Quantum Lightning, week 3: from code to a citable preprint

This week (7–13 Nov 2026) Quantum Lightning, my post-quantum risk platform, produced its first citable research output, and the platform gained three measurement-first features. The common thread: measure, don't promise.
Preprint + open benchmark suite
On 11 Nov I published "Exact-Answer Benchmarks for Quantum Cryptanalysis Resource Estimation: Known-Key Instances, Published-Estimate Regression Targets, and Closed-Form Reference Values" on Zenodo (DOI 10.5281/zenodo.24136745, CC BY 4.0), with the companion suite exactbench v1.0.0 (DOI 10.5281/zenodo.24136835, Apache-2.0, GitHub).
Published estimates of the quantum resources needed to break RSA and elliptic-curve cryptography differ by orders of magnitude, and the tools that reproduce them are rarely tested against anything with a known answer. Borrowing the plasma-physics habit of verifying equilibrium codes against closed-form exact solutions, the suite has four parts:
- Known-answer Shor order-finding instances. All 24 instances (up to 24 qubits) match the closed-form measurement distribution to a total-variation distance of at most 7.5 × 10⁻¹², in Qiskit statevector simulation and through an OpenQASM 3 round trip.
- Known-key ECDLP instances on toy prime-order curves, agreeing to machine precision.
- Regression targets from published estimates, recomputing each paper's headline numbers from its own stated inputs. The 897,864 physical-qubit total for RSA-2048 is reproduced exactly; several commonly quoted closed-form expressions deviate from their own papers' tables by up to 22%, and one typographical slip is flagged.
- 50-digit reference values for Weibull and lognormal arrival-time models.
Running an open-source estimator against the suite showed that a model calibrated at one key size can differ from the published per-run runtime by a factor of 2.3 at another.
Platform: three measurement-first features
- Bitcoin on-chain exposure census. UTXOs are classified by script type and labelled exposed now, exposed at spend or not determined — never "quantum-safe". P2PK and P2TR expose the public key from creation; P2PKH/P2WPKH only at spend unless an address is reused. It is a measurement instrument, not a rating.
- Overclaim lint (PT-BR/EN). Seven rules that flag narrative claims such as unscoped "formally verified" or "quantum-proof". QRC is a measured rating, never a seal; the lint also runs automatically when reports are generated.
- Evidence-of-assurance track A0–A5 on every CBOM component, from unknown to machine-checked proof at implementation level. It is reported separately from quantum exposure: a proof covers only its declared scope and adversary, and an A5 does not reduce harvest-now-decrypt-later risk.
I also unified the codebase into a single home with reconciled CI and APIs. Git balance for 7–13 Nov: 10 commits, 6 merged PRs, +10,806 / −1,750 lines. The platform repository is still private.
VQCC (Google Quantum AI on Kaggle)
The public score rose from 24.94081 (snapshot of 5–6 Nov) to 27.08908 on 13 Nov, a gain of 2.14827, with 47 submissions and 343rd place on this snapshot of the leaderboard. Ranks move as others submit. Part of the week also went into verification tooling for the competition. Compilation technique details stay private until the competition ends on 20 Jan 2027.
Takeaway
In post-quantum risk, the most valuable product is auditability. A number without a reproducible source is an opinion. Next update next Friday.

