GATE D24
Early cuts on reachable states
What changed
D23 proved that its early cuts are safe and showed, on a bounded pilot, that they prune strongly before reconstruction. The pilot numbers, every sampled candidate and every sampled cover removed, are measurements on a sample, not a statement about the whole search. D24 tests the proved cuts where it actually matters: on the real states the global master reaches, rather than on random candidates. The engine carries a single shared clock, one global residue pattern modulo eighty-four, the least common multiple of the level fourteen, twenty-eight and forty-two grids, and it never re-chooses that pattern separately at each prime, so the level constraints stay one common global object. On this real reachable-state stream it measures how much the proved cuts prune, the branching they introduce, the certificate cost and the projected campaign cost. No larger compute budget is authorized, and nothing about the conjecture is decided by opening this gate. LRC(13) remains open.
Evidence
Verification: double · I2 · 8 artifacts (reports, verifiers, corruption suites, manifest with SHA-256)