Published on

NotebookLM 2.0 Full Breakdown β€” Agentic Research and Code Execution Change Everything

"Half my day disappears just finding, organizing, and analyzing sources" β€” if you're a teacher or researcher, you know this feeling.

NotebookLM's June 8 update is a direct attempt to hand that "half" to AI. Code execution, automatic source discovery, and multi-format output generation β€” this is no longer a "chat-based summarizer." It's an autonomous research agent.


Table of Contents

  1. What Changed β€” The Start of Agentic Research
  2. A Notebook That Runs Code β€” The Cloud Computer Inside
  3. AI Finds Your Sources For You
  4. Expanded Output Formats β€” From PDF to PowerPoint
  5. Practical Use Cases for Educators

What Changed β€” The Start of Agentic Research

NotebookLM used to analyze only what you uploaded.

This update breaks that constraint. Gemini 3.5 is now the default model, and every notebook has its own sandboxed cloud computer. The AI can now search for sources, write and execute code, and export results as documents β€” all without your manual input. This isn't just "smarter chat." It's AI performing the research workflow itself.

On June 22, a memory retention feature was also added. NotebookLM now remembers previous conversations and task context.

NotebookLM 2.0 main interface after the June update


A Notebook That Runs Code β€” The Cloud Computer Inside

The most striking change is the code execution environment.

Each notebook now has an isolated cloud computer attached to it. Type "calculate the average score from this dataset and plot it" in plain English, and NotebookLM writes the Python code, runs it in the background, and shows you the result.

Transparent Reasoning β€” Seeing the Steps

Previous AI tools gave you answers. This update gives you a step-by-step view of how the AI reasoned through the problem β€” like showing the work, not just the answer. You can follow the AI's logic at every stage.

From an ed-tech perspective, this is significant. When students can see how an AI reasons through a problem, AI stops being a black box and becomes a thinking partner.


AI Finds Your Sources For You

Until now, using NotebookLM meant bringing your own sources first.

That's changed. Type your project topic in the chat, and NotebookLM uses Google Search to find relevant sources and adds them to the knowledge base for you.

Where This Shines

SituationOld WayNow
Starting new researchCollect β†’ upload β†’ analyzeType topic β†’ AI collects + analyzes
Preparing lesson materialsManual search + organizationAI recommends sources + structures them
Literature reviewManual curationAuto-discovery of related papers

Expanded Output Formats β€” From PDF to PowerPoint

NotebookLM's outputs have expanded significantly:

  • Charts and graphs: Visualize data analysis results
  • PDF documents: Export as formatted reports instantly
  • Spreadsheets: Structured data in tabular form
  • PowerPoint slides: Presentation decks generated automatically

The existing Audio Overview feature remains, with more format choices added.

"NotebookLM is no longer a research assistant. It's a research partner." β€” My takeaway after using the update firsthand.


Practical Use Cases for Educators

As both an EdTech CEO and a classroom teacher, here's how I'm thinking about integrating this into teaching:

  • Performance task design: Input a research topic and let NotebookLM gather sources. Cut the time teachers spend on materials curation in half.
  • Student projects: Have students evaluate AI-suggested sources critically β€” a powerful media literacy exercise.
  • Teaching reasoning: Show students the AI's step-by-step thinking as a model for logical reasoning.

Note: These features are currently available only to Google AI Ultra subscribers or Workspace customers with AI Ultra Access or AI Expanded Access.


Closing Thoughts

NotebookLM 2.0 marks a shift from "an AI that analyzes what I give it" to "an AI that thinks alongside me."

With code execution, automatic source discovery, and multi-format output, the question is no longer "how do I use it?" β€” it's "how much should I delegate to it?"


Related Posts

Are you already using NotebookLM? Which feature from this update are you most excited about? Let me know in the comments!


Sources:

NotebookLM 2.0 Full Breakdown β€” Agentic Research and Code Execution Change Everything | MINSSAM.COM