🏠 Home / Hub

AI Vibe 01 — Intro to AI-Assisted Dev

AI-assisted development — "vibe coding" — is the practice of using AI tools as a core part of your development loop to write, review, debug, and ship code faster.

1. What is AI Vibe Coding?

Vibe coding is not about letting AI write everything blindly. It is about using AI as a pair programmer that is always available, knows most syntax, and never gets tired — while you stay in charge of architecture, review, and judgment.

You remain the engineer. AI outputs drafts — you review, test, and decide what ships. "Vibe" refers to staying in flow state by delegating tedious boilerplate to AI.

2. AI Tools Landscape

ToolTypeBest for
Claude CodeAgentic CLIFull codebase tasks, autonomous file editing, bash commands
CursorAI-first IDEIn-editor chat, inline edits, codebase-aware suggestions
GitHub CopilotIDE pluginLine/function completions inside VS Code, JetBrains
v0.devUI generatorReact/Tailwind UI from text description or screenshot
Bolt.newFull-stack generatorScaffold entire apps (React, Vue, Node) from a prompt
Claude.aiChatGeneral-purpose: planning, review, explanation, writing

3. How AI Coding Works

AI models work with a context window — a limited amount of text they can see at once. Everything you send (code, instructions, conversation history) consumes that window.

# The basic loop:
You:  "Write a Vue 3 composable that fetches user data from /api/users
       with loading state and error handling."
AI:   [generates useUsers.js]
You:  [test it, find it doesn't handle 401]
You:  "It doesn't handle 401 responses — redirect to /login on 401."
AI:   [updates the composable]
You:  [test passes, commit]

4. What AI Is Good At vs Where to Be Careful

AI excels atBe careful with
Boilerplate & scaffoldingBusiness logic with complex domain rules
Common patterns (CRUD, auth, REST)Security-critical code (always review)
Syntax and standard library usageCutting-edge APIs (training data may be outdated)
Explaining and documenting codePackage versions (may suggest outdated ones)
Writing unit tests for pure functionsComplex integration tests with many dependencies
Refactoring for readabilityPerformance-critical optimisation (benchmark yourself)
Regex and SQL query generationKnowing when NOT to use a certain approach

5. Setting Up Your AI Dev Environment

Claude Code CLI

npm install -g @anthropic-ai/claude-code
claude   # starts an interactive session in your project dir

Cursor IDE

1. Download from cursor.com
2. Open your project folder
3. Cmd/Ctrl+K → inline edit
4. Cmd/Ctrl+L → open chat panel (codebase-aware)
5. Add .cursorrules file to root for project-wide instructions

VS Code + GitHub Copilot

1. Install "GitHub Copilot" extension
2. Sign in with GitHub account
3. Start typing — suggestions appear inline (Tab to accept)
4. Ctrl+I → inline chat for edits
5. Ctrl+Shift+I → sidebar chat

6. The AI-Human Collaboration Loop

  1. PROMPT
     Describe the task clearly with context and constraints
          |
          v
  2. REVIEW
     Read AI output carefully — don't blindly accept
          |
          v
  3. EDIT
     Fix issues, adjust style, fill in domain knowledge
          |
          v
  4. TEST
     Run the code — check all branches and edge cases
          |
          v
  5. COMMIT
     Write a meaningful commit message describing what changed
          |
     (loop back to PROMPT for the next task)
Smaller, focused prompts produce better results than one giant prompt. Ask for one function at a time, test it, then move to the next.

7. Getting Started Exercise

Try this exercise to feel the iteration loop:

  1. Open Claude Code or Cursor in a project folder
  2. Prompt: "Write a Vue 3 composable called useCountdown that counts down from N seconds and emits a 'done' event when it reaches 0."
  3. Read the output. Does it use ref and onUnmounted to clean up the interval?
  4. If not, follow up: "Make sure the interval is cleared when the component unmounts."
  5. Ask: "Now write a unit test for this composable using Vitest."
  6. Notice how each iteration refines the result based on your review.
The goal is not to get perfect code on the first try. The goal is to move faster through the draft-review-refine cycle than you could writing from scratch.

📌 Study Checklist