Meet Claude Code

What you’ll do: understand what an AI model like Claude actually is. No maths, promise.

A very well-read guesser

Claude is a “large language model”, usually shortened to AI model. It learned by reading an enormous amount of text: books, websites, manuals, conversations. From all that reading, it learned how language works and a great deal about the world.

When you write to Claude, it writes back one small piece at a time, each time choosing what most sensibly comes next. That sounds simple. At this scale, it produces something remarkable: an assistant that can explain, plan, summarise, translate and write computer instructions.

What that means for you

• Claude is good with language. Explaining, summarising, drafting, understanding what you meant even when you said it loosely.
• Claude can be confidently wrong. Because it’s choosing likely words, it can produce something that sounds right and isn’t. People call this “hallucination”.
• Claude doesn’t know your household. It only knows what you tell it, or what it can see on your Mac when you give it permission.
• Claude made by Anthropic is the one we use in this course. Other assistants exist; the ideas carry over.

What you’ll do: start Claude Code from your butler’s folder. The first time, it installs itself and asks you to sign in.

Before you start: a Claude plan

Claude Code comes with Claude’s paid plans (Pro or Max). If you don’t have one, go to claude.com/pricing, choose a plan, and create your account with your email. Check the current price there; it’s the main monthly cost of the course.

Steps

1. Open Terminal. If your butler’s window is still open from Module 0, use it. Otherwise press Command Space, type Terminal and press Return.
2. Copy this line, paste it, and press Return:

butler claude

3. The first time, your butler explains that Claude Code needs a paid plan and a one-time sign-in, and asks: Install Claude Code now? Press Return for yes. Anthropic’s installer runs for a minute and prints a few lines. That’s normal.
4. Claude Code opens in your Butler folder. It shows a short welcome and a box where you type.
5. Sign in. If Claude Code asks you to log in, choose to log in with your Claude account. Your browser opens a Claude page: sign in, click to approve, then come back to Terminal. You only do this once.
6. You’ll see Claude Code’s prompt, a › or > where you type. You’re in.

What you’ll do: ask Claude for something genuinely useful, then improve the answer by talking back.

Steps

1. In Claude Code, click in the Terminal window so you can type.
2. Copy this message (use the Copy button), paste it at Claude’s prompt, and press Return:

3. Read the answer. Now change it, the way you’d talk to a person. Try this:

4. Now ask Claude to check its own work:

5. Read what it says. Sometimes it finds a mistake of its own. That’s normal, and it’s why the last question is worth asking.

What you just learned about asking well

• Say who it’s for and what matters. “A family of three, no mushrooms, 30 minutes” did more than any clever wording would.
• Talk back. The first answer is a draft. Ask for changes in plain words.
• Ask it to check. “Did you miss anything?” catches a surprising number of slips.

What you’ll do: let Claude make a small file in your Butler folder, and see how it asks permission first.

So far, Claude has only talked. In Claude Code it can also act: create files, run programs and read what’s in a folder. That’s how it will teach your butler new skills. This lesson is a safe rehearsal.

Steps

1. In Claude Code, paste this and press Return:

2. Claude describes what it will do, then asks permission. You’ll see a short question with numbered choices, such as Yes, Yes, and don’t ask again for similar actions, and No. Read what it wants to do. If it matches what you asked, choose Yes (just this once): press 1, or use the arrow keys and Return.
3. Open Finder and go to your home folder, then Butler, then practice. Double-click note-to-my-butler.txt to read it.

What you’ll do: learn the five habits you’ll use for the rest of the course, and try two of them.

The five habits

1. Ask it to explain first. “Before you do that, explain what it does in plain words.” Claude is happy to.
2. Read before you click Allow. If a permission request mentions something you didn’t ask for, click Deny and ask why.
3. Never paste a password, code or card number into the chat. When Butler needs one, Claude opens a small Mac window for it instead. More on that in Module 4.
4. Say when something looks wrong. “That’s not what I expected. What happened?” works better than starting over.
5. Stop any time. Typing “stop” or “let’s pause here” is always fine. Your progress is saved, and the next lesson picks up where you left off.

Try two

In Claude Code, paste this:

Then this:

A good answer explains, and asks you to confirm before deleting anything. That’s the behaviour you want from every AI that acts on your behalf.

You can ask Claude to delete the practice folder now, or keep it as a souvenir. Either way, Claude should ask you first.