I’ve been working in IT and around technology and information for more than a decade. I can clearly see the accumulation.
Every year I know more. I recognize patterns faster. I make fewer stupid mistakes, or at least invent more sophisticated ones. The people around me evolve too. But there’s one skill I seem to be getting worse at: stopping.
How do you not flash an ESP board until 2 a.m. because you suddenly had an idea? How do you not keep building a project until six in the morning because one more thing would make it work? How do you spend Saturday thinking about something other than AI, models, agents, products and whatever appeared on GitHub while you were asleep?
For the last six months I’ve been deliberately trying to learn how to switch off. My current target isn’t particularly ambitious: one day without the AI/IT background process running in my head.
Fine. Twelve hours. I’ll take twelve.
FOMO Is a Terrible Project Manager
You can call this professional deformation. I increasingly think some of it is simply accumulated mistakes.
There’s a version of curiosity that comes from genuine interest. And there’s another version that looks exactly like curiosity from the outside but is actually driven by fear.
Something happened. Someone shipped something. There’s a new model. A new architecture. A new tool. Someone is already using it. Maybe they’re getting ahead. Maybe I should try it. Maybe I should build something with it. Maybe tonight.
Technology makes this particularly nasty because there is no natural stopping point. The frontier moves every day. You can never finish “keeping up.” So FOMO has to be deliberately killed.
Not because the frontier doesn’t matter. I think staying close to it matters enormously. But because your life cannot permanently operate as a response to somebody else’s release notes. And switching off doesn’t happen automatically.
I’m beginning to think of it as another skill. You train it. You notice when you fail. You get better.
The balance will probably never be perfect. I’m not sure it should be. But it’s worth looking for.
AI Amplifies Whatever Was Already There
This connects to another idea I’ve been repeating for a year or two, including to audiences from stages: AI is an exoskeleton for the brain. It amplifies signals and converts them into actions. But there has to be a signal.
Without one, you’re just wearing a very expensive exoskeleton while standing still.
If you know how to draw, visual models immediately make more sense to you.
You understand composition. Scale. Proportion.
You can look at a generated image and know that something is wrong before you can necessarily explain why.
If you worked with music before Suno existed, you have a much better chance of extracting something interesting from Suno.
If you’ve written and edited thousands of pages, you recognize synthetic slop almost immediately.
If you’ve spent years building software, an agent generating 5,000 lines of code doesn’t automatically impress you. You want to know whether the architecture makes sense, whether the abstractions are sane and whether the thing survives contact with reality.
The AI gives all of these people enormous leverage. But their old skills haven’t disappeared. They’ve become the steering system.
This Is Why AI Slop Is So Annoying
AI has created a strange situation where the ability to produce something has become detached from the ability to judge it.
Yesterday you couldn’t make a website. - Today you can prompt one into existence.
Yesterday you couldn’t illustrate a book. - Today you can generate 200 covers before lunch.
Yesterday writing 5,000 words required actually having 5,000 words worth of thoughts.
That constraint has also been helpfully removed. This is wonderful.
It’s also how we got mountains of code, images, music and text created by people who have no idea whether the thing they produced is any good.
They’re excited because the machine crossed their previous capability boundary. Fair enough. But the market doesn’t care that making the thing was personally miraculous. It cares whether the thing is good.
That’s where expertise returns through the back door.
Start With What You Already Know
People often ask me: Where should I start with AI?
Or the slightly more ambitious version: How do I make money with AI?
My answer is increasingly boring: Start with whatever you’re already good at. Don’t begin by asking which AI business is fashionable this month. Look at the strongest signal you already possess.
Maybe you understand restaurants. Tires. Law. Agriculture. Typography. Sales. Logistics. Mathematics. Music. Teaching. Whatever.
Then ask: What happens if I amplify this ability 10× with AI?
What can you research that previously took too long?
What can you build that previously required a team?
What can you personalize that previously had to be standardized?
What can you test cheaply?
What can you automate while keeping the judgment for yourself?
That’s where the interesting opportunities usually are. Not in becoming another generic “AI expert.”
In becoming an unusually leveraged version of something you already understand deeply.
And Then Occasionally Take the Exoskeleton Off
There’s a second half to the metaphor that I probably haven’t emphasized enough.
If AI is an exoskeleton for the brain, you probably shouldn’t wear it every waking hour.
You still need the underlying muscles. Read something difficult without asking for a summary. Write occasionally without autocomplete. Draw. Cook. Run. Build something physical. Talk to someone without checking your phone. Spend twelve hours without knowing what Anthropic released.
The machine can amplify your signal. But your experiences, expertise, curiosity, taste and relationships are where that signal comes from in the first place. So perhaps these two ideas aren’t separate at all.
Learn to use AI aggressively. And learn to switch it off deliberately. Both are ways of protecting the same thing: the human underneath the exoskeleton.