2 Comments
User's avatar
Lisa Raehsler's avatar

This is great, detailed information. I am definitely underutilizing Notebook. I do use Gemini for most of my Google Ads work because it seems to be best trained on its own products.

Pawel Jozefiak's avatar

The Gems workflow is the shift most people miss. They're treating NotebookLM like a search engine when it's closer to a context engine -> the difference between querying a library and having a librarian who already knows your project.

What's actually interesting here: the ROI isn't in the tool, it's in the clarity of what you're building toward. Repurposing content and simplifying documentation sounds obvious until you realize most people have the knowledge locked in notebooks they never revisit.

I've been running a similar logic with AI agents -> instead of chasing which model is smartest, I focused on which setup actually ships output. Turned out Haiku outperformed Opus on most real tasks once I stopped optimizing for impressiveness: thoughts.jock.pl/p/claude-model-optimization-opus-haiku-ai-agent-costs-2026