Claims outrun evidence
"Quantum advantage" appears in decks long before a named benchmark, problem size, or classical comparator exists to support it. Confidence is not the same as proof.
Home / Platform
The platform · how Veriqa grades a claim
Every quantum claim, graded against a classical baseline.
Veriqa is the evidence-grading engine beneath every agent. It turns claims, papers, patents, cryptographic inventories and theses into auditable, decision-ready outputs — each line carrying a source and a confidence, each verdict gated by a classical comparator before anything reaches a Proceed verdict.
01 · The decision gap
Claims move faster than evidence.
Quantum decisions land on desks every week — a pitch, a paper, a vendor roadmap, a board question. The language is confident. The proof is thin. Most teams have no in-house way to tell the difference.
No in-house bench
Few organisations keep a quantum physicist and a cryptographer on staff to read the primary sources. The expertise needed to grade a claim is exactly the expertise that is scarce.
Papers are not roadmaps
A peer-reviewed result on a toy instance rarely translates into a deployable, cost-competitive system. The gap between "demonstrated" and "useful" is where most value is lost.
The missing layer
Between raw sources and a real decision sits an evidence-grading layer that almost no one has built. Veriqa is that layer — a consistent, auditable way to score what is actually known.
02 · How a claim is graded
Five steps from source to verdict.
Veriqa runs every input through the same pipeline. The classical-baseline gate is the load-bearing step: a core advantage claim with no comparator is flagged needs-baseline, its maturity is capped, and a Proceed verdict is blocked in code.
- 1
Ingest
Company, paper, patent, URL, pasted memo, crypto inventory, or strategic question.
- 2
Extract claims + sources
Pull each claim, its evidence, maturity and assumptions — every line bound to a source ID and confidence.
- 3
Classical-baseline gate
Each advantage claim is checked for a named benchmark, problem size and classical comparator. Missing any →
needs-baseline. - 4
Score
Compute technical maturity, commercial urgency and hype / overclaim risk from the graded evidence.
- 5
Reviewer gate → verdict
The report stays in draft until an internal reviewer approves it; the software enforces the gate before delivery.
needs-baseline; maturity is capped and a Proceed verdict is blocked. The report still ships — as a monitor or further-diligence call, never a green light.03 · The scoring model
Three scores, each with a reason.
Veriqa never reports a single opaque number. It separates how strong the evidence is, how urgent the decision is, and how far the language has run ahead of the proof. Each score is a transparent, rule-based heuristic computed from features of the graded evidence — evidence density, hedging and superlative language, and the presence of benchmarks, baselines, dates and named entities — not a sentiment score and not a statistically validated model.
Technical maturity
Evidence versus hype. Raised by a named benchmark, a stated problem size, a present classical comparator, and reproducible or peer-reviewed results. Lowered by demonstrations on toy instances, missing baselines, and claims marked unverified. Capped when any core claim is needs-baseline.
Commercial urgency
Deadline pressure on the decision. Raised by a running clock — for example PQC migration against the harvest-now-decrypt-later exposure window — regulatory deadlines, and short asset shelf life. Lowered by speculative timelines with no external forcing function.
Hype / overclaim risk
How far the framing outruns the evidence. Raised by unbenchmarked "advantage" language, superlatives without sources, and missing classical baselines. Lowered by precise scope, stated limitations, and claims that cite their evidence with confidence levels.
Scores and any example values shown across this page are Illustrative — generated to explain the scoring approach, not drawn from a real subject. Decision-support and general information only — not investment, financial, legal, tax, or security advice, and not an offer or solicitation of any security.
04 · Claim-discipline charter
Five rules the software enforces.
Claim-discipline is not a style guide — it is encoded in the engine. The API blocks a Proceed verdict (internally the go key) while any core claim is needs-baseline, and routes every report through the reviewer workflow gate before it can be delivered.
- Benchmark required. No claim of advantage stands without a named benchmark and a stated problem size. Missing any → the claim is flagged
needs-baseline. - Show the classical comparator. Every quantum result is presented alongside the classical baseline it is measured against — never in isolation.
- Label illustrative data. Mock, sample and illustrative figures are marked as such, every time, so no reader mistakes an example for a finding.
- Cite every claim. Each factual claim carries a source ID and a confidence level. If it cannot be cited, it is marked
unverifiedand cannot raise the verdict. - Route through the reviewer gate. Every customer-facing report stays in draft until an internal reviewer approves it — the software enforces the gate. Independent expert review is recommended before high-stakes decisions.
Enforced in code. A Proceed verdict is invalid while any core advantage claim is
needs-baseline. The rules engine adds claim-discipline rows to the evidence table and gates the verdict before a report can leave draft — the same logic runs whether the report was produced offline or with AI enrichment.
05 · Four agents, one spine
Different inputs. The same auditable engine.
Each agent is a vertical front end on the same evidence-grading spine. They take different inputs and produce different deliverables, but every one inherits the classical-baseline gate, the citation rule, and the reviewer workflow gate. Follow the link on each card to the vertical it serves.
01
Readiness Diligence
Grades a company, claim, paper, patent or thesis into a board-ready decision memo with scores, an evidence table and exportable JSON.
For investors & corp-dev02
PQC Migration
Maps a cryptographic inventory against NIST FIPS 203/204/205, prioritises by exposure and shelf life, and returns a phased migration plan.
For security & CISOs03
Molecular Design
Separates AI-assisted design, physics simulation and selective quantum refinement into a conservative evidence package with wet-lab gaps flagged.
For pharma & R&D04
Market Intelligence
Tracks companies, papers, grants, policy and public signals into structured, machine-readable intelligence packs on a recurring schedule.
For investors & corp-dev06 · What a report contains
Built to be read by a board — and a machine.
Every Veriqa report is a complete, auditable artifact. Nothing in it is unsourced; nothing illustrative is unlabelled.
- Executive verdict — proceed, monitor, or require further diligence, in plain language.
- Score gauges — technical maturity, commercial urgency, hype / overclaim risk.
- Evidence table — every claim with a source ID and confidence level.
- Missing-evidence list — exactly what is absent and why it caps the verdict.
- Recommended validation plan — the next steps that would change the grade.
- Machine-readable JSON — the full report as structured data for your systems.
- Board-ready Markdown memo — a clean narrative summary for circulation.
- Reviewer status — draft until an internal reviewer approves the gate.
Illustrative evidence-table row Illustrative
Claim — "Helios Photon Labs solver beats classical optimisation on routing."
Status — needs-baseline · no named benchmark or classical comparator supplied.
Source — SRC-04 (vendor whitepaper, undated) · Confidence — low
Effect — maturity capped · contributes to hype-risk · Proceed blocked.
Claim — "Helios Photon Labs solver beats classical optimisation on routing."
Status — needs-baseline · no named benchmark or classical comparator supplied.
Source — SRC-04 (vendor whitepaper, undated) · Confidence — low
Effect — maturity capped · contributes to hype-risk · Proceed blocked.
"Helios Photon Labs" is a fictional subject used only to illustrate the format. Illustrative — Decision-support and general information only — not investment, financial, legal, tax, or security advice, and not an offer or solicitation of any security.
07 · Auditable & machine-readable
Reproducible by design.
A grade you cannot trace is a grade you cannot trust. Veriqa is built so any line can be followed back to its source and any verdict can be re-derived.
Every line in a report carries a source ID and a confidence level. The evidence table is the report — the verdict is a function of what is in it, not an opinion layered on top.
The full report exports as structured JSON conforming to a published schema, so it drops directly into your diligence pipeline, data room, or risk register without re-keying.
Grading runs on an offline, deterministic rules engine. The same inputs produce the same graded evidence and the same gate decisions — results are reproducible, not stochastic.
An optional AI-enrichment layer can summarise and surface context, but it never bypasses claim-discipline: it cannot lift a needs-baseline flag, uncap a maturity score, or unlock a Proceed verdict.
What "auditable" means here
- Source on every line — claims and confidence, not assertions.
- Schema-backed JSON — machine-readable export.
- Deterministic core — same input, same grade.
- AI never overrides — enrichment respects the gates.
- Reviewer gate — draft until approved.
08 · Questions
What Veriqa is — and is not.
Does Veriqa run quantum hardware?
No. Veriqa is Path A — an evidence and decision layer, not quantum execution. It does not design or run quantum circuits, integrate with hardware, or itself claim quantum advantage. It grades the claims that others make.
Who reviews the reports?
Every customer-facing report stays in draft until an internal reviewer approves it — the software enforces the gate before delivery. Independent expert review (for example, a domain specialist) is recommended for high-stakes decisions.
What is the difference between offline and AI-enriched?
The offline rules engine grades evidence and applies the gates deterministically — same input, same result. AI enrichment is an optional layer that adds summaries and context. It never bypasses claim-discipline: it cannot remove a needs-baseline flag or unlock a blocked verdict.
Can I export the data?
Yes. Every report exports as machine-readable JSON conforming to a published schema, plus a board-ready Markdown memo. Each line keeps its source ID and confidence so the export is as auditable as the report.
How current is the evidence?
Veriqa grades the sources you supply or point it at, as of the time of analysis. Each claim is timestamped and cited; where a source is undated or stale, that is reflected in its confidence. The grade describes what is known now, not a forecast.
What does a verdict actually recommend?
A verdict is decision-support — proceed, monitor, or require further diligence — never a directive to buy, sell or hold any security. It tells you how strong the evidence is and what would change the grade.
Decision-support and general information only — not investment, financial, legal, tax, or security advice, and not an offer or solicitation of any security.
Grade your first claim.
Open the console to run a claim through the baseline gate, the three scores and the evidence table yourself — or have us scope a brief against your own sources. Prefer a quick look first? Watch the live demo grade a claim.