I Built an App with AI — Here's What Nobody Tells You

So, yes. I built an app with AI.

It is called FoxDesk. It is public. It works. People can download it. Which is already enough to trigger two very predictable reactions online.

The first is: "See? This proves nobody will need developers anymore."

The second is: "This changes nothing. Real software engineering is still exactly the same."

As usual, both sides are a little too excited about being wrong in public.

The One-Prompt Fairy Tale That Wasn't

What actually happened was much less cinematic and much more useful.

This was not one of those stories where I typed a single brilliant prompt, took a sip of coffee, and watched a production-ready application descend from the heavens like the Second Coming of SaaS. It took more than a month of work. Real work. Thinking, testing, pushing back, redoing things, throwing things out, making tradeoffs, and repeatedly stopping AI from turning a simple app into a small technological empire.

Which, to be fair, it was more than happy to do.

At one point, what eventually became FoxDesk — a roughly 1 MB helpdesk package — could easily have ended up weighing hundreds of megabytes. AI kept suggesting dependencies, libraries, layers, abstractions, and enough moving parts to make it look like I was not building a helpdesk at all, but accidentally founding a new framework religion.

So I started over.

Why AI Doesn't Give You Taste

That experience taught me something I think a lot of people still miss: AI can help you build software much faster, but it does not automatically give you taste, restraint, or judgment. It does not tell you what should stay small. It does not tell you when something is becoming ridiculous. And if you have never built software before, that matters a lot, because you often do not even know what "reasonable" is supposed to look like.

AI will very confidently suggest a solution that could support interplanetary ticket routing if you let it. That does not mean your helpdesk needs it.

And that, I think, is the real split in the current conversation.

On one side, you have people posting one-prompt fairy tales on X. On the other, you have people insisting this is all temporary, developers are fine, and AI is just another overhyped phase that will pass like every other tech obsession.

I do not believe either version.

AI is not magic. But it is also not going away.

It is already useful. It is already changing how software gets built. And it is getting better fast.

More People Will Build — And That's a Big Deal

That means a lot more people will be able to build things they would never have attempted before. Tools, small apps, internal systems, weird experiments, personal products, ugly first versions of good ideas. That is a big deal. The old sentence, "I'd build it if only I knew how to code," is suddenly much less final than it used to be.

But there is a catch, because there is always a catch.

Building your own software with AI still costs time, money, energy, and attention. Sometimes a surprising amount of all four. And many people will discover, after proudly deciding to build instead of buy, that the boring old SaaS tool they were mocking last week is actually the cheaper option once they account for everything.

That is not a failure of AI. It is just adulthood.

Yes, AI is going to create an absurd amount of new software. Some of it will be genuinely great. A lot of it will be terrible. Vast, majestic amounts of nonsense will be produced at previously unimaginable speed. That is what happens when leverage becomes widely available. More things get built. More ideas get tested. More garbage appears. A smaller percentage remains useful.

Pareto will be just fine.

What Changes for Developers

For developers, though, this shift is real and personal.

AI feels a bit like a superpower. Not a replacement for skill, but an amplifier of it. A good developer can move faster, prototype faster, explore more paths, kill bad ideas earlier, and spend less time wrestling with the parts of the job that used to eat days for no good reason.

That part is great.

Because if everyone now gets access to similar coding leverage, the real differentiator shifts elsewhere.

Not to who can type more code from memory.

To who has better ideas, better taste, better judgment, and better execution.

That is where the game is moving.

The Part Nobody Talks About

Here is something most builders will not say publicly: there is a strange grief mixed into all this progress.

Coding used to have a quiet, personal rhythm. You would get stuck, sleep on it, and sometimes the solution just appeared the next morning out of nowhere. That tiny private victory was part of what made the craft addictive.

AI took that away. Not maliciously — just by being fast. Too fast, sometimes, for the satisfaction to keep up.

But nostalgia is a terrible strategy. The craft is evolving, so the enjoyment has to evolve with it. For me, the real thrill was never syntax or debugging anyway — it was watching a raw idea become a working product. That part got better, not worse. So I am leaning into it.

Management Still Hasn't Caught Up

Management has not fully caught up to this yet, which makes the whole thing even more entertaining.

One group now thinks every feature should take five minutes because "can't AI do it?" The other still behaves as if every new product decision requires half a year, three workshops, and a small support circle. One side believes AI can do everything instantly. The other acts like nothing has changed.

Both are going to have a confusing few years.

The truth, of course, is in the middle. AI is not a toy, but it is also not divine intervention. It is leverage. Extremely useful leverage. And like all leverage, its value depends mostly on the person holding it and whether they understand what they are trying to move.

Leverage Is the Real Story

That is also true outside the app itself.

A few years ago, I probably would not have written this article at all. Not because I had nothing to say, but because writing it properly would have taken time I would much rather spend with my family.

This time, the article started on a walk along the beach. I dictated the rough ideas, sent them from Telegram through my Watson, OpenClaw, and ended up with something I could actually publish instead of just another note that would quietly die in my phone.

The blog runs on Cloudflare for free, I manage it through a chatbot, and the whole thing is built with Astro. No admin panel. No plugin zoo. No backend drama.

AI is not just helping people code faster. It is helping them keep things smaller, simpler, and actually shippable. And that may be the most important part.

What Comes Next

In the near future, I do not think AI will do everything for us.

I think it will do something more practical than that. It will give a lot more people a lot more leverage.

Some will use it to build useful tools, real products, and things that would have taken months before.

Others will use it to build absolute nonsense at record speed.

Which, in fairness, is still progress.