What I'd Do Differently: Honestly, Not Much

People ask me, now that I've been building for a while, what I'd do differently if I could start over.

Honestly? Almost nothing. And I mean that in the least humble-brag way possible: I'd change almost nothing because I was in no position, on day one, to make better choices than the ones I actually made. I would have been the same clueless person, doing my clueless best.

The specific clueless things I did turned out to be the right specific clueless things.

Where I started

Zero coding background. Zero product background. Zero tech background beyond being a person who knows how to send an email and, on a good day, join a Zoom without needing help.

What I had was a bunch of ideas rattling around from years in industries where I'd noticed problems that nobody was solving well. That was it. That was the whole starting inventory.

The one thing I did that mattered most

Before I built anything, I opened ChatGPT and started talking. Not "help me code." Not "build me an app." Just:

"I have this idea. Here's what I think it is. Here's who I think it's for. Tell me if it makes sense. Push back. Be honest."

Then I let it push back. When it said something that stung a little, I sat with it. When it said something obvious, I noted it. When it said something that made me realize I hadn't thought about a whole angle, I chased that angle down.

That conversation, that first cheap conversation with an AI that has no reason to butter me up, was the single most valuable hour of the entire process.

The second thing that mattered

Once I had an idea that had survived being questioned — an idea I could describe in a paragraph without lying to myself about who it was for and what it did — I asked ChatGPT to help me write a really strong prompt to hand to Lovable to actually build it.

That's the whole trick. I didn't try to write the perfect prompt from scratch. I used the AI to help me articulate the request to the other AI that was going to do the building.

Then I copied that prompt over, sent it, and watched something appear on my screen that had a shape approximately like the thing I'd described.

It wasn't finished. It wasn't right. But it existed, which changed everything, because now the project was no longer an idea. It was a rough draft you could actually poke at.

What that first draft looked like

Not great, honestly. A first draft never is.

But once something existed, I could see what was wrong. I could see what was missing. I could see what I actually cared about that I hadn't even known to ask for. Every subsequent prompt was in reaction to something real, not to something imagined.

That is a completely different creative posture than starting from a blank page — and it's a much easier one for a non-technical person to operate from.

What I'd tell someone in my shoes right now

If you're staring at this exact starting line and want the shortest version of the advice:

Open a general-purpose AI. Talk through your idea until it survives real pushback or you find a better one.
Once you have an idea that holds up, use that same AI to help you draft a strong first prompt for a building tool.
Send the prompt. Look at the result. React to what's actually there.
Repeat, in small honest steps, for longer than you think you should have to.
Don't quit on a Tuesday.

That's it. That's the whole method. It's not a secret. It's not clever. It's not going to be on the cover of a business magazine. It just happens to work.

The thing I keep coming back to

You do not need to know what you're doing to start.

You need to be willing to not know what you're doing for a while, out loud, in front of an AI that doesn't care. You need to ask a lot of questions. You need to trust the process enough to keep going through the ugly stretch. You need to be curious more than you are ashamed.

Everything else — the prompts, the frameworks, the tools, the vocabulary — you'll pick up on the way. You do not have to earn permission to start. Nobody's issuing that permission. You just start.

The takeaway

I wouldn't do much differently. I'd probably drink more water and go to bed earlier and yell at my laptop a little less. But the actual method — talk to the AI about the idea, let it push back, write a good first prompt, look at what shows up, then keep going — is exactly the method I'd recommend to anyone in my position now.

Good or bad, that's how it starts. Good or bad, that's how it keeps going.

The only thing that matters is that you start.