[G] The context: gauge-coupling unification reduces (exact one-loop group theory [E]) to a single rational ratio B = (b₂−b₃)/(b₁−b₂) of beta-function coefficients. B is invariant under complete multiplets and moves only under incomplete weak-charged pieces. The Standard Model gives B = 0.528; the observed unification point wants ≈ 0.717. Moving it requires weak-charged incomplete multiplets — i.e., the menu is essentially superpartners.
The arrow in question: Akataleptos construction → a list of SU(3)×SU(2)×U(1) multiplets. Everything the construction natively supplies (the quotient topology, the retention ratios, b₁ growth, flat connections, the 48-element symmetry group) is not, and does not obviously enumerate, particle multiplets with spin types. The connector has to be invented — and any invented connector is where retrofitting sneaks in. Protocol discipline (write the mapping before scoring, hash it, count free choices, report every attempt) is the guard.
What the gap now has to accommodate [F]: bar-cocycles-incomplete-basis
established that the construction's H¹ is mostly complement flux — bulk-threading
classes the bar menu never touched (1190 of 1409 at K=3). The construction's
native supply of "incomplete pieces" is abundant but unmapped. A particle-content
mapping must first say which classes map to what, and why those.
Pre-registered proposal [S], not a result: the complement classes are the candidate "missing gears." The sign of the B-ratio miss (ΔB = +0.19, weak-charged incomplete pieces) is the information; a hit from the MSSM menu would only reproduce 1981, not predict.
Discriminator: any mapping with ≤2 free choices whose output is not MSSM-shaped would be evidence the connector is real.