4 YouTube videos that take you from zero to shipping with Claude Code
Hey, it's Cindy 🌱 Here are the four videos from the reel, in the order I would actually watch them, with the exact timestamps worth jumping to.
btw the waitlist for AI System Club is now open! If you're keen on using AI to automate your content, video editing, job search, business or build AI agents, feel free to sign up (only if you want to hehe). Ps. the waitlist gets first dibs:)
Read this before you start, so nothing surprises you. Claude Code needs a paid Claude plan. Anthropic's own setup docs say it plainly: "Claude Code requires a Pro, Max, Team, Enterprise, or Console account. The free claude.ai plan does not include Claude Code access." (setup docs). The four videos are free to watch. The tool is not. Video one is still worth watching before you decide, because ten minutes tells you whether this is for you.
why this order
Map, build, control, then habits
Most people start with the hour-long course, get twenty minutes in, and quit. The order below exists so each video answers the question the last one left you with.
🗺 First the map. Ten minutes that name every part of the tool, so the words in the longer videos already mean something.
🛠 Then build something. Nineteen minutes that end with a real thing deployed to a real URL. Momentum matters more than completeness at this stage.
🎛 Then the controls. The settings hour. This is what stops Claude Code wandering off, burning context and rewriting files you did not ask it to touch.
🔁 Then the habits. The team that builds Claude Code showing how they actually run it day to day, which is a different thing entirely from knowing the features.
The honest total is 1 hour 55 minutes. That is four evenings, not one sitting. Doing one a night and using the tool in between is the version that actually sticks.
before video one
Get it installed in five minutes
Do this first so you can follow along instead of watching. Everything here is from Anthropic's official setup page.
💻 Mac, Linux or WSL. Open Terminal and run curl -fsSL https://claude.ai/install.sh | bash. On a Mac, Homebrew works too: brew install --cask claude-code.
🪟 Windows. In PowerShell, run irm https://claude.ai/install.ps1 | iex. Or use WinGet: winget install Anthropic.ClaudeCode.
🖱 Not a terminal person? There is a Claude Code desktop app with a graphical interface, so you never touch a command line. The setup page links the Mac, Windows and Linux downloads at the top.
✅ Check it worked. Run claude --version. You should see a version number. If something looks off, claude doctor prints a read-only diagnostic of your install and settings without starting a session.
🚀 Start it. Open a terminal inside the folder you want to work in and type claude. It logs you in through your browser the first time.
video one
Learn 80% of Claude Code in 10 Minutes
Watch it here · Sajjaad Khader · 9 minutes 56 · free
The fastest complete picture of the tool anywhere. It is built as seven layers, and by the end you know what a CLAUDE.md is, what plan mode does, and what subagents are for. Watch this before anything else, because every longer video assumes those words.
2:33 · Turning AI into a real teammate Where CLAUDE.md is introduced. This is the single highest-value idea in the video and the reason the starter file below exists.
6:27 · Plan mode and ultra-thinking How to make it think before it writes. The habit that separates people who like Claude Code from people who fight it.
5:08 · Task queues Lining up work instead of babysitting one instruction at a time.
7:31 · Managing subagents Delegating parts of a job to separate workers. Skip on a first watch, come back after video three.
One thing is dated. Around 1:45 it compares model names that have since moved on. The reasoning, use the stronger model to plan and the faster one to execute, is still exactly right. Ignore the specific names.
Two real builds, and it does not stop at "it works on my machine". It connects the project to GitHub and puts it on the internet, which is the part most tutorials skip and the part that makes the whole thing feel real.
0:55 · Build one. Plan, build, iterate. Follow along rather than watching. Pause and do each step.
6:39 · CLAUDE.md. The second explanation of it you will have seen, from a different angle. That repetition is deliberate in this order.
7:52 · MCPs, skills and plugins. How Claude Code connects to other tools. You do not need this yet, but knowing the words helps.
13:33 · Connect to GitHub. Version control without understanding version control, which is genuinely what this unlocks for non-engineers.
17:17 · Deploy to the web. The end of night two should be a URL you can send someone. Do not skip this.
It is titled "for developers" and the opening is aimed at engineers, so start at 7:26. From there it is the most useful hour on this list for anybody, because it is entirely about control: what Claude Code is allowed to do, what it remembers, and how to stop it drifting.
17:30 · Modes The difference between letting it run and making it ask first. Learn this before you let it touch anything you care about.
19:15 · Settings and permissions The actual panel. Set it once and stop approving the same action forty times a day.
27:03 · Context, tokens and clearing Why long sessions get worse, and when to clear. The most common reason people think it "got dumber".
39:28 · Plan mode The proper version, in depth. Pairs with 6:27 of video one.
Later in the same video: skills at 48:05, MCPs at 53:27 and subagents at 57:51. Those three are worth a second visit once you have used the tool for a week, not on the first pass.
Three people from the team that builds it, showing how they actually work. This is the one that changes how you use the tool rather than what you know about it, which is why it goes last. Knowing the features and having the habits are different things.
4:48 · Their daily workflow. How they use questions, artifacts and the Slack integration inside real work.
6:41 · Loops and routines, run remotely. The idea worth stealing: work that runs without you sitting there watching it.
8:52 · How code review shaped it. Why a verification step in the loop beats a better prompt, which is the lesson under almost everything here.
14:04 · Feedback loops in practice. Building the thing with the thing, and what they check before trusting output.
paste this in
The starter CLAUDE.md
Three of the four videos land on the same point: the file that tells Claude Code about your project is what makes it good. Most people write three lines and wonder why it keeps guessing. Create a file called CLAUDE.md in your project folder and paste this in, then answer the bracketed parts.
📄 write my CLAUDE.md for me
Look through this project, then write me a CLAUDE.md file for it and save it in the root of the folder.
Work out what the project actually is before you write anything: what it does, what it is built with, how it runs locally, where the important files live, and any patterns that repeat across the codebase. Where you cannot tell from the files, ask me rather than guessing.
The file should contain: a two-line description of the project, the exact commands to install, run and test it, a short map of the folder structure with one line per important directory, the conventions you should follow when writing code here, and a short list of things never to touch without asking me first.
Keep it under a page. Write it as instructions to you, not as documentation for a human, and leave out anything you could work out by reading a single file.
Then keep it alive. Every time you correct Claude Code on something it should have known, add that correction to the file as one line. After a fortnight it stops making the same mistakes, and that is the whole trick.
night one
Your first real session
Watching without building is how this stays theoretical. After video one, open a new empty folder, run claude, and paste this.
🚀 build my first thing
I have never used you before, so go slowly and explain what you are doing as you go.
I want to build a small single-page web app that [describe the thing you actually want, in one sentence, for example: tracks which of my subscriptions renew this month and totals them up].
Start in plan mode. Tell me what you intend to build, which files you will create, and what it will look like when it works, then wait for me to say go before you write anything.
Once I approve the plan, build it, run it locally, and tell me the exact command I need to type to see it in my browser. If something breaks, show me the error and explain what it means in plain words before you fix it.
the four nights
A schedule that actually finishes
Night one, 10 minutes. Video one, then install it and run the first-session prompt above. End the night with something running locally.
Night two, 20 minutes. Video two, following along. End the night with a project on GitHub and a live URL.
Night three, 45 minutes. Video three from 7:26 to about 45:00. Set your permissions properly and write your CLAUDE.md while it is fresh.
Night four, 25 minutes. Video four, then pick one loop or routine from it and set it up on your own project.
Skip nothing on night one and two. The settings hour only makes sense once you have felt the problems it solves. Watching it first is how people conclude Claude Code is complicated.
the waitlist is now open
AI System Club is launching soon
Build AI systems with me through recorded lessons you can follow at your own pace. Join the waitlist for first access and the founding price.
btw this list gets first dibs the moment doors open :)