Start Smaller Than You Think

If another 50-year-old asked me today whether they should try building an app with AI, I would not say, “Yes, go change your life.”

That sounds nice, but it is not useful.

I would say this: start much smaller than you think. Then make it smaller again.

I’m building MyTradePilot, a stock-trading opportunity scanner, with no coding background. I do not say that as a humble brag. I say it because there have been many moments where I was staring at the screen thinking, “I have absolutely no idea what this means.”

Early on, I had a scanner button. Very exciting. I clicked it and it returned exactly zero candidates.

Not one.

The problem, I later learned, was something in a file called mockMarket.ts. There was a broken seeded random-number generator, and a hash function was returning a float where an integer was needed. I am writing those words now, but at the time they may as well have been written in smoke.

All I knew was: I clicked the thing, and the thing did not work.

That is where I think people like me have to get comfortable. You do not need to understand every technical word at first. But you do need to be honest about what is broken. “The scanner returns zero candidates” is a clear problem. “Make the app better” is not.

Another lesson: trust is harder than screens.

At one point the app showed a $124,000 phantom gain in the P&L. It was not real. It was just wrong math somewhere. But for a trading app, wrong math is not a small bug. It is the whole building shaking.

An app crashing is annoying. A trading app showing fake profit is terrifying.

That was the moment I stopped treating numbers like decoration. When the scanner finally returned real candidates, and the P&L math checked out against my spreadsheet, that felt better than any fancy feature. It was a tiny win, but a real one.

If you are starting today, I would tell you to build for those tiny wins.

Make one button work. Make one number correct. Make one page say the right thing. Then move on.

Also, pick the name early.

I did not. I started with QuantisAI, then changed to MyTradePilot. That alone was fine, but I ended up with brand mess all over the place. One page said TradePilot Academy. The homepage said MyTradePilot™. Other parts still had old wording. It made the whole thing feel less real, because it was arguing with itself.

And the logo? I wasted too much time there.

I paid for logo generation, rejected the first one, asked for five more, picked a paper-airplane-and-candlestick idea, then changed my mind and uploaded my own image. I can laugh about it now, but I was using logo tweaking to avoid harder decisions.

A logo feels safe. Fixing scary math does not.

Security was another area where I had to admit my limits. A scan flagged things with names like SUPA_anon_security_definer_function_executable and search_path_mutable. I did not know what those meant. I still would not pretend to explain them confidently.

So I asked the AI to fix them, read what I could, and sat with the uncomfortable truth: sometimes I had to trust the builder while knowing I did not fully understand the plumbing.

That is not a great feeling. But it is an honest one.

The first recurring bill was a strange milestone too. I agreed to pay monthly for the market data and infrastructure (the behind-the-scenes services that help the scanner run on real-ish data). That felt like crossing a line. Not because it was a huge amount, but because now this was not just playing around in a browser.

It was a commitment.

So my advice is not “believe in yourself.” My advice is more boring and more useful.

Do not try to build the whole app in one prompt.

Ask for one small thing. Test it. Ask what changed. Check the important numbers yourself. Keep a simple list of what is broken. Do not redesign the logo when you are scared of the next hard step.

And if your app touches money, health, legal issues, or anything that can hurt a real person, be extra slow. My biggest launch fear is not that a user sees an error message. It is that a real user loses real money because of a bug I did not catch.

That fear should not freeze you, but it should sober you up.

Small honest steps are not glamorous. They are just the only way I have found to keep going without pretending I know more than I do.

Takeaway: build the smallest real thing, then prove it works before you build the next one.