Skip to content

Exploration 03 — Goal rewrite

Status: session 3, 2026-07-26. The goal statement itself lives in plan.md (single source of truth); this memo records what changed, why, and what was deliberately kept or dropped. Inputs: 00-frame.md (diagnosis), 01-market-scan.md (waves 1–2), 02-personas.md (ratified ranking, engine plurality, product waves).

What changed in plan.md

  1. Extended End Goal replaced. The old goal described an open-source (Apache-2.0) platform for the development/research/regulatory market with revenue from managed hosting — written before any market contact, asserting a license posture that matched neither reality (all repos private) nor intent (commercial licensing under evaluation). The new goal states: a commercial, white-label, multi-engine risk platform; ranked personas (P1 insurer first); product waves (ORSA module → 5116/2024 pack → portfolio workbench); observable success criteria; licensing recorded as deferred pending P1 validation instead of asserted.
  2. Current Phase banner now points at the exploration track (sessions 0–3 done, session 4 = answer-layer RFC) instead of "No phase is active."
  3. Phase status table gains the exploration track row; Phase 7 marked pending re-cut — its adapter items (7.1, 7.3) are elevated by the engine-plurality commitment; its public-facing items (7.6 marketing site, 7.9 privacy banner) wait on the licensing decision.
  4. Open Cross-Cutting Points updated: the UI-license row generalized to the platform licensing posture (deferred, owned by the exploration track); a seismic capability row added (gate for #366); the pluggable-engine row upgraded from "settled architecture" to product-level commitment.

Kept deliberately

  • CLIMADA stance: cite, maintain goodwill, never position as a fork; clean-room engine unchanged. Engine plurality adds CLIMADA back as one selectable sidecar engine (Phase 7.3 design) — it does not reverse ADR-024's in-process ban.
  • The demoable MVP slice (CLAUDE.md) as the launchable-artifact health check. The role-agnostic product framing is retired (02-personas), but the slice remains the smoke-walk until persona surfaces replace it. Follow-up: re-cut the slice around the wave-1 ORSA flow once it exists.
  • Self-hosting and auditability as differentiators — they survive the goal change intact; what changed is that openness-as-license is no longer claimed as the mechanism (market scan killed that as a differentiator; provenance and reproducibility are the auditability substance, and they are already built).

Dropped or demoted

  • "Open-source (Apache 2.0) backbone + engine" as a stated fact → replaced by the deferred licensing decision.
  • Development/research/regulatory market as the primary segment → demoted to P3/P4 personas behind the insurer wedge.
  • Data marketplace → already parked; reframed as future "country/data packs."
  • Managed hosting as the revenue model → replaced by per-instance licensing + productized modules; hosting is a delivery option, not the business.

Decided / Parked / Killed (session 3)

Decided - plan.md Extended End Goal rewritten per above; committed to main. - Success criteria are observable user actions (per persona, from 02-personas), not issue counts.

Parked - CLAUDE.md MVP-slice re-cut — after wave-1 ORSA flow exists. - Phase 7 re-cut — session 4+ (needs the answer-layer RFC to know what 7 becomes).

Killed - "No phase is active" as a plan state — the plan now always points somewhere.