Skip to content

12-ROADMAP.md

Latest commit

 

History

History
138 lines (97 loc) · 7.64 KB

12-ROADMAP.md

File metadata and controls

138 lines (97 loc) · 7.64 KB
Raw

Roadmap to v1.0

Status: proposed 10-month / 40-week path
Owner: founder
Planning unit: elapsed weeks after funded start

1. Now, next, later

Now — validate and bound, months 0–2

Outcomes:

  • buyer, staging, and evidence-access hypotheses tested;
  • design-partner LOI and funded readiness work;
  • governance group established and paid;
  • pack 0.1 and receipt prototype reviewed;
  • intended-use and data-boundary legal memo;
  • ADR 0001 accepted.

Do not build adapters before the discovery gate passes.

Next — build and start pilot, months 2–7

Outcomes:

  • deterministic local runner and schemas;
  • twelve-case pack with governed core assertions;
  • file import, normalization, evaluator, review, receipt, and verification;
  • security, privacy, accessibility, EN/ES, and operations gates;
  • gated external-pilot setup and the six-week baseline.

Later — complete pilot, release, and learn, months 7–10

Outcomes:

  • v1.0 hardening and signed release;
  • annual-assurance conversion;
  • second-partner transferability test;
  • only then, P1 FHIR collection/delta/mapping improvements.

RIS/DICOM, pharmacy, CDS, hosted collaboration, and production canaries remain beyond v1 and require separate governance.

2. Milestones

Milestone Target Outcome/exit
M0 Discovery decision week 4 discovery continue gate passes
M1 Governance and pack alpha week 8 reviewers approve scope and pack 0.1
M2 Deterministic vertical slice week 12 one case imports, evaluates, reviews, renders, verifies
M3 Full reference pathway week 17 12 cases/36 assertions run on fixtures
M4 Trust beta week 21 security/a11y/i18n/ops gates and hidden faults pass
M5 External pilot baseline week 27 after six elapsed pilot weeks, at least 10 cases complete the partner pathway
M6 Remediation receipt week 33 after the full 12-week pilot, partner uses verified P0 delta and utility/control-value evidence in release decision
M7 v1.0 week 40 release checklist passes

3. Workstream sequence

Weeks Product/research Governance/evidence Engineering Trust/operations
1–4 interviews, workflow, buyer recruit reviewers paper schemas only initial threat/legal boundary
3–8 pilot design pack 0.1, oracle protocol architecture spikes privacy/a11y plan
7–12 usability prototype assertions reviewed vertical slice base CI/security
11–17 operator tests pack beta all file adapters/evaluator logs/signing
16–21 comprehension tests hidden fault authors hardening/P0 delta a11y/EN-ES/runbooks
20–21 contracting and calendar preparation only; no measured pilot activity final gate review no partner execution finish DG-04 evidence
22–27 pilot phases A–C: safety setup, mapping, dry run, baseline baseline adjudication partner-local execution and adapter fixes only incident/cleanup exercises
28–33 pilot phases D–E: remediation, rerun, closeout finding disposition delta and compatibility evidence closeout/cleanup evidence
33–37 release feedback or frozen extension measures pack 1.0 unaffected hardening; B-057 only in weeks 34–37 if invoked independent reviews continue
38–40 final decision pack 1.0 freeze release candidate independent dossier close

4. Decision gates

  • DG-01 week 4: build only with design partner, owners, evidence access, and legal path.
  • DG-02 week 8: encode only dual-approved assertions; experimental content cannot influence mandatory pass.
  • DG-03 week 12: continue custom engineering only if vertical slice is understandable to all core personas.
  • DG-04 week 21: start the 12-week pilot in week 22 only after safety properties, PHI canaries, and accessibility critical paths pass; pre-gate contracting and scheduling do not authorize a dry run or case execution.
  • DG-05 week 27: continue after the baseline only if evidence completeness is at least 75% and no boundary incident remains unresolved.
  • DG-06 week 33: release candidate only if the full 12-week pilot's technical/safety gates and predeclared utility/control-value gate pass, or B-057's single bounded, separately funded extension is formally invoked. The extension may run in weeks 34–37 only; if it consumes release-critical hardening capacity, DG-07 moves.
  • DG-07 week 40: release v1 only with external use, legal approval, and funded maintenance.

5. Capacity checkpoint

  • F product/delivery pool: 252.5 available days; 139 baseline assigned, 143 maximum with B-057.
  • E engineering pool: 170 available days; 124 baseline assigned, 130 maximum with B-057.
  • Specialists/participants: 380 budgeted hours; 345 maximum assigned.
  • Base B-001–B-056: 263 core days. Maximum with B-057: 273 of 422.5, leaving 149.5 days.
  • DG-01 temporal checkpoint: F 31 available / 27 assigned; E 20 / 3.
  • DG-04 temporal checkpoint: F 146.5 available / 127 assigned; E 105 / 99. The E pool is 94.3% loaded at this gate, so no added scope is permitted and a forecast consuming the six-day reserve adds capacity or moves DG-04.

No role exceeds total funded capacity, including the extension. Temporal gate capacity, not total-program slack, controls entry to the pilot. A scope change must remove equivalent work, add named capacity, or move the date.

6. P1 fast follows

Prioritize using observed pilot time:

  1. Advanced multi-receipt/cross-profile comparison if the mandatory P0 delta still leaves more than 2 hours of analysis per rerun.
  2. Mapping workbench if mapping exceeds 30% of delivery time.
  3. Read-only FHIR adapter if repeated collection is safe and broadly reusable.
  4. JUnit/SARIF if release-engineering integration drives renewal.
  5. Full Spanish execution guidance if operator demand is validated.
  6. Customer-local extension namespace if two partners need the same extension mechanism.

No P1 begins before v1 P0 evidence is complete unless it removes a pilot blocker without expanding safety scope.

6. Later exploration

RIS/DICOM

Requires DICOM Supplement 233 implementation research, imaging informatics lead, radiologist, modality/vendor staging access, and a distinct safety pack.

Pharmacy

Requires pharmacist governance, e-prescribing/label workflow mapping, patient communication risk analysis, and no assumption that EHR behavior transfers.

CDS

Requires fresh FDA/device analysis, clinical-specialty oracles, automation-bias evaluation, and strict separation from patient-specific advice.

Hosted control plane

Requires five annual customers or equivalent demand, security staffing, tenancy architecture, BA/HIPAA analysis, SOC 2 procurement evidence, and a separate ADR.

Production synthetic canaries

Requires proof they cannot affect billing, reporting, care, patient contact, inventory, or analytics; new legal, privacy, vendor, and patient-safety approval. This is a new product, not a configuration toggle.

7. Resourcing changes

Add a part-time interoperability engineer when partner mapping exceeds 20 F-pool hours/week. Add an operations/customer-success owner at five active annual customers. Add a dedicated clinical safety officer before any patient-specific, production, CDS, imaging, or pharmacy expansion. Governance/community review remains paid and independent at every scale.

8. Roadmap anti-goals

  • Do not count repositories, adapters, or assertions as success.
  • Do not broaden beyond four checkpoints to rescue weak buyer demand.
  • Do not build hosted collaboration for one procurement request.
  • Do not claim external validation from reference fixtures.
  • Do not turn indeterminate clinical questions into configurable passes.
  • Do not ship a “trans-safe score.”