This repository documents a reproducible protocol for testing whether AI reasoning remains structurally stable under repetition. The focus is not on answers, accuracy, or model internals, but on the ...