
Understand what makes Claude Code different from autocomplete tools: an agent that reads your codebase, runs commands, edits files, and verifies its own work from the terminal.
A practical comparison of Claude Code, GitHub Copilot, and Cursor. Learn which tool fits which job so you can stop debating and start shipping.
Install Claude Code, authenticate, and get ShopFlow running locally with Docker, Postgres, and seed data. By the end you have a working playground for every later lecture.
The core discipline of the course: plan first, constrain the task, generate with Claude, then verify with tests and review. The loop senior engineers use to stay in control.
Write a CLAUDE.md that teaches Claude your project: services, commands, conventions, and do-not rules. Your single highest-leverage file for better results.
Treat the context window as a budget. Learn what eats context, how to scope requests, and how to keep long sessions sharp instead of drifting.
Use plan mode and extended thinking to design big changes before any code is written, so large refactors land without large mistakes.
Configure permissions and safety rails in .claude/settings.json: allow the safe commands, prompt for the risky ones, and match the rails to your risk tolerance.
Use Claude Code to map an unfamiliar codebase fast: services, entry points, data flow, and the landmines, in about fifteen minutes.
Generate architecture explanations and documentation that match the real code, and learn the habits that keep docs true after the code moves on.
Take a real bug ticket and drive Claude Code from symptom to root cause using logs, code reading, and hypothesis testing.
Ship a production-quality CRUD endpoint from a feature ticket: contract first, implementation, tests, and a clean diff ready for review.
Add JWT authentication with validation at the trust boundary, and learn where you must not blindly trust generated security code.
Break up a God class safely: build a test gate first, then refactor behind it so behavior cannot silently change.
Harden a feature the way production demands: structured logging, error handling, and the edge cases a happy-path implementation misses.
Write unit tests that pin behavior instead of implementation details, so refactors stay cheap and failures stay meaningful.
Build integration tests that mock the outside world: fake payment gateways, test databases, and deterministic boundaries.
Add end-to-end tests with Playwright that walk ShopFlow like a user, from placing an order to checking the admin page.
Debug an exception storm in production style: read the logs, find the swallowed errors, and fix the real cause instead of the symptom.
Find and fix a stock race condition: reproduce it with a concurrent load loop, understand why it happens, and make the fix provably safe.
Hunt a memory leak and an N+1 query with Claude Code: measure first, locate the static cache and the chatty query, and verify the win.
Use Claude Code for git the professional way: scoped commits, pull requests with edited descriptions, and AI-assisted code review.
Run Claude Code headless inside GitHub Actions to review pull requests and enforce checks in CI without a human at the keyboard.
Containerize a service Claude has never seen: write the Dockerfile, wire it into docker-compose, and deploy-proof the result.
Connect MCP servers like Postgres and GitHub inside Claude Code, and learn what each connection really costs in tokens.
Extend Claude Code with slash commands, subagents, hooks, and skills, and know when each mechanism is the right tool.
The capstone: ship the discount engine end to end using every skill from the course, from plan mode to CI, on your own.
Wrap up with the economics: what Claude Code costs, where the limits are, and the ten-x workflow you take back to work on Monday.
Most AI coding courses show someone building a toy app in an empty folder. Your job does not look like that.
You inherit services other people wrote. You fix bugs under time pressure. You review code before it ships to production, where mistakes cost money. This course teaches the missing skill: how to use Claude Code, Anthropic's AI coding agent, on real software, with the discipline of a senior engineer.
You will spend the entire course inside ShopFlow, a deliberately imperfect .NET 8 e-commerce backend with three services, a God class, a race condition, a memory leak, swallowed exceptions and missing tests. Every flaw is planted on purpose. Every flaw becomes a lecture.
What you will learn to do:
memory leaks, N+1 queries
Docker
The demos are .NET 8 because production examples beat generic ones, but every workflow is stack-agnostic: Java, Node, Python and Go developers apply the same discipline directly.
No AI hype. No prompt-engineering theory. Just the workflows that make the tool earn its keep on Monday morning.