Objective
Your mission
Choose a capstone project, define the MVP, write the upfront context, and set up a context file. Output: project_plan.md.
You need: Everything you've learned in this course.
Time: 30 minutes.
Step 1: choose your project (5 min)
Criteria for a good capstone:
- You'll actually use it — a real problem you have
- Achievable in 1-2 sessions — not a social network, not an OS
- Requires a database — practices data patterns from Phase 6
- Requires auth — practices auth patterns from Phase 6
- Can be deployed — practices deployment from Phase 7
Project ideas:
- Habit tracker — track daily habits, streaks, completion
- Expense tracker — log expenses, categorize, see monthly summary
- Reading list — track books, reading status, ratings, notes
- Recipe box — save recipes, tag by category, search
- Workout logger — log workouts, exercises, sets, reps
- Idea journal — capture ideas, tag, search, revisit
Pick one. Don't overthink it. The goal is to ship, not to pick the perfect project.
Write in your file:
## Project Choice
Project: [name]
Why this project: [1-2 sentences — why you'll use it, what problem it solves]
Achievable in 1-2 sessions: [yes — why?]
Step 2: define the MVP (5 min)
Write down what's in scope and what's out.
The planning advantage
Planning before prompting is the cheapest speed-up in this course. The upfront context eliminates the most common AI failure: building the wrong thing. When the AI knows the tech stack, data model, feature list, and design direction from the start, it generates code that fits together.
Write in your file:
## MVP Definition
### IN SCOPE (MVP):
- [feature 1]
- [feature 2]
- [feature 3]
- [feature 4]
### OUT OF SCOPE (later):
- [feature]
- [feature]
- [feature]
Step 3: write the upfront context (10 min)
Before opening any AI tool, write the full upfront context statement:
Write in your file:
## Upfront Context
Project: [name]
Goal: [1 sentence — what it does for users]
Tech Stack:
- Frontend: [e.g., React + Vite + TypeScript + Tailwind CSS]
- Backend: [e.g., Supabase (PostgreSQL + Auth)]
- Deployment: [e.g., Vercel]
Features (MVP):
1. [feature with details]
2. [feature with details]
3. [feature with details]
4. [feature with details]
Data Model:
- [table name]: [columns]
- [table name]: [columns]
Design:
- [style direction, e.g., clean, minimal, mobile-first]
- [layout approach, e.g., card-based]
- [any specific UI requirements]
Security:
- RLS on all tables (users only see their own data)
- No service_role key in frontend
- Environment variables for all secrets
Step 4: choose your tool (2 min)
| Tool | Best For | This Project? |
|---|
| Lovable | Quick UI-first apps, no IDE setup | Good if you want speed |
| Bolt.new | Browser-based full-stack | Good if you want browser-only |
| Cursor | IDE-based, more control | Good if you want precision |
| Claude Code | Terminal-based, file-level control | Good if you're comfortable with CLI |
Use the tool you're most comfortable with. The habits matter more than the tool.
Write in your file:
## Tool Choice
Tool: [name]
Why: [1 sentence]
Step 5: set up your context file (5 min)
Create a context file before your first prompt:
- Cursor:
AGENTS.md in the project root, or rules under .cursor/rules/ - Claude Code:
CLAUDE.md - Lovable/Bolt: put it in the project's knowledge or instructions setting, so every chat starts with it
Write in your file:
## Context File
Context file created: [yes/no — where?]
Ready to use before first prompt: [yes/no]
Step 6: write your first prompt (3 min)
Draft the first prompt you'll send to the AI. It should include the full upfront context plus what to start with.
Write in your file:
## First Prompt (Draft)
[Paste your full upfront context here]
Start with: [e.g., "Set up the project structure, Supabase connection, and auth flow (signup, login, logout, protected routes)"]
Which capstone idea best fits the course's criteria?
A good capstone is something you'll use, done in one or two sessions, and needs a database, auth and a deploy, so it practises every phase.