Google announced on June 10, 2026, a major upgrade to its AI research and knowledge management tool NotebookLM. The new system has been officially integrated with the Gemini 3.5 Flash model and deeply incorporated Google's internal coding tool Antigravity.
The core of this upgrade is the configuration of an independent cloud computer for each user's notebook, enabling the tool not only to write and execute code in real time, but also to introduce deep research functions based on AI agents (Agent) to support more complex academic and engineering projects. Internal Google testing data shows that the new Agent architecture outperformed the previous version in approximately 65% of evaluation scenarios.
In terms of functional expandability, NotebookLM has added a "Zero-source" search option, allowing the system to dynamically access authoritative resources across the web through Google search and automatically complete annotations. In addition, the output format of research results has been significantly expanded, enabling users to export analysis reports as PDFs, Excel spreadsheets, PowerPoint presentations, and various image files with professional charts with one click. This update is now fully available to global Google AI Ultra users and Workspace enterprise users who have access to AI Ultra and AI Expanded.
As a representative product of early exploration into RAG (Retrieval-Augmented Generation), NotebookLM's evolution toward "AI agentization" and "dynamic computing" reflects the current trend of AI productivity tools evolving from simple text summarizers to end-to-end research assistants with active execution capabilities. By seamlessly combining the code execution environment with internet search, Google is further raising the technical barriers for deep research based on vertical knowledge bases.
