QGraph Scout

Latest result · 17 July 2026

19 seconds versus 6.89 million years

At the same 70-qubit width, IBM hardware returned 256 samples in 19 QPU-seconds. A local Aer fit projects 6.89 million years for one classical sample. This is a striking clock scenario, but not yet a quality-matched speedup.

Explore the clock comparison Original submission ↗ Read the article series ↗
IBM Kingston70 + 8 qubits
Schematic qubit network A chain of data qubits with eight connected check ancillas and three highlighted stabilizers. s₀ s₃₉ s₇₉ schematic · not the physical layout
19 sQPU time · 256 samples
6.89m yearsclassical projection · 1 sample
1.14 × 1013projected clock ratio
Nonequality-matched winner

Same width · different quality evidence

70 qubits: seconds versus millions of years

The quantum run is measured. The classical value is a log fit to five local Aer measurements at N=4, 6, 8, 10 and 12, evaluated at N=70.

IBM Kingston · measured

19 seconds QPU time for 256 complete 70-bit samples

Clock scenario

1.14 × 1013 classical projection divided by QPU time

Qiskit Aer · extrapolated

6.89 million years projection for one sample at N=70

Interim conclusion

A real hardware execution, a modest conclusion

The available data show a weak exploratory signal at a post-hoc one-sided 75% threshold, but no paper-matched or publication-ready confirmation.

Hardware execution

Passed

Predeclared 95% test

Failed

Post-hoc 75% analysis

Weak positive

Quantum advantage

Not demonstrated

Same dataset, three thresholds

Raw one-sided lower bound

zero = crossing point

Swipe horizontally for the full chart

Lower bounds at 95, 80 and 75 percent confidence The 95 percent lower bound is minus 0.01465, the 80 percent lower bound is minus 0.00146 and the 75 percent lower bound is plus 0.00125. −0.016 −0.010 −0.004 0 +0.004 95% · predeclared −0.01465 failed 80% · post-hoc −0.00146 failed 75% · post-hoc +0.00125 weak positive
LevelRaw lower boundOutcome
95%, predeclared−0.01465Failed
80%, post-hoc−0.00146Failed
75%, post-hoc+0.00125Weak positive

All values come from the same 80-stabilizer dataset. The distribution-free Hoeffding check remains negative even at 75%.

Not a one-to-one reproduction

The missing 19 checks change the experiment

Our circuit logic passed ideal checks. The main difference is the experimental stack: a different hardware window and 8 instead of 27 spacetime checks.

Tracker Submission #151

Boston · full protocol

Qubits
70 + 27 = 97
Postselection
0.10%
Point fidelity
0.37 ± 0.02
95% lower bound
0.092

Our full-check pilot

Kingston · 27 checks

Qubits
70 + 27 = 97
Accepted
0 / 8,192
Outcome
No DFE signal
IBM-job
d9bi5je6…

Our workable reconstruction

Kingston · 8 checks

Qubits
70 + 8 = 78
Postselection
2.45%
Point estimate
0.01217
95% lower bound
0 after clipping

Fire Opal · paired smoke test

Folding increased postselection probability, but not the signal

Three preselected stabilizers were measured folded and unfolded in the same six-circuit batch. The route worked technically; the effect test was negative.

Signed stabilizer expectation

Unfolded → folded

Unfolded Folded

Swipe horizontally for the full chart

Folding result per stabilizer Stabilizer 0 changed from minus 0.00971 to plus 0.04930; stabilizer 39 from plus 0.02655 to minus 0.01553; stabilizer 79 from plus 0.01923 to minus 0.06403. −0.08 −0.04 0 +0.04 +0.07 s₀ −0.00971 +0.04930 s₃₉ −0.01553 +0.02655 s₇₉ −0.06403 +0.01923
StabilizerUnfoldedFoldedDifference
0−0.00971+0.04930+0.05901
39+0.02655−0.01553−0.04208
79+0.01923−0.06403−0.08326

Evidence boundary

What has and has not been demonstrated?

Demonstrated

  • Complete bare 70-qubit circuit: 256 samples in 19 QPU-seconds
  • Explicitly labelled 70-qubit clock scenario of 1.14 × 10¹³
  • Direct IBM DFE execution with 70 data qubits plus 8 checks
  • All 80 stabilizers measured in two IBM Runtime jobs
  • Separate Fire Opal folding smoke test with 6 circuits
  • Weak post-hoc 75% proof-of-concept signal
  • Negative predeclared folding outcome

Not demonstrated

  • The original 95% fidelity claim
  • Publication-ready graph-state or rotated-state fidelity
  • Classical hardness of the complete 70-qubit instance
  • Quantum advantage

Reproducibility

Compact, sanitized artifacts

The site exposes the reported summaries for inspection. Large raw provider payloads, per-shot data and local credential provenance are deliberately excluded.

75% sensitivity dataJSON · download Folding summaryJSON · download Quantum-classical clock comparisonJSON · download Tracker Submission #151GitHub · external source
DFE run receipts · 80 stabilizers
IBM-job · s0–s9
d9bkverv6alc73ct7c90
IBM-job · s10–s79
d9bp6k66hjac73fg14m0
Backend
ibm_kingston
Transport
Direct IBM Runtime
Stabilizers
80 / 80
Raw shots
184,320
Accepted
4,519
Quantum time
58 s
Fire Opal folding-smoke · 6 circuits
Fire Opal action
2334093
IBM Runtime job
d9c8t7s1osis73bji0u0
Backend
ibm_kingston
Status
DONE
Circuits
6
Shots per circuit
2.048
Total requested
12,288
Quantum time
23 s

Next sensible step

Measure the same quality first

Extend the Aer measurements toward the submission's N=42 range and predeclare one shared output-quality or observable threshold. Only then can the seconds-versus-years clock scenario become a defensible speedup claim.