This is the moment
You started this course as a designer. You're finishing it as someone who can design a product, build it, ship it to the internet with auth and a database, and wire in AI features. Solo. In one day. Without needing a backend developer.
That's not a small thing. Most people who call themselves "technical" can't do all of that. This capstone is your proof of work.
💡 Pick a small idea
Scope aggressively — ship it today
The number one mistake in capstone projects is picking something too big. A great capstone idea has: one clear user, one core action, a way to see results immediately. Not a platform, not a marketplace, not an "app like Notion but better." Pick something you can ship in 3 hours.
Product ideas that work well
-
AI critique tool — paste a design description, get structured UX feedback from Claude. Simple form + API + display results.
-
Team mood board — team members sign in with Google, post what's on their mind (mood + text). Everyone sees real-time updates. Firebase.
-
Micro-learning cards — enter a topic, Claude generates 5 flashcard Q&As. Users can save their favorites to Firestore.
-
Job title generator — describe what you do, Claude generates 10 creative job titles. Share button copies to clipboard.
-
Sprint retro board — team members post what went well / what didn't / what to try. Real-time with Firebase Auth.
The 5-milestone structure
Follow these milestones in order. Each builds on the previous. Don't skip ahead — a half-built product is worse than a simple complete one.
🎯
Milestone 1 — 30 min
Define your product (one page, no Figma yet)
-
1Write 3 sentences: Who is this for? What is the one thing they do? What do they see when it works?
-
2List the screens (max 2): a landing/sign-in screen and the main app screen. That's it. No dashboard, no settings, no profile page.
-
3Decide: does this need auth (Firebase)? Does it need AI (Claude API)? Does it need persistent data (Firestore)? Only add what's essential.
🖊
Milestone 2 — 30 min
Design in Figma (wireframe, not high-fidelity)
-
1Open Figma. Create a frame at 1440px wide (desktop) and 390px (mobile). Draw the layout with rectangles and text — no colors, no final fonts.
-
2Define the components you'll need: a card, a form, a button, a header. Sketch each one.
-
3Screenshot the wireframes. You'll use these to brief Claude.
💻
Milestone 3 — 60 min
Build the UI with Claude
-
1Paste your wireframe screenshot into Claude with a prompt describing the product, design style (dark/light, colors, fonts), and what each part does. Ask for HTML + CSS as a starting point.
-
2Iterate at least 3 times. Each round: describe one specific thing that needs to change (not "make it better" — "the input should have more padding and the button should be yellow").
-
3Get the UI to a state where it looks right before adding backend. A good-looking static version is a better foundation than a half-connected mess.
🔥
Milestone 4 — 45 min
Add Firebase auth + data (if needed)
-
1Set up a new Firebase project (or reuse your one from Chapter 14). Enable Google auth. Create a Firestore collection for your product's data.
-
2Ask Claude: "Here's my HTML. Add Firebase using CDN imports. Add Google sign-in. On auth, show the app and hide the sign-in screen. Add [describe your data write operation]. Add [describe your data read/display operation]."
-
3Test it locally: sign in, post data, confirm it appears in the Firebase console. Fix any issues before deploying.
🚀
Milestone 5 — 15 min
Deploy and share
-
1Push your code to a GitHub repo. Import it to Vercel. Deploy. Add your Vercel URL to Firebase's authorized domains list.
-
2Share the live link with at least one real person who isn't a developer. Watch them use it without explaining anything. What confuses them? Note it for v2.
-
3When you're done — click the button below.
What you can do now
- Design a product in Figma, build it in HTML/CSS/JS, ship it to the internet — solo
- Add user authentication (Google sign-in) without writing a backend
- Store and retrieve data in real-time with Firestore
- Integrate AI into a product feature using the Claude API via serverless functions
- Deploy to production with CI/CD — every push goes live automatically