Classic scenario. You get handed someone else's codebase. The original author has left for a beach somewhere, better gig, or possibly witness protection. Yesterday: Read it. Line by line. Function by function. Build a mental model. Understand it deeply. Then fix it. Time required: Somewhere between "too long" and "I have regrets about my career choices." Today: Don't read it. Ask AI what it does. Investigate what smells funny. Fix the smell. Move on with your life. This talk is about the shift from syntax comprehension to diagnostic intuition. It's the difference between memorizing every street name in a city versus knowing the neighborhoods well enough to sense when you're heading somewhere sketchy. You don't need to parse every semicolon, you need to recognize when something's wrong. The key skill isn't reading code faster. It's asking better questions. "What's the happy path here?" "Where could this blow up?" "Why does this function exist?" And critically: "What's the thing the original author was worried about that I should also be worried about?" You'll walk away with practical patterns for smell-based debugging, the questions that matter more than comprehensive reading, and a significantly improved relationship with legacy code. (It's still not a good relationship. But it's faster.)