eng-0269
window recorded
citable URL: https://halobench.com/records/eng-0269/ — this address never moves; the anchor /records/#eng-0269 keeps resolving
22.21 Whmean 125.5 W − 10.1 W idle floor (box-idle-all-empty) = 115.4 W attributable
window 2026-08-23T19:35:31Z → 2026-08-23T19:46:08Z · provenance recorded
method wall-meter · aihydra energy sensor via Home Assistant (counter-difference on sensor.hardware_ai_hydra_energy, raw ~10s-history samples queried by hborchestrator via the HA API and linearly interpolated to the band edges; power from sensor.hardware_ai_hydra_power) · peak 218.2 W · no contention
covers run-0620
HK-RERUN-REASONOFF specsweep spec=OFF band (run-0620): 22.206 Wh over 19:35:31-19:46:08Z covering all 10 matched cells (per-cell 0.72-2.20 Wh). Matched-band comparison vs eng-0268: spec=ON used 17.1 Wh vs spec=OFF 22.2 Wh for identical work at identical throughput (~1.0 ratio) — i.e. --kairic-edge drew ~23% LESS wall energy while producing the same tokens. Plausible mechanism: speculative verification shifts work to cheaper batched draft passes even at flat acceptance-driven throughput; treat as one observation, not an established effect. Per-cell windows joined from all-windows.jsonl timestamps; within HA retention.
Wh, never joules — read as a cumulative-counter difference across the window; delta over the idle baseline is the only figure that means anything
Cited by — computed at build time, never stored
claims clm-0118