Start Smaller Than I Did

If you’re around my age, not technical, and thinking about building something with AI, I’d tell you this first:

Start smaller than your excitement wants you to.

I did not do that.

My first prompt was basically, “How would you suggest we build the greatest global soccer trivia game ever?” Not exactly a modest little warm-up project. The AI came back with a big, shiny idea: part Wordle, part Sporcle, part FIFA. I loved it. Within an hour, I had a working trivia version with a dark “Stadium Night” look and lime accents.

That first hour is dangerous, in the best and worst way.

Because once something appears on the screen, you start thinking, “Well, if it can do that, maybe it can do this too.” A few hours later I asked, “Should we build a fantasy league?” And just like that, my trivia game became a full platform with leagues, head-to-head matchups, salary caps, and a paid soccer data service.

That was the first lesson: AI makes adding things feel cheap. But every new thing adds weight.

The fantasy league worked, but it made every later decision more serious. Money came up. Legal questions came up. I asked whether a paid-entry cash pool might count as gambling. The answer was complicated enough that I decided not to touch the money directly and to deep-link to LeagueSafe instead. That felt like one of the more grown-up decisions I made.

So my advice would be: before you add the exciting second idea, make the first idea excellent.

In my case, I should have stayed with trivia longer. Not just 25-ish static questions sitting in a file. Real depth. Better categories. Image rounds. Audio rounds. Something that actually deserved the phrase “greatest trivia game.” I kept asking the AI if we had the best trivia game on the planet, and to its credit, it did not flatter me. It basically told me no. Not yet.

That stung, but it was useful.

I’d also tell you not to panic when something “breaks.” My first real bug felt ridiculous. I clicked “create account” in fantasy and nothing happened. So I clicked again. Then I got red text telling me to wait 49 seconds.

Naturally, I assumed I had broken the whole thing.

What had actually happened was the system was sending the confirmation email, but the screen gave me no clue. No spinner. No “check your email” message. Nothing. So I kept mashing the button until the app told me to slow down.

The fix was simple: add a loading spinner and a little message saying to check email. But the lesson was bigger: sometimes the app is working and the communication is broken. That is still a real bug. And yes, you are allowed to find those even if you don’t know code.

Another thing I’d say: don’t be embarrassed by tiny design obsessions.

I spent time making the word “football” magenta in a big fantasy headline. I asked for the landing page to be bolder and more exciting. I tweaked colors and the general feel. Maybe that sounds small, but those were moments where the project started feeling like mine, not just something the machine spat out.

At the same time, don’t confuse a polished screen with a finished product. I did that a few times. A nice hero section can make you feel further along than you are.

One more very practical thing: names are easier to change early. I started with Top Corner Entertainment, then worried it was already in use. The AI gave me a list of options, and I picked Hat Trick. It changed the name across the app in seconds — pages, tags, menus, even legal pages and translations. That would have taken me forever by hand.

So if the name feels wrong, fix it before the project gets too tangled.

And finally, accept that there will be parts you don’t fully understand. A security scanner flagged issues. The AI suggested fixes involving things like “RLS policies” and “server-only clients” — which I can explain only as “rules about who can see what” and “keeping sensitive stuff away from the public side.” I approved fixes I didn’t completely understand. That’s uncomfortable, and it should be. But it also taught me to slow down and ask what risk we were actually reducing.

If I were starting today, I’d build one narrow thing first and make it genuinely good. For me, that would have been soccer trivia only — deeper, richer, more fun — before fantasy leagues, payments, legal questions, and everything else.

Small is not less ambitious. Small gives you a fighting chance to finish.

What I learned: with AI, the hard part isn’t starting big — it’s choosing to stay small long enough to make something good.