What do you build when you can build anything?

#ai #discuss #productivity

My hot take is: Telling people that they have to build endlessly is very, very, very bad advice. If you have been remotely on LinkedIn over the last six months, you will have seen two different camps. The AI haters, the ones who don’t want to lay their eyes on your AI-generated code, article or post, and the AI defenders, the ones who say all day long that if we dont build with AI we will be left behind. But which camp is right? Both and neither, all together.

The Cost of Building Everything

I said building everything is a waste. A waste of resources (...poor water), a waste of money, a waste of time and a waste of bytes in a data centre, all for a dead product that maybe only one person will ever discover. The best skill we probably need to develop nowdays is knowing what not to build. In a world of endless possibilities, how do you choose what is a good idea? It is almost an impossible task.

It all started with a random idea I had for an app. I like photography, and i often go for photo walks in my cities. My idea involved creating an app that helps you create/attend/do a photo walk. Sometimes I am already out with my camera and want to find the next place to take photos, but I don't know exactly where to go. Or maybe I want to find a well-known photospot and try to recreate a particular photo. What sounds like a simple feature quickly becomes a set of different user needs. I started sketching the idea, using AI to explore the flows and turn the word vomit in my head into something tangible. But I had to stop and ask myself: is this actually worth building? When I go for my little photowalks, I actually like to getting lost, and find unexpected things. Is this even a good idea?

One would argue "why not" or even not think to question the idea at all. I could open my IDE, ask an agent to build the first version and have something running in a few hours. I could spend a weekend on it. I could launch it. I could even convince myself that I am doing something productive. But none of that answers the question. It only answers whether I can build it.

Welcome to Agency Hell

Engineers love to create their own personal versions of hell. First we had the tutorial hell, then we upgraded to the side-hustle culture hell. And now, apparently, we are in the 7th ring of founder and agency hell.

For a long time, there was a constraint between having an idea and building it. You needed to know how to code, find a designer, learn a new framework, figure out deployment, pay for infrastructure, convince someone to use it and somehow find the time to do all of that while having a job. So, most ideas died somewhere along the way, and that was probably a good thing.

AI removed a lot of those problems. You can have an idea in the morning and have a prototype by the evening. You can generate an app, the accompanying landing page, write the backend, create the db schema, produce the copy (with the compalsory em dashes) and deploy the thing without needing much more than some free time and a few API tokens (it could many $$, but that's an other discussion).

To some people that sounds like a freedom. But freedom without direction can become its own kind of trap.

When everything is possible, every idea starts to feel like an obligation. You see someone building an AI wrapper, so you build one. Someone launches a micro-SaaS, so you launch one. Someone makes €10,000 from a weekend project, so you start wondering why you haven’t done the same. This is what I mean by agency hell. Not the inability to build, but the constant feeling that you should be building.

What Do You Actually Care About?

When you can build anything, what do you build, though? That question is very personal for each of us. We can rephrase it: what do you care about? That is a much more difficult question to answer. But (in my humble opinion) that will break or make your relationship with AI.

I don't know anyone that grew up with aspirations on making the best accounting software, these kind of software comes from a need of understanding a complex domain. AI can help you build this kind of software, but it cannot give you a reason to care about the problem.

Most people care about their families, their hobbies, their friends and the things that make their lives meaningful. Those are the places where interesting problems hide. You might build something because your mother struggles with a particular task, because your dance studio has an annoying process, because your photographer friend struggles with their workflow, or because something in your own life keeps bothering you.

Not every hobby needs an app. Quite the opposite. The point is that caring about something gives you a reason to notice problems that other people don't see.

Possibility Isn’t Purpose

In a world of possibilities, we probably have to become more selective about what we build. AI has changed the economics of software creation, yes, but there are still only so many hours in a day, only so many people who will use your product and only so many problems worth solving.

Most people think AI unlocks some secret chest of money. The truth is that it doesn’t. It makes execution cheaper. It makes experimentation faster. It gives more people the ability to turn an idea into something real.

But it cannot tell you which idea deserves your time.

Because if you don't care about what you are building, congratulations, you just unlocked a whole new world of endless misery.

I would really love to hear about your thoughts about this topic. Yesterday I saw at least 2 posts of identical ideas being implemented by different people. Do we squeeze ideas to fit in the market in non-existing problems (create our own new kind of hell), or we solve problems people (or ourselves) have? I know that not all software are build for usefulness, sometimes being fun is enough, but i seem to struggle to recognize if my ideas are actually good and worthy to bring into life. So what is your take?

This post was written by human (me!) and passed through an AI grammar revision.