CLIMB status extract: actuator sensitivity, 8 September 2026

**Rung 2 of the deployment ladder was executed, and it refuted its own registered prediction.**
Design declared before the outcome: `plan/ACTUATOR_SENSITIVITY_2026-09-08.md`. Re-screening all
900 Phase-G clips at knee and hip-roll limits of 139, 120 and 90 N·m leaves 888 clips unchanged
and moves the flagged count from 99 to 98 of 900, with no clip newly flagged. The baseline
reproduces the published screen exactly on all 900 clips, maximum delta 0.0. The design had
predicted the flagged fraction could only rise; four clips fall, by up to 0.0236, because the
screen thresholds the translational component of a residual whose total the program minimizes.
The lower-bound sentence was withdrawn from the manuscript and the public page and replaced by
the measurement. Full record: `reports/actuator_sensitivity_2026-09-08/result.json`.


The status record also states: allocation is inconclusive; the H1 admission campaign completed ten training runs but stopped before the first evaluation cell. This is an editorial status summary, not new evidence.