So I mentor a high school kid in a STEM program. We met up recently and he came in all excited, full of ideas about what he wanted to do for his next project. He was bouncing between coding a sudoku game as a website or deploying a public Minecraft server.

I opened my laptop with the intent of simply trying to show him some ideas on the how. I wanted to show him how to write the soduku code all in JavaScript and deploy it via a cloud provider. How to launch a virtual server and download and install the server software, maybe even just find a container image and deploy it.

Within 15-20 minutes, I had literally implemented both of the ideas. Right there in front of him.

Now look. I want to be clear that I wasn’t trying to show him up. I think I was just excited and wanted to show him how all this was achievable. But I felt the mood in the room darken, I saw his face drop. I heard my own voice saying, “But these are valuable skills for you to learn how to do yourself,” and even I thought to myself that sounded lame.

I’ve felt bad about it ever since, and it took me a while to work out why.

I’ve written a lot lately about AI as a multiplier. That the gap between skill levels doesn’t close, it actually widens. That people that have skills and are adopting become exponentially more productive, and others are being left behind. In fact, it’s my opinion that the early-career are the most at risk, not the other way around. Because experienced people who are adopting are the ones who are basically beeing handed a team who will just do whatever they are told. And that is a scary force-multiplier.

But all of that was just theory. Just words I’d written in a blog, that I didn’t really see around myself at work because I was too busy orchestrating and launching the next fleet of agents to go write 97,000 lines of code.

Then I sat across a table from an actual beginner and showed, in about thirty minutes total, that those initial entry-level projects are now basically free. That you can literally in one or two prompts have an AI accomplish in seconds something that was typically a one-to-two week beginner project.

Here’s the part that really gets me. I once wrote about sweeping a restaurant walk-in, and going from forty-five minutes when I started to ten minutes a few years later. Nobody taught me that. I got there by doing the job badly, over and over, until my hands found the better way. There was no shortcut and nobody offered me one.

Experience comes from you DOING the thing, the grunt work.

So what happens when we hand all the grunt work to an AI?

I keep wanting to know what he was thinking. Was it, “Oh cool, this guy can teach me”? Or was it, “Oh, this is all easy and dumb, why should I bother?” I honestly don’t know, and I’ve decided that maybe (hopefully) I’m just projecting. The crisis might be entirely mine.

I did try to follow up with something else important, and here’s the key.

Look, I can have Claude write me a Terraform script that stands up an EC2 instance and deploys a Minecraft server in ten minutes. That part is genuinely solved.

But how do you make sure port 22 is locked down? Do you know why that matters? Is your security group CIDR scoped to something sane, or did you just open it to the world? Have you set up MFA so nobody hijacks your server and burns down every village your friends spent a month building?

And then the part almost nobody thinks about when writing code. How do you get anyone to actually play on this server? How do you tell people it exists? How do you make it fun enough that they come back net week?

None of that is code. All of it is the actual job.

The way I put it to him was this: an AI writing your code is like a spell checker. A spell checker will fix your spelling and tighten your grammar and make your writing better. What it will never do is have something to say. It can’t decide what the story is or why anyone should care. It makes your writing better. It doesn’t replace your writing.

I think that landed. I hope it did.

The lesson learned for me is that this changes the way I mentor. From now on, I’m not touching the keyboard. He builds it, no matter how much he struggles or how hard it is, and I sit next to him while he struggles.

Because the struggle isn’t a bad thing. The struggle is the learning. And I need to be careful about not taking that away from people, because it’s now easier than ever.