Open applied research · gated before production use

Open quantum research, benchmarked before adoption.

CoastalQuant evaluates emerging methods with versioned protocols, tuned classical baselines, and declared adoption gates. Publication alone is not evidence of production readiness.

Program status

A Module C diagnostic is public. No service is available and no adoption gate has been passed. Under the current service concept, production computation would initially be deterministic and classical; any research transfer would require the declared validation and adoption requirements.

Champion / challenger

One problem, two methods, declared criteria.

Classical baseline

Classical first, if a service launches.

Any production path must be tuned, validated, operationally reproducible, and costed end to end before customer-facing use.

Baseline locked before testing

Research challenger

Evidence before adoption.

The challenger uses the same instance, declared criteria, full overhead, and reproducible trials. Positive or negative results are publishable.

Diagram only · no performance claim

Concept diagram · bar widths are not performance data

State of the field

The uncertainty is the reason to evaluate.

Published benchmarks show no definitive quantum advantage in combinatorial optimization to date; the strongest early evidence sits in simulating quantum systems. That is why this program is designed around disciplined, pre-registered evaluation — not adoption by default. Negative results would be published.

Research modules

Research question

Portfolio optimization

Benchmark-only today

Baseline / challenger

Planned baseline: Tuned classical solvers
Proposed challenger: Quantum and hybrid optimization challengers

Adoption gate

≥5% total cost or ≥20% time on 30+ real instances, measured end to end (including encoding and queue time) against tuned classical solvers

Research question

Molecular simulation for candidate prefilter materials

Nearest scientific potential

Baseline / challenger

Planned baseline: DFT and classical screening pipelines
Proposed challenger: Quantum and hybrid chemistry methods (e.g., active-space simulation) beyond reliable DFT

Adoption gate

Independent experimental validation beyond the DFT baseline

Research question

Source attribution

Publication-drivenFirst diagnostic published

Baseline / challenger

Planned baseline: Classical ML with uncertainty
Proposed challenger: Quantum kernels and small-data models

Adoption gate

≥10% relative gain on a locked external holdout

Research question

Building-portfolio tail risk

Long-horizon · fault-tolerant era

Baseline / challenger

Planned baseline: Classical Monte Carlo
Proposed challenger: Amplitude-estimation methods

Adoption gate

Reproducible end-to-end advantage at operational cost

Publications & artifacts

A public record, not a promise.

Preprint · Module C · 22 Jul 2026

Preprint · not peer-reviewed

Statevector-referenced geometry survival of a four-qubit ZZ quantum kernel on IBM Quantum hardware

Rostyslav Sipakov · arXiv:2607.20377 · quant-ph, cs.LG

A fixed-subset hardware diagnostic asks whether a small quantum kernel preserves dataset geometry under error suppression on a Heron-class processor. Gate twirling helps; dynamical decoupling alone does not. No quantum-advantage claim is made.

Read on arXiv Code & data · Zenodo

Protocol · concept v0.1 · 02 Aug 2026

Pre-registration must be observable.

Endpoints, test sets, statistics, and baseline tuning budgets would be frozen in a versioned, timestamped public protocol before any run. Amendments would be logged, not overwritten.

Changelog

v0.1 · 02 Aug 2026 — Initial public protocol commitment and artifact ledger.

Boundary

The Registry records runs. It is not, by itself, pre-registration.

Production boundary

Five conditions for any production transfer.

Condition 01
Technical advantage on our specific problem
Condition 02
End-to-end economic advantage after encoding, queueing, error mitigation, and validation
Condition 03
Reproducibility across datasets and time periods
Condition 04
Enterprise-grade reliability and security from an available provider
Condition 05
Customers pay for the improved result — never for the word "quantum"

Illustrative benchmark registry · illustrative data

Every future run would leave a record.

Solver, backend, dataset, seed, shots, circuit depth, queue time, objective, baseline gap, cost, and result status would be recorded together.

Illustrative registry entries showing proposed baseline and challenger records. These are not completed runs or findings.
Concept ID	Instance	Method	Backend	Role	Status
A-001	Dispatch · sample portfolio	CP-SAT	Classical CPU	Proposed baseline	Design-stage
A-001-Q	QUBO mirror instance	QAOA	Simulator	Proposed challenger	Design-stage
B-001	Prefilter sample	DFT	Classical HPC	Proposed baseline	Design-stage
B-001-Q	Active-space mirror	VQE	Simulator	Proposed challenger	Design-stage

Anti-halo disclosure

What is not quantum.

The proposed sensor would not be a quantum sensor.
Under the current service concept, the planned device, machine-learning model, and cloud analytics would be classical.
Planned DFT material screening would remain classical computational chemistry.
Simulated annealing and tabu search—including any quantum-inspired variants—would be reported as classical methods.