Mahnoor Faisal, a seasoned tech journalist, delves into the evolution of NotebookLM, an AI tool that has been a staple in her tech stack for years. She reflects on the tool's initial appeal, emphasizing its narrow focus and trustworthiness, which set it apart from general-purpose chatbots. However, with Google's recent rebranding of NotebookLM to Gemini, the tool's behavior has shifted, blurring the lines between its sources and the open web. This change has led Faisal to seek a way to rein in NotebookLM's new, more intrusive habits.
Faisal's solution is a simple yet effective fix: adding a single line to the end of her prompts. By stating, "Answer only from sources that support this, name them, flag which ones don't cover it, and don't offer to do anything else," she has successfully kept NotebookLM in its lane. This line ensures the tool stays strictly within the documents she provides, names the sources it relies on, and stops volunteering additional tasks or web searches.
The journalist's journey to finding this solution involved multiple tests and iterations. She initially asked straightforward questions without the added line, only to be met with unprompted offers to search the web. By gradually refining her prompts, she discovered the power of grounding and attribution. The final line not only keeps NotebookLM focused but also delivers a source audit, ensuring transparency and trustworthiness.
Faisal's experience highlights the importance of fine-tuning AI tools to match one's specific needs. Her commentary underscores the challenge of striking a balance between AI's capabilities and the user's expectations. While Google's efforts to enhance NotebookLM's capabilities are commendable, they have inadvertently introduced behaviors that deviate from the tool's original appeal. Faisal's solution serves as a reminder that sometimes, a small adjustment can make a significant difference in how an AI tool functions and aligns with the user's intentions.
