You met Claude on Desktop in Level 2, and you’ve got the vocabulary from Level 1. Now the training wheels come off — but not in the way most people expect.
Claude’s first real superpower isn’t writing code. It’s reading it. For anyone, in any language, at the speed of a question typed into the chat.
Most people imagine “using Claude” as “tell it what to build.” That’s a small slice of what it’s good for. The bigger slice — the one this whole level is about — is using it to understand code that already exists, just by asking.
Why reading is where you start
A few reasons it’s the right first move:
- Most of programming is reading, not writing. Even experienced developers spend more time figuring out what existing code does than writing new code. Claude collapses that time into a chat reply.
- It works regardless of the language. A codebase in Rust, Go, Elixir, or some framework you’ve never heard of — it doesn’t matter. Claude reads them all. The language barrier that used to lock people out is gone.
- It works regardless of your background. Designer, PM, hobbyist, student, ten-year veteran in a different stack — reading with Claude is the leveler.
If you’ve ever opened a project at work, scrolled through a wall of unfamiliar files, and quietly closed the window — this level is the one you needed.
The thing that trips people up
People reach for “build me X” too early. They’ve heard “Claude can write code,” so they ask it to write code on day one — in a codebase they don’t yet understand.
That’s how you get changes you can’t evaluate. You can’t tell if Claude’s edit is right if you don’t know what the surrounding code does.
Read first. Change second. The order matters.
Zero blast radius — be fearless
There’s another reason reading is the right starting point: asking questions doesn’t change anything.
When you type “what does this file do?” into the chat, Claude reads the file and answers. No edits. No new files. No permission asks. Nothing on disk is different when you’re done. (Compare with Lesson 1.6 — that permission prompt only appears when Claude is about to do something, not when it’s about to say something.)
This means you can be fearless. Ask anything. Ask twenty things. Open a project you’ve never seen and poke around in the chat. The worst case is a wrong answer, not a broken project. We’ll come back to spotting wrong answers in Lesson 3.7.
What this level is going to do for you
Eight lessons. By the end you’ll be able to open any project in Claude on Desktop — yours, your team’s, an open-source one from GitHub — and ask, in plain English, questions like:
- What does this project do?
- What does this specific file do?
- Where is this function defined, and what calls it?
- Where does this value come from?
- What does this error mean?
- What changed here recently, and why?
You’ll also know how to tell when Claude is wrong, which is the skill that makes all the others trustworthy.
What’s next
The simplest, sharpest use of reading-with-Claude is asking about one specific file. Not the codebase. Not the project. That file, right there. That’s the next lesson.