This course teaches the agentic AI coding workflow most developers do not know exists. You will go from running your first claude command to spawning teams of agents that work in parallel, building your own MCP servers, writing hooks that fire on every commit, and routing work across nine different models including Claude Opus, GPT-5 Codex, Gemini 2.5, GLM-5.1, MiniMax M2.7, and the local Ollama stack with DeepSeek 671B, Kimi 1T, and Qwen Coder.
Why this course exists
Most AI coding courses you find right now are shallow Cursor tutorials or ChatGPT prompt galleries. This is the deep one. The one I wanted to find when I started, and could not. Every lesson shows real terminal output, real apps, real workflows. No slide decks. No theory. No filler.
What you will master:
- Claude Code installation, configuration, and the full CLI workflow on Windows, macOS, and Linux
- Building MCP servers in both Python and TypeScript with resources, tools, and prompts
- The Skills system: project-scoped and global, plus advanced patterns like multi-step and self-healing
- Subagents and parallel agent teams: 5+ workers running on a single project at the same time
- The full multi-model architecture: dual mode, triple mode, smart routing, cost-aware budgets
- Hooks and automation: PreTool, PostTool, Stop, Notification, and CronCreate scheduled agents
- Persistent memory through brain MCP, project profiles, and end-of-session lessons
- The rules system: Karpathy pillars, intelligence rules, red team rules, and the credentials model
- Real-world workflows: shipping a full feature, debugging a production bug, refactoring legacy code
- Module 13 — the Multi-Model Power Stack covering GPT-5 Codex, Gemini, GLM, MiniMax, Ollama, Cursor, Cline, OpenCode, Kilo Code, and Aider
