Planck 2018 measures the primordial tilt at n_s = 0.9649 +/- 0.0042; ACT DR6 measures 0.9743 +/- 0.0034, about 1.7σ higher, and joint fits with DESI BAO push toward 0.974 to 0.977. The two flagship CMB experiments disagree on the cornerstone parameter of early-universe physics, and each shift reorders the inflationary model landscape, from Starobinsky-favored to attractor-disfavored to renewed talk of n_s = 1.
In ΛCDM the spectral index is a free input, set by the shape of an undetermined inflaton potential. The model carries no expectation for where n_s should land, so it cannot say which measurement is wrong, cannot flag either value as surprising, and silently re-ranks an entire literature of potentials every time the central value moves.
SCT replaces the hot-dense-center with a superluminal collision cascade, and with it the spectral tilt stops being a free input. Successive collision stages spanning a finite range of scales produce a power spectrum with index n_s = 1 - 1/L, where L is the number of hierarchical nesting levels the cascade traverses (P36, P37, P38). The red tilt is a boundary correction from the finite dynamic range of a truncated geometric series: a counting of levels, not the shape of a field potential.
Read through this formula, the two measurements are statements about cascade depth: Planck's 0.965 corresponds to L of 29, ACT's 0.974 to L of 38, and the all-data 0.977 to L of 43. The independently estimated plausible band is L of 20 to 40, so both flagship values fall inside the same physical window, with the all-data value just beyond its edge, and the dispute shifts from which potential is correct to how deep the hierarchy runs. SCT requires only that n_s sit strictly below 1 by an amount 1/L for a physically reasonable level count; it does not need the two experiments to agree before the systematic floor is found.
The same cascade that sets the tilt sets the rest of the primordial sector: near-zero non-Gaussianity from central-limit averaging over thousands of collision events, isocurvature suppressed by complete thermalization, and no inflationary gravitational-wave background (r below 10⁻⁵). One mechanism, several testable corollaries, no inflaton.
This resolution is developed in Paper 4, From Chaos To Concordance Spectra.
The registered kill: a precise measurement of n_s inconsistent with n_s = 1 - 1/L for any physically plausible L of 20 to 40, that is, n_s below 0.950 or above 0.980 at 5σ, falsifies the cascade-tilt derivation. A confirmed exact Harrison-Zeldovich spectrum (n_s = 1) would do the same, since it requires infinite cascade depth. CMB-S4 reaches the needed precision.