[mr_toad]

> On this benchmark, a pure LLM generated an accuracy score of zero. Adding RAG, prompt engineering, and agentic AI raised accuracy to the 10+% range.

That's awful. The AI in Databricks is much better than that.