1
[2026-10-18] · src "Error-Corrected Memory and Logic on a Heavy-Hex Superconducting-Qubit Processor", arXiv:2610.11658v1 (preprint, 8 Oct 2026; ibm_boston, 156-qubit Heron)
On a 156-qubit monolithic heavy-hex device (median two-qubit fidelity about 99.7%) the authors run a dynamic compass code on a 7×7 patch with ZZ/XX distance 7. With BeliefMatching on ACES-extracted noise, exclusion of two defective qubits and leakage post-selection (cutoff 0.9 leak probability), the logical error per round is 4.11% (ZZ) and 2.56% (XX), which is 32.3% and 35.2% below the baseline; without post-selection it is 4.4% and 2.9%. A 5×5 stability experiment over 30 rounds falls from a failure probability of 0.346 to 0.21 with defect exclusion, ACES and soft information (up to 56% reduction with leakage post-selection), and keeps improving to 30 rounds only with defect exclusion (otherwise it plateaus near 12). Lattice-surgery Bell states between 3×3 and 4×3 patches have 58.6-78.4% fidelity unfiltered, and 95.8-97.8% only after logical-gap or belief-propagation-convergence post-selection that keeps 4.55% (3×3) or 9.64% (4×3) of shots. The paper compares against heavy-hex memory at up to about 6% (ZZ) and 12% (XX) per round, and against earlier d=3 surgery at 78% fidelity (6.3% yield) and 92.5% (0.5% yield). It gives no classical-simulation comparison, and none against a below-threshold result.
This is a hardware-engineering result, not a threshold crossing: a ~4% per-round logical error is far above the sub-0.2% regime that counts as useful memory, and the 7×7 patch only matches the average 5×5 patch, so there is no evidence here of error suppression with distance. What is useful is the diagnosis: a few badly coupled qubits (CZ errors two to three orders of magnitude above median) and leakage dominate, and excluding them recovers 30-40%. The lattice-surgery fidelities depend on discarding 90-95% of shots, so they do not demonstrate fault-tolerant logic. It is a single-group preprint without independent scrutiny.
2
[2026-10-18] · src —
Trumpet takes today: three days since it was last assigned, the longest gap of the three, and the last trumpet session was short. The target is interval accuracy, the neglected half of Arban's Scales & intervals section, played slurred in the middle register with no tonguing-speed work. The Art of Phrasing melody is No. 32, the continuation of the sequence.
5 minutes: long tones on written G, C and E, 8 counts each at quarter = 60, mp, soft attacks. Then 2 minutes of lip slurs, C to G and back, p. Listen for a tone that centres at once, not after a second.
Take one of the slurred interval exercises, thirds then fourths, in the middle register. Play at quarter = 60, then 72, then 84. Success criterion: three consecutive passes at each tempo where every interval speaks on the first try, with no pitch scoop and the same tone on both notes of the interval.
Play Art of Phrasing No. 32 twice: first straight, then with each leap placed as a singer would, the upper note arriving softly and the lower note not dropping out in weight. Aim for a line where the intervals sound like the melody rather than like exercises.
About 25 minutes. Stop when the tone spreads, the lips swell or the slurs start to crack, even if 84 is not reached; ending while the sound is still good beats finishing ragged.
[2026-10-18] · src —
hearMiles Davis Quintet — It Never Entered My Mind · Tidal · Spotify
A ballad played on a muted horn, where the interval work is on display: every leap lands softly and in tune, and the upper note arrives without a push. Listen for how little vibrato he uses and how each phrase is shaped toward one note before tapering. That plain, placed line is the model for the Art of Phrasing melody today.