Start Where You're the Expert

“Talk to it about something you already know well.”

I gave that advice near the end of The AI Mindset Shift and moved on. Easy to skim past, but it deserves more than a line because it’s not at all intuitive. Why would you ask about stuff you already know?

Most people don’t. They open their first AI chat and ask about something they don’t know anything about. Tax law. A medical question. Some technology they’ve been wanting to understand.

Here’s the problem: when you ask about something you don’t know, you have no way to grade the answer.

The response comes back looking polished and confident. Is it right? Mostly right? Subtly wrong in a way that matters? You can’t tell. You have to either trust it blindly or dismiss it entirely, and neither one teaches you anything about the tool.

Now flip it. Ask about something you know really well.

Your profession. The hobby you’ve had for years. The sport you’ve followed your whole life.

Now you can grade every sentence. You’ll see where it’s dead-on. You’ll see where it’s technically correct but generic, like someone who’s read about your field instead of working in it.

You can grade your questions, too. Ask the same question different ways. Add context one time, skip it the next. You’ll see immediately which version got the better answer, because you’ll know what the better answer is.

And eventually you’ll catch it saying something wrong with total confidence.

That last one is the most valuable. It doesn’t look like a mistake. It’s fluent, confident, and surrounded by the stuff it got right.

You need to see that multiple times in territory where you can catch it. That’s what builds the instinct for how far to trust it in territory where you can’t.

This is calibration, and there’s no shortcut for it.

So, pick a topic where you’re the person your friends call. Spend fifteen minutes asking the AI progressively harder questions about it. Start with basics, move to judgment calls, end with the tricky stuff only experience teaches. Then do it again in a few days. One session shows you a mistake. A handful of them builds the instinct.

Pay attention to where it starts guessing.

No article can hand you that instinct. You have to catch it being wrong yourself, on the one subject where nobody can fool you.