Vibe Coding with Cursor, Windsurf, and Lovable Build Apps Fast with AI-Assisted Software Development and Testing (Greg Lim)（Z-Library）

# Vibe Coding with Cursor, Windsurf, and Lovable — Reading Guide ## 【One-Line Pitch】 A hands-on, beginner-friendly guide to building real applications by describing ideas in plain English and letting AI agents (Cursor, Windsurf, Lovable) write, test, and deploy the code — perfect for non-programmers and experienced developers who want to prototype faster. ## 【Book Arc】 - **Opening (~0%–11%)**: Introduces the "vibe coding" mindset (coined by Andrej Karpathy) — describing apps in natural language instead of writing code line-by-line. Covers the tool landscape: browser-based tools like Lovable for beginners, IDE-based tools like Cursor and Windsurf for those wanting more control, and CLI tools (not recommended for starters). - **Early (~11%–28%)**: Walks through installing Cursor, setting it up (language, terminal, extensions, privacy mode), and building a first simple app with a single prompt. Demonstrates the iterative loop: ask the AI agent, spot a bug (e.g., a popup blocking gameplay), and have the agent fix it conversationally. - **Middle (~28%–44%)**: Maps traditional software development lifecycle (planning, design, implementation, testing, deployment, maintenance) onto vibe coding. Emphasizes writing a software specification document first — using ChatGPT to draft requirements in Markdown — then creating a phased to-do list to build features iteratively (HomePage, PracticePage, StatsPage, SettingsPage). - **Middle (~44%–56%)**: Covers Git version control fundamentals — initializing a repo, committing snapshots, and rolling back when the AI introduces bugs. Shows how to write custom rules (e.g., "run all tests after changes") so the AI follows your project conventions. - **Late (~56%–61%)**: Explains GitHub integration for cloud backup and collaboration, then walks through deployment to Vercel — linking the local repo, importing the project, and launching the app so anyone can use it via a shareable URL. ## 【Key Takeaways】 - **Vibe coding is a mindset shift, not just a tool** (Early): Instead of wrestling with every line of code, you guide AI agents conversationally — describing what you want, reviewing output, and refining through iteration. The skills transfer across tools, so don't skip sections for tools you don't plan to use. - **Choose your tool based on your experience level** (Early): Browser-based tools like Lovable are ideal for complete beginners — no setup, instant deployment. IDE-based tools like Cursor and Windsurf (both VS Code forks) offer more control but require some programming comfort. CLI tools are not recommended for newcomers. - **Start with a software specification document** (Middle): Jumping straight to "Build a Math Practice app" produces mediocre results. Instead, invest time up front drafting a detailed spec (using ChatGPT as a consultant) covering features, tech stack, and UI — this dramatically reduces back-and-forth later. - **Use a phased to-do list to build iteratively** (Middle): Break the spec into phases with individual tasks marked "not done." Build one feature at a time, test it, and refine. Because LLMs are non-deterministic, your app may differ slightly from the book's examples — that's expected. - **Git version control is essential when coding with AI** (Middle): AI agents can introduce bugs that break your app. Committing snapshots at every working state lets you roll back safely. Simply prompt the agent to "add Git to the project" and commit after each successful milestone. - **Write rules to shape AI behavior** (Middle): Create a rules folder with Markdown-like files (e.g., testing.mdc) specifying "Always Apply" rules like "run all tests after changes" or "ask before adding new tests." This keeps the AI aligned with your workflow. - **Deployment is just another prompt** (Late): With Cursor, you can say "deploy to Vercel" and the agent guides you through linking your GitHub repo, importing the project, and launching it — no credit card required. The full journey from idea to live app can happen in a single session. ## 【Reading Tips】 - **Skim the tool setup chapters** (Chapters 2–3) if you're already familiar with VS Code — the key insight is the Agent mode in Chat, which can read code, create files, and run commands. - **Deep-read the specification chapter** (Chapter 5) — this is the most transferable skill. The prompt template for drafting a spec with ChatGPT is worth copying and adapting for your own projects. - **Don't skip the Git chapter** even if you've used Git before — it shows how to prompt the AI to handle version control for you, which is different from manual Git workflows. - **Pay attention to the "non-deterministic" caveat**: The book repeatedly warns that identical prompts can produce different results. If your app doesn't match the screenshots, that's normal — focus on the workflow, not the exact output. - **The Lovable chapters (14–19) are skimmable** if you're using Cursor — but the concepts (spec documents, GitHub integration, deployment) mirror what you've already learned, reinforcing the transferability of skills. ## 【Coverage Limits】 This guide covers the core workflow (spec → build → test → version control → deploy) demonstrated primarily through Cursor. The excerpts do not cover the detailed Lovable no-code chapters (14–19), Supabase database integration specifics, or the full testing chapter content — those sections are summarized only at a high level. ##