A Readout Is Not an Idle: Data and Verification Code

Data, plans, analysis code and outputs for "A Readout Is Not an Idle: Phase, Coherence and Entanglement Costs to Untouched Qubits on IBM Processors" (K. Indahl). Covers every IBM run from 24 September to 23 October 2026 on seven IBM processors (ibm_fez, ibm_kingston, ibm_marrakesh, ibm_pittsburgh, ibm_aachen, ibm_boston: 156 qubits; ibm_phoenix: 120) and the Rigetti Cepheus-1-108Q runs of 25 September to 5 October. In the 21 and 22 October runs each mid-circuit readout is compared with an idle of exactly the published readout length, on the same qubits, in the same job (the 7 October copies are separate jobs with an idle within 0.24% of the readout length). On untouched qubits the readout turns the phase, costs coherence and costs Bell pair concurrence (selected shots can bias the coherence and concurrence costs upward, and excitation or leakage is not excluded as their whole cost on fez and kingston); also the two readout instructions compared, one coupled neighbour's static ZZ against IBM's calibrated value, Hahn-echo T2 against IBM's stated T2, the measured qubit's loss at equal elapsed time, the shape of idle decay, spaced X-gate train bookkeeping, an extra-gate test, and a Rigetti neighbour shift. Plans were pushed to a public server before the jobs where each page says so; checks that failed are reported as they came out. Version 6 carries the corrected paper (see the README for the changes) and adds the raw file of the 23 October ibm_marrakesh fresh-record Hahn run. Contents: one folder per run (plan and SHA-256, circuits, analysis and synthetic tests, IBM calibration at firing, job IDs, committed outputs); the 7 October recomputation and the inclusion rule; an index of every IBM job on our accounts; IBM's calibration records saved with the runs; the paper (PDF and LaTeX) and figures. raw_v6.zip holds the raw readout bits (.npz) of every run folder and the 11 October sweep. Write-up: https://heuremen.org Heuremen LLC. "We found it."