Steven GellerQuantum Computing, End to End

Book contents

Current section

Part X. Capstones

  1. Full-Stack Algorithm Trace
  2. Hardware-Constrained Lab
  3. Application Evidence Dossier
  4. Company Diligence Memo
  5. Public Playground MVP

Part X. Capstones · Chapter 85

Application Evidence Dossier

Most quantum application write-ups start with an algorithm and go hunting for a problem. The evidence dossier runs the other way — buyer, workflow, and classical baseline first, quantum method last — and this capstone produces one.

Artifact
In this chapter 9 sections

An auditable dossier binds one workflow claim to a dated source ledger, task and data contract, classical baseline, quantum resource boundary, reproduced result, integration path, counterevidence, decision, and measurable reversal condition.

An evidence dossier is a versioned packet for one decision, not a bibliography. It binds a claim identifier to workflow owner, task and data contract, comparator, quantum resources, result, reproduction, integration, counterevidence, decision, reversal condition, cutoff, and review date. A reader can trace every conclusion to a row.

Capstone brief: dossier one workflow claim

Sources receive roles: original result, official specification, reproducible dataset or artifact, independent comparison, synthesis, and contextual assessment. Current empirical and company facts require primary or official records near the claim. A review can explain a field but cannot silently promote the evidence level.

Evidence packet and source hierarchy

GAO and the National Academies supply contextual assessment. VQE is an original experiment. OpenQASM and official SDK docs define the dated implementation boundary. ACM and the benchmarking perspective specify artifact and comparator properties. The dossier records those distinct roles rather than counting citations.

Quantum application evidence dossier An application dossier from claim ID to reversible decision. Quantum application evidence dossier claim E-01 sourcerole reproduction counter-evidence avoidretest
Figure 85.1. Each decision arrow names an evidence cell. An expired source or failed reproduction breaks the path visibly.
Application Evidence Dossier: claim and source ledger, frozen 14 August 2026
Evidence IDRecordsClaim supported
E-85-1Timothy Proctor et al., Benchmarking quantum computers (2025)
U.S. Government Accountability Office, Quantum Computing and Communications: Status and Prospects (2021)
Application evidence must preserve task, scale, comparator, resource, and validation boundaries.
E-85-2U.S. Government Accountability Office, Quantum Computing and Communications: Status and Prospects (2021)
National Academies of Sciences, Engineering, and Medicine, Quantum Computing: Progress and Prospects (2019)
Alberto Peruzzo et al., A variational eigenvalue solver on a photonic quantum processor (2014)
Technology assessments are useful context but primary experiments and official records should support the load-bearing current claims.
E-85-3Association for Computing Machinery, Artifact Review and Badging (2026)Artifact review separates availability and reproducibility from the scientific conclusion.
E-85-4OpenQASM Technical Steering Committee, OpenQASM 3 specification (2026)
IBM Quantum, Qiskit documentation (2026)
Google Quantum AI, Cirq documentation (2026)
Official interfaces help specify the experimental boundary but do not establish end-user value.

Reproduce before interpreting

The model dossier evaluates a small optimization claim. It reproduces the result, retains a strong classical comparator, records queue and shot cost, and includes a counterexample instance where the candidate loses. The decision is avoid-now/retest-later because the result fails the representative-scale and baseline gates. The reversal trigger is a pre-registered held-out comparison, not a larger slide number.

Artifact contract. A dated dossier schema, populated example, executable completeness/freshness checker, and scoring rubric. All material claim IDs resolve to sources; current evidence has a cutoff and due date; task, baseline, resources, result, counterevidence, decision, and reversal fields are nonempty.

Executable reference fixture
TEXT_FIELDS = ("claim_id", "owner", "task", "baseline", "resources", "reproduction",
               "counterevidence", "declared_decision", "cutoff", "review_due")
def valid_iso(value):
    parts = value.split("-") if type(value) is str else []
    return len(parts) == 3 and [len(part) for part in parts] == [4, 2, 2] and all(part.isdigit() for part in parts)
def audit(dossier, as_of):
    errors = [name for name in TEXT_FIELDS if type(dossier.get(name)) is not str or not dossier.get(name)]
    sources = dossier.get("sources", {})
    if type(sources) is not dict or not sources:
        errors.append("sources")
    claim_sources = dossier.get("claim_source_ids", [])
    if type(claim_sources) is not list or not claim_sources or any(source_id not in sources for source_id in claim_sources):
        errors.append("orphan_claim_source")
    result, reversal = dossier.get("result", {}), dossier.get("reversal", {})
    for name, item in (("result", result), ("reversal", reversal)):
        if type(item.get("value")) not in (int, float) or item.get("unit") != "cost-fraction improvement":
            errors.append(name + "_schema")
    cutoff, review_due = dossier.get("cutoff", ""), dossier.get("review_due", "")
    if not valid_iso(cutoff) or not valid_iso(review_due):
        errors.append("date_schema")
    elif cutoff > as_of or review_due < as_of:
        errors.append("freshness")
    derived = "retest" if not errors and result["value"] >= reversal["value"] else "avoid-now"
    if not errors and dossier.get("declared_decision") != derived:
        errors.append("decision_trace")
    return {"status": "incomplete" if errors else "complete", "errors": sorted(set(errors)),
            "derived_decision": "invalid" if errors else derived, "observed_improvement": result.get("value")}
dossier = {"scenario": "held-out optimization evidence dossier", "claim_id": "QAPP-17",
           "owner": "operations research lead", "task": "20 frozen routing instances",
           "baseline": "tuned classical heuristic", "resources": "all tuning and sampling minutes",
           "reproduction": "fixture run R-07", "counterevidence": "infeasible candidate on instance 12",
           "declared_decision": "avoid-now", "cutoff": "2026-08-14", "review_due": "2026-11-14",
           "sources": {"E-04": "scenario:baseline-run", "E-07": "scenario:reproduction-run",
                       "E-11": "scenario:adverse-instance"}, "claim_source_ids": ["E-04", "E-07", "E-11"],
           "result": {"value": 0.00, "unit": "cost-fraction improvement"},
           "reversal": {"value": 0.02, "unit": "cost-fraction improvement"}}
base_result = audit(dossier, "2026-08-14")
bad = {key: value for key, value in dossier.items() if key != "baseline"}
bad_result = audit(bad, "2026-08-14")
no_counterevidence = {key: value for key, value in dossier.items() if key != "counterevidence"}
counterevidence_result = audit(no_counterevidence, "2026-08-14")
retest = {**dossier, "result": {"value": 0.03, "unit": "cost-fraction improvement"},
          "declared_decision": "retest"}
retest_result = audit(retest, "2026-08-14")
assert base_result["status"] == "complete" and base_result["derived_decision"] == "avoid-now"
assert bad_result["status"] == "incomplete" and "baseline" in bad_result["errors"]
assert counterevidence_result["status"] == "incomplete" and "counterevidence" in counterevidence_result["errors"]
assert retest_result["derived_decision"] == "retest" and retest_result["observed_improvement"] == 0.03
print(f"PASS: 85 application dossier status={base_result['status']} decision={base_result['derived_decision']} invalid={bad_result['errors'] + counterevidence_result['errors']} sensitivity={retest_result['derived_decision']}")
Quantum application evidence dossier: inspected record
Dossier fieldModel recordEvidence IDState
Taskheld-out optimization setE-01defined
Comparatortuned classical heuristicE-04present
Reproductionfixture hash and outputE-07small scale
Counterevidencecandidate loss on instance 12E-11retained
Decisionavoid now; retestD-01bounded

Exact validation command: python3 tools/validate_briefs.py --briefs data/editorial_briefs_64_87.json --from 64 --through 87 --check-rewritten-sources --execute-artifacts

Make the decision traceable to table cells

Make the decision point to cells: 'avoid now because E-07 lacks representative scale and E-11 shows a comparator loss.' This syntax reveals exactly what new evidence can change the conclusion. When a source expires or a reproduction fails, the checker invalidates the decision record rather than leaving the prose apparently current. Record null searches and failed replications as first-class rows, including protocol, date, owner, and the claim boundary each result weakens.

Validate dossier completeness and freshness

Prompt. Build a dossier for one quantum application claim and include one reproduced or independently checked result.

Deliverable. Claim ID, workflow owner, task/data contract, baseline, source ledger, quantum resources, result, reproduction record, integration map, counterevidence, decision, reversal trigger, cutoff, and review due date.

Pass condition. The checker finds no orphan claim or stale current source, the reproduced result matches its fixture, and the decision is no stronger than the weakest required evidence field.

Model dossier plus one-hundred-point rubric

Format. Model optimization dossier with an avoid-now/retest-later decision and 100-point rubric.

The reference dossier passes completeness and source mapping. Its score is analytic: 20 points for contract, 20 for source roles, 20 for reproduction, 15 for comparator, 10 for resources, 10 for counterevidence and 5 for review hygiene. It remains below the action threshold because representative evidence is absent. Removing the counterevidence row lowers the score rather than making the narrative more favorable.

Verification. The dossier checker returns incomplete when the baseline or counterevidence row is removed, and the declared reversal threshold changes the derived decision.

Analytic capstone rubric: 100 points
CriterionPointsEvidence rule
Task and workflow contract20Show the inspectable task and workflow contract record; an unsupported assertion receives no credit.
Source roles and claim mapping20Show the inspectable source roles and claim mapping record; an unsupported assertion receives no credit.
Independent check or reproduction20Show the inspectable independent check or reproduction record; an unsupported assertion receives no credit.
Classical comparator15Show the inspectable classical comparator record; an unsupported assertion receives no credit.
Quantum resource boundary10Show the inspectable quantum resource boundary record; an unsupported assertion receives no credit.
Counterevidence10Show the inspectable counterevidence record; an unsupported assertion receives no credit.
Currency and review hygiene5Show the inspectable currency and review hygiene record; an unsupported assertion receives no credit.
Total100All pass conditions remain mandatory regardless of point total.

Companion work

Artifacts for this chapter

These entries resolve to checked-in local source. Commands are reproduced exactly from the chapter manifest, and source-embedded fixtures are exported as direct downloads.

  1. Reproduce or test

    python3 tools/validate_briefs.py --briefs data/editorial_briefs_64_87.json --from 64 --through 87 --check-rewritten-sources --execute-artifacts

Provenance

Sources and review

  1. U.S. Government Accountability Office. Quantum Computing and Communications: Status and Prospects. GAO. 2021government technology assessment
  2. Association for Computing Machinery. Artifact Review and Badging. ACM Publications. 2026official reproducibility policy
  3. IBM Quantum. Qiskit documentation. IBM. 2026official documentation
  4. Google Quantum AI. Cirq documentation. Google. 2026official documentation
  5. OpenQASM Technical Steering Committee. OpenQASM 3 specification. Linux Foundation Joint Development Foundation. 2026official technical specification
  6. Alberto Peruzzo et al.. A variational eigenvalue solver on a photonic quantum processor. Nature Communications. 2014primary paper
  7. Timothy Proctor et al.. Benchmarking quantum computers. Nature Reviews Physics. 2025peer-reviewed perspective
  8. National Academies of Sciences, Engineering, and Medicine. Quantum Computing: Progress and Prospects. National Academies Press. 2019consensus study report

The load-bearing claims in the chapter are mapped inline to this registered source set. A citation supports only the bounded claim beside it.

Cite this chapter