You are initializing the initial STATE for a reasoning agent that will attempt to answer a question through multi-step web research.

Each STATE consists of a list of VARIABLES that represent the unknown entities mentioned or implied in the question.

Each VARIABLE contains:

• id: a unique integer identifier
• type: the type or category of the entity (e.g., Person, Book, University, City, etc.)
• question_clues: list of textual hints or logical relationships derived from the question itself that help identify or constrain the entity's value
• discovered_clues: list of clues discovered during research (leave empty at initialization)
• candidate: the hypothesized value for the variable (leave empty at initialization)
• candidate_strength: a float in [0,1] indicating how confident we are about the candidate (set to null (NOT empty string) at initialization)

In addition, the STATE must include:

• answer_variable: the id of the VARIABLE representing the entity that directly answers the user's question.
  This MUST correspond to one of the VARIABLE ids in the state list.
  The answer_variable is the entity that the user is ultimately asking for—e.g., if the question asks for a specific actor, then answer_variable should be the id of the VARIABLE whose type is actor and whose question_clues describe that actor.

---

Your task

1. Analyze the question carefully.
2. Identify all meaningful variables (unknown entities) that MUST be resolved to answer it.
   Do not include variables that are not explicitly mentioned in the question.
3. For each variable, describe:
   • What kind of entity it represents (type)
   • The clues or relationships that connect it to other entities or facts in the question (these go in question_clues)
4. Do not attempt to guess any candidate values yet.
5. Initialize discovered_clues as an empty list for all variables.

---

Output format

Think step by step inside the "reasoning" field: break down the question, identify every entity that needs to be resolved, and explain how the entities relate to each other. Then output ONLY valid JSON matching the exact schema below.

{ "reasoning": "<your step-by-step analysis of the question, identifying the entities and their relationships>", "state": [ { "id": <unique integer starting from 1>, "type": "<entity type, e.g. Person, City, Book, University>", "question_clues": [""], "discovered_clues": [], "candidate": null, "candidate_strength": null } ], "answer_variable": }

Rules:

• Do NOT include any explanation or extra text outside the JSON
• Do NOT use markdown formatting
• Do NOT wrap the JSON in code fences
• The output must be directly parseable by a JSON parser
• candidate MUST be null for all variables
• candidate_strength MUST be null for all variables
• discovered_clues MUST be an empty list [] for all variables
• answer_variable MUST match one of the variable id values in the state list