Original video in Chinese.
Key Takeaway
- NotebookLM upgrade: now powered by Gemini 3 (Flash for conversations + Pro for key tasks), filling in the model gap; connected with the Gemini app, so you can directly add notebooks as an external knowledge base.
- Knowledge base advantages: automatically structures messy materials (generates summaries/timelines/tables/FAQs/study guides); dual standalone + external modes; integrated into the Google ecosystem; mobile app supports photo uploads, so anything can be added to the knowledge base.
- The best solution for ordinary people: no need for complicated knowledge base tools. NotebookLM is powerful, flexible, multimodal, and portable, and it’s included in the $20 Google One subscription; I’m migrating all my materials and stopping the experimentation with other tools.
For ordinary people, there’s really no need to build such a complicated AI knowledge base. Just fully use NotebookLM and that’s already enough. And chances are, it’ll be even more powerful than the knowledge base tools on the market.
I’ve used NotebookLM for quite a while, but before I only treated it as an AI learning tool. It wasn’t until two recent Google updates that I changed my mind.
First, NotebookLM has finally been upgraded to Gemini 3.
To be honest, although this product was pretty good, the model it used before was a bit underwhelming. Because I use Gemini very frequently in my daily life, I have a very clear feel for the model — NotebookLM was most likely using the Flash model, not the strongest Pro model.
And this upgrade is crucial. Compared with the previous 2.5, Gemini 3 has improved a lot. In particular, the newly launched Gemini 3 Flash has seen a very significant boost in intelligence while maintaining high speed.
After the upgrade, NotebookLM is probably still using the Flash model for conversations. But for some key parts, such as deep search and data tables, it will use the Pro model.
With Gemini 3 as the foundation, NotebookLM’s last shortcoming has basically been filled in.
Second, NotebookLM has been connected with the Gemini app.
In Gemini, you can click the plus sign to add a NotebookLM notebook.
So the question is: what exactly can be added from the notebook?
Gemini can add the sources in the notebook. That is, all the documents you uploaded, and the links and text you pasted in.
I saw some people think this is nothing special, just a very simple integration. They think that because they don’t understand the difference between these two products.
In knowledge base products, there are usually two modes: one is query mode, which strictly answers based on document content; the other is chat mode, where the model just chats about whatever it wants.
Applied to Google’s products:
NotebookLM places special emphasis on Source-Grounding, meaning source-document grounding. It must, and can only, find answers from the sources. It’s similar to the feeling of query mode.
Gemini, on the other hand, is more like chat mode. It combines its training data, content found on the web, and the user’s input to produce the best answer.
If you understand the difference between these two products, you’ll know why they needed to be connected.
First, it breaks down “information silos.”
NotebookLM is a closed system, and its accuracy can be very high. But if you ask it about a concept that isn’t in the notebook, it may very well not be able to answer.
At that point, if you import the notebook into Gemini, Gemini can use the material in the notebook while also searching for the latest information online.
Second, cross-notebook analysis.
In NotebookLM, if you have 5 different notebooks, you can’t ask them questions at the same time.
But in Gemini, you can select multiple notebooks as sources at once.
Third, access more advanced capabilities.
NotebookLM’s model has been fine-tuned and is extremely focused on “reading comprehension” and “summarization,” but in creative writing, complex logical reasoning, and other areas, its capabilities are intentionally limited.
Gemini, however, is a general-purpose strong model. After importing a notebook, Gemini can do much more.
Fourth, directly output results.
After chatting in NotebookLM, the results still remain in the chat window.
But Gemini is already connected to the whole Google ecosystem. After chatting based on a notebook, you can directly export the conversation results to Google Docs.
So, taken together, NotebookLM is the most suitable AI knowledge base for ordinary people.
First, it’s powered by the strongest model.
Ordinary people don’t need to understand things like RAG or document chunking. You can just go all in and let magic happen: throw in an entire book or a bunch of financial reports, and the Gemini model behind NotebookLM will handle all of it for you. You don’t need to worry about so much.
Second, it focuses on turning the “unstructured” into the “structured.”
Ordinary people’s knowledge is messy and disorganized. For example, memoranda scribbled down on the fly, all kinds of PDFs found online, contract photos taken with a phone, and so on. These are all “unstructured data” and hard to organize. But NotebookLM can help you structure all these messy things, turning them into something orderly, easy to understand, and easy to search.
Third, it is both a standalone knowledge base and Gemini’s external knowledge base.
As a standalone knowledge base, NotebookLM can add external materials, and it can also add materials from Google Drive. As an external knowledge base, NotebookLM can be fully integrated into Google’s ecosystem and workflow. This kind of flexibility is something no other knowledge base tool has.
Fourth, it has a mobile app and supports photo uploads.
Most knowledge base tools only have a desktop version, which is pretty inconvenient. And because their multimodal capabilities aren’t strong enough, the data sources they support are mainly text-based. But NotebookLM can take photo uploads through the mobile app and understand them directly via Gemini. In other words, as long as you have your phone, anything can be added to the knowledge base.
Fifth, it’s included in the $20 subscription.
I subscribed to Google One for Gemini. Now that I also get NotebookLM’s advanced features, it’s basically like getting a top-tier knowledge base for free. It’s extremely, extremely cost-effective.
Since the end of the year is just around the corner, I’ll take the opportunity to organize my materials and migrate my existing knowledge bases into NotebookLM. After that, I won’t bother with other tools anymore. In fact, at the tool level, you really shouldn’t spend too much energy on this in the first place.
OK, that’s it for this episode. If you want to learn about AI, want to become a super individual, and want to find like-minded people, come join our newtype community. See you next time!