What is the primary function of "extended thinking" in a model like Claude?	It allows the model to reason step-by-step before generating a final response.
What is the key failure mode that extended thinking helps avoid?	Confidently getting a multi-step question wrong by answering immediately without internal reasoning.
When extended thinking is enabled, what internal process is visible in the response?	Internal reasoning tokens, often called a chain of thought.
With Opus 4.7, what is the key feature regarding "thinking"?	Thinking is adaptive, meaning the model decides dynamically when and how much to think without needing a token budget.
Which parameter is used to control the depth of thinking, and where must it be placed?	The effort parameter; it must be placed inside `output_config`.
Name three types of problems where extended thinking is most beneficial.	Math, multi-step logic, or regulatory analysis (anything involving trade-offs).
When should a user skip enabling extended thinking?	For simple tasks like classification, extraction, or boilerplate generation.
In a production setting, what is the major advantage of using extended thinking?	It allows the agent to connect problems and reason across different sections, rather than finding them one at a time.