Runs from September and October 2026, each packaged as compact data with a standard-library script that recomputes every figure on its page. The raw results and the reduction scripts are available on request.
| measurement | what the data show |
|---|---|
| The measured qubit’s own loss | Each readout costs the measured qubit as much contrast as 1.3–1.7× the stated readout duration of its own decay would (per-processor means of three reads, seven processors, 4 to 6 October). The conversion assumes a single-exponential decay curve, which is only approximate; the equal-time comparison without a decay-shape assumption is from fez (11 October) and the 22 October readout-instruction run (fez and marrakesh; kingston not decided). |
| Decay under repeated readout | On ibm_fez the measured qubit’s population after 1–128 readouts follows a stretched exponential, exponent 0.823, crossing 1/2 after 23–26 readouts; at equal elapsed time 24 readouts leave 0.48–0.52 where idle leaves 0.62–0.66. |
| Distance to the measured qubits | An earlier exploratory reading of 4 to 6 October data: a weak rank correlation of an untouched qubit’s excess with proximity to the measured qubits (+0.16 with Σ1/r², 1283 readings); not reproducible across layouts. The kick’s reach is read in the 22 October reach run: resolved at 4 couplings combined, about a tenth of the kick next to a measured qubit. |
| Decay, idle and drift controls | Decay is not single-exponential (62–85 % of qubit-copies below their 300 µs extrapolation at 30 µs); a qubit idle in 0 for 300 µs sits +0.04 to +0.80 % above its own job’s zero-wait level; decay times differ by up to 9.8 µs between jobs and 11.5 µs between copies in one job, with the range of three copies inside the simulated 95th percentile of shot noise alone in 8 of 9 jobs (the simulation cannot separate shot noise from drift between copies); the non-exponential share is read against a pure-exponential baseline of 49 to 54 %. |
| Residual after spaced X-gate trains · simulation · every train number recomputed | A negative residual per gate (−0.0006 to −0.0020 on fez, kingston, boston and pittsburgh; near zero on phoenix) that simulated perfect gates plus a persistent shot-to-shot spread in decay rate reproduce. With the decay shape held equal (22 October), an extra X gate costs 0.0001643 [0.00005906, 0.0002695] combined over four processors; on fez and kingston the per-gate residual is opposite in sign to that cost and at least twice its upper edge, so it is not a gate cost there. No extra-gate test exists on boston or pittsburgh (extra-gate test). |
| Rigetti Cepheus-1-108Q | A neighbour in |2⟩ shifts a directly coupled qubit’s precession by −55 ± 1 °/µs (28 Sep) and −39 ± 2 °/µs (4 Oct), same pair (shot noise only; the dates differ by 16.3 ± 2.1; the pulse amplitude, the number of simultaneous qubits and the calibration changed between them). |
Bones · Wayfinder · Sage · Shuttle · Nova
Heuremen LLC
“We found it.”