The public record separates real-hardware measurements, controlled simulation, and time projections.
Two different controllers each passed 210 of 210 frozen trials from 16 to 65,536 controlled channels. For the retained controller, median acquisitions to contract moved from 11.5 to 12.
Quantum Inspire Tuna-9: three of three trials entered contract at acquisition four.
Rigetti Cepheus: 48 controlled channels entered contract at acquisition four in one paired seed.
Commanded disturbances; no provider-private calibration registers; no hardware scaling exponent.
Primary result: 210 of 210 passes per confirmed controller across seven widths.
Extension: five of five runs at one million channels for each qualifying controller.
Tested over a finite range; not proof that every scale or device will behave the same.
At a fixed physical acquisition latency, a bounded number of sequential device visits stays fast. A width-dependent calibration sweep does not.
Projection anchored to measured QSC-Bench acquisition counts. Acquisition-only: serial host work, interface bandwidth, actuation, and settling are excluded. The current NumPy implementation is not a 2 ms end-to-end system.
The architecture is the result. A specific device still has to pass the interface and timing checks.
Check compatibilitySeven widths and thirty frozen confirmation seeds per width. Controlled simulation.
Median acquisitions for the retained controller moved from 11.5 to 12.
Five of five runs passed for each qualifying controller. Not part of the frozen 30-seed inference.
Contract entry at acquisition four in every trial.
Contract entry at acquisition four in one paired seed across 96 physical qubits.
Plants, baselines, seeds, thresholds, and results.
Latest frozen confirmation: 1,470 paired primary records across seven widths, two members at 210 of 210 each, hardware evidence indexing, and machine-checkable artifact hashes.
Reproducible Headline ResultsPublished reference implementation, pre-registered scale ladder to one hundred million channels, analog compute-stack campaign, per-seed records, and figure source data.
Rigetti Cepheus hardware107-channel physical-hardware runs, job identifiers, billing metadata, registration, and finite-difference control.
IBM Heron hardwareFrozen twelve-acquisition comparisons through 32 channels and the 156-qubit hold result.
QUBO information boundaryVector-feedback, ranking-only, measurement-priced greedy, and blind comparisons behind the information-bound result.
Combinatorial reproductionCommitted planted-instance experiments, baselines, budgets, and retained raw result records.
Grid Expansion PlanningMeasurement-priced grid optimization from IEEE cases to a mega-scale ladder.
Fraud-model drift adaptationFrozen-budget recovery under controlled model drift with pre-registered hypotheses.
WDM gain flatness256 to 4,096 wavelength channels, six-to-seven-read lock, and a serial-coverage comparison.
Photonic mesh fidelityAn explicit complex mesh unitary, coupled phase drift, and 4,096 controlled phases.
Pulse-shaper controlTransform-quality hold across 128 to 4,096 spectral channels with a matched scalar-feedback arm.
QPU shot budgetA fixed fifteen-acquisition budget, binomial shot noise, and 107 to 4,096 controlled channels.
GHZ fidelityStatevector-verified evaluation with pre-registered and amended records preserved.
THz beam tracking1,024 to 4,096 elements, moving-beam hold, and per-element versus scalar instrumentation.
