North star plan: SeldingerMed/vector — evidence spine to external reproducible benchmark (G2, to 2027-03-17) #69

Goal

G2 — USD 5,000,000 in binding pre-seed/seed commitments plus official grant awards by 2027-03-17 (binding), with Vector as the commercial/evidence spine. Also serves G1: a partner can inspect an independently scored evaluation rather than a self-reported metric.

Reality (observed 2026-09-20)

• SeldingerMed/vector last pushed 2026-09-11; zero open issues.
• docs/PLAN.md (master roadmap) and docs/NEXT_STATUS.md (roadmap charter 2026-09-08) exist in-tree but are not tracked as issues or milestones.
• docs/NEXT_STATUS.md states the next objective verbatim: "one real, reproducible benchmark that an external team can run unchanged and wants to run again", and warns that "more catalog entries, schemas, and cloud features do not satisfy that objective by themselves."
• operations/ceo/config/funding-ledger.json → binding_investor_commitments: [], official_grant_awards: [], submitted_applications: [].
• investor-pipeline.json gate is active: false and requires G1 progress plus Colin's approval.

Why this moves G2

G2 evidence is a binding commitment or an official award receipt. The fastest path to a binding technical/partner commitment is a third-party reproducible benchmark, which is also the only artifact that lets Vector be sold as an independent evaluator rather than a tool. The repo's own charter already names this objective; it lacks a tracked horizon and owners.

Phases to 2027-03-17

P0 — Internal evidence spine (weeks 1–3)

• Vector serves the research program: the E2/E3 evaluator contract is a Vector task-owned verifier, replayable from a stored trace.

P1 — Lumen evidence bundle (weeks 2–6)

• A Lumen evidence bundle exists as a Vector job with a content head and a reproducible replay (see SeldingerMed/seldinger-navigation#133).

P2 — One external reproducible benchmark (weeks 4–12)

• A single benchmark an external team can run unchanged on CPU, with a documented on-ramp and a public replay check.

P3 — Commercial and financing evidence (weeks 8–26)

• Paid pilot offer drafted; every G2-eligible artifact (commitment, award receipt) recorded in the funding ledger.

Next 2 weeks (derived work)

1. Confirm whether E2's "Vector task-owned metric contract" is already an authored Vector task; if not, author it and replay the stored trace.
2. Pick the single external benchmark candidate and write its on-ramp doc.

Acceptance for this plan

• The E2/E3 evaluator is a Vector task with a passing replay from a stored trace.
• One named external-team-ready benchmark exists with an on-ramp doc.
• Vector issues and milestones track this horizon; the in-tree PLAN/NEXT_STATUS remain the detailed status source.

Reality values above are read from the repo HEAD date, the open-issue list, docs/NEXT_STATUS.md, and operations/ceo/config/funding-ledger.json.