In the previous Chapter 2, we decided "what to build and for whom" in small scope. This chapter is Phase 2: Prepare — before you actually start working, you set up four foundations: tech, tools, design, and environment. The motto here is "don't fuss, don't agonize, pick the mainstream." Burning out on preparation defeats the purpose. Get AI-friendly choices done by the shortest path, and set yourself up to pour everything into the next "build" phase.
When unsure, pick "the one AI is good at" — that's all
What Phase 2 is about
The preparation phase is, in a word, "shopping before you build." The most common failure in indie development is melting days on "tool selection" and "environment setup" before you start building, then being too tired to reach the main event. Here, you don't buy the pro-grade high-end gear; you get only the basic set.
Research every trendy new technology, compare tools exhaustively, scrutinize paid plans. Run out of steam before building.
Decide a mainstream stack, one main tool, and the free tier within the same day, and start building the next morning.
The rule of thumb is don't spend more than a full day on prep. A perfect choice doesn't exist, and with AI it's easy to switch later. Set off with "this'll do for now" — that's the right answer.
Choosing tech — pick the mainstream and building is easy
A tech stack is the combination of language, framework, database, and infrastructure that builds your app. Not just on the hands-on route: even on the beginner route, knowing "what tech it'll be generated in" makes your instructions to AI sharper.
The conclusion on how to choose is simple. Pick the "mainstream" that AI has learned in bulk and writes fluently. Because AI writes more accurate code the more information exists about a technology. Niche cutting-edge tech has little information, and AI makes more mistakes. In indie development, "AI is good at it = building is easy = it gets finished" is the iron rule.
Plenty of information, so AI writes it accurately.
Examples: front end React / Next.js; back end Laravel (PHP) or Node.js; database PostgreSQL or MySQL; auth and DB via Supabase or Firebase.
Tech that just appeared, setups with little information in any language, experimental frameworks. AI makes mistakes and you get stuck verifying. Unless you're learning for its own sake, the mainstream is plenty.
💡 You don't have to decide the tech yourself. If you don't know which stack is best, it's enough to ask AI: "I'm building [what you want] as an indie project. Suggest a mainstream tech stack that's easy for AI to implement and has plenty of information." You can adopt what it suggests as-is. If you want to dig deeper, see AI-friendly tech stacks.
On the beginner route (no-code), you can even leave this selection to the builder. AI app builders like v0 and Bolt automatically use mainstream stacks such as React / Next.js under the hood, so you barely need to be aware of the tech at all.
Choosing tools — this is where routes diverge
The most important thing in this chapter is choosing tools. This is the one place where your main weapon clearly splits by the route you chose in Chapter 1. First, decide on one main tool.
Tell it "I want an app like this" in chat, and screens and code are auto-generated. v0, Bolt, and Lovable are the leading names. Watching the preview, you just converse — "change this to that" — and it takes shape.
Comparison → v0 vs Bolt vs Lovable
Handle code directly while leaving implementation, fixes, and testing to AI to go fast. Claude Code, Cursor, GitHub Copilot, and Codex are the mainstays. They support extension and maintenance, so they suit products you grow over the long term.
Comparison → AI coding tool comparison
If you're on the beginner route, first look at the app-builder comparison article, pick one, and it's fine to skip this section and jump to Design and UI. On the hands-on route, the next section compares AI editors in detail.
Hands-on route: compare AI editors
The hands-on route's main weapon, the AI editor, currently has four standards. All of them are mainstream, so honestly you won't fail whichever you pick. Just note each one's strengths and choose the one that clicks.
Implements and fixes autonomously across multiple files while conversing. Strong at large jobs you hand off end to end, from design to implementation to testing. Runs on the command line.
In a familiar VS Code-style screen, edit with AI while seeing the code. Completion, chat, and auto-fix are all in one. Ideal for your first AI editor.
The mainstay you drop into VS Code. Smartly completes code as you write. For people who want to add AI without changing their existing setup much.
Hand it a task and it advances the work autonomously in the background. An agent-style option for advanced users who want to run several jobs in parallel.
🧭 If unsure, choose like this. If you want to work with your hands on the code in an editor view, go Cursor (What is Cursor?); if you want to hand off whole jobs through conversation, go Claude Code. Start with either of these two and you won't go far wrong. For the detailed differences, see Cursor vs Claude Code vs Copilot vs Codex.
The knack is to pick one tool and use it deeply. Rather than trying this and that, get one to feel natural in your hands as a partner. Most operations are just "asking in plain words," so the cost of switching to another tool later is small too.
Design and UI — don't overdo it
The look of your app — logo, colors, screen layout (UI) — can now also be prepared with AI. But in the preparation phase, "don't overdo it" is an absolute rule. Get into design and time melts away endlessly. At the MVP stage, the mindset "pass if it's not ugly" is about right.
Generate logos, colors, and screen images from words. Even with zero design knowledge, just ask "in a clean blue palette." For a comparison, see AI design tool comparison.
There are also tools like Claude Design that let you go back and forth between design and code. You can drop the look you made straight into implementation.
Rather than insisting on building your own, riding on an existing UI template or CSS framework (like Tailwind) is smart. A polished look is yours in an instant.
⚠️ Don't melt a day on pixel-level tweaks. Wanting to fine-tune a button's corner radius over and over is a warning sign. As long as design keeps the three points — "decide 2–3 colors, use whitespace, tidy the fonts" — the amateur look disappears well enough. Fine detail can wait until it's launched and people start using it.
The beginner route's app builders tidy up the UI automatically too, so this section can be left almost entirely to AI. Even on the hands-on route, the fastest approach is to have AI produce a draft first, then fix only the parts that bother you. Don't try to design from scratch yourself.
A minimal environment setup
Last is the development environment. Think of this too as "signing up for accounts is about all." Unlike the old days, in most cases there's no need to spend hours building a local development environment. You can start everything on the free tier.
Create an account for the AI builder / AI editor you chose. Try the free plan first, and pay only if it's not enough.
On the hands-on route, get a GitHub account. The standard for saving and version-controlling code. Free to use.
Vercel, Netlify, Supabase, and the like. Generous free tiers. For now just know the names; we cover them in detail in the launch chapter.
✅ Get only what you need "right now." Buying a domain, a paid server, or signing up for a payment service is not needed yet. At the build stage, all you need is a "tool" and a "place for code." Whatever launch and monetization require, gather just what you need when you reach each chapter. Don't pay ahead of time.
Once you register a tool, most AI editors and builders will prepare a full set of the necessary files automatically just by pressing a "new project" button. If you get stuck on environment setup, paste the details straight to AI and ask; it usually solves it. From the prep stage on, AI is already your partner.
Chapter summary
- Pick mainstream tech. A setup AI writes fluently = easy to build = it gets finished. When unsure, let AI suggest it.
- One tool per route. 🌱 beginner = AI app builder (v0/Bolt/Lovable); 🔧 hands-on = AI editor (Claude Code, Cursor, Copilot, Codex).
- Don't overdo design. Make a draft with AI, and tidy just colors, whitespace, and fonts — that's a pass.
- Keep the environment minimal. Signing up for accounts is about all, everything on the free tier. Don't pay ahead.
The foundation is set. Next comes the phase where you finally work with AI to turn your product into something real. Spin the loop of "make it small, run it, fix it" fast — let's look at the concrete how in the next Chapter 4, "Build together with AI."