Four Gears and a Wall A Predictive Machine for the Prime Lattice, and a Proof of Its Boundary

Abstract

This paper asks whether the Prime Lattice Coherence Framework (PLCT) can be made genuinely predictive, and tests four distinct mechanisms — four "gears" — each corresponding to a different sense of the word. Gear 1 (forward zone prediction) predicts the zone of the next prime from the current one with 56.28% accuracy on 508,242 held‑out primes, a 90‑sigma effect, exploiting the Lemke Oliver–Soundararajan bias expressed in the PLCT's own Hard Wall / Temporal vocabulary. Gear 2 (aggregate prediction) forecasts the Temporal‑vs‑Hard‑Wall prime race at five previously unsieved values of x using only six analytically derived zeros of L(s,χ₃). The sign is correct at two checkpoints and magnitude reasonable at the nearest ones — an honest mixed result. Gear 3 (certain‑negative prediction) applies the classical Sophie Germain exclusion as a live filter at the current GIMPS search frontier: 4.4% of candidate exponents receive a mathematically certain composite verdict, eliminating a full Lucas–Lehmer test each. Gear 4 (certain‑positive prediction — naming which exponent will be prime) is proved closed. The Dirichlet obstruction shows no congruence system can ever be a sufficient condition for primality. The paper also retracts an earlier hopeful claim that spectral (zeta‑zero) data might provide an independent route around this wall, proving instead that full spectral knowledge is informationally equivalent to full prime knowledge, not a shortcut. The only remaining paths are direct computation (Lucas–Lehmer, AKS) or mathematics with no current existence