The Full Curriculum: 13 Modules. Two Halves. One Hinge

How the AI Engineering course is built and why the modules come in this order. The first half is learning to use AI properly: fundamentals (what an LLM is, tokens, why models hallucinate), a coding on ramp (terminal, git, project structure, Python and TypeScript), then inside the model (tokenization, embeddings, attention, transformers, context windows, reasoning models, evaluations), then models connected to code (APIs, streaming, structured output, tool use, caching, retries, token budgets), and a full chapter on context engineering (RAG, memory, context rot, compaction). The hinge: before it you learn to use AI at a high level, after it you build with it. Four chapters take Claude Code from level one to level four, from using it well to configuring it, orchestrating several agents, and building skills, MCP servers and agent systems. Then underneath the tools: how an agent harness works, and building your own coding agent from scratch, the agent loop, file tools, shell access, permissions, persistence, compaction, subagents and an MCP client. The final chapter is production reality: observability, security guardrails, cost, deployment. Do the exercises in order and keep the repository open while you learn. The course lives inside The Rabbit Hole, free to join: https://www.rupayanism.com