The Window's Leak: One Process

The cost is one process, and this paper names it before dividing it into an order and a divisor. The window's leak: the held grain leaks through the strained hold's own structure, into the substrate, at the substrate's rate. Its three parts each have an address.

The rate is the substrate's, and the locus is canon. AP30, Lemma 3: "the substrate's leakage rate is α_em (AP06: the leakage constant)" — "the substrate speaks at α_em" — and in AP30's own summary, α_em "provides both the dynamic maintenance rate and the perturbative expansion parameter." The wind that wants the window shut is the same substrate that presses on every break; it speaks at one rate everywhere. The flip's hold sheds against it at that rate: one factor of α. One voice at every door, file-verified: AP46's shed rate is "one part in α⁻¹⁸ per electron-tick" because, by its own thread ruling, escape means "threading eighteen sequential gates — one α-transparent door each." The door is α everywhere, and each site's structure sets only how the rate assembles: parallel channels multiply it by a count (α × 21, the proton's maintenance — AP30: "each channel contributes α_em"), sequential gates exponentiate it (α⁻¹⁸, the eighteen threaded faces of the cosmic shed), and a held grain multiplies rate by worth (α², the flip). Three assemblies, one base rate. No second base rate exists anywhere in the family.

The depth is the grain's worth. What leaks is the held ε, one grain of α-worth — AP45's own words in the draft in hand: "one ε-grain, α-worth, dimensionless" — the rider unchanged, since a draft's wording is not a lock. The denomination has a canon face — AP30 itself writes "α_em … identified as ε — the leakage constant, the substrate's fundamental coupling strength (AP06)" — and its forcing is carried honestly, file-verified at the canon audit: ε = α_em in the quotient was SELECTED among AP06 section 0.7's domain readings (the AP06 file holds that field open against the canonical corpus-wide ε = α_em, exactly as cited), the selection argued in AP44 section 5 and booked as D-CCC.4 — AP44's own section 9 saying it plainly: "the value is not [derived], and the paper says so" — its closure route graded DERIVED-CONDITIONAL-ON-AP45, the route's target being precisely AP06 section 0.7's held-open field. This paper inherits that conditional and says so. Worth times rate: α².

The geometry belongs to section 5, where the same process is counted through the structure it leaks through. One process; section 4 is its depth, section 5 its shape.

And the cost enters lawfully. It never books: it is a dimensionless fraction of the flip's own booking, which is exactly what a possibility-side cost looks like from the energy face under the family's clause. No reference is carried anywhere — not on the energy face (that would be a booking, double-counted at order one) and not on the possibility face (dimensionless quantities do not compose with mass). The ground exerts its pull for free; the grain is held against it; the substrate's press does the rest.

The one claim sections 4 and 5 share, stated as itself: AP30's maintenance mechanics — the substrate's rate, expression-leakage, normalisation by the leaking structure — read at a strained hold instead of a ground's upkeep. That transfer is the paper's central structural claim. It is not derived from anything deeper here; it is stated, located, and switch-guarded (KS-FLIP.2, KS-30.2), and every grade in this paper rides it in the open.

Grade of the order, worn on the face: DERIVED-AT-THE-STRUCTURAL-REGISTER, riding — the stated transfer above; T-b (balance-impossibility, a registered conjecture); D-CCC.1 (the formalisation); and the ε = α_em closure conditional on AP45's lock.