1 The shift: Code is free now. Understanding isn't.
This chapter covers
- Comprehension debt: the gap between code you ship and code you understand
- Coder mode vs engineer mode: generating code vs understanding it
- Using AI to understand a codebase, not just generate more code
- Architectural thinking: the 10% skill AI can't replace
- Turning AI's flaws into reps that sharpen your judgment
AI tools have made writing code significantly easier, a practice often called vibe coding. In vibe coding, you describe a feature to an AI, let it generate the code, and then ship it with minimal review. This approach is common: GitHub reports that Copilot writes nearly half the code in files where it's enabled. Boris Cherny, who developed Claude Code, even completed a month of work (259 pull requests and roughly 40,000 lines) without opening his editor, generating every line with AI (https://fortune.com/2026/06/11/anthropic-claude-boris-cherny-doesnt-write-code-by-hand-anymore/). This book, however, focuses on the other side of the equation: using AI to understand code, especially systems you didn't write and can't yet explain, rather than just generating more of it.