Vibe Coding Workflow
The Loop
This is the workflow you’ll use every day. Students generate a fresh build every day from their enriched PRD—not iterating on yesterday’s build. Each day adds a new PRD layer that makes the build richer. By comparing daily builds, you see directly how richer planning yields richer apps.
Planning IS the skill. The better your PRDs, the better the AI builds. Your clinical expertise, design judgment, and structured thinking are what make the apps good—not coding ability.
Your PRD Files
PRDs (Project Requirement Documents) are Markdown files that serve as your project’s “source of truth.” The AI reads them before every build. You add new files as you learn new dimensions.
| PRD Layer | Purpose | When You Add It |
|---|---|---|
| Masterplan | Vision, users, goals, outcomes — the foundation | Day 1 |
| User Journey | Personas, jobs-to-be-done, friction & delight, focal features | Day 2 |
| Design Guidelines | Visual identity, tone, component patterns, accessibility | Day 3 |
| Feature Specification | Detailed feature behavior, data needs, integrations, reach features | Day 4 |
| Polish & Launch | Bug fixes, UX refinements, presentation prep | Day 5 |
Think of these like a chart note: structured, updatable, and authoritative. The AI treats them as its instructions.
How the Tools Connect
- Claude Chat is your thinking partner—for brainstorming, synthesizing, structuring, and critiquing
- GitHub stores everything—PRDs, handoff notes, reflections, app links
- Lovable builds apps from your PRDs (Week 1 primary, Week 2 option)
- Claude Code is the Week 2 evolution: from Chat to Agent—AI interacts with real codebases
- PRD files are the context layer—they tell every tool what you’re building and why
Intellectual Foundations
This workflow draws from two podcast episodes that serve as the course’s anchor texts:
Lazar Jovanovic
- PRDs as sources of truth—Maintain Markdown files the AI reads before every build
- Parallel prototyping—Spin up 2–3 versions simultaneously to find the best direction
- 80% planning, 20% building—Invest in documentation; let AI handle implementation
- Treat AI like an engineer with limited memory—Give it perpetual context through up-to-date files
Nathaniel Whittemore
- Vision-first, not task-first—Start with the big picture before specific features
- Dump first, organize later—Speak messily; let AI structure your thoughts
- Push back and demand critique—AI sounds confident when wrong; ask it to challenge you
- AI as mirror, not oracle—Use AI to reflect your ideas back and see gaps
- Handoff documents—Write a brief note at the end of every session to preserve context
The 80% Challenge
Here’s a challenge for Day 1: think about the apps you use every day. By the end of this course, you should hesitate before signing up for the next one—because you’ll realize you could build 80% of it yourself, customized to your exact needs, for free.