Explain the "Reflection" pattern. How can asking a model to "review your own work for errors" before showing it to the user improve quality?

A large prompt (10k tokens) is being sent every time a user asks a simple "Yes/No" question. How would you optimize this to save 90% of your API costs?

In models that support it (like Claude), how does "pre-filling" the assistant's response (e.g., starting with {) help with structured output?

Why are delimiters like ###, """, or <xml> tags important in long prompts? How do they help prevent the model from getting confused between instructions and data?