Going deeper with Claude
Once you have built a thing or two, Claude has a few extra layers worth knowing about. None of this is needed to start, so come back when you are ready.
The different ways to use Claude
Confusingly, "Claude" shows up in a few places, and they each feel a little different. It helps to know which is which.
Claude in a browser or app
The chat at claude.ai, and the desktop and phone apps. You type, it answers. Best for thinking out loud, drafting, and learning.
Claude Code
The version that lives in the terminal and actually builds things on your computer, reading and writing real files. This is the one you set up in Stage 3.
Claude working on your computer
The desktop app is steadily gaining the ability to do more on your machine too, not just chat. The line between "chatting" and "doing" is blurring.
The honest part: it takes a little while to get used to switching between "just chatting" and "Claude Code actually doing things in the terminal." That wobble is normal. It settles.
Getting comfortable with the terminal
The chat is a friendly box that feels like texting. Claude Code lives in the terminal, which looks barer and behaves differently: it shows you what it is about to do and waits for your nod. The first few times it feels alien. Give it a handful of sessions and it stops being scary. If something on screen confuses you, you are allowed to literally ask Claude "what does this mean?" right there.
CLAUDE.md: your standing instructions
This is the most useful thing to learn early. CLAUDE.md is a plain text file that Claude Code reads automatically every single time, so you do not have to keep repeating yourself. You write how you want it to work and what it should know, once, and it remembers.
There are two levels, and that is the clever bit:
Global (you)
One file in your home folder that applies to everything you do. Good for personal preferences, like "explain things simply, I am not a developer."
Project (this build)
A file inside a specific project folder that only applies there. Good for context about that one thing you are building.
You do not have to write these by hand. Ask Claude to set up a CLAUDE.md for you and tell it what to put in.
Slash commands
A slash command is a shortcut you trigger by typing a "/" followed by a word. Claude Code comes with a set built in, and you can make your own for things you do over and over, so a whole routine happens from one short command instead of a long re-explanation. Think of them as saved buttons for your most common jobs.
Hooks (for later)
A hook is a small instruction that runs automatically when something happens, for example every time Claude is about to change a file, or when you start a session. People use them for things like a notification when a job finishes, or a safety check before something runs. This is genuinely advanced, and you can build plenty without ever touching it. Just know the word exists so it does not throw you later.
These are power tools, not requirements. Reach for them when you notice yourself repeating something or wanting more control. Until then, ignore them with a clear conscience. And whenever a setting looks baffling, ask Claude to set it up and explain it in plain words.
The short version
- 1Claude chat is for thinking; Claude Code (in the terminal) is for building.
- 2The terminal feels odd at first and gets comfortable fast.
- 3CLAUDE.md saves your instructions so you stop repeating yourself, globally or per project.
- 4Slash commands are saved shortcuts; hooks are advanced auto-actions for much later.
- 5None of it is needed to start. Ask Claude to set any of it up for you.