# n120 — the push: three barrels fired at the topological branch
# (2026-07-11; the instruction was "push the exercise" — so the branch
#  got the full treatment: its founding conjecture tested, its best
#  structural claim tested quantitatively, and one of its own readings
#  killed. Script: n120_push.py + the B-redo fits.)

## BARREL A — the audit certificate: PASS (the founding conjecture holds)
The winding state's violations vanish IDENTICALLY: max|div| = 8e-16,
max|curl| = 3e-17 — and q = -1 is the same zeros. Therefore a
superposition |+1> + |-1> has Gamma_audit = 0 EXACTLY: THE MONITOR
CANNOT SEE THE CLASS. Matter is audit-dark, as conjectured in n115.
But a probe coupling to raw FLUXES decoheres it at
Gamma_flux = gamma * q^T M q (= 0.578 gamma on the test lattice):
CHARGE SUPERSELECTION IS RE-DERIVED — enforced by the particle's own
far field (matter watches matter), NOT by the vacuum. The framework
needed superselection (observed) and indestructibility (the audit
cannot erase a class) simultaneously; the branch delivers both,
through different channels. Composition, not tension.

## BARREL B — mouths are monopoles: PASS after the method fought back
First shots FAILED (pure power-law fits: exponents -1.38, -2.60) —
the n112 lesson applied: the second mouth's field + periodic images
enter as a constant offset that a log-log fit reads as a steep slope.
Two-parameter fits (a/r + b), box 48, mouths 24 apart:
  B1: phi(r) = +0.0869/r - 0.0055, residual 1.1%
      (continuum point charge: 1/4pi = 0.0796 — 9%, lattice-small-r)
  B2: R(D) = 0.504 - 0.154/D, residual 2.0%
      (continuum Coulomb: 2/4pi = 0.159 — 3.2% agreement)
A MOUTH OF A LONG HANDLE IS A POINT CHARGE; OPPOSITE MOUTHS BIND BY
COULOMB. Wheeler's charge-without-charge (1957) runs, quantitatively,
in this framework's own statics. Consequences:
- CHARGE CONSERVATION IS TOPOLOGICAL: mouths come in pairs; there is
  no way to make one.
- PAIR ANNIHILATION = HANDLE CONTRACTION: e+e- were always the two
  ends of the same object; annihilation is the object shortening
  until it vanishes. n115's K2 (the annihilation kill) partially
  DISSOLVES: pair events need no surgery. Only SINGLE-particle
  destruction needs surgery — and the cold vacuum cannot do surgery.
- 2D handles (n116) showed dipoles because their mouths were close:
  the dipole was two monopoles unresolved. Consistent, in hindsight.

## BARREL C — the internal kill: headstone 30
If generations were excitations of one handle, the first-excitation
ratio is BOUNDED: m1/m0 = sin(3pi/2l)/sin(pi/2l) in [2.000, 3.000]
for every l >= 3 (computed, the whole table). Measured mu/e =
206.768: OUTSIDE, by two orders. GENERATIONS ARE NOT HANDLE
EXCITATIONS — the reading is dead on contact with the muon. Buried
as headstone 30. What survives: generations as different handle
TYPES (circumference, twist number, linking) — unpriced, honest.
The excited tower (2-3x the ground state) remains a PREDICTION of
the branch: every twisted handle has its first internal resonance
below 3m. Where that shows up in nature, if anywhere, is open.

## What the push forces onto the record
PREDICTION 25 — THE PROTON NEVER DECAYS. If baryon number is a
topological class, single-particle decay requires surgery, and the
cold vacuum cannot perform it (surgery temperature = the n117
baryogenesis epoch; pair events are contractions and conserve the
class). Proton-decay searches keep returning null — Super-K,
Hyper-K, forever — in sharp opposition to GUTs, which need decay.
KILL: one confirmed proton decay. (Symmetric bet: frozen mu vs
GUT-covariation, n94 — the framework keeps betting against grand
unification and keeps being specific about it.)

## Balance
The founding conjecture certified (audit-dark matter, superselection
from the field). The branch's boldest structural claim confirmed
quantitatively (Coulomb mouths, 3%). One reading killed by the muon
(headstone 30). One forever-prediction registered (25). The method
fought back once and was caught by the n112 lesson. The branch is
now load-bearing: it holds charge conservation, annihilation,
superselection, and indestructibility in one topological hand.