A decision-making playbook for solo product builders — 20 principles across 5 dimensions,
extracted from the full development conversations of 6 shipped products.
独立开发者的产品思维手册 · 从 6 个已上线产品的全过程对话中提炼 · 20 条原则 / 5 个维度
中文 · English · Quick Reference ↓
Stuck on a decision? Find the symptom, open the principle.
卡住的时候，对照症状直接打开对应原则。
| Symptom · 症状 | Principle · 原则 |
|---|---|
| "Don't know what to do first" · 不知道先做什么 | 01 Scenario-first · 02 Spec before code |
| "Features piling up, can't stop" · 功能越加越多收不住 | 05 Subtract, don't add · 08 Infrastructure mindset |
| "Looks the same as competitors" · 和同类产品看起来差不多 | 03 Aesthetic as stance · 10 Borrow cognition |
| "Just using whatever AI gives me" · AI 给什么我用什么 | 11 Commander posture · 15 Experience before review |
| "Tech says can't do it, thinking of downgrading" · 技术说做不到想降级 | 06 Vision non-negotiable |
| "Bugs keep resurfacing" · bug 修了又出 | 13 End-to-end fixes · 14 Design vs execution error |
| "Content quality is just off" · 内容质量总差一口气 | 12 Content standards > technical standards |
| "Don't know if the product is any good" · 不知道产品好不好 | 17 Real engagement metrics |
| "Shipped, what's next?" · 上线了不知道下一步 | 20 Distribution is product |
| "Users complain, can't pin it down" · 用户反馈不对劲说不清 | 07 Product boundary · 18 Persistence = trust |
👉 Full cheat-sheet: quick-reference.md — with 5 daily self-checks and the three-line compression.
AI coding tools can now write almost anything. The bottleneck has moved from "can you build it" to "do you know what to build, where to stop, and how to tell it's done".
Most product writing is written by teams for teams. This playbook is written for one person making every call alone: what to build, what to cut, when to ship, what metric actually matters.
- Indie hackers shipping with AI (Claude Code, Cursor, v0, Lovable, etc.)
- Product people who code, engineers becoming product people
- Anyone who has ever asked: "AI just built me 10 things — but is any of this the right thing?"
- Read
SKILL.md— the entry point, with a trigger table mapping symptoms to principles. - Read
principles/— one file per principle. Each has a real dialogue, anti-pattern, and thinking formula. - Read
case-studies/— 6 complete project retrospectives. The principles are derived from these. - Use
quick-reference.md— a one-page checklist for when you're stuck.
As a Claude Code / AI agent skill:
> Using product-thinking-playbook: should I add this feature now, or wait?
| # | Principle | Dimension |
|---|---|---|
| 01 | Scenario before feature | Intent |
| 02 | Spec before code | Intent |
| 03 | Aesthetic as stance | Intent |
| 04 | Subtitle as promise, not description | Intent |
| 05 | Subtract, don't add | Boundary |
| 06 | Vision non-negotiable, implementation flexible | Boundary |
| 07 | Product boundary = user's decision path | Boundary |
| 08 | Infrastructure mindset — restraint | Boundary |
| 09 | Decisive on direction, meticulous on detail | Execution |
| 10 | Borrow existing cognition | Execution |
| 11 | Commander posture with AI | Execution |
| 12 | Content standards > technical standards | Execution |
| 13 | End-to-end fixes, not surface patches | Quality |
| 14 | Design error ≠ execution error | Quality |
| 15 | Experience before review | Quality |
| 16 | Security = product quality | Quality |
| 17 | Core metric bound to real engagement | Closed Loop |
| 18 | Persistence is the foundation of trust | Closed Loop |
| 19 | Cross-platform parity + honest permissions | Closed Loop |
| 20 | Distribution is part of the product | Closed Loop |
MIT — use it, fork it, translate it, turn it into your own playbook. Credit appreciated, not required.
AI 已经能写几乎所有代码。瓶颈从"能不能做出来"转到了"知不知道该做什么、什么时候停、怎么算做完"。
市面上的产品文档大多数是团队写给团队的。这个 playbook 是写给一个人的 —— 做决定靠自己、代码让 AI 写、判断必须自己下。
- 用 AI 工具（Claude Code / Cursor / v0 / Lovable）独立发产品的人
- 会写代码的产品人、想往产品走的工程师
- 任何曾经问过 "AI 一口气给我写了 10 个功能，可这些是对的吗？" 的人
- 先看
SKILL.md— 入口文档，有"症状→原则"对照表 - 读
principles/— 20 条原则各自独立成文，带真实对话 + 反模式 + 思考公式 - 读
case-studies/— 6 份完整项目复盘，原则就是从这些对话里提炼的 - 用
quick-reference.md— 卡住的时候一页速查
作为 Claude Code skill 使用：
> 用 product-thinking-playbook 判断：这个功能该加吗？
I. 立意 Intent
- 场景先于功能 · 2. 先定性，再动手 · 3. 气质即立场 · 4. 副标题是承诺
II. 边界 Boundary 5. 做减法而非加法 · 6. 愿景不妥协，实现可妥协 · 7. 产品边界=用户决策路径 · 8. 克制：Infrastructure 心智
III. 执行 Execution 9. 决策果断，细节挑剔 · 10. 借力已有认知 · 11. 指挥者姿态 · 12. 内容标准 > 技术标准
IV. 品质 Quality 13. 端到端修复 · 14. 设计错误 vs 执行错误 · 15. 先用再评 · 16. 安全=产品品质
V. 闭环 Closed Loop 17. 核心指标绑定真实投入 · 18. 持久化=信任基础 · 19. 跨端一致+权限诚实 · 20. 分发是产品的一部分
- Contra Code · VS Code 内嵌游戏浮层
- AudioLib · 开发者用的音频库服务
- Gospel Hub · 福音音乐聚合站
- Spybook · 互动侦探叙事小说
- Tarot · AI 塔罗咨询
- Xianxia · AI 仙侠小说生成
MIT · 随便用、随便 fork、随便翻译、随便改成你自己的 playbook。致谢欢迎，不强求。
作者 @yangyue1974 · 43.group