blakeandersonw/ai-app-playbook

This is a learning resource, not a library: it hands a coding agent a staged playbook for taking a consumer AI app from niche selection to a first thousand downloads. Read it as one operator's opinionated default stack with cost estimates, not as neutral reference material.

The problem

A solo builder who wants a paid mobile or web app has to make dozens of interlocking decisions at once — what problem to pick, where to host, which auth or payment provider, how to keep model spend bounded. The usual path is to stitch that together from scattered tutorials, then discover the choices conflict. This playbook replaces that search with a single ordered path: niche scoring, a build phase with "done when" checks, a launch sequence, then a weekly growth loop. The result is a chosen stack and a timeline instead of an open research task.

What's different

Two choices stand out. First, it is written to be consumed by an agent: clone the repo, point Claude Code or OpenCode at AGENTS.md, and the chapters act as an operating brief for scaffolding and building project instructions. Second, it commits to one stack — Cloudflare Workers, Better Auth, D1/PlanetScale, OpenRouter with DeepSeek, Superwall/Stripe, Tinybird — with cost comparisons against what it calls "the popular stack" stack chapters.

That specificity is the benefit and the trade-off. You get concrete templates for a Worker, auth and Wrangler config rather than abstract advice. But the recommendations are the author's own estimates at list prices as of October 2026, and the README explicitly names products to avoid — Vercel, Supabase, Firebase, Next.js, PostHog and others cost chapter. If you already run one of those, the playbook argues against your existing setup rather than meeting you where you are. The revenue model is also a benchmark: 30% trial conversion and 30% paying $30/year, roughly $2.70 per download. The author reports having scaled five apps to $100k+ MRR, but the numbers in the repo are projections, not audited results.

Who should try it

Good fit: a solo developer or very small team building a consumer subscription app who wants a complete, opinionated default and is willing to adopt Cloudflare's platform and mobile-first payments. The phased checklist and the money section are useful for deciding whether the economics work before writing code. Poor fit: teams already standardized on another stack, backend-heavy products, or anyone needing a vendor-neutral comparison — the playbook is a prescription, not a survey. It also assumes a coding agent such as Claude Code or OpenCode is available, since the primary entry point is pointing one at the repo agent chapter.

Requirements and limitations

You need an agent harness — Claude Code or OpenCode with a model such as DeepSeek — plus accounts on the named services; the repo supplies templates, not a one-command scaffold templates. Committing to the stack means learning Cloudflare's Workers, D1, R2, KV and Queues rather than a conventional Node framework, and the playbook deliberately avoids ORMs and Next.js backend chapter. That is a real migration cost if your team's skills lie elsewhere.

The cost tables are estimates at 100k users and list prices dated October 2026; they are not a quote and will drift as pricing changes. The first-chapter validation method — ten short videos and a waitlist — is a marketing tactic with uncertain conversion, not a guarantee. Ongoing upkeep means tracking provider pricing and the agent brief as the chapters evolve.

HotRepo's take

HotRepo observed this repo at #5 on the NEW-REPOSITORY board for TypeScript on 2026-10-27, with 207 total stars — a single snapshot, not a growth series, so persistence cannot yet be judged. The decision-relevant check is the cost comparison in 12 · Costs: verify each provider's current pricing and whether the DeepSeek-by-default assumption still fits your model needs before adopting the stack.