Ø Frozen Predictions · the document of record
The 420 Code · Studio G · the frozen file is never edited; corrections supersede, dated, in the open
RIGIDITY.md was frozen on 2 August 2026. Four of its statements have been overtaken by events or found wrong since. The document is not edited. This note stands beside it.
Its argument is untouched by all four: 21 is built from the face count and the dimension count, both fixed upstream and before any number was computed; the exponential in G closes the window to a single integer; and rigidity is not uniqueness. verify_rigidity.py reproduces the table on demand and its figures are correct as printed.
"G lands at 0.69 % — about 31σ against the measurement's formal bar…"
That is wrong by a factor of ten. CODATA 2022 gives G = 6.67430(15) × 10⁻¹¹, a bar of 1.5 × 10⁻¹⁵. AP28's structural value 6.7206 × 10⁻¹¹ sits 4.63 × 10⁻¹³ above it:
4.63 × 10⁻¹³ / 1.5 × 10⁻¹⁵ = 309σ
The correct figure is 309σ, as the scoreboard at /prereg/ prints it. The document's "31σ" implies a bar ten times CODATA's, which nothing supports.
The bullet's argument survives the correction and is strengthened by it: the point was that a fitter would have done better than 0.69 %, and 309σ makes that sharper, not softer. The neighbouring claim in the same bullet — that the gap is "roughly 14 times the historical scatter between G experiments" — is right at the order (6937 ppm against a lab-to-lab spread of up to 540 ppm, about 13×).
The bullet rests on the 0.69 % never having been touched. On 11 August 2026 AP44 (The Snap) priced the arena's once-held ε and realised the coupling as α²¹(1 + 1/π)/(1 + α), giving 6.6719 × 10⁻¹¹, −0.036 % against CODATA. AP28's value stands beside it as the structural value and the record of the omission; neither is edited.
What this does to the argument: the correction did not come from tuning against G. It came from a structural omission exposed when the neutron's bare row fired at 7.24σ on 2 August — the same omission, paid at three sites (AP44, AP47, AP48), moving three derivations toward measurement at once. A framework built by fitting has no mechanism that produces a correction of that shape. The bullet's conclusion holds; its evidence is now the correction rather than its absence.
Read the realised value against the laboratory measurements rather than the adjustment and the picture the document was pointing at becomes explicit: the atom-interferometry value (LENS 2014, 6.67191(99) × 10⁻¹¹) sits 1 ppm from the realised value, while CODATA's adjusted mean sits 360 ppm away. Both readings are on the scoreboard; the switch on G, KS-CCC.3, tests the direction the consensus moves, not the magnitude against CODATA.
Both sections list KS-45.1 (H₀ = 74.3 ± 1.2, from the floor inverted) among the forward predictions that would carry the weight the retrospective numbers cannot. KS-45.1 fired on 3 September 2026, on the corpus's own derived rate, as AP46 §9 had pre-registered — 5.7σ on the width it registered. It is shown as a corpse and never repaired (ERRATUM-2026-09-03-H0-KS45.1.md).
This is not a loss to §5's argument; it is §5's argument executing. It is, however, a change to the list: of the forward predictions §4(b) names, the proton at order α³, KS-42.6 and KS-41.1 remain live, KS-45.1 is dead, and four new registrations were frozen at the wave (KS-ASM.1, KS-STRETCH.3, KS-FLIP.1, KS-CCC.3).
"H₀ converging above or below 71.9"
That threshold belonged to KS-45.1, which has fired. The corpus's expansion rate is now AP48's closure: 67.45 km/s/Mpc, window 64.4–70.8, carried by KS-ASM.1, which dies if the settled rate lands outside that window in either direction — including if the Cepheid ladder's 73 is confirmed as the expansion rate.
Frozen files are never edited; corrections supersede, dated, in the open. Copyleft 2026. Don't be a cunt. Be kind.
← back to Ø Frozen Predictions · Copyleft 2026. Don't be a cunt. Be kind.