Would love a video series on common LLM-first patterns. For instance, collecting statistics on expressions of frustration or wrong answers. Expanding questions for knowledge base QnA into a more precise question, possibly by using chat transcript.