Control NotebookLM From ChatGPT, Grok, Mistral, Gemini and more. IT'S CRAZY 🔥
Introduction to Jacht Gro Gemini
Overview of the Tools
- Jacob introduces the concept of controlling a notebook through chat interfaces on desktop using Jacht Gro Gemini.
- He mentions the release of Notebook LM MCP, thanking contributors for their feedback and support.
- Jacob discusses ongoing updates to the MCP, specifically addressing authentication issues that can last between 10 to 20 minutes.
Authentication Improvements
- Automated background authentication is introduced, which will run Chrome in headless mode when token expiration is detected.
- The goal is to ensure that Chrome does not remain in memory after use, enhancing performance and user experience.
Setting Up MCP Super Assistant
Installation Process
- Jacob explains how MCP Super Assistant injects elements into chat UIs and supports multiple platforms but notes its clunky usability.
- He demonstrates adding the tool to his Chromium-based browser and confirms successful installation.
Configuration Steps
- A server needs to be run for connection; Jacob navigates through setup instructions without entering developer mode.
- He creates a configuration file (config.json), utilizing an example from Notebook LM for guidance.
Running Commands and Testing Functionality
Command Execution
- After setting up, Jacob runs commands in a terminal where the config file is located, confirming successful execution.
Interaction with Chat Interfaces
- He emphasizes educating the chat interface on interacting with MCP by inserting necessary information each time it loads.
Testing Notebooks with ChatGPT and Grock
Initial Test Attempts
- Jacob attempts to list notebooks using ChatGPT but encounters issues due to untested functionality prior to this session.
Successful Interaction with Grock
- Switching to Grock yields better results as he successfully retrieves notebook information despite initial formatting surprises.
Researching Nvidia Rubin Notebook LM
Initial Instructions and Model Selection
- The speaker requests a fast web research on the Nvidia Rubin notebook LM, expressing frustration with the current model's performance.
- They instruct to use MCP tools for executing the task, indicating confidence in the model's ability to run without further input.
Execution of Research Task
- The speaker checks the status of the research but encounters limits due to free usage constraints; however, they confirm that instructions were executed.
- Transitioning to Gemini, they highlight its capability to attach multiple notebook LMs and ask questions, noting this feature inspired their own ideas about MCP.
Management and Configuration Challenges
- The speaker discusses needing an interface for end-to-end management of notebooks and attempts to configure settings for checking research status.
- They express hope that using a different thinking model will yield better results while manually initiating processes.
Results from Notebook Interaction
- After refreshing their notebook page, they expect all sources to be added as confirmation of successful execution.
- Acknowledging limitations with free versions of Grock compared to paid ones, they note differences in performance between models like ChatGPT and others.
Final Attempts with Paid Account
- The speaker shares their preference for other AI models over ChatGPT due to perceived inefficiencies and mentions hitting usage quotas.
- Switching to a paid account allows them another attempt at running commands successfully within ChatGPT’s environment.
Successful Completion of Tasks
- They successfully configure commands for identifying notebooks and generating infographics, demonstrating improved functionality with paid access.
- The process culminates in creating an infographic within the notebook, showcasing effective multi-tasking capabilities of the AI when properly configured.
Conclusion on Performance Observations
- The speaker reflects on their experience over 20 minutes without issues or expiration notices during tasks performed under a paid subscription.