launch-checklistdraft

Launch Checklist: Curator pilot (trustworthy session loop)

Launch Overview

Field Value
What Pilot of the trustworthy session loop with 3–5 coaches
Launch Date TBD
Launch Type Pilot / private beta
Launch Owner Founder (TBD)
Go/No-Go Decision Maker Founder (TBD)

Key Stakeholders

Role Name Contact
Product/Founder TBD
Engineering TBD
Legal/compliance TBD
Pilot coaches (3–5) TBD

Engineering Readiness

Item Owner Status Notes
[ ] Session intake + consent capture Eng Not started Hard gate
[ ] ASR + diarization integrated Eng Not started Vendor TBD
[ ] Depersonalisation-before-LLM Eng Not started ADR-001
[ ] Reproducible analysis pipeline + evidence quotes Eng Not started The moat
[ ] Editable draft + explicit send Eng Not started No auto-send
[ ] Minimal client dossier Eng Not started P1

QA & Testing

Item Owner Status Notes
[ ] PII-leak audit = 0 raw identifiers external Eng Not started P1 blocker
[ ] Reproducibility experiment (F1 ≥ 0.70, inter-run agreement) Eng Not started P1 blocker / kill criterion
[ ] No-consent refusal verified QA Not started P1
[ ] Out-of-scope alert behavior verified QA Not started P1
[ ] Re-run does not overwrite coach edits QA Not started P1

Design & UX

Item Owner Status Notes
[ ] Dossier-first review flow Design Not started Human-in-the-loop visible
[ ] Low-confidence flags on uncertain diarization Design Not started
[ ] Error/empty states Design Not started See 16-deliver-edge-cases.md
Item Owner Status Notes
[ ] Per-session consent flow reviewed Legal Not started CCPA/CPRA
[ ] Privacy policy + data-processing terms Legal Not started
[ ] Retention policy (audio/video default 90d, configurable) Legal/Eng Not started Per studio page
[ ] Access logging Eng Not started

Analytics & Monitoring

Item Owner Status Notes
[ ] Instrument NSM + reproducibility + minutes-saved Eng Not started See 23-measure-instrumentation-spec.md
[ ] Pilot baseline established Founder Not started First job of pilot

GA-only rows (N/A for pilot — honest scoping)

Marketing campaigns, app-store listing, support staffing/training, CDN/SSL/scale, on-call rotation, cross-browser/mobile/load testing — N/A at pilot stage; revisit at v1/GA.

Go/No-Go Criteria

Must Have (Blockers)

  • [ ] Consent gate enforced (no consent → no processing)
  • [ ] PII-leak audit = 0
  • [ ] Reproducibility experiment meets target (else this is a STOP/PIVOT signal, not a launch)
  • [ ] Human-in-the-loop invariant verified (no auto client delivery)

Should Have

  • [ ] Client dossier minimal version
  • [ ] Out-of-scope alerting

Nice to Have

  • [ ] Micro-tracking bot

Rollback Plan

Trigger Conditions

  • PII leak detected → immediate processing stop + purge + root-cause (per studio page).
  • Reproducibility below target → halt scaling, do not promote output to "sendable."

Rollback Steps

  1. Disable processing for affected sessions. 2. Notify pilot coaches. 3. Preserve protected-perimeter data; purge external. 4. Root-cause + fix before resuming.

Rollback Owner

Founder/Eng (TBD).

Check-in Schedule

Checkpoint Date Attendees
Pre-pilot readiness review TBD Founder, Eng, Legal
Weekly pilot sync weekly Founder, pilot coaches
Post-pilot go/no-go TBD Founder (+ advisor/investor)

Open Issues

Issue Owner Status Impact
Payer model unresolved Founder Open Business model (Risk)
ASR vendor selection Eng Open Blocks build
Reproducibility threshold validity Eng Open Defines "sendable" (Risk)