Getting started with ChatGPT

Most people's first hour with ChatGPT is wasted. That is normal. The first hour is for figuring out how to be wrong with it before you can be useful with it.

Here is what usually happens.

The first thing you will do

You will ask it something you already know the answer to. You cannot help it. You do not trust it yet, and the natural way to test whether something is trustworthy is to check it against ground you already know.

Watch what it gives you back. The parts that match what you expected are reassuring. The parts that are subtly off — a date, a name, a small fact — are the more useful information. Those are the places it might be wrong about things you do not already know.

The second thing

You will ask it something you wish you knew. A topic you have been avoiding. A skill you have not had time to learn. An explanation of a concept that always seemed too dense.

This is when it earns its keep. A patient explainer who never makes you feel stupid for asking, who will rephrase the same idea five different ways until one of them lands, who will quiz you afterward to make sure you got it. A good tutor on demand. Most adults have not had one of those since school.

The third thing

You will try to make it do your job for you, and you will be disappointed.

The disappointment is the most useful thing that will happen in your first hour. It teaches you the actual shape of the tool. ChatGPT is not a colleague. It does not have stakes in your project. It does not know what you decided last quarter. It will produce something that looks like the right answer until you check it against the world.

The disappointment is not the failure of the tool. It is the discovery of where the tool ends and you begin again.

What nobody tells you

ChatGPT gets dramatically better the more context you give it. The most common cause of "AI is overrated" is people typing one-line prompts and concluding that the tool cannot read minds. It cannot. But it can read paragraphs.

The other thing nobody tells you: it is not Google. Google retrieves things that exist on the open web. ChatGPT generates things that look like they could exist. The relationship between what it gives you and what is true is not the same. You learn to read its output the way you read a smart but new employee's first draft — useful, often correct, never the final word.

The thing worth aiming for

The first hour is for play. Try things. Ask it weird questions. Ask it to explain something you understand to see if it explains it the way you would. Ask it to disagree with you. Ask it to act as different kinds of people and see how its answers shift.

The people who get the most out of AI are not the ones with the cleverest prompts. They are the ones who took an hour, then another, then another, building the intuition for what it is and is not.

Spend that hour. The rest follows.