawa couplings,
- the flavor mixing angles (e.g. the Cabibbo angle \(|V_{us}|\)),
- the cosmological constant \(\Lambda\).
These are soft. They carry units, they run with energy under the renormalization group, and they depend on boundary conditions. A geometry that is dimensionless and deformation-invariant is, almost by definition, the wrong kind of thing to pin a running, dimensionful number. So we take them as measured inputs — a small set of anchors, declared in full on Foundational Constraints — and we say plainly that they are measured, not derived. (You also cannot get a scale from nothing: dimensional analysis forces at least one measured dimensionful anchor, for any theory.)
Why the line falls exactly there
Topology is rigid; magnitudes flow. That single sentence predicts the whole pattern — and it predicts our own near-misses, which is the real test of an idea. A clean-looking inflaton slope that looked like \(1/6\), and a mixing angle that looked like it might be a simple geometric ratio, both turned out not to be forced — exactly as "continuous ⇒ soft" says. Meanwhile the discrete count, three families, is forced — exactly as "discrete ⇒ rigid" says. The framework's honesty here is not a retreat; it is the pattern working in both directions.
The one candidate we tested — the top quark
There is a single continuous number that might have broken the rule, and for an interesting reason. The top-quark Yukawa coupling \(y_t\) has a candidate forcing mechanism that is not topological: the renormalization-group infrared quasi-fixed-point. If the top coupling were pulled toward its value by the flow of the equations themselves — a dynamical attractor — rather than being a free input, it would have been the first continuous magnitude the framework forces: a derivation from dynamics, not shape.
We ran the computation — and it does not reduce \(y_t\). At one loop, the infrared attractor sits near \(y_t \approx 1.34\), while the measured value is \(0.9665\); reproducing the measurement requires a high-scale coupling below the natural range, so the measured top Yukawa is simply not the fixed-point value. (Checked target-blind, and independently reproduced.) So \(y_t\) stays an honestly measured anchor — and the pattern holds with no continuous exception: the single best candidate was put to the test, and the constants remain measured.
The honest bottom line
The framework's real, checkable wins are the discrete structure of the Standard Model — three families as a topological integer, the gauge group as a set of symmetries, the dimension count. Its continuous constants are measured inputs, stated as such. That division — forced counts, measured constants — is itself one of the clearest and most falsifiable things the framework says: if a future result showed a family count that wasn't an integer index, or a coupling that was geometrically forced with no dynamical mechanism, the picture would have to change.