How I built this site with AI in a day
From a template I liked to a live site with a database, an admin portal and real security, in one working day with Claude as my engineer.
I wanted a home for everything I build with AI. Not a portfolio that sits there, but a place where people can ask me questions, watch me answer them, and hire me when they're ready. Here's how it came together in one day, with Claude Code doing the engineering and me making the calls.
It started with a template I liked
I found a design called Departures: an airport board with flip-letter tiles that spell out your name. I loved the feel. But I didn't want to buy a template and swap in my name. I wanted it to be mine.
So I gave Claude the template, my old AimBox repo, and the vision: welcome people into my AI journey, let them ask questions I answer on video, and give businesses a way to book an AI audit or training.
Before writing a single line, Claude asked me four questions:
- Accent color? Signal orange.
- Database now or later? Build first, connect it later.
- Show prices? No. Every offer starts with a call.
- Where do the videos live? Instagram.
That's the part people miss about building with AI. The tool is fast, but it still needs you to decide.
Reusing what I already built
My AimBox store already had login, an admin area and spam protection that work in production. Instead of rebuilding those, we lifted them straight over. Same with a security lesson from PG Customz: we once found that a regular user could make themselves an admin. This site has that hole closed from day one.
The best part of building a lot of projects is that every new one starts further ahead.
Building it
In a few hours the site had:
- A home page with the flip-letter board, my selected work and my AI journey
- An Ask page where you can send me a question
- A Work with me page for audits, training and builds
- This journal
- A private admin where I answer questions, track consulting requests and write posts
It even ran without a database at first, saving everything to a file on my computer, so I could click through every page and form before anything was live.
Then I changed my mind (twice)
Seeing it on screen changed things. I reordered my projects so AimBox leads, cut two that didn't belong, and added real links so every project is something you can actually visit.
Then the big one: I liked the airline look, but not the airline words. "Fares", "Gate C3" and "Tear and send" felt like a costume. We kept the flip tiles and the board, and rewrote everything else in plain language. The design stayed, the gimmick went.
For my journey timeline, Claude went through my own files and GitHub history to find real dates: my first AI agents in January, the n8n automations in February, the AimBox Golf AI drill library in March, and my first real apps in April. Nothing made up.
Going live
Pushing to GitHub put it live on Vercel. Then we connected Supabase for the database. Before calling it done, Claude tested it the way a stranger would:
- Can someone read questions that aren't published? No.
- Can someone see the email of whoever asked? No.
- Can someone skip the site and write straight to the database? No.
Then it sent one real test question through the live form, confirmed it saved, and deleted it.
What I learned
1. Decide first, build second. The four questions at the start saved hours.
2. Reuse everything. Every past project made this one faster.
3. Look at it, then change it. The first version is for reacting to.
4. Test like a stranger. Live doesn't mean safe until you've tried to break it.
Got a question about any of this? Ask me and I'll answer it on video.