Features that used to take me weeks, an agent now builds while I make a coffee. And what does my industry do with that? Doomsday mood. I think we should start building really big things instead.
I come from the pre-AI era
Before this sounds like LinkedIn hype: I know what it was like before. I built my first websites at 16, moved on to no-code tools after that, and at 18 we founded Relationflow. Our first product was a CRM, built entirely with a no-code tool. It took 15 months to finish.
And that was no rocket science. A proper permission system was a project for weeks. An export that didn’t choke on umlauts ate days. Every little thing felt like running a marathon with a backpack on.
I later built the second version of that CRM on my own, with AI. Seven weeks.To be fair: the second time around I already knew what I was building. That makes a real difference. Just not a factor of nine.
You used to need twenty people
If you wanted to build something genuinely new a few years ago, a platform that turns a whole field upside down, you needed a team of twenty. Frontend, backend, mobile, infrastructure, design, testing, product. Plus the money to pay all of them for a while, so a few million, fast. For most people the idea was dead before it started.
Today a small team of one or two people pulls that off in a few months. That’s not a thought experiment, it’s how I work every day.
When building is no longer the bottleneck, the idea is.
And then everyone builds to-do apps
Really, this forces us to think much bigger. And what do I see every day? People using the most powerful tools that have ever existed to build the hundredth to-do app. Or the next chat wrapper around a language model. You can do that. It’s just a bit like renting a crane to put a flower pot on your windowsill.
Meanwhile the projects nobody could afford before are suddenly realistic. Software for industries that have been running on Excel and fax machines for twenty years. Tools only big corporations used to have, for the business with thirty people. Whole platforms instead of single features. There’s so much lying around that simply nobody has built yet.
Why is everyone so scared?
What honestly annoys me is the mood in my own industry. On Reddit, on X, in every dev community it’s the same: a new thread every day asking whether programming is dead now. Whether you should still tell your kids to study computer science. Whether anyone will need developers in two years.
I get the worry, I really do. But for the first time ever, we can build projects we used to only talk about. That’s a great thing. And anyone who thinks this doesn’t need developers anymore has never steered a project for three months.
The agents type, you steer
A big project still takes weeks or months, agents or not. Hundreds of decisions get made along the way. What does the data model look like? Where does one service end and the next begin? What happens when two people save at the same time? What do we deliberately not build? An agent gives you an answer to every one of those questions instantly. Whether it was the right one, you often only find out weeks later, when everything is already built on top of it.
That’s exactly why we still need people who understand the architecture underneath and steer a project like that in the right direction. And even as the models get better every year, developer tasteI looked for a better word. There isn’t one. makes a difference like night and day. Give two people the same agents and you get two completely different products. One runs. The other is one you still want to touch in two years.
So: build something big
If you’re a developer: less fear, more ideas. Your job is getting bigger right now. You’re now the person who can replace an entire team, as long as you know where it’s supposed to go.
And if you’ve got an idea in a drawer that you never dared to touch because it would have needed twenty people and a few million: run the numbers again. It may well have just become realistic. And if you’re looking for someone to build it with, you know where to find me.
