From Zero React Native Experience to a 6-Day MVP: What Vibe Coding Actually Looked Like
It only works when the division of labor is explicit.
重點摘要
The author built the MVP with AI as a pair, not by blindly prompting it.
He kept domain modeling, API/server design, and verification, while delegating RN screens and Expo questions.
The app reached App Store review on day four, but later exposed hidden technical debt and risky shortcuts.
A backend engineer with nine years of experience describes building a React Native MVP in six days with AI as a pair programmer. Rather than letting the model do everything, he split the work clearly: he owned the domain model, server, API design, and verification, while the AI generated screens, navigation, and framework-specific code. The post reconstructs the sprint from commit logs, showing how context documents mattered as much as code and how quickly the app reached App Store review. But the story is not a hype piece: AI-produced code also introduced shortcuts, loose exceptions, and hidden risks that only surfaced later. The central lesson is simple—use AI for speed, but keep accountability human.
从零 React Native 经验到 6 天 MVP:所谓“vibe coding”到底是什么
只有在分工明确时,它才真正有效。
重點摘要
作者把 AI 当作搭档来做 MVP,而不是盲目让它包办一切。
他负责领域模型、API/服务端与验证,AI 负责 React Native 界面和 Expo 问题。
项目在第 4 天提交 App Store,但也逐步暴露出捷径和潜在技术债。
一位拥有九年后端经验的工程师,讲述自己如何在六天内与 AI 搭档完成一个 React Native MVP。他并没有把所有事情都交给模型,而是清楚地划分职责:领域模型、服务端、API 设计和验证由自己负责,React Native 界面、导航以及 Expo 相关问题则交给 AI 生成。作者根据 commit log 回顾这段冲刺,指出前期的架构文档和上下文比单纯写代码更重要。虽然项目很快提交到 App Store 审核,但 AI 代码也带来了便利性的捷径与隐藏风险;最终结论是,速度可以借助 AI,责任必须由人承担。
從零 React Native 經驗到 6 天 MVP:所謂「vibe coding」其實是什麼
只有在分工明確時,它才真正有效。
重點摘要
作者以 AI 作為搭檔完成 MVP,而不是把所有事情交給它。
他負責領域模型、API/伺服器與驗證,AI 則處理 React Native 畫面與 Expo 問題。
專案在第 4 天送審,但也逐步暴露出捷徑與潛在技術債。
一位擁有九年後端經驗的工程師,分享自己如何在六天內與 AI 搭檔完成一個 React Native MVP。他並沒有把工作全丟給模型,而是明確切分職責:領域模型、伺服器、API 設計與驗證由自己把關,React Native 畫面、導覽與 Expo 相關問題則交給 AI 產出。作者根據 commit log 回顧這段衝刺,指出前置的架構文件與上下文比單純寫程式更關鍵。雖然專案很快送進 App Store 審查,但 AI 程式碼也帶來方便性捷徑與隱藏風險,最終教訓是:速度可以交給 AI,責任必須由人承擔。