Notes: See Claude Code in Action

In This Section, You Will Master:

• What separates an agent that does the work from a chatbot that hands you text
• The agentic loop, the five steps Claude Code moves through on every task you give it
• Where Claude Code sits among the other AI coding tools, and when each one is the right reach
• The vocabulary developers use every day, so none of it lands as jargon later
• The trust but verify mindset that keeps the work honest

How to Use These Notes:

These notes are designed as quick-reference companions to your video learning.

• Use them to review key concepts after watching lectures
• Reference them when you need a quick reminder of frameworks and tools
• 🎯 Core Concepts = Essential terms and frameworks
• 🚀 Actionable Items = Practical implementation steps

The Growth Payoff:

Nothing on your machine changes today, and that is deliberate. What changes is that you can explain what this tool actually does, name the loop it runs, and tell which situations call for it. Tomorrow, when you type your first real instruction, that understanding is what lets you ask for what you want instead of repeating what somebody else said.

---

Lecture: What changes when AI stops advising and starts doing? (Start Here)

The difference between AI that advises and AI that acts, shown by building a working app in front of you rather than describing one.

🎯 Core Concepts:

• The AI you have used in a browser is a brain that can see but cannot open its own eyes. Everything it works on has to be carried in, and everything it decides has to be carried back out
• Claude Code is that same brain given eyes and hands, sat down at your computer, working through a job from one end to the other
• Until now, AI has been a phone call with somebody brilliant. You describe the problem, they tell you what to do, and then you put the phone down and do all of it yourself
• The dramatic change comes from reshaping your workflow around Claude Code, not from fitting Claude Code into the workflow you already have
• You do not need to be a developer for this, though if you already build things you can move faster

---

Lecture: What actually changes about the way you work? (More Than a New Tool)

Which parts of your day stop being yours, including the kind of work nobody has ever been able to hand off.

🎯 Core Concepts:

• You stop carrying the steps. You describe what you want in your own words and it gets done
• The obvious half is repetitive work. The half nobody expects is work that is nearly the same each time but not quite, the version with one odd exception in it
• That kind of work has never been possible to hand to anyone else, because explaining the exceptions took longer than doing it yourself
• Work where you understand the problem completely and have no idea how to fix it can also be handed over
• Your day shifts from executing to deciding what should be done, judging what came back, and pointing it somewhere better. You are still the only one who knows whether it is any good

---

Lecture: Why does this course live in a black window? (The Terminal)

What a terminal actually is, why it is safe, and why this tool was born inside one.

🎯 Core Concepts:

• A terminal does the same job as clicking an icon, a different way. You type the name of the thing you want, press enter, and the computer goes and does it
• A terminal hands you no deeper power over your machine. Whatever you type is checked by the same permission system as anything you double-click
• Mistyping cannot break anything. You get a line of red text saying it did not understand you, and nothing else happens
• What a terminal gives you is reach. Your computer is full of programs with no icon anywhere, and you can run one straight into the next
• Claude Code lived only inside a terminal for its first nine months. It arrived in February 2025, and the desktop app did not exist until that November
• You will hear this window called a console, a command line, or a shell. Terminal is the word used here

---

Lecture: What are HTML, CSS and JavaScript? (What a Website Is Made Of)

The three files sitting under every website you have ever opened, and why they are kept apart.

🎯 Core Concepts:

• HTML holds what the page says. Every word and every box you can point at
• CSS holds how it looks. The colours, the spacing, the headline being large while the small print stays small. Also called a style sheet
• JavaScript holds what the page does when you touch it
• They are separate so you can change every colour without touching a word, and rewrite every word without moving anything out of place
• All three sit together in a folder. Developers tend to say directory, and both words mean the same thing

---

Lecture: Can Claude Code Really Build an App From One Sentence? (Live Demo)

One sentence, three working files, and a moment where Claude corrects its own code without being asked.

🎯 Core Concepts:

• The output was three real files written straight into the project folder, not a block of text for you to copy and paste somewhere
• Claude read the folder first, including a CLAUDE.md file holding project preferences, which is why it chose plain web files over a framework
• Partway through it hit a security warning on a risky pattern and switched approach on its own, choosing the safer route without being asked
• The verify step was a summary of what got built, handed back so you could decide what to do next
• Take the shape of the conversation from this demo rather than the time on the clock. Your own first run may be faster or slower

🚀 Actionable Items:

• Keep a CLAUDE.md in your project folder so Claude reads your preferences before it plans anything

---

Lecture: What Just Happened Under the Hood? (The Agentic Loop Explained)

The five steps underneath that build, named, so you can always tell where in the cycle you are.

🎯 Core Concepts:

• The agentic loop runs Read, Plan, Execute, Adapt, Verify
• Read comes before anything else. Reading your project first is what lets Claude make decisions that fit it, instead of guessing
• Plan is stated openly on screen before any file is created, which is your chance to let it run or redirect it
• Adapt is what makes this a loop instead of a straight line. Execute and Adapt braid into each other every time the project pushes back
• Verify on that take meant stating what was built and handing the decision to you, rather than running the app itself
• Claude does not always make the right call when it adapts. Knowing when to trust an adaptation comes with practice

---

Lecture: Can you explain Claude Code in one sentence? (Your 60-Second Challenge)

A one sentence answer you write yourself, because a concept only becomes yours once you can say it in your own words.

🎯 Core Concepts:

• Claude Code reads your entire project, makes its own decisions about how to build, writes across multiple files, runs commands, and fixes its own mistakes
• A chatbot can only hand you text to paste somewhere. The work of turning it into files stays with you
• The moment you can put a concept into your own words, it stops living in someone else's head and starts living in yours

---

Lecture: How to Get the Most Out of This Course

How the fifteen days are laid out, and the small habits that decide whether you finish.

🎯 Core Concepts:

• Fifteen days, one section a day, and every day hands you something you did not have the morning before
• Every lecture teaches one thing. That is why they are short, and why you can find that one thing again three weeks later when you need it
• Nothing here has to be memorised. Every word is explained where it is used and why somebody invented it
• Questions posted get an answer within six to eight hours

🚀 Actionable Items:

• Take the free readiness test today. Today is the only day the score means anything, because by tonight you already know things you did not this morning
• Decide tonight what you want Claude Code to build for you first

---

Lecture: What do developers mean by commit, branch and pull request?

The words that turn up in every conversation about software, each one invented because somebody got burned.

🎯 Core Concepts:

• A commit is one saved moment, the difference between what a file was and what it became, with a line from you saying what you did
• Git is the program on your own machine that captures those moments. GitHub is a website holding a copy so it can be shared and survive a dead laptop. Two different things, and one is not short for the other
• A codebase is a folder of files that remembers everything that has ever happened to it
• A branch is a second version you can experiment in while the real one stays exactly as it was
• A pull request lays both versions side by side, every line about to change, before any of it is real. A ticket asks for work, a pull request offers work already finished
• A bug is the wrong thing. Debugging is the hunt for it, not the mistake itself

---

Lecture: What is actually changing in the industry right now? (The AI Engineer Role)

What the real numbers say about how developers are working now, and the shift the headlines keep getting wrong.

🎯 Core Concepts:

• 73% of engineering teams now use AI tools every day, Claude Code alone processes over 100,000 GitHub commits a day, and 46% of developers name it their most valued tool
• Two developers start at the same 9 AM. One writes function by function. The other describes three features, presses enter, and reviews three pull requests by lunch
• The second one moved away from defining themselves by lines of code, and toward the systems they design and the calls they make
• The industry calls that role the AI Engineer, and it works more like a tech lead directing a team than a coder typing every line
• The developers who made that shift got more technical, not less, because directing intelligent tools well needs a deeper grasp of architecture than writing each line by hand ever did

---

Lecture: Claude Code vs Cursor vs Copilot vs Windsurf? (Honest Comparison)

A map of the territory, so you can tell which moment calls for which tool.

🎯 Core Concepts:

• Copilot suggests the next line inside the file you are typing in. Fast and unobtrusive, and it does not understand your project as a whole
• Cursor is for fast inline changes on a file you already have open, with that pair-programming feeling
• Claude Code is the autonomous end. It reads the whole codebase, writes across dozens of files, runs commands, and keeps going until the task is done
• Windsurf sits between the editor experience and the autonomous agent
• Claude Code uses roughly 5.5 times fewer tokens than Cursor for identical tasks, and Cursor gives inline feedback that a terminal cannot
• Comparing them directly is like asking whether a hammer beats a screwdriver. The answer depends on the job in front of you

---

Lecture: What two skills does an AI Engineer need? (Agentic Coding and Context)

The two skills the role is built on, and why one of them decides what your sessions cost.

🎯 Core Concepts:

• You are the architect and Claude Code is the builder. You decide what to build and what quality bar to hold, and it handles the implementation
• Agentic coding is describing what you want at a higher level, having it implemented autonomously, then reviewing the decisions that were made
• Context engineering is setting your project up so Claude understands it deeply: configuration files, rules that guide behaviour, memory between sessions, and connections to outside tools
• Context is the difference between a developer burning six hours and forty dollars on a task, and the same developer finishing the same kind of work in fifteen minutes for under a dollar
• This is a different job from machine learning engineering, which is about building the models themselves

---

Lecture: When should you trust Claude's code, and when check it? (Trust but Verify)

Where the work is reliably good, where it deserves a second look, and how to tell the difference quickly.

🎯 Core Concepts:

• Claude Code behaves like a talented colleague who is occasionally a little overconfident
• Reliable areas: project structure, standard framework patterns, and common configurations
• Areas deserving closer attention: edge cases, security-sensitive logic, and integrations with external services
• Trust but verify. You are not reading every character, you are learning where your attention actually pays off
• A real example: a session timeout that refreshed on every request, so a logged-in session never expired. Ninety seconds to catch in review, weeks of damage if it had shipped

---

Lecture: That is day 1 done (What You Can Now Explain)

What changed today is in you rather than on your laptop, and that was the point.

🎯 Core Concepts:

• Claude Code is not installed yet, and nothing on your machine is different from this morning
• Every word you picked up today arrived the same way. Somebody hit a problem, and the word turned up as the answer to it
• The words came first on purpose. Installing in the first ten minutes leaves you with a powerful tool and nothing of your own to ask it
• Tomorrow the first thing you ask for is yours rather than a copy of somebody else's prompt

🚀 Actionable Items:

• Decide what you want Claude Code to build for you first, before you install anything

This resource is part of

Claude Code: From Beginner to Pro Agentic AI Engineer

by Arun Nagarathanam

Take your skills to the next level with

AI GEO 2026: Get Recommended by ChatGPT, Google, Perplexity

by Arun Nagarathanam