I've been reading everything I can find about how schools should handle AI.

My daughter starts her senior year in a few weeks. My son is three semesters from a mechanical engineering degree. So I read the arguments, and I worry about the right way to teach them in a world that changed underneath the curriculum.

Then last week it landed on me that I've been worrying about the wrong people.

My kids aren't the problem.

They'll be native to this

When I graduated, internet careers didn't exist. Not as a category, not as a title, not as anything a guidance counselor could have described to me.

I've spent the twenty-nine years since working on nothing else.

Nobody taught me that. There was no class. I didn't study for it. The world grew a new layer and I ended up standing on it, the same way my kids will end up standing on this one without anyone teaching them a thing.

The proof that they're going to be fine is me.

The people actually on a clock are the ones who already finished their education. Whose skills were built on a wave that's now just the floor. Who have twenty more working years and a set of competencies that quietly stopped being the differentiator.

That's me too. It might be you.

What actually happened

Web team at a utility company. Then a web content management software company. Then an online community platform.

Look at the direction. Infrastructure, then content, then people. Each time, the layer underneath got easy enough that everybody had it, and the value moved up one. Web architecture was a specialty until it was a commodity, and then the money was in managing what sat on top of it. Content management was a specialty until it was a commodity, and then the money was in the people using it.

That's the pattern, and it works in any field:

The skill that pays is never the new one. It's the one layer above whatever just got easy.

Which is why I keep saying the prompt is the floor. Not because prompting stopped working. Because it started working for everybody.

I didn't see any of it coming

Here's the part I'd rather leave out.

I didn't choose this. I didn't survey the landscape in 1997 and position myself in internet technology. I took a job at a company that happened to be building for the web, and the web happened to be where everything went. Three times in a row I was carried rather than steering.

I didn't see it coming. I was just young enough that it didn't matter.

That's the trouble with using my career as advice. Luck did the work, and luck doesn't repeat on demand. I was twenty-something with nothing to lose and a forty-year runway. Neither of those is true for me now, and I'd guess neither is true for you.

Some people get brought in by the waves. Others paddle out for the bigger ones.

I got brought in. Three times. It worked, and it won't work again.

Paddling out

I wrote Human Traits to put a point of view into the world about what we're becoming alongside these machines. That view has held. Through every hype cycle and every correction since, the thing I believed hasn't needed revising.

A point of view that holds is one thing. Building something other people can stand on is the layer above it.

So that's what I'm doing now, on purpose for the first time. A studio. A second book. Teaching this to twelve people at a time and helping companies adopt agentic workflows.

"Position yourself for the future" is a sentence written by people standing on the beach. I'm out past the break right now. There's no guarantee I catch this wave. There's no guarantee that if I catch it I get up and ride it.

I paddled out anyway. That part is the only part I control.

Praxis

This week, run the Prompt Audit. Ten minutes, three steps.

First, the question that works in any field, whether or not it involves AI: what part of your work got noticeably easier in the last three years? Now name what sits one layer above it. That's where the value went.

Second, open the last ten things you asked a model to do. Count how many were a single question with a single answer, and how many were a process you'd trust to run without you watching. The ratio is your answer.

Third, and don't skip this one. Take the single request you make most often. Write down everything the model would need to already know to do it well without you explaining yourself again. Your standards, your context, the things you'd tell a new colleague on their first day.

That list isn't a prompt. It's the layer above one. You just did the work.

Going beyond the prompt, one layer up, is what the next cohort spends three weeks on. It starts September 8. Details here.

I still don't know if I catch this one.

But when it happens, when the wall stands up behind you and time slows all the way down, that's what you paddle out for.

– Michael Founder, The Drop In & Author of 'Human Traits — a novel exploring humanity's relationship with AI'