Is reasoning better handled by models or prompts? Vaibhav and Dex tackle this question by contrasting reasoning models with reasoning prompts. They explore the use of reasoning in AI without defaulting to specialized models, drawing on practical examples like querying a Neo4j database with Cypher.

• Illustrates how to convert a basic AI agent into a reasoning model using structured prompts, instead of relying on advanced models.
• Examines real-world cases, such as complex queries about movie datasets, to identify where reasoning models excel.
• Discusses the trade-offs between reasoning models and prompt-based reasoning in terms of accuracy, cost, and prompt engineering effort.

The session primarily uses BAML functions and focuses on efficient reasoning techniques, assuming a strong understanding of LLMs and AI systems.