“What if…?”: how AI got the ideas out of my head

“What if...?”

That was the title of the talk I gave yesterday as a guest speaker in Antonio Iturra N.'s Leadership + AI course at the University of Santiago, Chile

They're questions I ask myself every day. What if this didn't have to be done by hand? What if the data just showed up on its own, instead of me having to go hunt for it? What if I made a video of this thing I'm showing? What if something read my emails and messages and told me out loud which ones matter? I ask them about everything: work, the house, my kids. For everyone else it's exhausting. For me it's the most fun thing there is.

The problem is that those questions used to just stay there. I could see the whole solution, built, working perfectly, inside my head, and no way to get it out.

But when I went all in on AI and automation, everything changed. Three years ago the first thing I made was a game to help my son learn geometry. It was suspiciously easy. He liked it so much he started asking me for more, and he was learning a lot. But that was just the beginning...

Then came surveys, websites, images, presentations, videos, animations. Every thing I tried opened up three more. Until I got to data, which is what fascinates me, and I never left: dashboards that are insanely complicated on the inside and simple to look at on the outside (that's the whole trick), 20 agents that work with me, skills I custom-built for myself. Today the things I make are used by people I don't even know, to decide things that matter.

And every thing I automated gave me back the hours to build the next one.

At the end of the class, someone asked me if I wasn't becoming dependent on AI. Partly, yes, I told them. But the agents I built think the way I think, because I built them around how I reason. If tomorrow everything goes down — I run out of credits, the internet cuts out, whatever — I lose speed. I don't lose the way I solve things.

Because the questions are still mine, and so are the answers. I'm a hopeless questioner: I don't trust a single thing it gives me, and I check it with the same unbearable persistence I use to check myself. What AI does give me, by far, is speed (Thought it, done)

When people talk about human in the loop (a human right there inside the circuit, checking.) I'm not in the loop: I am the loop haha. And on top of that I torment my work tools a little: at 4 in the morning, the one sending me to bed is Claude or Codex because I've worn them out!

Every time someone asks me, “hey, could this actually be done?”, I give the same answer: if it's about whether it can, it can. So if you have one of those ideas you shelved years ago because it couldn't be done.
Chances are that today it can.