Research Ledger · Self-graded
Verified, re-scoped, or retracted — the same scorecard we sell, applied to our own results first. Grades change when evidence changes. Retractions stay on the ledger permanently, because a record you can edit is not a record.
What was claimed: A 14-bit toy-curve ECDLP solve using Regev's algorithm on IBM QPUs — one bit under the submission that won Project Eleven's Q-Day Prize. Privately confirmed within the prize; consolation prize awarded.
What killed it: A Gidney-style null test — randomized data substituted into the classical post-processing — reproduced the "solve." The pipeline was finding the key; the hardware wasn't. The retraction invalidated twenty months of post-processing methodology, published in full as the first public falsification of a prize-confirmed submission.
What survived: The null-test protocol itself, now a standard preflight on every Firebringer result.
ARTIFACT · Quantum Post-Mortem (PDF) · job IDs in the published repository
What was claimed: Doubling the prior public ECDLP record, 6-bit to 12-bit, on IBM hardware.
What the null test showed: Unlike the 14-bit result, these runs survive randomized-input substitution — classical post-processing alone does not reproduce the correct key. But honesty cuts both ways: the correct key ranked outside the top 10 of the hardware-derived candidate pool in most runs (worst case ≈ rank 25).
Current claim: Hardware runs enrich the correct key into the top of the candidate pool; final identification is classical search over that pool. "Solved" is withdrawn. "Enriched" is what the evidence supports.
ARTIFACT · methodology & null tests in Quantum Post-Mortem (PDF)
What was claimed: A full HHL linear-solver — not just the QPE stage — executed end-to-end on IBM Torino, inverting a spectral operator built from elliptic curve 11a1. The blocker was never the algorithm: measuring a qubit the unitary never touched fails in the stack (the "orphan-qubit rule"). Hanging the eigenvalue rotation on a data qubit already inside every controlled-U, and pinning the transpiler to only the qubits the problem needs, removes the failure.
Scope, stated plainly: 5 logical qubits, depth 210, 1,024 shots, 2-bit clock precision — a toy instance. Clock-register phases 1·2·3 dominate, matching the non-zero eigenvalues of the operator. This demonstrates executability under measurement hygiene, not utility. Limits remain: depth near decoherence, coarse precision, hardware noise.
IBM TORINO · JOB D4VI91SG… · 1,024 SHOTS · GATES 358 · NISQ HHL End-to-End (PDF)
What was claimed: Register-specific reorganization of IBM QPU outputs around synchronized meditation windows, scaling with participant count, with apparent negative controls.
What killed it: Five independent artifacts, each caught by a specific tell — two backends merged into one timeline, a decoder analyzing compressed bytes as bitstrings, a "control" session with no time axis, independence-assuming statistics on drifting hardware, and metrics whose names didn't match their formulas. Under the statistically valid test — circular-rotation permutation with false-discovery-rate correction across all 36 endpoints — zero significant results.
What survived: The 15-rule NISQ falsification checklist, published below. Two years of being fooled, converted into the instrument that prevents it.
ARTIFACT · Quantum Post-Mortem (PDF) · full autopsy, all five artifacts, job IDs included
Timeline integrity, pipeline integrity, design integrity, statistical integrity, epistemic integrity. Each rule names the specific artifact it kills, from "use QPU execution timestamps, never submission times" to "write the kill condition down before the first job is submitted." None require a PhD. All of them would have prevented a specific failure in our own work.
ARTIFACT · included in Quantum Post-Mortem (PDF)
The broadest claim in the market, run through the five-category scorecard. Verdict: reject the phrase, fund the migration — crypto-agility is the property you can actually buy. Includes the ten questions to ask before accepting any "quantum safe" pitch, and why a live "quantum resistance demo" is a negative signal.
ARTIFACT · Scorecard 001 (PDF) · Audit one-pager (PDF)
Why publish retractions? Because an experiment is trustworthy exactly to the degree that it is built to survive your own preference for its outcome. A validation practice that has never falsified its own work is asking you to believe it would falsify yours. This ledger is the proof of instrument. Yours gets graded in private — the audit applies this same scorecard to your vendor's claim.