Ten minutes to something you can send someone

10 min read·Lesson 1 of One terminal, done properly

By the end of this lesson Claude Code is installed on your machine, you have run it once, and there is a finished page on your desktop that you made by asking for it in plain words. You can double-click it, and you can send it to someone. Ten minutes, one terminal window, nothing to configure.

You are probably here because you already use Claude Code, and you use it the way almost everyone does: one window open, one question at a time, watching it work and waiting for it to finish. That is a perfectly good way to start and a bad way to spend a day.

This course rebuilds that into a working day. But a day is built out of hours, and every hour is built out of one session that behaves itself. So module 1 is the single session, done properly — installed, understood, named, and recoverable when you break something.

Start with a win you can show someone.

Install it

One line. Pick your machine.

Open Windows Terminal or PowerShell from the Start menu and run:

PowerShell. This is the recommended Windows install and it updates itself.

irm https://claude.ai/install.ps1 | iex

Close the terminal and open a new one. Installers add Claude Code to your PATH, and a window that was already open does not know that yet. This one step accounts for most of the "it says it isn't recognised" messages we get.

claude --version

You should see

A version number, something like 2.1.240. If you get 'not recognized' or 'command not found', you are in the old window — open a fresh one.

Make something

Give yourself a folder to work in. Claude Code reads and writes files in the folder you start it from, so starting it in a fresh empty folder means it cannot touch anything you care about.

mkdir first-day

cd first-day

claude

The first run asks you to sign in and pick a text style. Then you get a prompt, and that is the whole interface.

Four things about this box, because they are the four that trip people up:

You type, you do not click. There are no buttons. It is a conversation.
Esc interrupts. If it starts doing something you did not mean, press Esc once. It stops. You have not broken anything.
Ctrl+D twice exits. Once is a warning, twice is out.
/help lists everything. Every command starts with /. You will meet about eight of them in this course and ignore the rest.

Now ask for the thing. Type this, exactly as it is, and press Enter:

Make me a single HTML page called card.html — a one-page "what I do" card with my name, one sentence about my work, and three things I can help someone with. Make it look good on a phone. Use one file, no build step, no internet. Then tell me how to open it.

Claude will read the empty folder, tell you what it plans to write, and ask for permission before creating the file. Say yes. It writes card.html and stops.

You should see

A short list of what it did — one file created — and a line telling you how to open it. Not paragraphs of explanation. If you got an essay and no file, you are in a chat window, not Claude Code.

A terminal running Claude Code, writing card.html and reporting one file created — The whole first run. It reads the folder, says what it will write, waits for a yes, then writes one file and stops.

Open it

Leave Claude with /exit, then open the file.

PowerShell. Opens it in your default browser, and it resolves the name against the folder you are standing in.

Invoke-Item card.html

That file is yours. It has no dependencies and no server behind it. You can email it, put it in a message, or drop it in a folder and forget about it for a year, and it will still open.

Send it to someone. That is the point of this lesson: ten minutes in, you have a thing rather than an opinion about a thing.

Check it

claude --version prints a version number in a fresh terminal window
card.html exists in the first-day folder
Double-clicking it opens a readable page with your actual name on it
The page still looks right when the browser window is narrowed to phone width
You know how to interrupt Claude without closing the window

If you see 'claude' is not recognized as an internal or external command

You are in the terminal window that was open during the install. Close it, open a new one, and try claude --version again. If a fresh window still fails, the installer put Claude Code somewhere your PATH does not cover, and claude doctor in the next lesson will tell you where.

If you see It wrote the file somewhere I cannot find

Run pwd on macOS or Linux, or cd alone on Windows, and it prints the folder you are actually standing in. That is where the file went. Claude Code writes relative to where you started it, always.

If you see On Windows, start card.html opens the wrong file or nothing at all

start hands the name to Windows to resolve, and Windows does not always resolve it against the folder your prompt is showing. Use Invoke-Item card.html instead, or .\card.html with the leading dot-backslash — both are explicitly relative to where you are standing. If you are not sure where that is, run cd on its own and it prints the folder.

What you just learned

Claude Code installs in one line and needs a fresh terminal window afterwards, which is the single most common first failure.
It works on files in the folder you start it in, so an empty folder is a safe place to experiment.
Esc interrupts, Ctrl+D twice exits, /help lists the rest.
A plain-language request with a named output file produces a real artefact, not a description of one.

practice

Open Claude Cowork on your own computer and try this on real files. Each module ends with a checkpoint where you submit what you made.

Open in Claude

What this actually costs