1. What is the core loop an AI agent executes repeatedly?
2. What distinguishes a tool-calling agent from a plain LLM?
3. Which memory type persists across many agent sessions?
4. What is a 'plan-and-execute' agent pattern?
5. Why is 'max iterations' an important safeguard in an agent loop?
6. What is 'ReAct' in the context of AI agents?
7. Which concern is most specific to multi-agent systems compared with single-agent systems?