A fully deterministic scheduler running on GPU by expressing the entire control logic as tensor ops scheduler that runs like a tiny ML model.Turning a branch-heavy OS scheduler into a static GPU compute graph (program-as-weights experiment).