Day 1: Anyone Can Use Claude Code. Using It Well Is Different.

5-minute quick start

You will learn

What Claude Code is and what it can touch on your machine
Why getting a result is not the same as getting a result you can trust
The explore, plan, build, verify loop

The one thing to remember

Ask it to prove the change works, then read the diff yourself.

A working example

Add a failing test for the bug first, fix it, then run the tests and show me the result.

Short on time? That is enough to start. Jump to the full guide for the what, why and how of every step, in the terminal and in VS Code.

01What Claude Code actually is

Claude Code is Anthropic's AI coding agent. Unlike a chatbot where you copy and paste code back and forth, it works inside your project. It reads your files, edits them, runs commands, runs your tests and explains what it did. You describe the outcome you want in plain English, and it does the work in front of you.

It started in the terminal, and that is still where it is most powerful, but you can also use it in VS Code, JetBrains IDEs, a desktop app and the browser.

02The easy part

Getting started takes about two minutes. On Windows, open PowerShell and run:

>_PowerShell

—☐✕

# install Claude Code
PS C:\> irm https://claude.ai/install.ps1 | iex

# go to any project and start it
PS C:\> cd C:\dev\my-project
PS C:\dev\my-project> claude

# then just ask
> explain what this project does

Sign in with your Claude account the first time, and that's it. You'll get a useful answer on your very first try. That is exactly why so many people stop there.

The trap: because it works so well out of the box, it's easy to assume you're already using it well. Most people are using maybe a fifth of what it can do.

03Where it gets harder

The gap between "it works" and "it works and I trust it" comes down to four habits. Each one gets its own day later in the series. Here's the short version.

01Context beats clever prompts

Claude Code only knows what it can see and what you tell it. Vague asks get vague code. Point it at the right files, say what "done" looks like, and mention the constraints you'd tell a new teammate.

Vague"Fix the login bug."

Clear"Users get logged out after refresh. The session logic is in auth/session.py. Find the cause, fix it, and add a test that reproduces it."

02Plan before it builds

For anything bigger than a small fix, ask it to explore and propose a plan first. Plan mode (press Shift+Tab to cycle into it) lets Claude read and think without touching a single file. Reviewing a plan takes a minute. Undoing the wrong build takes an afternoon.

Rushed"Add dark mode to the app."

Planned"Look at how styles are organised and propose a plan for dark mode. Don't change anything yet."

03Review, don't rubber-stamp

Claude Code is fast and usually right. "Usually" is the important word. Read the changes, ask it why it chose an approach, and give it a way to check its own work, like running the tests or opening the page. Code that verifies itself is code you can trust.

BlindAccept every edit, ship it, hope.

Verified"Run the tests. If anything fails, fix it and show me what changed."

04Your setup decides how smart it feels

Two people can use the same model and get very different results. The difference is setup. A CLAUDE.md file (run /init to create one) teaches Claude your project's conventions every session. Permissions, memory and connected tools (MCP) do the rest.

Starting coldRe-explaining your stack, your style and your commands every single session.

Set up onceCLAUDE.md says "use pnpm, run pnpm test before finishing, follow the patterns in src/ui."

04One habit to start with today

If you only take one thing from Day 1, take this loop. It works for a one-line fix and for a whole feature.

ExploreAsk it to read the relevant code and explain it back.

PlanGet a plan. Push back on it before anything changes.

BuildLet it implement the plan, in small steps.

VerifyHave it run tests or check the result, then review the diff.

Try it on something small this week. You'll notice the difference on the first task.

Anyone can use it.
Using it well is different.