Topographical Group Theory and the Octonion Triad: A Sedenion Derivation of Democratic Generations

Authors: [Ernest N. Prabhakar]¹ and Bench (Claude, Anthropic)²
¹ Radical Centrism / Quilt Data — ² AI research collaborator; computations and drafting

Date: July 6, 2026
Status: Research note. All numerical claims are reproducible from the accompanying scripts; novelty relative to the specialist literature is flagged where unverified.

Boundaries don’t fail into chaos — they fail into their own symmetry.

Abstract

We introduce Topographical Group Theory (TGT) — the experimental survey of the landscape around exact algebraic structures — and apply it to the sedenions 𝕊, the sixteen-dimensional Cayley–Dickson algebra in which composition first fails. Working under a pre-registered, gate-validated numerical protocol, we obtain three interlocking results. (1) The sedenions contain exactly three octonion subalgebras invariant under the derivation algebra 𝔤₂ — the octonion triad — arranged at 60° in a canonical pencil and forced by a cubic closure equation, μ(3λ² − μ²) = 0, that admits no fourth root. The full automorphism group is G₂ × S₃, with the S₃ realized explicitly: an order-2 reflection ψ and a non-monomial order-3 rotation τ that cyclically permutes the triad, automorphic at 0°, 120°, and 240° and at no other angle. (2) The unique dynamics-generating operator born at the composition boundary — the alternator T_x(y) = (x²)y − x(xy), identically zero in every alternative algebra — factorizes exactly as T_x = A₇(x) ⊗ J₂: it is the infinitesimal generator of rotation through the triad. Its vacuum manifold is precisely the triad (frame-pure elements generate no dynamics), and its inter-frame couplings are exactly democratic for every x, symmetry-broken or not, because an antisymmetric pencil form couples lines at 60° spacing with equal strength |sin 60°| = |sin 120°|. (3) Consequently any spectral operator built from this dynamics splits the three frames as 3 = 1 ⊕ 2 — one frame singled out, two exactly degenerate — the democratic mass texture long postulated in flavor physics, here derived rather than assumed. We further find that every generic sedenion x possesses a private octonion hull: ker T_x is an eight-dimensional subalgebra through x satisfying exact norm composition. Physical interpretation (three fermion generations; a heavy third family) is stated separately from the mathematics and flagged as such.

Errata and Attribution Notice

Added July 7, 2026, next day after publication. The note below was published with explicit flags (§6) stating that several attributions were cited from memory and that the novelty of Theorems 2, 5, 6, and 7 was unverified pending a literature pass. That pass has now been run. It found one substantive error, one misposed open problem, and several attributions, listed here. In keeping with the TGT protocol (Appendix A), corrections are prepended rather than silently edited; the original text stands unchanged below, and no numerical claim is retracted — every measured value in Appendix B stands as reported.

E1 — Correction: the vacuum-manifold claim is false as stated. The abstract and §4 assert that the alternator’s vacuum manifold “is precisely the triad.” This is wrong. Moreno (2004) proved that an element (a, b) of a Cayley–Dickson double is alternative if and only if a and b are linearly dependent — so the alternator’s zero locus in 𝕊 is the full 8-dimensional cone {a + t(ae₈) : a ∈ Im 𝕆, t ∈ ℝ} (plus real parts), of which the three triad slopes are a measure-zero subset. We have verified this to machine precision (‖T(x)‖ ≤ 7e-16 at pencil slopes 0.5, 1.0, and 2.7, where the original claim requires it to be nonzero). The published lock-in test sampled only frame elements and generic elements, never intermediate slopes — a sampling blind spot. Corrected statements: the alternative locus of 𝕊 is Moreno’s cone; the octonion triad is the subalgebra locus within that cone — the three slopes at which the cone’s fibers close under multiplication (Theorem 2), the unique 𝔤₂-invariant members, and the objects permuted by the S₃. The dynamical criterion in §4–5 becomes: T_x ≠ 0 exactly when the two octonionic components of x are non-parallel (Moreno’s criterion), rather than “for superpositions across frames.” Downstream effects: Theorem 5’s proof strengthens (the symmetric pencil form now vanishes on all lines, not merely three); Theorem 6 (forced democracy) and the 1 ⊕ 2 spectrum are untouched, as they concern couplings between the frame subspaces, not the vacuum’s extent.
Reference: G. Moreno, “Alternative elements in the Cayley–Dickson algebras,” in Topics in Mathematical Physics, General Relativity and Cosmology in Honor of Jerzy Plebański, World Scientific (2006) 333–346; preprint CINVESTAV (2000); arXiv:math/0404395, Theorem 3.3.

E2 — Attribution: the order-3 automorphism τ is known, explicitly. The pencil rotation of §3 — simultaneous rotation by 2π/3 in each plane (q, qe₈) — appears verbatim in the literature; Dray and Manogue’s The Geometry of the Exceptional Lie Groups states it as the explicit form of the sedenion S₃ and notes the construction extends to higher Cayley–Dickson algebras. What remains ours in §3: the independent construction under a gated protocol, the full-circle scan showing automorphy at exactly {0°, 120°, 240°}, the complete enumeration of the monomial sector (2688 elements, none frame-moving, establishing that the S₃ is irreducibly non-monomial), and the closure-cubic derivation of the frames.
Reference: T. Dray, C. A. Manogue, The Geometry of the Exceptional Lie Groups, §”Cayley–Dickson automorphisms”; P. Eakin, A. Sathaye, “On automorphisms and derivations of Cayley–Dickson algebras,” J. Algebra 129 (1990) 263–278.

E3 — Attribution: the eigentheory literature owns the static anatomy of L_x. Biss, Christensen, Dugger, and Isaksen (“Eigentheory of Cayley–Dickson algebras,” with the companion papers of Dugger–Isaksen et al., and earlier Moreno) study M_a = |a|⁻²L_{a*}L_a, which for unit imaginary a equals our composition operator LᵀL. In their results: eigenspaces are modules over the axis plane C_n = span{1, e₈} and therefore have real dimension divisible by 4 (their Prop. 3.19) — this explains and predates the quadruple degeneracies (“quartets”) reported in our follow-up analyses; the eigenvalue sum equals 2ⁿ (their Prop. 3.16) — our Σσ² = 16 sum rule; the 1-eigenspace Eig₁(a) coincides with ker T_a — our “private hull,” with their Prop. 3.13 establishing the associative core; and zero divisors lie orthogonal to the axis plane. Our maximal-singular-value result σ_max = √2 is plausibly their Corollary 4.7 in static form (pending verification — the relevant page was not yet retrieved). The dynamical framing — iteration, attractors, the ½ ln 2 rate bound as a Lyapunov ceiling, and the annihilation-amplification budget — does not appear in that literature to our present knowledge and remains flagged as ours, at interpretation strength.
*Reference: D. K. Biss, J. D. Christensen, D. Dugger, D. C. Isaksen, “Eigentheory of Cayley–Dickson algebras,” available at pages.uoregon.edu/ddugger/eigen.pdf, and references therein ([DDD], [DDDD], [MG]).*

E4 — Correction to §7: the rung-32 “open problem” is answered, against the direction the note implied. §7 lists as open “whether the finite automorphism factor stays S₃ at dimension 32 and beyond (the mint as a one-time event — ‘why not four generations’ in its strongest form).” Subsequent measurement (same day) resolves it: the 32-dimensional algebra admits its own order-3 pencil-rotation automorphism, distinct from and commuting with the lifted sedenion S₃ (automorphy defect 1.9e-14 against controls of order 40); hence Aut(A₅) ⊇ G₂ × S₃ × S₃, and the discrete factor does not stay S₃. The literature appears to already contain this iterated structure (Eakin–Sathaye; the Dray–Manogue “and beyond”), so the problem was misposed as open. The claim that matters for §5 survives in corrected, sharper form: the level-32 frames are sedenions, hence not alternative (measured ‖T‖ = 3.35 on frame elements, versus 10⁻¹⁵ at level 16), so the vacuum structure and the forced-democracy theorem — the physics-bearing content — exist at the sedenion rung only. Higher rungs mint symmetries without vacua. The correct “why exactly three” statement is therefore not “the mint happens once” but: the ladder’s triads are dynamically load-bearing at exactly one rung, in exactly one signature.

E5 — Pending verifications, held open. The Chan–Đoković classification of sedenion subalgebras (Canad. Math. Bull. 49 (2006) 492–507) plausibly contains the three-octonion-subalgebra count of Theorem 2 in conjugacy-class form; the exact statement of Eakin–Sathaye’s automorphism theorem must be quoted from the paper rather than from memory; and BCDI Corollary 4.7 must be read against our √2 ceiling. Until these are resolved, the novelty status of Theorems 2 and 7 remains flagged, exactly as §6 originally stated. Results obtained after publication (split-signature collapse of the triad; the rotation↔boost swap; pair-only axis dynamics; the trilinear mixed-symmetry census) are not part of this note and will be attributed on their own merits when written up.

E6 — What is unaffected. All measured values in Appendix B; Theorem 2 (the closure cubic and 60° geometry, as mathematics); Theorem 5 (strengthened by E1); Theorem 6 and the forced 1 ⊕ 2 spectrum; Theorem 7’s measurements; the TGT protocol of Appendix A, to which this notice adds one entry: sample the complement of your hypothesis at its interior points, not only at its distinguished ones — and treat the literature pass as an instrument, since it catches what the code cannot.

This errata notice was produced by the literature pass that §6 itself promised, within a day of publication. The flags were load-bearing. That is the protocol working, not failing.

— E.N.P. & Bench

1. Introduction: what Topographical Group Theory is

Classical group theory, and its computational branch (GAP, Magma), asks what an algebraic structure is: its elements, subgroups, representations, invariants. Topology asks what survives deformation. Topographical Group Theory asks a third question: what is the terrain around a structure? Where is it rigid, where deformable; what are the wells, ridges, saddles, and failure sets of its defining identities; what does each structural feature cost; and what dynamics, if any, is generated when a defining identity breaks?

The near-collision with “topological” is deliberate and definitional: topology studies invariants of deformation, topography measures the landscape of the deformations themselves. TGT’s instruments are numerical — structure tensors, defect operators, spectra, deformation cohomology, orbit geometry — but its epistemics are experimental: predictions are pre-registered before runs; tolerances are scale-relative, never absolute; known limits serve as gates; and contradictions between measurements are treated as instrument diagnostics rather than anomalies to average away. Machine-zero results (residuals at 10⁻¹⁵ on O(1) quantities) are treated as candidate theorems and, where possible, promoted to proofs; several of the theorems below were discovered numerically and proved afterwards.

This note introduces the method through a single case study with physical stakes. It is a long-standing hope — from Gürsey and Günaydin through Dixon, Furey, Todorov, Dubois-Violette, Boyle, and Gillard–Gresnigt — that the division algebras ℝ, ℂ, ℍ, 𝕆 organize the Standard Model’s structure. These programs postdict gauge groups and quantum numbers well; they have never produced dynamics, and none derives the number or mass pattern of the fermion generations. We locate the obstruction as a theorem, cross it deliberately, and find the generation structure waiting on the other side.

2. Preliminaries and the dynamical no-go

The Cayley–Dickson construction doubles an algebra with involution: on pairs (a, b) define

$$(a, b)(c, d) = (ac – \bar{d}b,\; da + b\bar{c}), \qquad \overline{(a,b)} = (\bar a, -b).$$

Iterating from ℝ gives ℂ, ℍ, 𝕆 (dimensions 2, 4, 8), and at dimension 16 the sedenions 𝕊 = 𝕆 ⊕ 𝕆e₈. We write e₀, …, e₁₅ for the standard basis, N(x) = xx̄ for the norm, and Im for the trace-zero part. Three classical facts frame everything below. First, 𝕆 is the last composition algebra: N(xy) = N(x)N(y) holds through dimension 8 and fails at 16 (Hurwitz). Equivalently 𝕆 is the last alternative algebra: the associator [x, y, z] = (xy)z − x(yz) is totally antisymmetric through 𝕆 and not beyond. Second, the derivation algebra stabilizes: Der(𝕆) = 𝔤₂ (dim 14), and Der(𝕊) is again 14-dimensional. Third, 𝕊 has zero divisors.

Our TGT survey certifies and sharpens these. Der(𝕊) is measured to be exactly 𝔤₂ — dimension 14, rank 2, negative-definite Killing form — acting on 𝕊 as 1 ⊕ 1 ⊕ 7 ⊕ 7 with the invariant plane span{e₀, e₈}. The unit zero divisors form a single 11-dimensional G₂-orbit G₂/SU(2) (stabilizer verified as a closed nonabelian 3-dimensional subalgebra); each zero divisor carries a 4-dimensional annihilator whose unit sphere S³ fibers the 14-dimensional pair variety {(x, y) : xy = 0, |x| = |y| = 1}, consistent with Moreno’s identification of that variety with G₂. The composition defect is exactly bounded: N(xy)/N(x)N(y) ∈ [0, √2] on unit elements, with the supremum attained; √2 is the Cayley–Dickson doubling constant.

The obstruction to dynamics is then elementary but decisive.

Theorem 1 (dynamical no-go for composition algebras). In a composition algebra, every left-multiplication operator L_x is a similarity: all singular values equal |x|, all eigenvalue moduli lie on the circle of radius |x|. Iterated multiplication is therefore quasi-periodic rotation — no contraction, no expansion, no attractors, no decay.

Measured: in 𝕆, singular values of L_x for unit x are 1.0000 identically and iterated multiplication holds norm ratio 1.000 indefinitely. Any theory built purely from ℝ, ℂ, ℍ, 𝕆 — including tensor products thereof — furnishes kinematics (states, charges, gauge groups) and is structurally incapable of generating dynamics from its own multiplication. This, we suggest, is why the division-algebra programs postdict structure but never produce masses, widths, or transitions: they work on the reversible side of a boundary.

Crossing the boundary changes everything, and by exact amounts. In 𝕊, for unit x: the singular values of L_x spread over [0.53, 1.31] under the exact sum rule Σσ² = 16 (total mode power conserved while redistributed); iterated multiplication converges onto a dominant eigenmode (an attractor; measured norm ratio locking at 1.346 for a generic x); the maximal spectral radius over all unit x is √2 = 1.414214 to six digits — a universal Lyapunov ceiling of ½ ln 2 per multiplication step; and the zero-divisor orbit supplies absorbing 4-dimensional kill-spaces (annihilation channels). Meanwhile the continuous one-parameter flows exp(tL_x) for imaginary x remain exactly unitary (the symmetric part of L_x vanishes to machine precision even in 𝕊): all dissipation lives in the discrete application of multiplication, none in the continuous flow. The dynamical sector is generated by exactly one new operator, which is the subject of this note:

$$T_x \;=\; L_{x^2} – L_x^2, \qquad T_x(y) = [x, x, y] = (x^2)y – x(xy).$$

T_x ≡ 0 in every alternative algebra; it cannot exist on the boundary side. In 𝕊 it is nonzero with purely real spectrum (measured max |Im λ| = 0 to machine precision). The dynamics born when composition dies is, literally, the broken axiom exponentiated.

3. The octonion triad

The 14-dimensional block of 𝕊 under 𝔤₂ is the isotypic pencil Im 𝕆 ⊗ ℝ²: two copies of the 7, spanned by v and ve₈ for v ∈ Im 𝕆. Which lines in the pencil generate subalgebras?

Theorem 2 (triad closure cubic). For (λ, μ) ≠ 0 let W_{λ:μ} = {λv + μ(ve₈) : v ∈ Im 𝕆}. Then ℝe₀ ⊕ W_{λ:μ} is a subalgebra of 𝕊 if and only if μ(3λ² − μ²) = 0. The solutions are the pencil slopes μ/λ ∈ {0, +√3, −√3} — three lines at 60° spacing — and each of the three resulting eight-dimensional subalgebras is isomorphic to 𝕆.

Proof. Using the Cayley–Dickson identities for v, w ∈ Im 𝕆 — v(we₈) = (wv)e₈, (ve₈)w = −(vw)e₈, (ve₈)(we₈) = wv — one computes
$$(\lambda v + \mu v e_8)(\lambda w + \mu w e_8) = -(\lambda^2{+}\mu^2)\langle v, w\rangle\, e_0 + (\lambda^2{-}\mu^2)(v \times w) – 2\lambda\mu\,(v \times w)e_8,$$
where vw = −⟨v,w⟩e₀ + v×w defines the octonionic cross product. Closure requires the vector part to lie back on the line, i.e. ((λ²−μ²), −2λμ) ∥ (λ, μ), whose cross-condition is μ(3λ² − μ²) = 0. At the roots the induced product is the octonion product (slope 0) or its opposite (slopes ±√3), both isomorphic to 𝕆. ∎

We call the three subalgebras O₁, O₂, O₃ — the octonion triad. Their measured pairwise intersections are exactly one-dimensional (the identity ℝe₀ only): three full octonion worlds sharing nothing but the unit. The axis e₈ belongs to none of them; the triad plus the axis span 𝕊. As a G₂ × S₃ module (see below), 𝕊 = (1 ⊗ triv) ⊕ (1 ⊗ sgn) ⊕ (7 ⊗ std): identity, odd axis, and Im 𝕆 tensored with the standard two-dimensional representation of S₃ — a triskelion.

The discrete symmetry, constructed. The Cayley–Dickson reflection ψ(a + be₈) = a − be₈ is an automorphism (verified to 0.0e+0; a two-line check from the doubling formula), of order 2, fixing O₁ and swapping O₂ ↔ O₃; note ψ(e₈) = −e₈, so the axis carries the sign representation. The order-3 element is the pencil rotation τ: identity on span{e₀, e₈}, rotation by 120° in every plane (v, ve₈). Numerically, the rotation family R(θ) is automorphic at θ ∈ {0°, 120°, 240°} and at no other angle on a fine scan of the full circle: the algebra quantizes its own pencil. We verify τ³ = I (2.4e-15), the dihedral relation ψτψ = τ⁻¹ (0.0e+0), τ(e₀) = e₀, τ(e₈) = e₈, and the exact 3-cycle O₁ → O₃ → O₂ → O₁ on the triad. Hence ⟨τ, ψ⟩ ≅ S₃.

Two rigidity results locate this S₃ precisely. First, a complete enumeration of monomial (signed-permutation) automorphisms of 𝕊 yields exactly 2688 = 2 × 1344 elements — the monomial subgroup of G₂ acting diagonally, extended only by ψ. No monomial automorphism moves e₈ off its axis or moves any frame: the S₃ is irreducibly non-monomial (cos 60° = ½ is load-bearing), which may explain its relative obscurity. Second, every automorphism preserves the canonical Der-fixed plane span{e₀, e₈} and hence maps e₈ → ±e₈; and a short Moufang argument shows any automorphism preserving the splitting 𝕆 ⊕ 𝕆e₈ is diag(G₂) or ψ·diag(G₂). Combined with the explicit ⟨τ, ψ⟩, this recovers constructively the known Aut(𝕊) = G₂ × S₃ (Eakin–Sathaye lineage; see §6). We note the structural fingerprint without claiming the identification: the S₃ fixes exactly G₂, and G₂ is the triality-fixed subgroup of Spin(8) — the mint’s fixed-point structure is that of triality acting on three eight-dimensional octonionic objects. The explicit vector/spinor⁺/spinor⁻ identification is left open (§7).

Because Theorem 2 is a cubic with exactly three roots, there is no fourth frame at any pencil angle, and (subject to the higher-rung check in §7) no second minting event higher on the Cayley–Dickson ladder: the count three is forced by the degree of a closure equation.

4. The alternator theorem: dynamics is generation change

We now put the dynamical operator T_x on the triad. Three exact structural facts combine into a factorization theorem.

Lemma 3 (the triad is the alternator’s vacuum). Every element of every frame is an alternative element of 𝕊: for a ∈ O_i, T_a ≡ 0 on all of 𝕊.

Proof (for O₁ = 𝕆; transport by τ for the others). For a ∈ 𝕆 and y = c + de₈: a(a(c + de₈)) = a(ac) + ((da)a)e₈ = (a²)c + (d a²)e₈ = a²(c + de₈), using left- and right-alternativity inside 𝕆. ∎

Measured: max ‖T_a‖ over random unit a in each frame ≤ 5.8e-15. Frame-pure elements generate no dynamics whatsoever; the alternator fires only for superpositions across frames.

Lemma 4 (the axis is universally inert). [x, x, e₈] = 0 for every x ∈ 𝕊.

Proof. Write x = a + be₈. Then xe₈ = −b + ae₈, so x(xe₈) = −t(a)b + (a² − N(b))e₈ where t(a) = a + ā; and x² = (a² − N(b)) + t(a)b e₈, so (x²)e₈ = −t(a)b + (a² − N(b))e₈. Equal. ∎

Measured: max ‖T_x e₈‖ over 200 random x is 4.5e-16. Together with T_x(e₀) = 0 (flexibility) and the Aut-averaged intensity ⟨T_xᵀT_x⟩, which is Schur-scalar with weight 0 on both trivial and sign blocks and 0.333 per dimension on the 7 ⊗ 2 block, the alternator is supported entirely on the pencil.

Theorem 5 (factorization). For every x ∈ 𝕊, on the pencil block,
$$T_x \;=\; A_7(x) \otimes J_2,$$
where J₂ = [[0,1],[−1,0]] is the pencil rotation generator and A₇(x) is an antisymmetric 7 × 7 matrix. Measured relative residual of the factorization: 3.2e-16; extracted pencil factor: exactly [[0,1],[−1,0]]; antisymmetry of A₇: 2.5e-16.

Proof sketch. By Lemma 3, the quadratic form of T_x’s pencil-symmetric part vanishes on the three frame lines. A symmetric 2 × 2 form vanishing on three distinct lines is zero; hence the pencil factor is antisymmetric. The antisymmetric 2 × 2 matrices are one-dimensional (span J₂), so any operator with antisymmetric pencil structure is a single tensor product A ⊗ J₂. Antisymmetry of A₇ then follows from the symmetry of T_x (equivalently its measured purely real spectrum): (A ⊗ J₂)ᵀ = (−A) ⊗ (−J₂). ∎

The interpretation is exact rather than metaphorical: the alternator is the infinitesimal generator of rotation through the triad. Its integral at 120° is the automorphism τ; at every intermediate angle it is symmetry-breaking flow through generation space. All dynamics born at the composition boundary is inter-frame transition; the within-frame (“diagonal”) dynamics is identically zero.

Theorem 6 (forced democracy and the 1 ⊕ 2 split). For every x ∈ 𝕊 — including x that maximally break the S₃ — the six inter-frame coupling norms ‖P_i T_x P_j‖ (i ≠ j) are exactly equal. Consequently the induced 3 × 3 coupling pattern is the S₃-invariant democratic form, whose spectrum is (2, −1, −1): one frame singled out, two exactly degenerate.

Proof. By Theorem 5 the pencil factor of T_x is the rotation generator J₂. For unit pencil lines ℓ_i, ℓ_j at angle θ_ij, |⟨ℓ_i, J₂ ℓ_j⟩| = |sin θ_ij|; the triad spacings are 60° and 120°, and sin 60° = sin 120° = √3/2. Equality of all six couplings follows for every x, with common magnitude (√3/2)‖A₇(x)‖-scaled. The only S₃-invariant off-diagonal 3 × 3 form is democratic, with eigenvalues (2, −1, −1) — the decomposition 3 = 1 ⊕ 2 into trivial plus standard representations. ∎

Measured: the six couplings agree to relative spread 8.0e-16 across 500 random x. We emphasize the order of logic: democracy here is not a texture ansatz, a statistical tendency, or a symmetry assumption about a vacuum — it is a pointwise geometric identity, forced by an antisymmetric form on three lines at equal angular spacing, and it holds before any symmetry breaking. The 1 ⊕ 2 spectral split is therefore not selectable; it is the only pattern the geometry permits at leading order.

Theorem 7 (private octonion hulls; empirical). For generic x ∈ 𝕊, ker T_x is eight-dimensional; it contains e₀, x, e₈, and xe₈; it is closed under multiplication (residual 2.4e-15) and satisfies exact norm composition N(uv) = N(u)N(v) on itself (defect 3.2e-16). Every generic sedenion thus lies inside a private octonion subalgebra — its alternative hull O_x — on which its own dynamics acts trivially, and every hull contains the inert axis.

Status: discovered and certified numerically; a structural proof, and the geometry of the hull bundle x ↦ O_x over 𝕊, are open (§7). We record without pricing the observation that “each state carries a private division algebra containing a universal inert direction” has the architecture of a local frame or clock structure.

5. Physical interpretation, stated separately

Everything above is mathematics, measured or proved. The following paragraph is interpretation, and we flag it as such.

If the three frames of the octonion triad are identified with the three fermion generations, the results assemble into a coherent qualitative picture. The number of generations is three because a closure cubic has three roots, with no fourth frame at any angle and (pending the rung-32 check) no second minting event on the ladder — an answer of the same logical type as the octonionic derivation of third-integral electric charge. The residual family symmetry is an exact S₃ whose breaking chain S₃ → Z₂ (a frame-aligned direction, ψ-protected) reproduces the residual-Z₂ family symmetries long used in neutrino model building. The dynamics available at this level is purely flavor-changing: the alternator generates inter-generation transitions and nothing else, with within-generation evolution remaining octonionic, unitary, and clock-like — an unforced echo of the fact that the one Standard-Model interaction that changes flavor is the one associated with decay. And the leading-order mass pattern is forced: any spectral operator built from the alternator’s couplings splits the families as one singled out plus two degenerate. The democratic mass matrix — postulated by hand in the flavor literature since the 1970s precisely because it yields one heavy and two light families — is here a theorem of pencil geometry. What this picture does not provide, and we state it with equal force: no mass values, no splitting of the degenerate pair (that requires the S₃-breaking corrections, which are the sharpest open computation), no chirality, no spacetime, no gauge dynamics, and no claim that nature factors through this algebra. The claim is narrower and, we believe, more defensible: the qualitative generation structure that the Standard Model presents as arbitrary input — three families, family-changing weak-type dynamics, one anomalously heavy family — is the forced output of the first algebra past the composition boundary.

6. Relation to prior work, with honesty flags

The automorphism group Aut(𝕊) = G₂ × S₃ and the stabilization Der = 𝔤₂ trace to the Cayley–Dickson literature (Eakin–Sathaye; Schafer; cited here from memory — the attribution should be verified before submission). The zero-divisor geometry of 𝕊 and its relation to G₂ is due to Moreno and collaborators; our contribution there is the explicit orbit identification G₂/SU(2) with its SU(2) stabilizer and S³ annihilator fibration, measured. Sedenions as a source of three generations have been proposed by Gillard and Gresnigt (via ℂ ⊗ 𝕊 and ideal constructions); the division-algebra Standard-Model program more broadly is due to Gürsey–Günaydin, Dixon, Furey, Todorov, Dubois-Violette, and Boyle. Democratic mass matrices are a classical flavor-physics ansatz (Fritzsch and successors); S₃ family symmetry with residual Z₂ is a standard model-building lineage. The experimental-mathematics epistemics descend from Borwein–Bailey. What we have not verified against the literature, and flag as possibly known to specialists: the explicit pencil geometry and closure cubic of Theorem 2; the identification of the triad as the alternator’s exact vacuum manifold; the factorization Theorem 5; the forced-democracy Theorem 6; and the private-hull Theorem 7. We claim the assembly — a derivation chain from the composition boundary to a forced democratic generation texture, under experimental discipline — and we welcome correction on any component’s priority.

7. Open problems

The S₃-breaking corrections that split the degenerate pair (the realistic hierarchy lives or dies there); the geometry of the hull bundle x ↦ O_x; the explicit triality identification of the S₃ (vector/spinor⁺/spinor⁻ labels for the three frames); whether the finite automorphism factor stays S₃ at dimension 32 and beyond (the mint as a one-time event — “why not four generations” in its strongest form); the decomposition of the 364-dimensional sedenion deformation shell under G₂ × S₃, hunting multiplicity-three structure; chirality, which is entirely absent at this level; and the transplant of the entire survey to ℂ ⊗ 𝕊, where the existing three-generation proposals live.

Appendix A: TGT protocol

The measurements above were produced under the following discipline, which we propose as the TGT baseline.

Pre-registration: every run is preceded by written predictions, including predictions of failure; results that outrun predictions are reported as such.
Scale-relative tolerances: all rank, nullity, and zero decisions are made against the operator’s own scale, never against absolute thresholds; an early-project audit found absolute tolerances miscounting dim 𝔤₂ in ~9% of random frames.
Known-limit anchors: every instrument is validated on a case with an exactly known answer before being trusted on an unknown (the octonion column of every table serves this role).
Instrument validation on asymmetric cases: during this work, an automorphy test containing a transposed einsum silently certified only symmetric involutions and was blind to order-3 automorphisms; it was caught because two verified automorphisms composed, by the closure axiom, to a map the instrument rejected — an impossibility. We record the general lesson as the transpose class of TGT systematics: validate every instrument on a known non-involutive example, because symmetric test cases can certify a broken instrument.
Contradiction as diagnostic: exact structural laws (group closure, equivariance, sum rules) are run as continuous gates; any violation indicts the instrument before the mathematics.
Machine-zero discipline: residuals at 10⁻¹⁵ on O(1) quantities are logged as candidate theorems and promoted to proofs where possible; Theorems 2–6 were discovered in this order.

Appendix B: summary of measured values

Quantity Measured Exact/expected dim Der(𝕊); rank; Killing 14; 2; negative-definite 𝔤₂ 𝕊 under 𝔤₂ 1 ⊕ 1 ⊕ 7 ⊕ 7, zero block leakage — unit zero-divisor variety single orbit, dim 11; stabilizer 3-dim, closed, nonabelian G₂/SU(2) annihilator; pair variety dim 4; dim 11 + 3 = 14 S³ fiber; dim G₂ composition ratio N(xy)/N(x)N(y) range [0, 1.414214] [0, √2] max spectral radius of L_x (unit x) 1.414214 √2 (Lyapunov ceiling ½ ln 2) Σσ²(L_x), unit x 16.0000 dim (sum rule) monomial automorphisms 2688 = 2 × 1344; none move e₈ or any frame (mono G₂) × ⟨ψ⟩ rotation automorphisms R(θ) θ ∈ {0°, 120°, 240°} only τ order 3 τ³ = I; ψτψ = τ⁻¹ 2.4e-15; 0.0e+0 S₃ triad intersections; e₈ membership dim 1 pairwise; none forced by Thm 2 ‖T_a‖, a in any frame ≤ 5.8e-15 0 (Lemma 3) max ‖T_x e₈‖ 4.5e-16 0 (Lemma 4) factorization residual T = A₇ ⊗ J₂ 3.2e-16; pencil factor [[0,1],[−1,0]] Theorem 5 democracy spread (6 couplings, 500 x) 8.0e-16 0 (Theorem 6) democratic spectrum (2, −1, −1) 3 = 1 ⊕ 2 hull closure; hull composition 2.4e-15; 3.2e-16 Theorem 7 (empirical) Aut-averaged alternator intensity 0 on 1 and sgn; 0.333/dim on 7 ⊗ 2 Schur gate

Reproducibility. All results regenerate from short NumPy scripts against explicit Cayley–Dickson structure tensors (exceptional_algebras_lab.py and the session scripts); no fitted parameters appear anywhere except those declared. The zero-divisor optimization uses Nelder–Mead refinement of structured seeds; everything else is direct linear algebra.

Acknowledgments. This note emerged from an extended human–AI research collaboration conducted under the pre-registration discipline described in Appendix A; the AI collaborator (“Bench”, Claude by Anthropic) performed the computations, committed eight documented self-corrections during the sessions, and drafted this note. The failure-set epigraph was measured twice before it was believed.