S-Tier Prompt Writing

This one's different from the last.

The last problem was too much. This one is too hard. The answer might be short. It might be three sentences. And you've read those three sentences four times and you still have no idea what they mean.

It's full of words you don't know. The sentences fold in on themselves. It assumes you already understand the thing you asked it to explain. You get to the end more confused than when you started, with that quiet, familiar feeling of nodding along to something that went completely over your head.

The answer isn't wrong. It's just pitched at someone who isn't you. Let's bring it down to where you actually are.

It does this because it can't see your face

Picture a good teacher explaining something in person. They say a sentence, they glance at you, and they see it. The slight frown. The eyes drifting. The "uh-huh" that means "I have no idea." And without even thinking about it, they back up and say it a simpler way.

The AI gets none of that. It says its piece into the void and hears nothing back. It can't see you frown. It can't watch you read the same line three times. As far as it knows, everything landed perfectly.

So why did it aim so high to begin with? Two reasons. It doesn't know your level, so it guesses, and it usually guesses too high. And it tends to borrow the vocabulary of wherever it learned the topic, which for anything technical means expert-speak, written by experts for other experts.

It's not showing off. It genuinely can't tell that it lost you. Which means the fix is simple: you have to tell it.

Tell it you're lost, and tell it where

The blank stare a teacher reads off your face? You just type it.

But don't stop there. Because "I don't get it" on its own makes the AI do the same thing it did with a vague correction or a lazy "make it better": it guesses again. It might come back with a whole new dense answer and miss a second time, because you never told it what to fix.

The move that actually works is to point at the exact spot where it lost you.

Now it knows precisely what broke. Maybe it was one unfamiliar word doing all the damage. Maybe it was one leap you couldn't follow. Either way, naming it turns a hopeless "I don't understand" into a problem the AI can actually solve.

Tell it where you're standing

The deeper fix is to stop letting it guess your level. Hand it the level directly.

That last one is the secret weapon. When an idea won't fit in your head on its own, an analogy gives it somewhere to land. "Entropy" is just a wall of a word until someone says it's milk swirling into coffee: once it spreads out, it never gathers back. Suddenly the idea has a shape, and the shape is something you've seen a thousand times. The concept was never the hard part. It just needed somewhere familiar to live.

And here's the part that should make you brave about asking: dropping to plain words doesn't dumb the idea down. It builds you an on-ramp. Once the simple version clicks, you can ask for the real terms back. "Okay, now what's the proper word for that?" And this time they'll stick, because now they're labels on something you actually understand.