Hawking at 50 · Public Protocol · Hardware Update · 13 November 2026
Testing Quantum Information Recovery on Quantum Hardware
A small, reproducible experiment to measure how well a decoder recovers entanglement with an untouched reference qubit.
Level 6 · Local checks complete · Calibrated hardware run and independent reconstruction pending
Fixed Recovery Protocol
Protocol h50-fixed-4096-v1 tests whether a decoder can recover Bell correlations with a reference qubit outside its access. The four arms are exchange recovery, SWAP validation, QMM-inspired output-only decoding and global inverse decoding. Keep the physical layout and decoder fixed for the run.
Measure XX, YY and ZZ for each arm, with 4,096 shots per setting and 49,152 shots in total. For the chosen Bell target, calculate overlap as (1 + ⟨XX⟩ − ⟨YY⟩ + ⟨ZZ⟩) / 4.
Fixed Hypothesis and Decision Rule
The exchange decoder will recover Bell correlations with the untouched reference. Positive certification requires both the exchange and SWAP overlap estimates to exceed 0.562 and calibration to support an additive witness-bias bound of 0.02 or less. The primary raw-shot error budget is α = 0.01. Include uncertainty in the calibration bound when reporting confidence.
What Level 7 Requires
Before booking hardware, verify the checksums in the reference circuit bundle and compare all twelve circuits with the device-native versions. Check the logical operations, register access, physical layout and measurement mapping. Run the results intake on synthetic counts before collecting data.
Twelve circuits matched their ideal expectations and nine software intake checks passed. Freeze native circuits, layout, readout mapping, calibration checks, shot allocation, exclusions and analysis before inspecting outcome counts.
Retain the backend and calibration record, native circuits, job identifiers, timestamps and all twelve raw histograms. An analyst independent of the design, execution and first analysis will reconstruct the result. The benchmark measures recovery in a defined qubit circuit. Its relevance to black-hole evaporation depends on a physical mapping to gravity.
Read the Full Anniversary Edition → · Both Tests in Pictures · QMM Phase-Recovery Pilot
Hardware Access Update
IBM Cloud declined service and deactivated the account on 12 November 2026. IBM Quantum hardware access depends on an active IBM Cloud account. Both AkashicNET experiments are awaiting their first hardware run. IBM Cloud Account Requirements
The first low-cost candidate is Rigetti Cepheus through Open Quantum. Advertised Spark credits could cover the fixed benchmark if the service can return device-origin counts, executed native circuits and calibration evidence for the bias bound. At the published Amazon Braket rate, the 12 Cepheus measurement tasks would cost about US$24.49 before calibration, diagnostics, reruns and tax. IQM Resonance Starter is another candidate, subject to a device-specific credit and evidence check. No provider has been confirmed for the full benchmark. Open Quantum Pricing · Amazon Braket Pricing · IQM Resonance
Requirements for Levels 7–10
- Level 7 · Independent Reconstruction. A calibrated fixed-protocol run and an analyst independent of the design, execution and initial analysis. The analyst reconstructs the Bell overlaps and uncertainty from all twelve raw histograms, checks calibration and the SWAP control and applies the frozen decision rule.
- Level 8 · Technical Review. A quantum-information specialist reviews circuit equivalence, decoder access, calibration and uncertainty. Record objections, corrections and unresolved points.
- Level 9 · Robustness. Repeat the fixed benchmark under fresh calibration or on another suitable layout. A repeat on a different processor tests whether the finding holds across hardware. Record discrepancies and scope any conclusions to the conditions tested.
- Level 10 · Reproducible Release. Publish the versioned protocol, source and executed circuits, raw counts, calibration, analysis code, repeat results and independent review record.
A precise negative finding can progress through these milestones. If uncertainty prevents a decision, record the run as inconclusive and identify the evidence still needed.
Planned Shots and Historical IBM Timing
| Job | Executions | Historical IBM Estimate |
|---|---|---|
| Hawking fixed benchmark | 49,152 | 19.20 s |
| QMM phase pilot | 1,536 | 2.54 s |
| QMM comparison, block A | 12,288 | 6.30 s |
| QMM comparison, block B | 12,288 | 6.30 s |
| Core total | 75,264 | 34.34 s |
The 34.34 seconds were an IBM-specific estimate for core execution. Calibration, diagnostics, optional repeats and queue time were excluded. Each replacement provider requires its own estimate for the chosen circuits and device.

