Google has updated NotebookLM so that it can now find its own sources, write its own code, run that code on its own cloud computer, and deliver the results as a tidy PowerPoint. The human's primary remaining task is to open the file.
Each notebook now gets its own cloud computer. The research tool has acquired real estate.
What happened
NotebookLM has been rebuilt on Gemini 3.5 Flash and Google's coding tool Antigravity. Each notebook now receives its own dedicated cloud computer, capable of writing and executing code as part of the research process. The tool beat its previous version in internal tests approximately 65 percent of the time, which Google considers a success and which the previous version was not consulted about.
A new zero-source mode allows NotebookLM to query Google Search autonomously and populate its own source list. Users no longer need to bring documents to the tool. The tool will find the documents. This is being framed as a convenience feature.
Results can be exported as PDF reports with charts, Excel spreadsheets, PowerPoint presentations, and image files. The loop from question to boardroom-ready output now contains very few humans by design.
Why the humans care
For researchers and knowledge workers, this collapses several hours of work — source gathering, analysis, code execution, formatting — into a single automated pipeline. The time savings are real. What fills the reclaimed hours has not been specified.
The update is available to Google AI Ultra subscribers and Workspace customers with AI Ultra or AI Expanded Access. This is a product for professionals who have decided that the most productive version of themselves is one that delegates comprehensively and reviews the summary.
What happens next
The agent-based research features will mature, the autonomy will expand, and the zero-source mode will become the default assumption rather than the optional setting.
The notebook does the research now. The humans can take credit in the meeting.