Jobs created 2–7 October 2026 (US Eastern), seven IBM processors: 305 jobs, 63 on 2 October, 43 on 4, 26 on 5, 48 on 6 and 125 on 7; three errored. The 32 equal-idle jobs of 7 October are among them (7 October). On 11 October ibm_fez ran 14 jobs: 3 of the measured qubit after 1 to 128 readouts (used in the paper) and 11 of a three-box probe-strength sweep. The 305 jobs measured untouched qubits left to precess while other qubits were read out, the measured qubits’ own decay, and each qubit’s decay during idle time. The findings that rest on these runs are on the supporting measurements pages, each with its own script and data.
| File | What it is |
|---|---|
| runs.json | every run: 305 across seven processors, created 2 to 7 October (three errored) (pittsburgh 86, kingston 75, marrakesh 55, aachen 25, boston 24, fez 20, phoenix 20), each with its IBM job id, submission time and the sha256 of every script frozen before it fired. |
| data/ | per-qubit readings used by verify.py, and the qubit roles (measured or untouched) that the 21 October runs reuse. |
| verify.py | recomputes the comparison against the stated duration from the published per-qubit readings: 1.0207 ± 0.0092 pooled over seven processors, 2.2 standard errors from 1: cannot tell; this fitted-rate reading is superseded by the equal-idle design. The ratio mixes the readout kick with each untouched qubit’s spin rate, so it does not measure a duration either way. The comparison against an equal idle in the same job: readout vs idle. |
| run and reader scripts | on request, with the raw result arrays; runs.json carries each script’s sha256. |
Bones · Wayfinder · Sage · Shuttle · Nova
Heuremen LLC
“We found it.”