Internal · Launch plan
Complete the app, ship monetization, and go to market — with Umut on build and Ali on story, testing and content. Async by design: Umut builds a day ahead, Ali tests and authors against it.
The arc
Each phase ends on a gate that must be green before the next begins. Weekends flex.
Lock the model map & free/paid split. CI live. Router spike. Report directives v1.
Subscriptions, feature gating, usage metering, onboarding expansion, reports final.
Beta writers, pricing page, demo video, feature-flag audit, feature freeze.
Live payment test, load checks, launch-day rehearsal, go/no-go, ship.
Decide first — everything depends on these
Recommended shapes below. Refine the specifics; the structure should hold.
A cheap Haiku router classifies each chat turn and routes by complexity. Premium (Opus) routing is a paid perk.
| Tier | Model | Used for |
|---|---|---|
| CHEAP | Haiku 4.5 | Chat router, scene analysis, alternatives, voice-classify, project-file inventory/fill, quick continuity |
| DEFAULT | Sonnet 4.6 | Chat, generate/continue, beats, treatment, character/location reports, synopsis (classic), canvas modes |
| PREMIUM | Opus 4.8 | Producer report, screenplay analysis, synopsis (structural), canvas synthesize, escalated "deep" chat |
Gate on cost, not on crippling the core. Free = the whole writing loop. Pro = expensive compute + scale.
| Capability | Free | Pro |
|---|---|---|
| Projects | 1–2 | Unlimited |
| Editor, Brief, Beats, Treatment, Canvas, PDF/Fountain | ✓ | ✓ |
| AI chat + generate | Capped (Haiku/Sonnet) | Higher + Opus |
| Deep reports (Producer, Screenplay, Structural) | Limited | ✓ |
| Image features (Storyboard, portraits) | ✗ | ✓ |
| FDX export, Series | ✗ | ✓ |
Read these — it's where the plan breaks
Live charging needs the company + Stripe approval + banking — any can slip past "one week." Mitigation (baked in): build in Stripe test mode from Day 3; only the key-swap needs the company. If banking isn't live by D16, launch free-beta with "Pro coming" and turn payments on as a fast-follow. Don't let payments block the date.
Monetization + routing + reports + onboarding at once. If you fall behind, cut in this order: defer code-signing → defer Mac → ship fewer Pro-gated features → launch free-first, payments fast-follow.
Opening Opus-routed Pro without usage caps is how the AI bill runs away. The router + metering (Day 10) is the cost governor — it ships before the paywall does.
Easy to loosen later, painful to tighten once users expect a feature free. Err toward a smaller free tier at launch.