01 · The core idea
Three things to understand first
Claude Code is an agentic coding tool. Each word in that phrase matters.
It lives in your terminal
You run it with the
claude command in any project directory. It reads and writes your files, runs shell commands, uses Git, all from the command line, not a browser window.It understands your entire codebase
Instead of indexing everything upfront, it uses "agentic search": it hunts for context as needed across all your files and directories. Your code never has to be bulk-sent to any server.
It takes actions, not just suggestions
Claude Code can edit files, run tests, commit to Git, and iterate until something works, autonomously. You review and guide; it handles the iterative loop.
"It's not a better autocomplete. It's a different workflow entirely."
Shawn Kanungo, on what makes Claude Code different from CopilotThe one-sentence mental model
Claude Code is like having a junior developer sitting next to you who can read your entire codebase, run commands in your terminal, and autonomously complete multi-step tasks, while you review the diffs and steer direction.
01 / 05
02 · The difference
Three tools, three philosophies
Most confusion about Claude Code comes from comparing it to tools it isn't trying to be.
Chatbots
Example: Claude.ai, ChatGPT
Lives: Browser / app
Sees: What you paste in
Does: Suggests text only
Memory: Per conversation
Best for: Questions, explanations, drafting
IDE plugins
Example: GitHub Copilot
Lives: Inside your editor
Sees: Current open file
Does: Inline completions
Memory: Immediate context
Best for: Line-by-line autocomplete
THAT'S THIS ONE
Claude Code
Lives: Your terminal
Sees: Entire codebase
Does: Runs commands, edits files, commits
Memory: Across sessions
Best for: Multi-file tasks, full workflows
The key distinction
Traditional tools analyze what's visible in your editor. Claude Code reads entire codebases, understands file relationships across directories, executes commands to verify changes work, and iterates until tests pass, without you manually orchestrating each step.
Where Claude Code fits in your stack
It doesn't replace your IDE. It runs alongside it. You keep VS Code or JetBrains open. Claude Code runs in the terminal and you review its changes as diffs in the editor. Both windows open at once is the normal setup.
02 / 05
03 · What it actually does
Real-world use cases
From everyday tasks to ambitious multi-hour sessions. Here's what people are actually doing with it.
Routine Everyday tasks
Bug fixes across multiple files, writing unit tests, explaining unfamiliar code, handling Git workflows (branching, PRs, commit messages), refactoring functions, and generating boilerplate from patterns already in your codebase.
Transformative Deep codebase work
Large-scale refactors across many files, implementing new features end-to-end (touching routes, models, tests, and docs in one go), migrating between frameworks or languages, and doing thorough code reviews of entire PRs.
Agentic Autonomous long-horizon tasks
Multi-hour sessions on genuinely hard problems. Rakuten gave Claude Code a 12.5M-line codebase and it worked for 7 hours with minimal guidance, completing a complex ML implementation at 99.9% accuracy. A Google principal engineer noted it replicated a year of architectural work in one hour.
Know the limits What still requires you
Engineers report using Claude for ~60% of their work with a 50% productivity boost, but only 0-20% of tasks can be fully delegated without review. Strong test-driven development practices amplify results significantly. You're the architect; Claude Code is the implementor.
03 / 05
04 · Your reading list
Official docs & best resources
Curated links for today's orientation. Start with the first two. They cover 80% of what you need.
Official Claude Code overview
code.claude.com: The canonical documentation. Start here.
↗
Introduction to agentic coding
Anthropic blog: How the paradigm shift from autocomplete to agents works.
↗
DeepLearning.AI free course
Video lessons (Claude Code: A Highly Agentic Coding Assistant).
↗
Why everyone is talking about it
Shawn Kanungo: Context, growth story, and what sets it apart.
↗
Claude Code product page
Anthropic: Real demos, use-case videos, and subscription options.
↗
💡 Don't try to read everything today. The official overview + the blog post give you the full mental model. Save the course and the deep-dive article for when you're a few days in and have hands-on questions.
04 / 05
05 · Your outcome
Write your 3 workflows
The most important thing you'll do today. Be specific: not "coding help" but "refactoring the auth module" or "writing tests for functions I'm scared to touch."
1 Workflow one
2 Workflow two
3 Workflow three
Day 01 checklist
- Read the official Claude Code overview documentation
- Understood the difference between Claude Code, Copilot, and chatbots
- Explored at least one demo or blog post about real use cases
- Wrote down 3 workflows I want Claude Code to help with
- I have a clear mental model of what Claude Code is, and what it is not
05 / 05
Day 01 complete
You now have the mental model. Tomorrow is Day 02: Installation & First Session. You'll set up Claude Code and run your first real commands.
Save this file. It's your Day 01 reference. Come back to it if the mental model ever gets fuzzy.