The anchor example is a real pitch draft produced with this methodology (project: pqc-migrate, a post-quantum cryptography migration analyzer). Study the moves, not just the words.
Before (product framing — the kind NSF tends to decline):
We built a multi-language scanner that finds quantum-vulnerable cryptography in source code and generates a migration plan.
After (research framing — what NSF asks for):
The unsolved problem is detection completeness: existing scanners report what they find and stay silent on what they miss. We will measure and bound recall for cryptographic inventory, calibrate per-finding confidence, and mechanically verify that a migration preserves behavior. The working scanner is feasibility evidence; the science is the measurement.
The move: a feature ("finds crypto") became a measurement problem ("bound what we miss").
Because migration correctness cannot exceed inventory correctness, a single missed RSA usage defeats an entire migration — yet no method today states its own miss rate.
This single line answers "why does recall matter?" and "why is completeness the research target?" Find the equivalent load-bearing sentence for every project.
Weak (attackable): "No vendor does cryptographic discovery." (False — IBM, Keyfactor, SandboxAQ all do. A reviewer kills this instantly.)
Strong (defensible): "Existing offerings from IBM, Keyfactor, SandboxAQ, DigiCert, Entrust, Venafi, AWS, and Azure address discovery, inventory, CBOM generation, and migration workflows. But to our knowledge no public materials report empirically validated recall bounds, calibrated per-finding confidence, or a mechanical migration-equivalence check. That gap is a measurement-and-verification problem, not a missing feature."
The move: credit the real capability, then narrow to the precise residual no one occupies.
Overscoped (red flag): Phase I will deliver discovery + migration + governance + enterprise deployment + pilots.
Disciplined (properly scoped): Phase I proves three falsifiable questions — can completeness be measured, can confidence be calibrated, can equivalence be verified. Governance, deployment, and pilots are explicitly reserved for Phase II. Overscoping Phase I is the surest way to miss the milestones a Phase II award depends on.
Technology Innovation (excerpt, ~research-framed):
Organizations facing post-quantum migration deadlines must first answer a hard question: where is all our quantum-vulnerable cryptography, and can we prove the inventory is complete? Today that is manual, non-reproducible consulting. The innovation we propose is an automated method whose recall is measured and bounded, whose per-finding confidence is empirically calibrated, and whose migration plans are mechanically verifiable against a reference implementation... [continues through: the unsolved indirection problem; concede-and-narrow on incumbents; the central hypothesis; working systems as de-risking evidence; the Phase I-vs-vision boundary].
Notice the order inside one field: problem → innovation → why it's unsolved → why incumbents don't close it → hypothesis → feasibility evidence → scope boundary.
- A hard, specific problem stated as a question completeness/correctness can't be assumed away.
- The innovation named as measurement/verification, not feature.
- The unsolved part made explicit (what is undecidable, unmeasured, uncalibrated).
- Honest competitive boundary (concede-and-narrow).
- One central hypothesis.
- Working artifacts as feasibility evidence, not as the point.
- Falsifiable Phase I objectives with metrics + a clean Phase II boundary.
- Honest team posture with a concrete gap-mitigation plan.