Original Summary

Ok so I keep seeing pretty much every "how to build your SaaS" post recommend the exact same combo: Lovable + Supabase + Stripe + Resend + Vercel. And honestly, if you're just testing an idea or trying to get something live by the weekend, that stack is genuinely fine. No shade at all, but if you're planning to keep building on top of it for the next year or two, I think it's worth starting a bit more structured than that. The pain usually doesn't come from any single feature being hard. Multiple workspaces, roles and permissions, paid plans, usage limits, transactional emails, file uploads, testing, security: none of these are unusual on their own. The problem is they all start depending on each other pretty fast. Your subscription decides your plan, your plan decides your limits, your workspace decides your permissions, and your permissions decide what data you're even allowed to touch. Your backend has to enforce all of that at once. And with AI generating more code, more abstractions, more glue around everything, the "simple" stack stops being so simple pretty quickly. Anyway, here's the stack I personally use. Not claiming it's the ultimate SaaS stack, it's actually pretty boring on purpose, but it works for me, and I figured it might help people getting into vibecoding who want something a bit more solid than a demo. Coding: Codex + Claude Code I'd rather have Claude Code or Codex working directly inside my actual repo than build the whole app inside Lovable, Replit or whatever. You keep control of the real codebase: git history, branches, terminal, dependencies, tests, deployment. Claude or Codex can still read the project, touch multiple files, run commands, catch errors and test what it just changed. You get most of the upside of AI-assisted dev without the app living behind some other platform. Language: TypeScript + Tailwind TypeScript is the obvious pick for me. You can use it across almost the whole app, the ecosystem is huge, and models are really good with it at this point. I wouldn't bother introducing a second language unless the product forces my hand. Same logic for Tailwind: AI is surprisingly solid with utility classes because changes tend to stay local instead of quietly breaking some global css rule somewhere else. App: Next.js React on the front, server logic when you need it, a huge ecosystem, good AI support. There are cases where I'd pick something else, but for a normal SaaS I don't see the point of splitting frontend and backend into two separate projects on day one. Start simple, split later if the architecture actually needs it. UI: Untitled UI React Honestly the one thing I still don't trust AI with is designing a full interface from scratch. You can ask it to make a dashboard look premium for the 200th time and it still comes out looking like an AI dashboard. So I start from a real component system instead. Untitled UI gives you solid React co

  • 情报分类：商业与市场研究
  • 分类依据：内容涉及商业、投资或市场动态
  • 信息来源：Reddit · SideProject
  • 发布时间：2026/9/22 17:03:42