Ask someone to describe the app they've always wanted to build and watch what their hands do. They start drawing in the air - a box here, an arrow going over there, "and then this thing talks to that thing" - and if there's a napkin anywhere near them, it's gone. Nobody I know thinks in paragraphs. We think in shapes.

And yet the way we've decided people should talk to the most powerful software ever made is a text box. A blinking cursor. Type what you want.

Here's the honest truth about prompting that the rest of us like to skip over. It's a skill, and a strange one. The people getting magic out of these models are mostly people who already write for a living, or code for a living, or spend far too many hours talking to chatbots (guilty). They know you have to set the scene, name your constraints, say what you don't want, and then go back and forth a few more times until it clicks. Most people type one line, get something mediocre back, and quietly decide AI isn't for them. That isn't a failing on their part. We built the door at a height only some people can reach.

But anyone can sketch. You don't need a vocabulary for it, you don't need to know what a database is to draw a little cylinder and point an arrow at it, and you definitely don't need to know how a language model likes to be spoken to. Kids draw before they write. Every meeting that ever got anywhere had someone get up and grab the whiteboard marker, and the moment they did, the room went quiet and everyone finally looked at the same thing.

That's the part I find fascinating. A drawing isn't just a different way of saying the same words. It carries things that are genuinely hard to say. Where something sits on the page tells you how important it is. Two boxes close together belong together. An arrow is a relationship you'd need a whole sentence for, and a crossed-out arrow is an entire argument. You can look at a sketch and hold all of it in your head at once, which you can never do with a paragraph - a paragraph you have to read in order, one word after another, and by the end you've forgotten how it started.

So a picture really is worth a thousand words. It's also a thousand words the person never had to find.

I've been spending my days building exactly this - a board you can draw on, where the drawing becomes the instruction - and it's changed how I look at every AI product I use. The model on the other side has gotten good, really good. The bottleneck has quietly moved. It's not what the machine can do anymore, it's whether a normal person can even ask it.

I keep thinking about the gap between people who have ideas and people who can build them. It's always been a gap of tools and training and time, and for most of history the people on the wrong side of it just didn't build. My parents have ideas. Half my friends who've never written a line of code have ideas. They're never going to learn to prompt, and they shouldn't have to.

If the input becomes a sketch, a lot of that gap closes. Not all of it, I'm not naive. But enough that someone with a napkin and a good idea gets a real shot.

Most people just never had anyone to hand the picture to.