Questions, hypotheses, methods, baselines and limitations—not promotional conclusions. Long-running records may remain active for weeks or months while data, replications and hardware evidence accumulate.
A repeatable statistical signal survived audit, but its implementation did not pass the preregistered operational cost gate.
A preregistered early-prediction model performed at random discrimination on held-out data.
One million checks confirmed exact modular relations around the W3-dependent state.
An exact R56 primitive decided part of a B8 target using 392 of 512 semantic bits under its model.
More than 120 billion hashes were evaluated; no candidate survived the frozen grammar.
A stock Bitmain Antminer S9 is the real-hardware validation target for compatible research strategies.