Blake Anderson open-sources ai-app-playbook, using a Coding Agent to build consumer-grade apps on a fixed stack…
English text is being prepared.
Blake Anderson has open-sourced "ai-app-playbook," an operating playbook for coding agents to execute, covering product selection, building, launch, and first-month profitability. It has two layers: a human-facing playbook/00–13 chapters and an Agent-facing AGENTS.md (Phase 0–14, with each step marked [AGENT] or [HUMAN]); before starting, one must pass the Gate A scoring. The tech stack is fixed on Cloudflare Workers, D1, Better Auth, OpenRouter routing to DeepSeek, Superwall plus Stripe, etc., and explicitly lists a long set of non-choices such as Vercel, Supabase, and Next.js.
Why read this
AI score 64
This open-source playbook locks down product selection, tech stack, and cost assumptions, offering a complete reference workflow for human-AI division of labor in building consumer-grade AI apps.
Topics
View original postHere's my current tech stack for scaling consumer apps in 2026. It's by far the best & cheapest stack for agent-native development ⬇️ https://t.co/0yqdwDeCp5
View original
Here's my current tech stack for scaling consumer apps in 2026. It's by far the best & cheapest stack for agent-native development ⬇️ https://t.co/0yqdwDeCp5
Body
用 Coding Agent 按固定栈做出能赚钱的消费级 AI App
@blakeandersonw 开源的「ai-app-playbook」,是一份给 coding agent 执行的操作规程:人机分工走完选品 → 搭建 → 上架 → 首月盈利;技术栈固定在 Cloudflare 全家桶,并明确列出一长串「别用」。 https://t.co/G1zD5Yikef
它是什么 - 两层结构: · 给人看的章节 playbook/00–13:从找 niche、用智能体写代码,到后端、库、登录、模型、客户端、付费、数据、设计、留存、分发、成本、合规。 · 给 Agent 看的 AGENTS.md:Phase 0–14,每步标 [AGENT] / [HUMAN];动手前有 Gate A(评分够、有需求信号、单位经济说得通才开工)。仓库还带 templates/(Worker、Better Auth、wrangler、应用级 AGENTS.app.md)。用法很直白:clone,空目录里把 Claude Code / OpenCode 指到 AGENTS.md。
产品方向 默认做「对人有用、人愿意付钱」的消费端:健身、专注、学习、平静等。一句模板:对 [谁],[App] 把 [输入] 变成 [结果],几秒内完成。 七维打分(痛感、频率、可演示、AI 是核心、愿付费、能触达、低风险),≥25/35 且无单项 1 分才考虑做。明确避开:泛聊天机器人、平台很快免费化的能力、以诊疗建议为核心、没供给的双边市场、靠应用商店硬啃企业销售。
赚钱假设(作者写明是对照基准,不是目标) 默认约 $30/年 + 试用。漏斗粗算:下载 × 30% 开试用 × 30% 付费 × $30 ≈ 每下载 $2.70(约 9% 下载变付费用户)。成本表标注 2026-10-26 牌价;作者称自家栈(Cloudflare + Superwall + Tinybird + DeepSeek 等)在各规模下约占收入 2.5%(不含商店抽成与投放)。对比表里把 Vercel、Supabase、Clerk、前沿大模型等标成「Popular」,倍数差很大。
技术栈(强制)与反栈 · 用:Workers(普通 fetch,不套框架)+ D1 纯 SQL(放大后 PlanetScale via Hyperdrive)+ R2/KV/DO/Queues 等;Better Auth;AI 默认 OpenRouter → DeepSeek(经 AI Gateway),量大可上 https://t.co/7pS44wKG7o / RunPod;支付 Superwall(端)+ Stripe(Web);分析与错误 Tinybird;端:SwiftUI、RN 或 Kotlin、React + Vite;包管理 bun。 · 不用:Vercel、Netlify、Lovable、Bolt、v0、Firebase/Supabase、Clerk/Auth0、Sentry/PostHog、Next.js、Hono、Drizzle/Prisma 等一长串。原则是:栈做不了就停下来问人,别偷偷加依赖。
阶段节奏(摘要) Gate A 过关后:Brief → 账号与仓库 → 后端与库 → 登录(匿名优先)→ AI(流式、超时兜底、成本记账、evals)→ Web / iOS / Android → 支付与权益校验 → 漏斗事件进 Tinybird → 引导打磨 → 留存(推送、连续打卡、邮件)→ 上架合规 → 冲首批约 1000 下载(短视频、创作者平台 SideShift/Trybe、私信)→ 每周一个漏斗实验 + 一个产品修补。人负责账号、真机签字、发版与生产;智能体不碰生产数据与生产部署。