1.1
What Claude Code Actually Is
Why this exists
You've been using Claude in a browser. It's good, and you've probably already got value out of it. So the obvious question when someone hands you a different version is: what's actually different, and why does it need its own eight-week course?
The answer is one thing, and it's not a feature. It's a change in who does the work.
The idea
In the browser, you are the courier. Claude Code fires you.
Watch what actually happens when you use Claude in a browser to do something real. You describe the problem. It tells you what to do. You go and do it — open the file, run the thing, check the result. Then you copy the result and paste it back so it can tell you the next step. And again. And again.
Every single round trip goes through you. You're the hands. The thinking happens in the browser tab and the doing happens on your machine, and the only connection between them is you, copying things back and forth.
Claude Code deletes that job. It opens the file itself. Runs the thing itself. Reads what came back itself. Notices the error itself, and tries again. You describe the outcome once and it works through the loop on its own until it's done or it needs you.
That's the whole difference:
| Browser | Claude Code | |
|---|---|---|
| Who reads your files | You, then paste | It does |
| Who runs things | You | It does |
| Who reads the result | You, then paste | It does |
| Who decides the next step | It — from your summary | It — from what actually happened |
| Where the work lands | Your clipboard | Real files on your machine |
That last row is the one that changes what's possible. In the browser, the output of an hour's work is text you have to do something with. In Claude Code, the output is the thing itself — files created, changed, working.
And the fourth row matters more than it looks. In the browser it decides the next step from your summary of what happened. You'll leave things out, because you don't yet know which details matter. It sees the actual output — the full error, the real file, the thing you'd have skipped. It's not just faster. It's working from better information.
The cost is real and you should hold it from day one: it can change and delete things on your computer. The browser could only ever give you text. This one has hands. That's the entire reason permission modes exist, why you'll learn to undo before you build anything, and why this course spends as long on not-breaking-things as on making them.
Do
No install yet — that's 1.3. This one's a baseline you'll want later.
- Open Claude in your browser.
- Pick a real task from your week that involves a file or a document. Something you'd actually do.
- Work through it properly with Claude, the normal way.
- Count every time you copy or paste to move information between Claude and something else. Keep a tally as you go, not from memory.
- Also note: how long did it take, and how many times did you have to summarise something back to it?
- Write it in a file called
baseline.md— the task, the number of copy-pastes, roughly how long.
That number is the thing you're eliminating. Look at it again in week three.
Done when
Never do
Don't skip the trade-off. Everything from here assumes you're holding both halves: it can act on your machine, which is the entire point, and it can act on your machine, which is the entire risk.
People who only hold the first half move fast for about two weeks and then delete something they needed. You'll learn how to make that recoverable in Level 4 — until then, work in folders where nothing matters.