The 420 Code · Prediction Freeze

AP49 — The Hold, locked

Locked 2026-09-06, at the author's own direct confirmation — not merely delivered. Pays AP30's KS-30.3: the sixteen read, the series closed to all orders, the proton predicted a second time, sharper.

Locked — 2026-09-06

This is the state of record for Artist's Proof 49, The Hold. It was verified independently (manifest, a full re-derivation of the chain and both kill-switch thresholds from CODATA 2022 alone) before being locked, and locked only once the author confirmed directly — not merely delivered — that he stood behind it. Anchored on the420code-proof at tag ap49-locked-2026-09-06. Kill switches KS-HOLD.1–3 are armed from this lock; AP30's KS-30.3 is paid.

AP30 counted the proton standing still, and the cost of keeping it, and the cost of keeping that cost — and stopped, with a coefficient of sixteen it couldn't explain. This paper reads the sixteen (the four conditions of the record's repair, held through the four dimensions) and notices the keeping never ends: each keeping is kept by the next, at the same small share, forever. Summed in closed form, that is the proton's weight to the digit the best measurement has — with the next digit predicted, and its sign, before any measurement that can see it.

What's being closed

AP30 derived m_p/mₑ = 1836 + 21α·(1 − 1/(84π)) + α²·21·16/1836 + O(α³) — the static structure, the repair, and a second-order term whose coefficient, sixteen, was pinned by the data (15.97 ± 0.05) but never derived. AP30 owed that derivation as KS-30.3. This paper pays it two ways: the sixteen reads as 4 × 4 — the record's four conditions, each held through the four dimensional expressions, one rule already used for AP30's own 84 (the proton's 21 channels × 4) and AP47's 8 (the neutron's pair × 4) — and the series, once that reading fixes the ratio, closes at every order.

The chain, closed

16α/1836r, the repair's share of the whole — 6.359×10⁻⁵
1836.152673445D_closed, all orders summed
−0.59σvs. CODATA 2022 (−10 parts per trillion)
+6.2×10⁻¹⁰the third order, sign given blind
m_p/mₑ = 1836 + 21α·(1 − 1/(84π)) + 21α·r/(1 − r), r = 16α/1836
= 1836 + 0.152663699 + 9.7459657×10⁻⁶ = 1836.152673444951

The rule — every order beyond the first is the previous order times r — was given by the author before any arithmetic was shown to him (§4). It closes AP30's open O(α³) as a geometric tail, 21α·r/(1 − r), which is what pays KS-30.3's convergence claim. Against CODATA 2022 (1836.152673426(32)): residual −1.895×10⁻⁸, −0.59σ. Against the independent H₂⁺ value (Alighanbari, Schenkel, Korobov, Schiller, Nature 644, 69, 2025): −3.10×10⁻⁸, −0.66σ. The chain adds at every order and has no negative term — it sits 6×10⁻¹⁰ further from the measurement than the truncated D of 2 August did (0.57σ → 0.59σ), and does not try to close the present gap.

Supersedes, by reference only, the corpus's own pre-registration of 2 August 2026 (digest 6ea78dac…): D = 1836.152673444331 stands beside D_closed, never in its place. The third-order coefficient, c₃ = 21·16²/1836² = +1.595×10⁻³, is stated with its sign positive, in advance of any measurement that can reach it — the pre-registration's own constraint on c₃ is −0.047 ± 0.082, which holds the registered value at 0.6σ.

The reading, and what the neutron rules out

From {21, 3, 4}, sixteen can only be 4 × 4 — the corpus's own grammar for a second-order support count is an object through the four dimensional expressions, already used twice: 21 × 4 = 84 (AP30, the proton's channels) and 2 × 4 = 8 (AP47, the neutron's pair). The hold's four conditions through the same four dimensions gives 4 × 4 = 16 — a share of the whole, not a leakage divisor like the 84 and the 8. A second reading gives the same sixteen for the proton (the four conditions closing on each other, 4 × 4 as a grid) but predicts 2 × 2 = 4 at the neutron, where AP47 already locked 8 — dead on canon. The author refused it independently, on the axiom: a record does not close on itself; held, it can be reconfigured; closed, the universe stops.

Kill switches

SwitchClaimStatus
KS-HOLD.1The closed form: D_closed = 1836.152673444951, c₃ = +1.595×10⁻³ sign stated in advanceLIVE — EMPIRICAL
KS-HOLD.2The reading: the sixteen is 4×4, one rule with AP30's 84 and AP47's 8LIVE — STRUCTURAL
KS-HOLD.3The base: the chain holds against the bare coupling 21α, not the repair net of its own leakLIVE — EMPIRICAL
KS-30.3 (AP30)The series converges; the explicit higher-order forms existPAID · structural register

KS-HOLD.1 dies at 3σ once the measurement's uncertainty reaches 3.4 parts per trillion with the central value unchanged (about five times today's 17.4 ppt precision), 5σ at 2.1 ppt; the third-order term itself is resolved at about 52 times today's. KS-HOLD.3 fires if the measurement settles on the leaked base, 1836.152673408, 3.7×10⁻⁸ below D_closed — the two are twelve sigma apart at 1.7 ppt precision, about seven at 3 ppt; today's measurement sits between them and decides nothing yet.

What this paper does not claim

  • Does not claim the sixteen was derived before it was known — it was carried in AP30 from the data; its reading was found on 6 September 2026, after, and the paper says so.
  • Does not claim the third-order term is testable now. It is registered, with its sign, for a measurement that does not yet exist.
  • Does not claim the present −0.59σ residual means anything — it is noise until it isn't.
  • Does not claim the chain's rule governs any site but the proton's series — the neutron's and G's higher orders are held open, not claimed.
  • Does not touch AP30's static layers, its first-order term, its isotropic-leakage lemma, or α's value; does not repair anything fired.

Run it yourself

Reproduces the repair term, r, every order of the chain, the closed-form tail, D_closed, c₃, the residual and σ against both CODATA 2022 and the independent H₂⁺ value, and both kill-switch thresholds — from CODATA 2022 alone, stdlib only, no network.