Independent claim audit for the QC industry.
T5.4 Claims Audit takes any quantum performance claim — a vendor blog, an arXiv preprint, a partnership announcement — and returns a structured, reproducible assessment. N0–N4 noise-evidence level. 9 truth rules. Citable verdict ID.
T5.4-AUDIT-2026-001 — IonQ 1.032 kHz heterogeneous photonic link
The claim: IonQ + Qubitekk + NIST + NRL demonstrated a 1.032 kHz heterogenous photonic link between a Ba-138 trapped-ion node and a SiV-diamond node, with a Bell-state fidelity of 87.9% and a Sagnac-PPLN quantum frequency converter. arXiv:2610.10705 (9 Oct 2026, 39 authors).
The audit scope: re-derive Bell-state fidelity from raw counts (or use stated 87.9% as ground truth), cross-reference with 2024 Monroe group 250/s baseline, trace citation chain to MIT-led 2025 study, cross-vendor benchmark against IBM Quantum Heron r2 218-qubit and Google Willow claims. Apply all 9 T5.4 truth rules. Issue verdict by 20 Nov 2026.
// Process
Five steps. One verdict. Citable as eig-t54-...
Submit
POST claim + EQL spec + benchmark results to /v1/claims/audit
Classify
Engine assigns N0–N4 noise-evidence level based on submitted evidence
Apply
9 truth rules evaluated against claim text and supporting evidence
Verdict
One of 7 verdict codes assigned. Confidence 0.0–1.0 with justification
Cite
Audit ID eig-t54-{16-hex} returned. PDF + JSON + plain-text report formats
// Verdict taxonomy
7 verdict codes
Every audit returns exactly one verdict from this enum. The verdict is citable as eig-t54-... + the specific code.
supportedsupported_under_conditionspartially_supportedinsufficient_evidenceunsupportedpending_reproductionrejected_misleading// Noise-evidence levels
N0 → N4 evidence taxonomy
N4 evidence = real hardware execution with calibration. Anything below N4 must be presented with the appropriate qualification.
N0N1N2N3N4// 9 truth rules
The criteria every claim is measured against.
Based on SOC2 / HIPAA audit conventions adapted for the QC industry. Each rule produces a pass / fail / partial result with one-line justification.
GET /v1/claims/truth-rules- 01Claim specifies test methodology
- 02Claim specifies baseline definition
- 03Claim reproducible from public information
- 04Hardware execution, when claimed, actually occurred
- 05Statistical significance reported (n, error bars, p-values)
- 06Comparison to classical reference solver included
- 07Claims scope matches evidence scope
- 08Vendor-specific framing does not mislead
- 09Acknowledgement of limitations present
// Pricing
$2,000 to $50,000 per year.
Per-audit pricing for one-off claim validation. Annual subscription for organizations running frequent audits. Founding-customer terms available.
Single-claim audit. N-level classification. Verdict + citable ID.
Single-claim with reproducibility framework. Methodology review. Suitable for press citation.
Multi-claim research study. Cross-vendor reproducibility. Benchmark suite.
Multi-claim enterprise study. UCS cross-vendor portability bundled. White-label rights.
Unlimited T5.4 audits + QORCH Photon Transfer + UCS pilot access. Founding-customer terms available.
“Our confidence-gated decoder reduces logical error by 40% vs MWPM baseline on IBM Brisbane.”
What you receive
- PDF audit report (signed, branded)
- Plain-text version (for email embedding)
- JSON response (machine-readable)
- Verdict ID for external citation
- Reproducibility framework document
- Marketing recommendation (what to claim publicly)
- N-level classification with evidence inventory
- 9-rule truth-rule scorecard