Status: Rewritten 2026-07-09; row statuses refreshed 2026-10-06 against the code, see Progress: September 24 to October 6. Original framing: launch the companions behind a paywall, then iterate: more tricks, more companions, better UX everywhere.
The vision in one paragraph
Companions are the product's differentiator: a coach in your pocket for every action sport, with a body, a voice, and a memory. Kaori (snowboard) is built — 3D stage, live voice, speech-synced trick demos, unified memory. Next: ship her to users with a free-sample paywall and voice-token allocation, then iterate the companion experience (more tricks, snowy environment, outfit/board unlocks, stance-aware coaching), add the skateboarding companion, refactor Spots into a flagship-quality UX, and instrument the whole app with usage analytics so decisions stop being guesses. Surfing companion after skate + snow are strong.
Priority Matrix
P0 — Ship the companions (now)
The feature is merged and prod-deployed; nothing installable contains it yet. Full audit: Companions Launch.
| # | Task | Repo | Effort | Status |
|---|
| 1 | Secure the voice WebSocket (JWT on upgrade, session caps, text-length cap) | Backend | 1–2 days | Partial: session, per-IP and message caps shipped 2026-09-30; JWT on upgrade pending |
| 2 | New EAS dev-client build + physical-device validation (MToon, voice, choreography) | Mobile | 1 day | Done (v3.2.0, v3.3.0) |
| 3 | v3.2.0 TestFlight + Play internal builds (version bump, store pipeline dusted off) | Mobile | 1 day | Done: 3.3.0 live 2026-09-21 |
| 4 | Monitoring: Sentry both services, kith-voice in ecosystem.config, ElevenLabs credit alarm, uptime checks | Backend | 1–3 days | Partial: /health shipped 2026-09-25; Sentry, spend alarm, uptime checks pending |
| 5 | Release notes docs/releases/v3.2.0.md when the build ships | Docs | 0.5 day | Done |
P1 — Monetization foundation (before public launch)
Full model: Monetization: Paywall & Tokens.
| # | Task | Repo | Effort | Status |
|---|
| 1 | COGS levers: Flash-class TTS on stage + OpenRouter prompt caching | Backend | 0.5 day | Pending |
| 2 | Token metering: user.wallet.voiceTokens, usage_events, requireVoiceTokens middleware, lazy daily grant | Backend | 3–5 days | Partial: text and voice metering with free/Plus tiers shipped 2026-09-30 (kaoriUsage); wallet and purchases pending |
| 3 | Free-sample gating: Kaori-only for free tier, locked companion states + upsell UI | Backend + Mobile + Web | 3–5 days | Pending |
| 4 | Web purchases: extend Stripe payments.js with tier + token-pack prices, site paywall/wallet UI | Backend + Web | 3–5 days | Partial: Plus checkout activates since 2026-10-06 (webhook fix, verify-session, reconcile); tiers and token packs pending |
| 5 | Mobile IAP via RevenueCat (fold into the companion EAS build) | Mobile + Backend | 5–10 days | Pending |
| 6 | Basic rate caps on bot-chat (interim until metering lands) | Backend | 1 day | Done 2026-09-30 |
P2 — Companion iteration + the skate companion
| # | Task | Repo | Effort | Status |
|---|
| 1 | Regular/goofy stance onboarding — early profile question for boardsports users; feeds stance-aware choreography + coaching | Mobile + Backend | 2–3 days | Pending |
| 2 | Better snowboard 3D design (shape, materials, deck art) | Mobile | 1–2 days | Pending |
| 3 | Refine FS360 + expand the snowboard TRICKS registry (bs 180, ollie, butters, grabs…) | Mobile | ~0.5–1 day per trick | Pending |
| 4 | Snowy stage environment for Kaori (skybox/terrain/props; per-sport environments pattern) | Mobile | 3–5 days | Pending |
| 5 | Companion models → CDN with device cache (kills the 14MB bundle tax; prerequisite for outfits + roster growth) | Mobile + Backend | 1–2 days | Pending |
| 6 | Unlockables v1: boards + colorways, XP/tier/purchase unlocks collection | Mobile + Backend | 3–5 days | Pending |
| 7 | Tony — skateboarding companion: VRM avatar, persona corpus (skate media), voice, skate TRICKS registry (ollie, kickflip, shuv…), street/park environment | All | 2–3 weeks | Pending |
| 8 | Analytics events on all companion surfaces (demo opens, voice minutes, trick views) via usage_events | Mobile + Backend | 2 days | Pending |
| # | Task | Repo | Effort | Status |
|---|
| 1 | Spots UX refactor — browsing/filtering/detail flows re-designed to flagship quality; the map overlay work was the start | Mobile + Web | 1–2 weeks | Pending |
| 2 | Full-app UX audit (web + mobile): intuitiveness/value pass on every flow, informed by analytics | All | 1 week audit + fixes | Pending |
| 3 | Usage-analytics instrumentation app-wide (PostHog exists on web; add mobile SDK + event taxonomy) | Mobile + Web | 3–5 days | Pending |
| 4 | Surf companion (Rico) — after skate + snow are strong | All | Later | Backlog |
| 5 | Engineering-standards carryover: tests on critical paths, Zod API schemas, structured logging, API versioning | All | Ongoing | Partial |
Engineering-standards status (October 2026): Biome + Husky + CI ✅ (all four repos) · Node 20 on prod ✅ · Google Play closed alpha ✅ · Push notifications ✅ (v3.1.0) · health endpoint ✅ (2026-09-25) · unit tests ✅ in every repo (backend node:test, web and mobile Jest) · schema validators ✅ (warn mode) · rate limiting ⚠️ partial (auth, registration, bot-chat, MCP) · Sentry and graceful-shutdown verification still pending, folded into P0#4 above.
Quarterly Plan
Q3 2026 — Launch the companions
Q4 2026 — Iterate + second companion
2027 — Scale the roster
Metrics to Track
Companion & business
| Metric | Why |
|---|
| Stage opens / DAU | Is the flagship discoverable? |
| Voice replies per user per day | Token-tier tuning input (from usage_events) |
| Free → paid conversion | The paywall's whole job |
| Voice COGS per user per month | Must stay under tier price; alarm on drift |
| Demo → "Learning" trick status lift | Does coaching actually work? |
| D1/D7/D30 retention, companion users vs not | The differentiation hypothesis |
Engineering health
| Metric | Current | Target |
|---|
| Crash-free sessions | Unknown (no Sentry) | >99.5% |
| Mean time to detect error | Days (user report) | Minutes (Sentry alert) |
| CI pass rate | Green on all repos | >95% |
| Test coverage | ~0% | 40% on critical paths |