EExtractionpriority #1
Completion
67% · 4/6 milestones
Investment
43 stories · 219 commits · $35.74
S-E-67 (Make silent extraction repair LOUD)
milestones to done
✓Deterministic offline eval + golden corpus (S-E-30, S-E-22)
✓Transcription coverage + extraction truncation guards (S-E-31 / 32)
✓Decomposition / applicability / self-contained learnings (S-E-25 / 36 / 37)
✓Knowledge clustering → journal (S-B-20 Phase 2b)
○Agreed extraction quality bar formally set and met on real pilot transcripts (the andon)
○Pipeline durability + observability complete (silent-failure hardening)
GoalThe #1 product risk — keep tuning, and take on the architectural changes (multi-level passes, clustering/aggregation, discard-as-signal, encounters-as-a-lens) that go beyond prompt iteration.
Done (MVB)Extraction output on real pilot transcripts clears the agreed quality bar (the andon-cord metric), and the pipeline is durable/observable enough to catch silent failures. Tuning continues past this but no longer gates MVB.
BProduct architecture for MVBpriority #2
Completion
43% · 3/7 milestones
Investment
52 stories · 286 commits · $578.69
S-B-110 (Fix the two live silent-data-loss defects — `DEF-603`, `DEF-604`, and `DEF-597` which `DEF-604` subsumes)
Sub-initiative Recording pipeline — 23 stories · 8 open defect(s) · 51 fixed
Latest: Continuous foreground drain (lift the per-tick cap for the active app)
milestones to done
✓Session / recording data model matches reality (S-B-20)
✓Native audio capture replaces Plaud, on device (S-D-53)
✓Server pipeline + Supabase data sync foundation (S-B-20, S-B-15)
○Multi-user auth in production (S-B-17 OTP slice only — not full onboarding)
○Per-user data isolation live for a real second user
○Location-encryption before any sharing
○Consumer sharing
GoalReplace the Plaud-fed, local-first, single-user foundation with a capture model that isn't Plaud, a session model that matches reality, and multi-user.
Done (MVB)Native capture replaces Plaud in production; the session/recording data model matches reality; and multi-user is fully shipped — auth + per-user data + location-encryption + consumer sharing all live (a second real user onboards and captures independently).
AOperating model & agent orchestrationpriority #3
Completion
80% · 4/5 milestones
Investment
71 stories · 346 commits · $63.66
idle — no active story
milestones to done
✓Doc-collision model + PM / Ops / Release-Manager serializers live
✓Build-loop skills codified (promote / build / reconcile) + Review agent live
✓Agent roster operational (10 live)
✓Run automation (runs + sprint-plan + readiness gate)
○Self-healing bug loop fully autonomous (triage live; auto-fix+merge + CRM verify pending — S-A-95/96)
GoalRun TaleWater as a one-human business via a team of agents — roles, budgets, handoffs, audit trails, scheduled reviews — and close the acute process wounds (colliding docs, an amnesiac build loop) that would sabotage any multi-agent setup before it starts.
Done (MVB)The agent roster runs the full build loop end-to-end; Nate's only inputs are scope-gating, subjective acceptance, and physical actions; no process wound (doc collision, amnesiac loop, an unwired self-healing arc) requires a human patch. (Paperclip platform = V3, out of MVB.)
DConsumer product surfacespriority #4
Completion
33% · 2/6 milestones
Investment
142 stories · 761 commits · $2,199.29
S-D-177 (Water summary description: own-recording primary, general-knowledge fallback, manual last resort)
milestones to done
✓Map & Knowledge screen: map, pins, floating cards, tab bar (S-D-57 / 58 / 59)
✓Field Notes recap / write-up (S-D-44 / 45)
○Map / pin polish to a paying-pilot bar (DEF-161 open — zoom-to-first-pin)
○Observation notes
○Consumer sharing
○Thumbs up / down feedback on Knowledge entries
GoalWhat V2 ships to pilot users on top of the new architecture — a rising quality bar because they're (eventually) paying.
Done (MVB)The pilot-facing SPEC §7 surfaces (map/pin polish, observation notes, sharing, thumbs-up/down feedback) are built to a paying-pilot quality bar.
COperator back-officepriority #5
Completion
75% · 3/4 milestones
Investment
38 stories · 155 commits · $204.52
idle — no active story
milestones to done
✓Cost telemetry / cost model live
✓CRM: customers directory + per-user session drill-down (S-C-23)
✓Extraction-review surface — every line item reviewable (S-C-20 / S-C-22)
○Per-user pilot extraction-output review off-device (rides B3 multi-user)
GoalSee costs immediately, and see each pilot user's extraction output for quality and training. Early CRM, cost telemetry, pilot-data review, and the sharing view converge on one user-keyed, off-device surface.
Done (MVB)Nate sees per-user cost telemetry and can review every pilot user's extraction output from an off-device surface.
FBusiness readinesspriority #6
Completion
0% · 0/8 milestones
Investment
1 stories · 10 commits · $ accruing
S-F-01 (Form WA LLC + EIN + business bank account)
milestones to done
○Entity formed (WA LLC — S-F-01 in progress)
○EIN
○D-U-N-S number
○Trademark filed
○Apple Developer org transition
○App Store subscription products configured
○Privacy policy + Terms of Service
○App Store listing assets
GoalClear the non-code blockers to operating as a real (pilot-stage) business and taking money.
Done (MVB)TaleWater can operate as a viable (pilot-stage) business via the App Store, not Stripe: entity → EIN/D-U-N-S → trademark → Apple-org → App Store Connect subscription products → privacy/ToS → listing.