nx_deltaclass_lib.nx

nx_deltaclass_lib.nx -- WHY a headline ratio moved: SCOPE vs EXECUTION, decomposed EXACTLY. THE DEBT THIS EATS: the ecomat headline went 500 -> 418 permil over two weeks and READ AS DECAY. It was not decay. Delivery rose 34 -> 51 levels (+250 permil at a FIXED bar) while the bar itself rose 68 -> 122 (-332 permil of pure requirement growth). Net -82. Nobody could tell those apart without deriving it by hand, so a marathon had no defined start and no defined end. THE DECOMPOSITION IS EXACT, not a heuristic. For ratio = cur/bar: exec_effect = permil(cur1,bar0) - permil(cur0,bar0) <- what DELIVERY did, bar held FIXED scope_effect = permil(cur1,bar1) - permil(cur1,bar0) <- what the REQUIREMENT move did total = permil(cur1,bar1) - permil(cur0,bar0) residual = total - exec - scope <- integer truncation, EMITTED not hidden Both effects are computed INDEPENDENTLY and the residual is PRINTED rather than derived away, so truncation stays visible. Deriving one term from the other would force residual==0 by construction and hide the very rounding the integer-truncation law says goes silent. TWO ORTHOGONAL AXES, deliberately NOT collapsed into one label: a sample can raise the bar AND deliver in the same window, and flattening that into a single enum is exactly how real progress gets misread as thrashing. license_tier: ORIGINAL No hw writes (Rule 26).