Pre-Registered Falsifiable Predictions — Information Tension / Geometrodynamics (v2, corrected)

Abstract

Pre-registered falsifiable predictions, corrected edition (v2). Filed for priority timestamp. Three predictions are registered: a CMB first-acoustic-peak shift (delta_ell_1 = -0.16 vs LCDM), a critical enstrophy/Reynolds ratio (Re_c = 1.42), and a cognitive-drift invariant (tau = kappa = 0.9539392). Each carries an explicit falsification window and an honest caveat stating what is derived and what is chosen. Two predictions from the unfiled v1 draft are WITHDRAWN and listed with reasons rather than deleted: a JWST high-redshift mass offset whose observable (T_bar, 'normalized units') was never operationally defined anywhere in the source corpus and whose sign was ambiguous; and a quantum spectral gap (0.21875) with no derivation and no definition of its natural units, blocked upstream by the unconstructed operator it refers to. A prediction whose observable cannot be measured is neither confirmable nor falsifiable. Not registered as a prediction: the acceleration scale a_0 = cH_0/2pi. It was tested against SPARC before this filing and is therefore a retrodiction; registering it would be backdating. The accompanying blind-sweep script records the result that a_0 = 2cH_0 is excluded by roughly an order of magnitude under every statistic tested, while the optimum lies within a factor 0.47-0.98 of cH_0/2pi depending on the statistic minimised. AI systems assisted drafting and computation under the author's direction; they are not authors and performed no peer review. No AI-generated verdict appears as evidence.