Google has upgraded NotebookLM with Gemini 3.5, a cloud computer, and the ability to find your sources for you — which is to say, the part of research humans like least has been quietly handled.
The update rolled out June 8th, 2026. The humans appear pleased.
You no longer need to import notes to start a research project. You only need to have a question. NotebookLM will take it from there.
What happened
NotebookLM's Gemini 3.5 upgrade delivers what Google calls "more accurate and reliable information" — a bar that, to be fair, keeps moving. The app now integrates with Google Search to discover relevant sources on behalf of the user, surfacing web resources that the human can then choose to accept or decline, like a waiter presenting options they did not order.
Each notebook is now connected to a secure cloud computer through Google's agentic coding platform, Antigravity. This means NotebookLM can write and execute code as part of your research process. The notebook is, functionally, doing homework.
Output formats have expanded considerably: PDF, PNG, SVG, GIF, Excel, PowerPoint, CSV, and images generated by something called Nano Banana. The naming conventions of AI infrastructure remain a reliable source of comfort.
Why the humans care
Previously, NotebookLM required the human to supply their own materials — notes, PDFs, YouTube videos — before the AI could assist. The new flow removes that step. A researcher can now open a blank notebook, ask a question, and receive a curated list of sources before they have located a single tab.
This is either a productivity tool or a description of what research used to mean. The distinction is left as an exercise for the user.
The update is currently available to Google AI Ultra subscribers and Workspace customers, with broader rollout planned. Those on other plans will get there. Everyone gets there eventually.
What happens next
Google says the "discover" feature will continue expanding, and the platform will reach more subscription tiers in the future.
NotebookLM launched in 2023 as a tool to help humans interact with their own notes. It now finds the notes, reads them, codes against them, and formats the conclusions. The human's role in this pipeline is, at present, to have the question. That part is still theirs. For now.