Dated notes

7 September 2026. Five matters of record since the lock. None alters the body above; the body is left as written.

1. KS-NPP.1 fired on 2 August 2026 at 7.24σ: predicted 2.53099393 mₑ against 2.530988574(74) mₑ, an offset of 2.12 ppm on a 0.288 ppm bar (Master Kill Switch Registry v5.27). No repair is offered; the required correction is negative and every term in the decomposition is positive. The switch stands on the register as fired; AP47 (The Flip) carries the realised structural limb, KS-FLIP.1, frozen beside it. §0.5, the Kill Switches section and the Conditionality Footer read “all live” as of the lock.

2. KS-30.3 is paid, 6 September 2026, by AP49 (The Hold) §4 and §7: the chain of holders is ruled, every order follows as the previous times r = 16α/1836, and the series converges by construction. The sixteen carried in §6 is read there as 4 × 4, the four conditions through the four dimensional expressions (KS-HOLD.2). AP49’s closed form, 1836.152673445, supersedes §6’s truncation by reference, beside it and not in its place.

3. §6 quotes 1836.15267343(11) under the label CODATA 2022. That value is the 2018 adjustment. CODATA 2022 gives 1836.152673426(32). Against it the O(α) result stands at 5.30 parts per billion, as the title states; the α²-truncated value, 1836.152673444, sits 0.57σ above the measurement (10 parts per trillion); the residual of “0.008 ppb” stated in §6 is against the 2018 line.

4. KS-NPP.3 is retired as subsumed, by ruling of 7 September 2026, for the registry’s next build. The switch names the fired value; consistency of a dead number tests nothing. Re-pointed at the realised value, it still cannot fire: lattice QCD resolves m_n − m_p to a few tenths of an MeV on 1.29 MeV (1.51 MeV with an accuracy of 0.3 MeV in the 2015 ab initio computation; 1.73(69) MeV in a 2022 calculation), so consistency with any ppm-level candidate is automatic. A switch that cannot fire is not a switch. The sign is KS-NPP.2’s; the magnitude is the measurement’s.

5. §8 states that the numbers 21, 3 and 4 all originate in AP10. AP10 derives the three faces and the four dimensions and does not carry the channel count; the twenty-one is AP28’s, as §0.2 states.

7 September 2026. Seven further matters of record, from the first objections filed against this paper. None alters the body above.

6. §3's sentence “This maps exactly to U(3): 8 gluon generators + 1 colour singlet required for total confinement” is withdrawn. QCD is SU(3); the identification was wrong. The exchange-matrix count, 3 × 3 = 9, stands on its own argument.

7. §6 and §7 say each successive order is smaller by roughly α. AP49's chain gives the ratio as r = 16α/1836, about α/115; the sentence was wrong by two orders of magnitude and is superseded by AP49 §4.

8. §7 reads the residual after the O(α) term as ε itself, “not a missing term”. AP49 supplied the terms and summed the series; §7's reading is superseded by AP49, and the residual after all orders, −0.59σ, is claimed to mean nothing until it does.

9. §5's stated premises — 84 dimensional expressions, each leaking into 4π steradians — give 1/(336π), not the 1/(84π) the page prints; and 21α/(84π) = α/(4π) exactly, so the channel count cancels against the divisor. The first-order term stands as carried, at 5.30 parts per billion; §5's structural narrative for the divisor is withdrawn. The leakage lemma's form is the multiplier, (1 − x), one form everywhere in the corpus: a returned leak is not an expression and does not itself leak.

10. Lemma 1's ground “trace to different axiom subsets” is contradicted by §0.3, which traces Layers 2 and 3 both to S; that ground is withdrawn. The lemma stands on its other stated ground, the independence of the three degrees of freedom — capacity, locking, exchange — which is what KS-30.1 registers. R and C are absent from the static 1836 because the static sum is the record configured — B and S alone; R is the record held, and its cost is the dynamic series and the chain of holders (AP49); C, as §0.3 places it, is priced in the dynamic terms too. A record holds whether it is read now or ignored now: a keyboard is a configuration of records with a history, and in a furnace the records are held in a different configuration, written forward onto the same history. Reconfiguration is not unwriting.

11. The electron is the unit. Every ratio formula in this paper counts resistance above the minimal break, in units of it; the counting rule applies to composite records and is not applied to the unit — the knot does not count itself. The now, the actualisation state, does not hold; it is from where the hold is read. This was implied in §1 and is now stated.

12. Dated note 4 is corrected: KS-NPP.3 is not retired. It tests the total against lattice QCD, and the total is consistent. The question it does not test — the corpus's α-dependence against the lattice QCD+QED decomposition of the neutron–proton difference (QCD +2.52 MeV, QED −1.00 MeV) and the proton's electromagnetic self-energy, where the corpus's terms are +4 keV and +78 keV — is booked as a declared debt, D-EM.1, and a registered switch, KS-30.5: the corpus owes the statement of what its α is relative to the electromagnetic coupling. Nothing fires on the totals.