I wasn't proud of the projects I built with AI

I'm going to be honest: for a long time I would finish a project and feel no pride at all. The code worked, the deploy was live, people were using it. And still there was that nagging feeling that I wasn't the one who built it. The AI was.

I'd look at the result and think: "if anyone can ask a tool for this, what was my merit here?". That question stuck with me for quite a while.

The day it clicked

Then, after building enough things, I started to notice something: it's not because I didn't type every line that I didn't take part. A lot of it depended on me. Actually, all of it depended on me.

Without someone to orchestrate it, the tool is useless.

The one who decided what to build was me. The one who picked the stack, designed the database, noticed the AI's answer was wrong and made it redo the thing: me again. The one who connected the parts, tested everything in the real world and racked my brain when nothing made sense was also me.

AI writes fast. But it has no idea where it's going. I'm the one holding the map.

The math I do today

After a lot of projects, the way the work splits for me looks more or less like this:

• 50% is knowing what to say. Giving the right context, asking for the right thing, the right way. A bad prompt generates bad code, and you won't even notice if you can't read the result.
• 40% is knowing what you're doing. Architecture, database, security, the why behind each decision. That's what makes you able to say "this is wrong" straight to the AI's face.
• 10% is code. The part everyone thought was the whole job.

And there's a detail: the 10% only comes out right if the other 90% exist. There's no shortcut.

"So did tech degrees die?"

Quite the opposite. I still think a tech degree is extremely relevant, maybe even more relevant now than before.

Because if half the job is knowing what to say, someone has to give you the foundation for that. It's the degree that teaches you what exists, how things connect and why you ask one way and not another. Without that foundation you're not orchestrating anything: you're just pushing buttons and hoping.

An AI answering is worth nothing if you don't even know what to ask.

In practice

Today, at work, I use AI for pretty much everything: automations, code, studying, architecture decisions, review. It's in every step of the process, and the work didn't stop being mine because of it.

Man, the pride came back. Because the project is mine: the vision, the decisions, the responsibility when it breaks. The AI just types faster than I do.

That's it.