It All Started with a Spreadsheet
I wasn't planning to build a website.
I just wanted to make a spreadsheet.
As a financial advisor, I deal with numbers every day. Client asset allocation, insurance planning, retirement calculations — all in Excel. But Excel has one problem: It's really hard for clients to understand.
You throw a spreadsheet full of numbers at a client, they give up after three seconds.
So I thought: Can I make these numbers look better?
Seven Days, Six Pivots
The birth of this product wasn't a straight line. It was six pivots.
Day 1: "I Want to Make a Spreadsheet"
I opened Gemini, used Canvas mode, and told it: "Make me a financial calculator spreadsheet."
It gave me a basic table. Functional, but no different from Excel.
Day 2: "Can We Visualize This?"
Looking at that table, I thought: Can the numbers become charts? Bar charts, pie charts, so clients can instantly understand their asset allocation?
Told AI, it made them.
That's when it started getting interesting.
Day 3: "Can This Become a Web Page?"
If this thing could become a web page, I wouldn't have to email Excel files to clients anymore. Just send a link, client opens it on their phone.
"Can this become a web page?"
AI said yes. Gave me a bunch of HTML files.
I didn't understand them, but they worked.
Day 4: "How Do I Make This a Website?"
The web page was done, but I could only see it on my computer. How do I let clients see it too?
That's when I first heard the word "deployment." First time encountering Vercel. First time understanding what "putting something on the internet" means.
Deployed it. Opened the URL — there it was.
Day 5: "The URL Is Ugly"
Vercel's default URL was a random string. something-random-123.vercel.app.
I can't send this kind of URL to clients.
Spent a day on custom domains. Buying a domain, setting up DNS, waiting for propagation. I'd never heard any of these terms before.
Day 6 and 7: "Ugly → Fix → Ugly → Fix"
All the features were there, but it looked ugly.
These two days were basically one loop:
Look → "This part is ugly" → Tell AI to fix → Fixed → Look → "That part is ugly too" → ...
Wrong font, too much spacing, clashing colors, mobile layout broken — fixing them one by one.
That Moment
On the evening of day seven, I picked up my phone, typed my URL into the browser.
The page loaded.
I saw something I built. On my own phone. At my own URL. Anyone could open it.
In that moment, I knew everything was different.
Not because the website was impressive. It was rough, basic features, design was just "okay."
But because — I did it.
A financial advisor who can't code, using AI, in seven days, went from a spreadsheet to a website.
If this was possible, what else couldn't be done?
I Almost Gave Up in the Middle
Sounds like it went smoothly? It didn't.
Between day three and day four, I almost quit.
Because of Firebase.
When I wanted the website to "save data," AI started talking about Firestore, Collections, Documents, API Keys, environment variables — a bunch of things I completely didn't understand.
Every time I followed AI's steps, I got errors. Fix one place, another place breaks.
I spent an entire day stuck on the same problem.
That night, I seriously considered: Forget it, I'll just use Excel.
But the next morning, I opened AI again. Not because I had some epiphany — because I wasn't willing to give up.
How did I eventually solve it? Just fed the error messages to AI one by one, let it fix them one by one. No eureka moment — just grinding through it.
The Chain Reaction After Finishing
After UltraAdvisor went live, my first reaction wasn't "I did it."
It was: "I need to get this out there."
I started thinking about how to list it, how to promote it, how to get more people to see it.
Then I realized: the promotion process itself needs tools. Need automated posting, scheduling, tracking data.
So I built more things.
Then I thought: the process of building these things is itself valuable. What if I documented every step and turned it into a deliverable service?
That's how Ultra Lab was born.
A spreadsheet
→ A visualization tool
→ A website
→ A brand
→ An automation system
→ A service company
None of it was planned. Each step was "finish the last one, naturally see the next one."
Why Your First Product Matters
Not because it'll succeed. It probably won't. My first version of UltraAdvisor was rough too.
It matters because: It changes how you see yourself.
Before building your first product, you think "building products" is for other people. For engineers, for entrepreneurs, for people with resources.
After building it, you know: I can do this too.
This shift in self-perception is worth a hundred times more than the product itself.
Because once you have the first one, there'll be a second, a third. And each one will be faster and better than the last.
How to Start Your First Product
Don't think too big. Don't think "I want to build a world-changing app."
Think of something you need. Something small, specific, something you actually use today.
Something you do in Excel every day → Can it be automated?
A tool you think sucks → Can you make a better one?
Something you repeat at work → Can AI do it for you?
Then open AI and say: "I want to build this."
It'll ask you for details. You answer. It'll start building. You check the result. Wrong? Say it's wrong. Right? Keep going.
Seven days later, you might also pick up your phone, open your own URL, and see something you built.
In that moment, you'll know what I'm talking about.
This is part five of the "Getting Started" series. Previous: Why You Don't Need to Learn to Code
Want more free resources? Join the Solo Lab Discord.
Discord: https://discord.gg/ewS4rWXvWk
Originally published on Ultra Lab — we build AI products that run autonomously.
Try UltraProbe free — our AI security scanner checks your website for vulnerabilities in 30 seconds: ultralab.tw/probe
