Week 1 — Set Up
Get their environment built so they can start from day one.
By the end: a live landing page on the internet, a database connected to Claude, and a working login. They know they can build software.
Week opener — say this up front
- Pick your path: know what you're building? Start Set Up. Not sure? Do the 5-Step Validation in Resources (a lunchtime activity, don't camp there).
- It's not hard, it's new. Unfamiliar is normal. Follow one step at a time.
- Approve everything. Claude will ask permission constantly. Say yes.
1 Accounts + repo Recorded
- Create the GitHub account
- Create the repository
- Add the README
2 Your workspace Recorded
- Install VS Code · install GitHub Desktop · sign into GitHub
- Clone your project onto your computer
- Open the project in VS Code
- Install the Claude Code extension (Anthropic) · sign in
- Say hello to Claude and watch it come alive
Milestone: "your own AI assistant running on your computer." You only do this once.
3 Build + deploy Recorded
- GitHub authentication (the
gh login, ~2 min) so Claude can save and push your work
- Ask Claude to build your landing page
- Set Claude up to commit directly to main
- Let Claude commit and push your changes
- Deploy your site on Vercel → live URL
- Make any other change just by asking
Say it here and keep saying it: a change on your computer is not a change on your site. If you can't see it live, you haven't pushed it.
Add this: "import your repo once." If a deploy fails, fix that project, don't re-import. Re-importing creates a duplicate Vercel project and every push then deploys to all of them.
4 Supabase — the brain Recorded
- Connect Supabase straight to Claude (MCP) → ask for a table, it builds it live. No SQL, ever.
- A table is a smart spreadsheet your app uses
- Users table (who signed up) · progress table (what they finished)
- Runs on PostgreSQL — real, professional database tech for their resume
Missing step, this breaks everyone: their Supabase keys live in .env.local, which is gitignored, so Vercel never gets them. Site builds fine, database unreachable. Either add the keys to Vercel here, or teach the Marketplace Connect Account path (never "Create New Project," that makes an empty database).
5 The login Recorded
- "The moment your website becomes a web app"
- Built with one single prompt
- Sign up, log in, log out yourself
- Watch your test account appear in your database
Not launching to the world yet. Just proving your app can recognize people.
Week 2 — Build
Decide what you're building, then build the product behind the login.
By the end: they can answer the only question that matters, "what have you built?" A real platform they can pull up on screen and show people.
Week opener — say this in Video 1, before anything else
- Buy your domain this week. You won't connect it until Week 3, but get it now. It's cheap and DNS takes time.
- Start your Stripe account and your business/EIN now. You don't use Stripe until Week 3, but these take days to clear. Start them today so they're processing in the background.
- If your state's LLC is slow, a DBA can get you running in a day or two. This is what I did, not legal advice. (Point to the entity-types resource.)
- Already validated? Good. If not, go do the 5-Step Validation in Resources first, then come back.
- Here's the week: decide what you're building → learn what makes people come back → build it and put it live.
1 Reverse-engineer & decide what to build Recorded
- Find proven ideas + your competitors on the marketplaces (TrustMRR live; Acquire, Flippa, IndieHackers in Resources)
- Filter by revenue, study what sells and why, reverse-engineer it better
- Capture designs with SkillUI: capture 2-3 sites, then compose (layout from one, colors from another)
- Pick your platform type: course · quiz · practice · tool
- Nail your unique proposition
- API keys: content/pre-made = none needed. Live AI feature = key + cost per use
- Write your spec, then build it out — hand Claude everything you just decided, read the spec, approve it, let it build
Deck: /week2/video1 · They end this video with a platform, not a plan. The design files describe how it looks; the spec describes what it does. Claude needs both. Ask for the spec before the build so they read and approve it first.
2 What great platforms do (and why) Recorded
- Netflix → persisted progress state ("resume"). Removes the decision.
- Spotify → aggregation query (progress recap). Invisible progress makes people quit.
- Duolingo → counter with a reset rule (streaks). Loss aversion.
- Brilliant → interactive component + scoring. Doing beats watching.
- A copy-paste prompt per mechanic, swap in your expertise
- It's a menu, not a checklist. Pick 2, maybe 3. Not 4.
Deck: /week2/video2 · Underneath, all four are just reading and writing user progress to a database.
Why this lands better after V1 builds: they now have a real platform on screen, so "does a streak actually fit this thing I just built?" is concrete instead of hypothetical. They're deciding what their platform needs, not picking from a menu in the abstract.
3 Brand it & make it legit To record
- Have Claude create your logo, favicon, and OG image → commit + push, they go live automatically (no manual GitHub)
- Terms & conditions · privacy policy · cookies notice
- Refund policy + disclaimer (you'll need these once you sell)
- Study a similar platform's terms to spot what you're missing
- Commit, push, live
Why the actual building isn't its own video: the platform gets built at the end of Video 1, straight from the approved spec. No separate "build" video, and no reintroducing SkillUI, they already have the files.
Why these two pair up, and why legal is here not Week 3: logo/favicon/OG and legal pages are both fast finishing touches that make it feel like a real company. Week 3 is already carrying the domain, Stripe, and the entire Systeme.io funnel build, which is extensive on its own. Legal is quick and belongs with the polish. Week 2 ends on "my platform is live, branded, and legal."
Week 3 — Launch
The connection week. Turn the platform into a real business that can take money.
By the end: live on their own domain, able to accept a real paying customer, and in front of actual people.
Week opener
- You built it. Now we connect everything and put it in front of people.
- Nobody wakes up on Monday and joins a platform they've never heard of. You have to intentionally put your expertise in front of people, and there's a process for that.
- This is where you stop consuming AI content and start owning an AI-powered business.
Connect your domain
- Vercel → Settings → Domains → add the domain you bought in Week 2 → update DNS
Update Supabase URLs Do it with the domain
- Supabase → Authentication → URL Configuration
- Set Site URL to your real domain
- Add your domain to the Redirect URLs allow-list (keep localhost so local work still runs)
Why here and not earlier: during Weeks 1-2, localhost is the correct setting. It only breaks once real people confirm emails on a live site. Connecting your domain is the exact moment this must change, so teach them together or everyone hits it.
Stripe + pricing
- Connect the Stripe account you started in Week 2
- Pricing: monthly vs annual. Lead with monthly for an MVP, grow from there
- Wire subscriptions into the login you already built
Why pricing lives here: pricing and payments are one conversation. Splitting them makes no sense.
Opt-ins + email marketing
- Funnel + email platform (Systeme.io)
- Lead magnet → opt-in → tag → automated follow-up sequence
- Export all your emails and add everyone. Don't sift, don't overthink who sees it
- Link-in-bio agents / chatbots as lead magnets (Chatify example)
Test a real signup, end to end
- Sign up on the live domain as a real user
- Confirm the email lands and redirects to your site, not localhost
- Confirm the account appears in your database
- Run a real test payment through Stripe
A green build proves nothing. The only proof is a real person completing the full flow on the real domain.