What This Project Taught Me

If I were talking to someone my age who was curious about building something with AI, I would tell them this first:

It is both easier and harder than it looks.

That sounds unhelpful, I know. But it’s the most honest thing I can say after building Beads of Hope T1D, a little ecommerce and community site to raise money and awareness for Type 1 Diabetes through handmade bead products.

On the first day, I gave the AI one enormous prompt. I asked for everything at once: the shop, the customize page, the community area, checkout, the whole thing. Looking back, that was probably like walking into a construction site and saying, “Please build me a house, furnish it, landscape it, and make sure the plumbing works.”

And somehow, about 20 minutes later, there it was. A real-looking storefront. Google login. Stripe checkout connected. A working shop page. The AI had created all sorts of things behind the scenes that I didn’t fully understand.

I remember feeling thrilled and completely disoriented. Like, “Wait. Did I just make a website? Or did I just accidentally open a door I don’t know how to close?”

The first lesson was that AI can give you momentum before you feel ready.

That is wonderful. It is also unsettling.

Then came the walls.

That same first day, none of my product images showed up on the published site. The shop had duplicate category buttons. Every time I clicked where a product image should have been, the page jumped back to the top. I kept telling the AI what was happening, and it kept trying things, and for hours I felt like I was stuck in a loop with a very confident intern who could work incredibly fast but did not always understand what I meant.

A few days later, I hit an error on the customize page: Cannot find module '@/assets/blue-ribbon.png.asset.json'.

I did not know what an asset.json was. I still barely do. I only knew the page was broken, and the error message looked like it belonged to someone else’s life.

So that is lesson two: you do not need to understand every word, but you do need to keep describing the problem in plain English.

“I clicked this and it jumped.”

“The image is still missing.”

“This page is broken.”

That counts. It may not feel technical, but it is useful.

I also learned that tiny design details can feel weirdly emotional.

I spent time moving a blue ribbon graphic down a quarter inch. I asked for the T1D Hope text to stand out more with white shading. I wanted the badge reel photos to zoom when someone hovered over them, then I wanted 3X zoom, then 2X, then only on the badge reels.

Were these life-or-death details? No.

Did they matter to me? Absolutely.

Because I wasn’t just making “a site.” I was trying to make something that felt real. Something that did not look like a hobby I had abandoned halfway through.

Another lesson: caring about the small things is not silly. Sometimes that is where the project starts feeling like yours.

The hardest part, though, was security.

A scan found problems I would never have known to look for: guest cart hijacking, prices that could be controlled from the customer side, and database rules that were too loose. Row-level security, I learned, means the rules that decide who can read or change what in the database.

Reading those warnings made my stomach drop. I felt out of my depth immediately.

The AI fixed them, but I did not walk away thinking, “Well, I’m a security expert now.” I walked away thinking, “Thank goodness this got checked.”

So I would tell someone like me: ask for scans. Ask what could go wrong. Do not assume that because a page looks nice, the important invisible parts are safe.

And then there was the business side.

By June 9, the AI told me the code-side work was essentially complete. That sounded like the finish line, but it wasn’t. I still had to deal with the real Stripe account, real bank information, real shipping, real tax.

That was a humbling moment.

The app being built is not the same as the business being open.

The AI can help wire things together, but it cannot be me. It cannot decide whether I am ready to take the first real order. It cannot feel the nerves of wondering whether anyone will buy. It cannot carry the reason I started.

For me, that reason is my son.

When we worked on the launch video, the first version felt too basic. I told the AI, honestly, that I could have made it on my iPhone. The next version was much better, but one card said, “She was four when everything changed.” I changed it to, “He was two when everything changed.”

Because that is the truth.

That is why I kept going when the images broke, when the errors made no sense, when the security scan scared me, and when the last business steps felt heavier than the building part.

What this project taught me is that AI can build fast, but meaning still comes from a person.

Small takeaway: you do not have to feel technical to begin — you just have to stay in the conversation long enough to learn the next step.