Your First Day with Claude Code, If You’ve Never Opened a Terminal

I was just on a call with a young woman not long out of college. She’s creative, she’s comfortable with data, and she is exactly the person I expected to be all in on AI. She doesn’t use it at all. She didn’t use it in college either, which I still can’t quite get over, and she never said she was scared of it, which surprised me too.

I have some version of that call all the time, and it surprises me every time. So this guide is for her, and for you if you’re in the same spot: you don’t write code, you’ve heard of Claude Code, and you haven’t opened it. I’ll tell you how my first day went, what I’d do with your first hour, the two mistakes I made, and where to go from there.

I didn’t know what a terminal was

I mean that literally. I had never opened one, and a friend of mine, a software engineer I later started an AI consulting business with, kind of laughed at me when he found out. Then I installed Claude Code, and there it was: a mostly empty window with a cursor blinking at me, asking what I wanted to do. I typed a question the way I’d ask a person, and it answered, and I realized I was just talking with it, and things changed.

Start in the chat, then move to Claude Code

If you’ve never used Claude at all, start at Claude.ai, the regular chat. Get familiar with it, get comfortable talking to it, and try things. Anthropic has its own courses and tutorials, and in my opinion they’re the first place to go; just work through them.

I wouldn’t stay in the chat for long, though. The chat talks with you about your work. Claude Code opens the files on your computer and does the work, and that’s where you’re headed.

The first job: my desktop

With that cursor blinking at me, I honestly couldn’t imagine what was possible, so I asked for help with the one mess I knew I had. My desktop was covered in screenshots. I’d been saving files wherever, and the chat had been saving its own files wherever, and my computer felt like that scene in Love Actually where the writer’s manuscript blows off the table and into the lake, every page of it. I asked Claude Code to clean up my desktop and organize my folders, and then I watched it do it.

It gave me a sense that I could organize myself, and watching it work made me understand how strong it was, because it had its hands on my actual files.

I’d give you the same first job. It’s small, you know the files well enough to judge the result, and you end the hour with a clean desktop.

Two mistakes I made

The first is about context, a word you’re going to hear constantly, and all it means is what Claude knows when you start talking. In the chat, that’s whatever you’ve typed or uploaded. In Claude Code, it’s the folder you opened it in, because it can read what’s in that folder and it starts out knowing nothing else. My mistake was opening it wherever I happened to be, with my files all over the place, so it never knew my project and I had to explain the project again every time I started. Every one of those explanations spends tokens, and tokens are what your plan gives you a limited supply of.

The fix is to give each project one folder and put everything about the project inside it, including a short note that says what the project is and how you like things done. The note is a plain text file called CLAUDE.md, and if you type /init, Claude writes the first draft of it for you. It reads that note every time it starts, and it keeps notes of its own as you work, and that’s what people mean when they say it has a memory.

The funnier mistake came when I started building landing pages. I was so excited that I could make a real web page just by asking for one that I’d copy the link and send it to people. The link pointed to a file on my own computer, and nobody could open it but me. It’s the most embarrassing mistake you can make, and I made it more than once. A page you build lives on your machine until you publish it somewhere, which is how I ended up learning Netlify, and Vercel, and Claude’s own artifacts, all of them ways of putting a draft where somebody else can see it.

Where to go after day one

1. Talk to it. I leave things out when I type, and when I talk I end up giving it the whole situation, which is what it needs to help me solve the problem. My setup is in Talk to Your Claude Code.

2. Build the folder. The folder matters most when you have a client, or a brand of your own. It’s where I keep the strategy, written down so Claude can read it: the rules I want followed, and samples of my own writing for it to match. I call those samples my fingerprint. The Company Is the Context is the whole system.

3. Teach it your jobs. Once you’ve explained the same job three times, have it write the job down so you never explain it again. That file is called a skill, and the skills guide is written for people who don’t code.

4. Later, run more than one. A session is one conversation working on one project, and nothing stops you from having several going at once. I call mine the fleet. Managing multiple sessions covers how I keep track of them, and the Mac mini guide covers the machine that keeps them running when my laptop is closed.

What my days look like now

My days right now are kind of incredible. I go back and forth between clients and my own businesses and projects, with several sessions working at the same time, and I can look across all of it and keep everything moving. I used to need a lot of people to do what I’m able to build mostly on my own now. The best way I can describe it is a team of problem-solvers that is always there, for whatever the problem happens to be.

I don’t know whether the young woman from that call has opened it yet. I hope she has, and I hope she started with something as small as her desktop.

— Joe