5 resources to master Claude Code in the exact order I do them
Hey, it's Cindy 🌱 You commented CLAUDE, so here are all five, with the exact lessons and chapters worth your time, what you can skip, and one thing to try with each.
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:)
before you start
Get set up once
💻 Install it On Mac or Linux, run curl -fsSL https://claude.ai/install.sh | bash in your terminal. On Windows PowerShell, run irm https://claude.ai/install.ps1 | iex. Check it worked with claude --version.
💳 The plan you need Claude Code comes with Pro ($20 a month), Max and Team plans, or an API account. The free plan doesn't include it. Plans.
💡 Keep one small project open the whole way through, like a simple to-do page or a tool for your own work. Every resource below is more useful when you try it on something real.
resource one · about 1.5 hours
Anthropic Academy: Claude Code 101
Start here because it explains what you're actually directing: what Claude can see, what it's allowed to do, and the workflow everything else builds on. 12 short lessons, a quiz and a badge at the end. Open the course.
🎯 Focus on these four lessons. How Claude Code works, the explore, plan, code, commit workflow, context management, and the CLAUDE.md file.
⏭️ Skip or save for later. Skip the install lesson if you're already set up. Hooks and MCP can wait until the basics feel easy.
✋ Try this. Press Shift+Tab to switch into plan mode, ask for a plan for one small feature, approve it, let Claude build it and commit. Type /context before and after to see how much of Claude's memory the task used.
resource two · about 4 hours
Nick Saraev's Claude Code full course
This is the deep one, so split it across a few sessions. You don't need all four hours after Claude Code 101, so jump straight to the parts below. Watch it on YouTube.
🎯 Watch these five parts. 0:34:40 for getting Claude to screenshot and check its own work, 1:44:30 for planning and building a full proposal app, 2:13:53 for context and tokens, 2:35:29 for skills (and why he turns connected tools into skills), and 3:11:08 for subagents.
⏭️ Skip or skim. The first 24 minutes and 1:31 to 1:44 repeat what 101 covers. Plugins and the pitch at the end can go too, and agent teams at 3:26 are optional because they get expensive fast.
🔨 Build along from about 3:02. He turns an email-labelling task into a skill, then runs it across 10 subagents at once. Copy it on a test inbox: label 15 emails with Claude, turn that into a skill, then split a bigger batch across subagents that only report back the counts.
⚠️ Know what's changed. It was recorded in February on older models and the Antigravity editor, so use any editor and whatever model Claude Code picks now. Agent teams needed an experimental setting at the time, so check the current docs before trying them.
His two best lines: "a minute of planning saves you 10 minutes of building" (1:41:30), and without a way for Claude to check its own work you lose most of the value of AI (0:40:00).
resource three · 27 minutes
Nate Herk's Claude Code skills video
Once you've built something, skills are how you stop repeating yourself. A skill is a saved set of instructions Claude runs the same way every time. Watch it on YouTube.
🎯 Pause on these moments. 5:24 for what's inside a skill, around 11:56 for how Claude decides when to use one, around 14:18 for making skills faster and cheaper, 17:13 for his six-step framework, and around 24:12 for the list of common problems and fixes.
⏭️ Skip. 18:02 to 19:04 is installing VS Code, and the ending is a promo. The skill-builder he demos comes from his paid community, so watch that part for the process rather than the file.
✋ Try this. Do one task you repeat with Claude by hand, then say "turn this into a skill and ask me questions first". Test it on a second example by asking in plain words, without the slash command, to check it triggers on its own.
Two small corrections to the video, from Anthropic's skills docs: you don't need to register a skill in CLAUDE.md, Claude finds it automatically. And only the description line is really needed at the top of the file; write it the way you'd naturally ask for the task.
resource four · about 2 hours
DeepLearning.AI's Claude Code course
Taught by Elie Schoppik from Anthropic, this one shows Claude Code on three different real projects: a chatbot, a data dashboard and a web app built from a Figma design. Open the course, free for a limited time during their platform beta.
🎯 Focus on these four lessons. Setup and codebase understanding, testing and debugging, adding multiple features at the same time, and creating a web app from a Figma mockup.
⚠️ Know what's changed. The course notes install Claude Code with npm. Use the installer at the top of this guide instead. The chatbot project also needs an Anthropic API key to actually answer questions.
✋ Try this. Grab the chatbot starter from the course files, ask Claude Code "give me an overview of this codebase", and compare its answer with the sample in the lesson 2 reading notes.
resource five · keep it open
Anthropic's best-practices guide
This isn't one to read once. Keep it open while you build, because it's the difference between Claude doing okay work and great work. Open the guide.
✅ Give it a way to check Tests, a build or a screenshot to compare. Ask Claude to show you proof it works, not just say it does.
🗺️ Plan first, unless it's tiny Use plan mode for anything real. If you could describe the change in one sentence, skip the plan.
🧹 Keep context clean Type /clear between unrelated tasks. If you've corrected Claude twice and it's still wrong, clear and write a better prompt.
📄 Keep CLAUDE.md short Start with /init, then for every line ask "would removing this cause mistakes?" A long file gets ignored.
✋ Try this. For your next feature, use the "let Claude interview you" section: have Claude ask you questions and write a SPEC.md, type /clear, build from the spec in a fresh session, then ask a subagent to check the result against it.
💡 Skip the "automate and scale" section for now. Running lots of Claude sessions at once only helps after the basics are second nature.