INTERACTIVE LAB · PART 2

Claude Code: the brain gets a keyboard

Diploma in IT · Interactive lecture
Every chapter has something to try.

CHAPTER 1 · WHAT IS CLAUDE CODE?

Same brain. Now with hands.

Last lecture: Claude is a frozen brain that guesses the next word, and MCP gives it helpers. Claude Code takes that idea all the way — it puts the brain inside your terminal, where it can open your files, type into them, and run programs. Chat gives you advice. Claude Code does the job. Watch both get the same request:

The job: “My app crashes when I click Save — fix it.”

CLAUDE IN A CHAT APP

My app crashes when I click Save — fix it.

That sounds like a null error! Common causes: ① the title field is empty, ② the folder doesn't exist yet, ③ a typo in the save path. Could you paste your save code and the exact error message? I'll take a look 🙂

Gives advice. Your hands do the typing.

CLAUDE CODE IN YOUR TERMINAL

                {{ l.t }}

                $ waiting for a job…

Does the job. Your hands stay in your pockets.

The take-away: it's the same frozen brain from last lecture. What changed is hands — Claude Code can read your files, edit them, and run programs. That combination is what people call an agent.

CHAPTER 2 · THE LOOP BEHIND THE MAGIC

Look. Plan. Act. Check. Repeat.

Claude Code isn't magic — it runs a loop, exactly like a careful human developer. The best part is what happens when the check fails: it doesn't give up or lie. It loops back and fixes its own mistake. Step through a real job: “make the login page remember my email.”

Look

read the code

Plan

decide the change

Act

edit & run

Check

run the tests

↩ check failed → back to Act. No drama, just another lap.

Why this matters: last lecture Claude guessed a football score would be wrong. Here, the loop means it never has to guess — it runs the code and looks at the result. Reality is its fact-checker.

CHAPTER 3 · YOU'RE THE BOSS

It asks before it touches anything

An AI with hands on your computer sounds scary — until you learn the rule: Claude Code proposes, you approve. Every risky action pops up a permission request first. Your job right now: handle three real requests. Read carefully — one of them is a trap.

PERMISSION REQUEST {{ permNum }} OF 3

{{ permDetail }}

The pattern: reading is safe, edits deserve a glance, commands deserve a careful read. The permission prompt isn't a formality — you are the safety check. Never approve a command you don't understand.

CHAPTER 4 · CLAUDE.MD — THE HOUSE RULES

A sticky note it reads before every job

Remember the hidden note (system prompt) from last lecture? Claude Code lets you write one for your project: a file called CLAUDE.md that sits in your folder. House rules, written once, obeyed every time. Toggle some rules, then give it a job:

📌 CLAUDE.MD (LIVES IN YOUR PROJECT FOLDER)

{{ mdText }}

                {{ l.t }}

                $ set your rules, then press the job button…

Try this: run the job with no rules, then flip them on one at a time and run again. Same brain, same job — different behaviour. That's the power of a written rule over a spoken wish.

CHAPTER 5 · TALKING TO IT WELL

The better the brief, the better the work

Claude Code is like the smartest intern you've ever had: brilliant, tireless… and it only knows what you tell it. Same bug, three ways to ask. Tap each one and compare what happens:

WHAT HAPPENS NEXT

The recipe: say what's wrong (the symptom), where (file, error message), and how to prove it's fixed (run the tests). Describing a job clearly enough for a machine — that's the skill of the decade, and it works on humans too.

CHAPTER 6 · RIGHT TOOL FOR THE JOB

Chat window or terminal?

Rule of thumb: if the job is words and ideas, chat is perfect. If the job needs hands inside your files, that's Claude Code. Tap your pick — wrong guesses are free. {{ quizScore }}

CHAPTER 7 · THE BIG PICTURE

Claude Code on one napkin

If you can redraw this, you've got it. The brain from last lecture, sitting inside a terminal, working in a loop — with you at the gate.

🧑‍🎓

You

Describe the job in plain English — symptom, place, proof

🛡 The gate

Every risky step waits for your ✓ — you are the safety check

🧠

The same frozen brain

now living in your terminal

🔍 Look → 📝 Plan → ✏️ Act → ✅ Check ↩

📌 CLAUDE.md — your house rules, read before every job

📖 Read files

safe — it just looks

✏️ Edit files

glance at the diff first

▶ Run commands

read these carefully

1 · Claude Code = the same frozen brain, plus hands on your computer. That's an agent.

2 · It works in a loop — look, plan, act, check — and fixes its own failures.

3 · It asks permission for risky steps. Never approve what you don't understand.

4 · CLAUDE.md is your project's hidden note — house rules it reads before every job.

5 · Clear briefs get better work: the symptom, the place, and the proof.

LECTURER BAR

Chapter {{ chapNum }} of 7 · ← → keys work too