Try it today, for nothing

Everything in this section makes more sense once you have used a model for yourself. Here are three ways to do that in the next ten minutes, none of which costs anything.

1. A chat website's free tier

The fastest path. Every major lab runs one, and most have a usable free tier with no card required.

You will get a general-purpose assistant. Type a question, get an answer. That is the whole interface, and it is the right place to start.

2. A model playground

Sites like Google AI Studio and Hugging Face Spaces let you try specific models directly — including open-weight models you could later run yourself.

This is where you discover that different models genuinely behave differently. The same prompt to three models produces three different answers, and the personality differences are real rather than marketing.

3. An aggregator's free models

Services that route to many models usually keep a few available at no cost. It is the cheapest way to compare several models side by side without making accounts anywhere.

The ten-minute exercise that actually teaches something

Do this rather than asking it for a recipe.

Step one — ask it something you know well. Your job, your hobby, a subject you studied. Read the answer critically. You will find it is mostly right, confidently phrased, and wrong in one or two specific places. That is the single most useful experience in this whole section, and it is why you should do it with a subject you know.

Step two — ask it for a fact you can check. A date, a quotation, a statistic. Then check it. Repeat a few times. You are calibrating your own trust, which is a skill rather than a setting.

Step three — give it something long and ask for a summary. Paste in an article or a document and ask for five bullet points. This is what most people actually use AI for, and it is the task where it is most reliable — because the material is right there in front of it.

Step four — ask it to explain something at two levels. "Explain this as if I am twelve. Now explain it as if I have a degree in it." Watch the same knowledge reorganise. This is genuinely useful and impossible to do with a search engine.

What not to put in

Before you paste anything, remember: on a free tier, your text is often used to train future models.

Do not put in anything confidential — client material, medical information, unpublished work, other people's personal data, passwords, or anything you would not email to a stranger.

If the answer to "would I be comfortable if this were published?" is no, do not paste it. That rule covers essentially every case, and it is the reason local models exist at all.

What to notice while you use it

Three things, and they are the seeds of everything else here:

1. It sounds certain even when it is wrong. 2. It forgets. Earlier parts of a long conversation start to slip out of view. 3. It changes. Ask the same question tomorrow and you may get a different answer — sometimes better, sometimes worse.

Where to go next

• How to check what it tells you — the habit that matters most
• The words people use — so the vocabulary stops being a barrier
• What can your machine run? — if you would rather not send anything anywhere at all