v100: 5 residual terms are one Django settings cluster; four scope variants used a positional model disproved in v84; corrected gist drop-set variant improves both axes by 0.14 sigma; 513 tests

v99: ACON directional scopes map the frontier; doc-unique proven a strict subset of forward-unique; fwd/bwd union covers 20/25 lost terms; 501 tests

v96: silent offset bug found and fixed (31 to 127 spans); two implementations converge to 10dp; dedup hypothesis falsified; decoder saturated; 478 tests

v92: structural_threshold built default-off; +0.275 char at zero retention cost at thr15; bare-tag regex fault caught by all-identical check; 438 tests

v85: no official harness exists; pathkeep retention NULL (retracts v84 +0.0009); 89.2% of restored basenames already present; 3 compressors registered; 373 tests

v17: per-axis cap sweep - thinking is the efficient axis; default is a local F1 optimum

v14: order-sensitive coherence metric + COACH configs (adaptive_structural 83.44% @ 0.8723)

v10: structural_cot_compressor - 62.58% @ 0.909 retention, new joint-F1 leader

v9: wide retention eval over 150 challenges/1500 QA; zero-retention prevalence quantified