Updated May 28, 2026
Daycare for Teenage AI: A Calm Plan for the Frontier-Model Era
Frontier AI is brilliant in flashes and confidently wrong the rest. Treat it like a gifted teenager: supervise it, build the skill, use your runway wisely.
Hello, hello! 👋🏻
Lately I’ve been having a lot of interesting conversations about AI.
Maybe it’s people watching their stock portfolios get rattled by the SaaS Apocalypse. Maybe it’s the steady drumbeat of AI in the news and on TV. Maybe they finally got some time and played with a frontier model like Claude Opus 4.6. Or maybe it was an article like this one from the Wall Street Journal on the downsides of AI at work.
Whatever it is, suddenly a lot of people want to talk to me about AI.
And most interestingly, it’s people who don’t normally talk to me about AI. People who were never going to be touched by it are realizing they might be after all.
Maybe their high-margin software company won’t exist in the same form. Maybe the way they rolled out AI at work isn’t really working. Maybe that hodgepodge of AI tools is just burning their people out.
I can’t tell you exactly what shifted, but it did.
So if your boss is suddenly listening to you about AI (or your spouse, or whoever the big AI skeptic is in your life), let’s review the facts with them.
Let’s review the facts
A frontier model like Opus 4.6 or ChatGPT 5.2 is likely already better than 50% of your current human team at a wide range of tasks.
The average gap between frontier AI model releases has collapsed. It used to be roughly 77 days a few years back. Now it’s around 16 days. The labs are leapfrogging each other almost constantly.
In the next two or three releases, the models will likely be smarter (math smart) than 99% of the people on this planet, while still hallucinating something like 10% of the time and needing considerable adult supervision.
Think daycare for teenage AI.
Brilliant in flashes. Wildly capable. Also fully willing to confidently tell you something completely wrong. You don’t hand it the keys. You supervise it.
What are you going to do with the runway?
Call it a few more releases. So you’ve got the next year or so before AI is, on paper, smarter than most of the people you and I know.
What are you going to do with that runway?
For me, it’s two things: getting better with AI, and going to the beach with my family.
Because it’s easy to let a few startling data points stampede you. The reality on the ground is calmer than the headlines. Only a minority of folks are using AI daily for real work, which means we’ve got a long way to go before Salesforce goes out of business.
As a reminder: AOL only ended its dial-up service in September of 2025. 🙂 The future arrives slower than the press release.
So chill out, but have a plan
Head to the beach. Hang with your family. But have a plan.
The plan is simpler than the AI-industrial-complex wants you to believe. Treat AI as a skill, not a one-time tool you “implement” and forget. It’s closer to a utility than a gadget. And if you can talk, you can use it. The hard part isn’t the buttons. It’s the habit.
If you want the map of which tool does which job before you go all-in, that’s exactly what I keep updated over on the AI marketing hub. Start there, then pick your lane.
The real question isn’t which AI. It’s this:
What are you actually using AI for?
That’s the whole game. The people pulling ahead aren’t the ones who collected the most tools. They’re the ones who built a behavior. I broke down exactly that split in The AI Divide: Operators, Experimenters, and Holdouts. Worth a read if you’re trying to figure out which one you are right now.
See you on the beach,
Alec
I write one email a week for marketers who’d rather skip the hype and actually get better with AI. Real workflows, real reviews, zero fluff. If that’s you, come hang out.