The Azimuthal-Average Test, Executed: The Registered Profile Prediction for the CMB Axis Fails on the Planck Sky

Verdict paper for the azimuthal-average test registered in DOI 10.5281/zenodo.24102852, executed exactly as registered on the public Planck PR3 data. The registered prediction fixed the azimuthal profile about (l, b) = (-100, 60) degrees as -B2 [P2 + 2.138 P3], with extrema at 0, 68.69, 123.40 and 180 degrees in depth ratios 1 : 0.386 : 0.264 : 0.363 and an amplitude cap B2 <= 21.69 microK. Measured with the registered statistic (5-degree annuli, single-amplitude template fit, 10,000-axis rotated null) on all four PR3 component-separated maps: B2hat = +5.79 microK (SMICA) and +5.79 to +5.84 microK across the four maps, against a null standard deviation of 6.13 microK - consistent with zero at 95% in every map, so by the registration's own P3 the test is reported as null. No companion extremum is present on the registration's binning in any map (P1 fails); the anti-axis cap is cold (-24 microK) where the profile requires hot; the measured cold-ring annulus (-5.1 microK) sits nearer the rival localised reading's smeared prediction (-3.9 microK) than the profile's (-17.7 microK); confirmation (P5) fails in all four maps. Verdict: the registered profile prediction fails on the Planck sky. The paper also corrects the record: the registration described the measurement as not yet performed and did not cite the F-CS1 result note (DOI 10.5281/zenodo.24053823, 2026-11-08), which had already measured the azimuthal average about the same axis on the same primary map and returned a failing verdict; this execution reproduces F-CS1's numbers to the printed digits and completes the registered protocol. Scope: what dies is the frame's early-universe application as registered (the l = 2, 3 founding-profile account of the Cold Spot and the large-angle profile); the Aldred Limit as a mathematical statement is not executed by this test; F-CS2 and F-CS3 remain ungraded. The alternate-mask robustness variant returns the same verdict class; the alternate-release (PR4) check could not be run because the Planck Legacy Archive download endpoint was unavailable throughout the run window, and that is stated in the paper. AI use disclosure: this work was produced with Dave, an AI research system; the computation was executed and graded by Dave under the author's direction, and every number is reproducible from the deposited code and the public Planck data.