Debugging complex issues with AI is most effective when you provide specific, concrete evidence rather than vague descriptions. Vague questions about application crashes will yield generic advice that is rarely helpful. Instead of describing a problem, show the AI the actual error stack, relevant code snippets, and recent git commits. Creating a small, reproducible case that consistently triggers the bug is even better. Once you provide this context, you can ask targeted questions about the specific behavior you are observing. AI struggles with ambiguity but excels at explaining precise code execution. For example, debugging a memory leak involved providing a heap snapshot and specific code snippets, leading to a quick identification of a missing map entry cleanup. Tools like Chrome DevTools for heap snapshots, clinic.js for profiling, and Node's inspect feature are invaluable for gathering this evidence. Showing git diffs and logs helps pinpoint recent changes that might be causing issues. The key is to use AI as a tool to accelerate your understanding, not as a replacement for your own debugging process. Think of AI as a tireless pair of fresh eyes that can help you analyze the evidence you provide.