Show, don't claim

The Evidence Ledger

The public record of Rosetta's verdicts. For every parameterized quantum recipe and problem class, this ledger will state whether it beat the best classical solver — at what size, with what raw, reproducible evidence. Every verdict will include seeds, library versions, and a downloadable notebook. Negatives included: the verdicts everyone else buries are the ones we publish.

Recipes in pipeline

—

Verdicts published

First real verdict

in progress

Illustrative preview

These are the types of entries the ledger will contain — not measured results yet. Every row below is a structure demo. When the first real fight completes, its verdict replaces a demo entry here, with raw data attached.

How to read this ledger

Recipe. Fight. Verdict.

01 — RECIPE

A candidate algorithm

A parameterized quantum algorithm (QAOA, VQE, annealing) harvested from arXiv or built in-house, mapped to a class of problems — not a cherry-picked instance. Each gets a stable RQ-NNNN ID.

02 — FIGHT

Quantum vs the classical champion

Both sides run the same instance with the same compute budget. The classical corner is the classical champion, and every run logs a fixed seed and library versions.

03 — VERDICT

Win, not yet, or negative

The judge publishes an honest outcome with the estimated crossover point — the size where quantum is projected to start paying. That number is the product.

ID	Recipe	Problem class	Vertical	Status	Evidence

All entries flagged is_demo: true — illustrative structure preview, not measured results.

The system at work

rosetta://foundry · activity stream illustrative feed — what the live console will stream

demo lines · once the harness runs, this feed streams real events with raw-data links

Methodology — how the judge works

Rigged against ourselves.

A ledger is only worth what its weakest rule allows. These are the rules every published verdict will have passed.

01 · The classical champion

Quantum never fights a strawman. The classical corner is the best available solver — OR-Tools, Gurobi, well-tuned simulated annealing — kept strong on purpose. A weak baseline would invalidate the entire ledger.

02 · Same instance, same budget

Both sides solve the identical instance with the same compute budget, scored on a double axis: solution quality and time. No hand-picked instances, no asymmetric hardware.

03 · Reproducible by construction

Every experiment logs its fixed seed, library versions and raw outputs. Each published verdict will ship with a runnable notebook — re-run it and get our numbers, or catch us.

04 · Negatives get published. Always.

Vendors can't publish negative verdicts about the industry they sell. We can — and we red-team our own judge: every way we find to cheat it becomes a new protocol rule.

Questions worth asking

Why would anyone publish "not yet"?

Why publish negative verdicts?+

Because a buried negative is a decision someone else makes blind. Every "quantum doesn't win here yet, and here's exactly where the data says it might" saves a team months of exploration — and the archive of honest negatives compounds into an asset no latecomer can buy or accelerate. If quantum advantage arrives late, the referee's archive gains value. If it arrives on time, we hold the map of where it crossed first.

How do we keep the classical baseline honest?+

The classical champion is maintained as seriously as the quantum side: strongest available solvers, tuned, versioned. Both sides get the same instance and compute budget. And we periodically red-team our own judge — weakened baselines, convenient instances, cherry-picked metrics — and turn every exploit we find into a published protocol rule.

What exactly is a "crossover point"?+

The projected problem size where the quantum side matches classical solution quality within tolerance while the classical solver's runtime becomes prohibitive. It's the number a buyer actually needs — not "quantum is coming," but "re-test at ~2,400 assets." It is always labeled a projection until it is measured. Nobody else publishes this number.

When does the first real verdict land?+

The first fight — QAOA vs OR-Tools on constrained portfolios — is in progress. Whatever the outcome, it gets published: win, not yet, or negative. That first honest report will replace a demo row on this page, raw data attached.

No hype. Just the number.

The first real verdict is in progress.

When it lands — win, not yet, or negative — it publishes here with seeds, versions and a reproducible notebook. Be the first to read it.

Get notified when it lands →

routes via email while the ledger backend is being built — no fake subscriber counters here