AP47 — The Flip, locked

What's being priced

The neutron is not the proton plus a thing — same three couplings, nothing added. One of the three simply sits in the other orientation: not an object held, but a distinction held open against the ground state's settled balance. Holding that distinction is tension, and tension sheds. AP47's claim is that this held distinction costs a continuous second-order tax on the booking of the difference itself, paid at what the paper calls the window's leak — and that the tax is exactly Δ_real = Δ_bare/(1 + α²/8π), a quotient on the ideal gap, never a new energy term. No flip, no cost: the proton, holding nothing open, is excluded by definition, not by a check.

The frozen row (once locked)

Δ_real = Δ_bare/(1 + α²/8π) = 2.53098857035 at CODATA-2022 α, against the measured Δ_meas = 2.530988574 ± 7.4×10⁻⁷ (0.29 ppm) — a residual of −0.005σ.

Shown beside it, forever, never repaired: the original bare-difference prediction — no hold-cost correction — fired dead at 7.24σ on 2 August 2026 (KS-NPP.1). Pricing the missing structure did not revive that row. It produced a new row beside the corpse; nothing got fixed.

The same cost applied to the ground-state proton — which holds no distinction and so should reject it — moves the proton by 121,576σ at the named bar. (The origin note's first pass read 223σ against an unnamed coarser bar; the errata's corrected figure only strengthens the elimination.)

The elimination

Required divisor 25.1499; window at the audited bar (7.4×10⁻⁷ mₑ): [22.096, 29.183].

Support 2 is the only survivor: the flip's structure is an orientation pair (support 2) expressed through all four spacetime dimensions under AP30's own leakage-normalisation construction, giving 2 × 4 = 8 dimensional expressions and a divisor of 8π. Everything else either fails the arithmetic window outright, undercounts the holding dimension (the three-dimensional reading), or has no construction behind it at all (9π — no half-circle warrant, no support of eighteen).

Dependency chain — what this leans on

• AP30 (The Resistance) — the parent construction and its leakage lemma, published on the420code.org. KS-30.2 (the isotropy assumption) is shared, not re-derived.
• AP44 (The Snap) — locked — source of the ε = α_em selection (D-CCC.4). Its state of record is on this site: /prereg/ap44/.
• AP46 (The Stretch) — locked 2 September 2026 at composite 9.0. Source of the “holding is tension, and tension sheds” mechanism and the pairs-balance ruling. State of record: /prereg/ap46/.
• AP45 (The Blink) — locked 2 September 2026 at composite 9.0. The ε = α_em closure rides this lock. State of record: /prereg/ap45/.
• AP06 (The Leakage Constant) — the published corpus, the domain field the ε selection was made among.

AP45 and AP46 locked 2 September 2026 on the420code-proof (tag ap45-ap46-locked-2026-09-02). AP47 itself locked 3 September 2026 (tag ap47-locked-2026-09-03).

Kill switches

Nothing on this table is resolved. Five are live surface; KS-NPP.1 is a corpse kept on the board, not a switch waiting to fire again.

Impact of the derivation

AP47 prices a held distinction — the neutron is the proton with one coupling flipped, not a proton plus a thing. The cost is a quotient on the ideal gap, never a new energy term: Δ_real = Δ_bare / (1 + α²/8π). Same α, no fitted parameter. The landing is not the claim.

G grades −0.005σ as weak evidence by the family's own standard. The paper is worth having for the structure and the kill surface, not for the seventh decimal. KS-NPP.1 stays FIRED. The new row stands beside the corpse; nothing got fixed.

The other three locked costs either predict no number (Blink), wait on metrology's migration (Snap), or freeze a coarse window (Stretch, ±0.66 Gyr). The Flip is the one that meets a seventh-decimal measurement and either lives or dies there. That is the impact: a nature-facing row at 0.29 ppm, killable by the next mass-difference tightening, with the original miss kept in public so the landing cannot be mistaken for a patch.

Deuteron — standing record

The deuteron as test bed was withdrawn on 10 August after 12.72 against predictions of 1, 2 and 6. AP47 §8 does not revive it. Binding is ruled; the mass face is a declared null. The author remains free to return, on his clock.

8 — Binding: the Ruling

Source: AP47_The_Flip_FINAL_v3.md · 1 September 2026 · §8

A bound neutron is still udd. The orientation is intact; nothing in the nuclear well touches the flip. The ruling, spoken and standing: the distinction is held against the structural alternative; binding closes the exit, not the window. The ε is a structural residue, not a dynamical option — availability in the ledger is structural, and reachability is the exit's business. The bound neutron pays. Its stability needs no second mechanism: the release is blocked (for the deuteron, energetically — there is no bound diproton; Pauli blocking proper does the work in larger nuclei), and a blocked exit over an open window is a hold that sheds forever, which the family already knows how to say.

Proposition (the capture-chain identity). The gamma plus recoil in n + p → d + γ equals the free neutron's booking plus the proton's minus the deuteron's, identically, by energy conservation, whatever happens inside. Therefore the tabulated neutron mass is the free booking always; the frozen row of section 6 tests the free neutron regardless of this ruling; an independent free-neutron mass would match the table either way; and a closed window's 2.7 eV would hide inside the deuteron's binding bookkeeping — B_d = 2,224,566.362 ± 0.373 eV from the CODATA-2022 mass triplet — where theory is a thousand times too coarse to look.

Declared plainly, because honesty counts edges rather than collecting them: this ruling carries no mass-face observable, under either answer. No switch is armed on it because no measurement can reach it. The only conceivable future handle is dynamics — the lifetime physics of the window's snap — and that is forward-register, held open in section 0.7.

KS-FLIP.3 — the truncation. ARMED. The construction terminates at second order: no α³ term exists. The absent third-order piece sits at 3.9×10⁻⁸ — 5.3% of the audited bar, 0.020 eV — so adjudication waits on roughly twentyfold better metrology on the deuteron binding energy (B_d currently ±0.373 eV, CODATA 2022). The claim's silence at third order is itself a prediction, and nature will eventually rule on it.

ERRATA item 5 (ERRATA_FLIP_2026-09-01.md): Section 7's bound-neutron edge ("an order of magnitude and more below current binding-energy precision on the deuteron") is wrong twice — the bar direction, and the observable itself. Superseded by AP47 section 8.

T3 — The deuteron wall (WP-BUMP.1 §6): The 12.72 scaling ratio killed every accumulating per-configuration cost. Transience is the shield: the bump must provably never book, or KS-CCC.2 fires. Deferred with the deuteron programme by the author's ruling — months out, on his clock, nobody else's.

PROV-CCC.7 §4: he is not done — the neutron, the proton, the deuteron, and whatever else he chooses remain open to him without condition. The corpus's switches bind the corpus's claims.

Anchor status

Locked. AP47 is anchored on the420code-proof at tag ap47-locked-2026-09-03. A receipt/provenance anchor (ap47-freeze-anchor-2026-09-03) was made first, on the same day, timestamping the bundle at receipt without asserting a lock — the same two-step pattern AP44 used. The lock itself followed once the author confirmed directly, in his own word, that he stood behind it; Studio G's own attestation of that walk was checked, not simply trusted, before the lock anchor was made. File-level digests the anchor covers:

• AP47_The_Flip_FINAL_v3.md — 6e7fe6486a16…
• neutron.docx — 2eba753397e9…
• ERRATA_FLIP_2026-09-01.md — 4688019d570e…

Every check below has run — as code, right here, no clone or dependency needed, and again as part of an independent canon audit before the lock:

It reproduces Δ_bare, Δ_real, the firing (still FIRED — this script doesn't repair it), the full elimination table, and the proton check, from CODATA 2022 alone.

• Δ_bare and Δ_real recomputed at CODATA-2022 α, matching 2.53098857035.
• The firing recomputed: (Δ_bare − Δ_meas)/u = 7.24σ.
• The elimination window [22.096, 29.183] and the support pinch [1.76, 2.32] checked.
• The proton elimination recomputed: ≈1.2×10⁵σ.
• The chain order verified — words before arithmetic — across neutron.docx, the draft lineage, and the audits.
• The paper's canon quotations cross-checked against AP30/AP06 (published) and AP44/AP45/AP46 (all locked) — clean, with one citation-hygiene note on two AP46 attributions, logged in CANON_AUDIT_v1_3_2026-09-03.md on the proof repo.