Are You Building Because You Can

I have built somewhere between two and five webapps or tools with AI every month for the last three years. Three of them are still active. Everything else is gone.

I am exhausted from vibe coding. Not because the work is hard. The work has never been easier. I am exhausted because building has become cheap, and figuring out what to actually build has become harder than ever.

The blog was not just a build

There were nights in 2023 when I would wait for my Claude usage to reset. Close the laptop. Walk around for an hour. Come back. Refresh. Wait some more. I was building this blog. It was the first thing I had ever made with AI.

I had wanted a place like this for years. Not a Substack page. Not a Blogger account. I had used both. They worked. They let me publish. But they never felt like mine. Someone else's design. Someone else's defaults. Someone else's idea of what a blog should look like. I wanted a space I could shape. A place that looked like how I think.

The catch was that I had no idea how to make a website. I did not know what a frontend or a backend was. I did not know what a database did, where one lived, or why anything needed compute to run. The word hosting was vague to me. I had been telling myself I would build a blog one day for years, but I had no real way in.

What changed in 2023 was not that I suddenly learned to code. AI did not hand me a finished product either. What it gave me was someone to talk to. I would ask basic questions and get real answers. Then better questions. Then questions I would have been embarrassed to ask a person. Each conversation taught me a piece of how software actually fits together. By the time the blog was live, I understood what I had built. I had not just shipped a thing. I had learned a domain.

Most of what I built later is gone

Some CNAMEs expired quietly, without me noticing for weeks. Some apps I stopped opening after the first few days. There is at least one I genuinely cannot remember building. The blog is still here. Two other tools survived. Everything else disappeared.

For a long time I told myself this was just how building goes. Not everything sticks. Prototypes die. That is normal. But that explanation always felt convenient. The honest reason is that those later builds were missing most of what made the blog work.

The blog had a complete cycle behind it. Something sparked, the want for my own writing space. I followed the spark and started exploring what I would even need to build. I checked what I already knew, which was almost nothing, and saw where the gap was. I used AI to learn my way through the gap, conversation by conversation. I built something with that understanding. And then I put it on the internet under my name.

Spark, explore, gap, learn, build, share. Each step did real work.

Most of what I built later was only the build. I had a passing idea, opened a prompt, generated something, deployed it. No real spark. No exploration. No gap to close. Nothing meaningful to share. Just an artifact. The cycle was missing four out of six steps.

You can produce code that way. You cannot produce anything that lasts.

The same pattern shows up at work

Someone builds an internal dashboard in a week and is proud of how fast it came together. Two months later the team has stopped opening it. Someone else adds an AI feature to a product because the competitor has one, not because customers asked for it. The feature ships, gets a line in the release notes, and quietly stops being used.

There is no spark, no exploration, no learning, no real adoption. Just a build. The cycle was collapsed into one step because the build itself became cheap.

The AI builds at work that actually stuck share something with the blog. Someone was already doing the thing in some form. Someone took the time to understand it. Someone learned something they did not know before. Then they built. Then they shared it well enough that other people started using it. Full cycle. The ones that died had the same problem mine did.

My spark for software is fading

I have done that loop now. I know how a frontend connects to a backend. I know why an old MacBook in a corner of my apartment in Munich is enough to serve this blog. I know what compute is and what hosting costs and how to break a deployment in five different ways. The mystery has thinned out. The thrill of building software for its own sake is mostly gone.

That is fine. The spark is meant to fade once the thing it pointed to is no longer mysterious.

But the mind is an ocean. Other things excite me now, and the same cycle applies to all of them. When something genuinely interests me, I let myself stay with it long enough to see if it holds. I look at what I already know and where the gap is. I use AI to learn through the gap, not just to produce things on the other side of it. Eventually I make something with that new understanding. Eventually I share it.

This is why AI matters far beyond building. It can be the partner across every stage of that cycle, not just the last one. Most people only meet the build version of it. They never find out what it is like to think with it, learn with it, explore something with it that they had no entry point to before.

Before you open the next blank prompt

Slow down for a moment and ask which step of the cycle you are actually on. Not what can I build. But what sparked me here, and have I let myself stay with it. What do I already know about this and what is missing. Can I learn it before I start producing. And when I produce something, do I have any real plan for sharing it.

At work, the same questions apply. What is the team genuinely curious about. What do we not understand yet. What would we have to learn to do this well. Who needs to use the thing we build, and how do we bring them along.

The shortcut is to skip all of that and ship something this afternoon. The shortcut is what is making most of us tired. The cycle is what makes building feel meaningful again.

Most of what I built in the last three years is gone because I forgot that. The blog is still here because I did not.