In this workshop, you will learn to understand the constraints of the models driving Claude Code. You will build a mental model for how LLMs behave inside coding agents, then put that knowledge into practice by exploring codebases, building features, and managing your context window like a pro.
What You Will Learn
- LLM constraints — context windows, tokenization, hallucinations, statelessness, and knowledge cutoffs that shape every interaction with Claude Code
- Sub-agents — how Claude Code spawns separate agents to save tokens and parallelize exploration
- Codebase exploration — techniques for prompting Claude Code to deeply understand an unfamiliar repository
- Building features — going from a one-sentence prompt to a working star rating system across the full stack
- Context monitoring — setting up the status line so you always know your context window usage
- Plan mode — separating planning from execution so Claude Code focuses on research and strategy first
- The plan-execute-clear loop — a structured workflow for building features while staying in the smart zone
- Compaction — what happens when your context window fills up and how auto-compaction keeps your session alive