I Built 8 Apps With AI and Somehow That's Not Even the Dumbest Thing I've Done This Year

If you spend enough time online, you'd think AI app development is basically a money printer. Every day there's some guy on Twitter posting screenshots about how he built an app over a weekend and now makes more money than a surgeon while working three hours a week from a beach in Bali.

Meanwhile I'm over here trying to figure out why Apple rejected my app again.

Over the last year I've built RaceIQ, HydroPal, DoughBuddy, Athlo, DiscMode, ClaimCheck, BeanQuest, and a handful of other projects that seemed like brilliant ideas at two in the morning. When people talk about AI app development, they usually focus on the building part. That's because building is fun. Building is exciting. Building lets you feel like a genius for a few hours.

Nobody talks about what happens after that.

Nobody talks about spending an entire evening trying to understand why a build that worked yesterday suddenly doesn't work today. Nobody talks about certificates, provisioning profiles, metadata requirements, subscription configurations, privacy disclosures, screenshots, or the fact that Apple can reject your app for reasons that sometimes feel completely reasonable and other times feel like they were selected by spinning a wheel.

The funny thing is that I genuinely thought building would be the hard part.

I was wrong.

Building is easier than it's ever been. That's the entire reason we're having this conversation. A year ago I couldn't have built most of these products. Today I can sit down with Claude, ChatGPT, Cursor, Codex, or whatever tool we're all obsessing over this week and get something working shockingly fast.

The hard part is getting another human being to care.

That's it. That's the whole game.

You can build an app in a weekend. Convincing somebody to download it is harder. Convincing them to open it twice is harder than that. Convincing them to pay for it is harder still.

RaceIQ is probably the best example. I spent months building features, refining flows, fighting Apple, tweaking onboarding, fixing bugs, and thinking about training plans. Then I launched it and discovered the most humbling fact in software: nobody wakes up in the morning hoping a stranger launches another app.

People care about solving their problems. They don't care about your roadmap.

That's why one real user teaches you more than a thousand views on a launch post. One paying customer will expose flaws in your assumptions faster than any analytics dashboard. One confused person trying to use your product will immediately find the thing you thought was obvious but absolutely is not.

And then there's Reddit.

God help me, Reddit.

I have probably learned more from Reddit than any other platform. I've also been banned from enough subreddits that I'm starting to wonder if there's a secret leaderboard somewhere. Every community has its own rules, its own culture, and its own interpretation of what constitutes self-promotion. Sometimes I deserved it. Sometimes I genuinely have no idea what happened. Either way, nothing will humble you faster than spending an hour writing what you think is a thoughtful post only to watch it disappear because you violated Rule 14 subsection B regarding promotional content on alternate Thursdays.

The frustrating thing is that Reddit is also where some of the best feedback lives. Buried somewhere between the trolls, the arguments, and the guy explaining why every startup idea is doomed, there's usually somebody telling you exactly what's wrong with your product.

They're often right.

The biggest lie in the AI app gold rush is that the bottleneck is coding. Coding is becoming less of a bottleneck every month. Distribution isn't. Trust isn't. Retention isn't. Product-market fit definitely isn't.

Everybody is posting screenshots of revenue. Nobody is posting screenshots of retention.

Everybody is posting launches. Nobody is posting what happened three months later.

Everybody wants to talk about how fast they built something. Very few people want to talk about whether anybody actually came back and used it again.

I still think this is the best time in history to build software. I just think the conversation around it has become a little detached from reality. AI didn't magically eliminate the hard parts. It mostly eliminated the excuses. The hard parts are still there waiting for you the moment you're done building.

Apple is still waiting.

Users are still waiting.

Marketing is still waiting.

And if you're unlucky, Reddit is definitely waiting.

I've built eight apps with AI so far. Most of them still need work. Some are further along than others. Only one has survived Apple's gauntlet. I haven't become rich. I haven't discovered a secret growth hack. I haven't unlocked passive income.

What I have done is learn more in the last year than I did in years of talking about building things.

Turns out shipping teaches lessons that dreaming never will.