Transcription
Most people use Notebook LM and Gemini separately. You dump some docs into Notebook LM, ask it a few questions, then switch over to Gemini when you need something more. But Google connected these tools in a way that most people are completely sleeping on. And when you put them together the right way, what you end up with is something I've been calling a Notebook LM Gemini agent, a specialist that's permanently grounded in your own knowledge base [music] that already knows your context before you ask a single question. So, let me show you what that actually looks like.
So, let me explain what I mean by that term because it's not just a catchy name. It describes something structurally different from a regular AI agent. When most people use AI, the pattern goes something like this. You open a session, you explain your context, you get something useful, you close it. Next time you come back, you start from zero again. You're spending a chunk of every session just getting it back up to speed before you can do anything. People try to patch this with better prompts, longer system prompts, more detailed setups, and that helps at the margins, but it doesn't solve the core issue. Your knowledge, your documents, your frameworks, your research, none of that is actually connected to the AI in a persistent way. You're uploading files per session, pasting context manually, hoping nothing important falls through the cracks.
A Notebook LM Gemini agent solved this by treating the problem as two separate layers that need to be connected. The first layer is knowledge. Everything you know, everything you research, everything relevant to the domain you're working in. The second layer is behavior. How the AI should think, what role it plays, how it structures its responses. Notebook LM handles the knowledge layer. It can hold up to 300 sources per notebook, PDFs, Google Docs, web pages, transcripts, audio files, and critically, when it answers questions, it's grounded in those sources. It's not improvising from general training. It's citing your actual material. That's the whole design philosophy behind it. Gemini Gems handles the behavior layer. A gem is a custom AI assistant where you define the role and logic once, and every time you open it, that's already baked in. No setup, no re-explaining. You just show up and work.
Before this integration, these two layers were disconnected. Your behavior lived in the gem. Your knowledge lived somewhere else. Now you can attach a Notebook LM notebook directly to a gem, and that's what creates this agent: knowledge plus behavior, persistent, grounded, and automatically updated every time your notebook grows. That's Notebook LM's Gemini agent, and once you build one, standard AI chats start to feel pretty limited by comparison.
Before we get started, allow me to introduce today's video sponsor and a longtime supporter of the channel, Scribba. It's the only coding platform I've found where the video and code editor are combined into one. That means with Scribba, you can hit pause anytime, edit, and run the code right inside the lesson like a real pair programmer beside the teacher. Around 30% of their courses are free. And if you're a web dev who wants to include AI into your apps, Scribba's AI Engineer path takes you from the basics to building real AI-powered apps, from chat assistants to recommendation engines and computer vision tools. And if you're brand new to coding, I highly recommend to start with their Full Stack Developer path. A 100-hour complete journey through JavaScript, React, Node, databases, and even AI too. Build actual projects from what you learn. Scribba isn't just another video course platform. It's an interactive, hands-on learning platform that gets you building today, not someday. Learn to code with Scribba, from full stack to AI. Start now for free and get 20% off their pro plans with my link. Visit the link in the description below to claim your pro plan.
So, the Gemini agent we're going to build today is called a personal investing research analyst. The idea is simple. Instead of asking Gemini generic questions about a stock and getting generic answers, we're building an agent that evaluates anything through your personal investing framework, your criteria, your personal risk tolerance, your logic. And the reason I picked investing for the demo isn't because this is an investing channel, it's because the grounding is really easy to see. You can immediately feel the difference between a generic AI answer and one that's working from your actual material.
So, we'll start in Notebook LM. So, this is going to be our knowledge layer. This is where the agent's brain lives. We'll create a new notebook. So, let me click on create new notebook. Now, here's where something called Deep Research comes in. And this is one of the more underrated features in Notebook LM. Most people don't have a fully written investing framework sitting in a document somewhere. Deep Research lets you generate one. Instead of building your knowledge base manually from scratch, you give Notebook LM a prompt and it goes out across the web, pulls from multiple sources, high-quality sources, and builds you a structured, cited research document automatically.
So, I'll give it a prompt something like: "Generate a comprehensive personal investing framework for a long-term growth-focused investor who prioritizes companies with strong AI exposure, avoids high debt balance sheets, and wants to understand both the bull and bear case before making any decision. Include evaluation criteria, red flags, sector preferences, and a consistent template for assessing any new stock." So, that's the prompt I'm giving it. It might sound a little bit contradictory because I'm asking for high AI exposure and not a lot of debt, but, you know, for this case, we'll just do this. And all you have to do now is just press enter. So, what it's doing right now is going across the web, finding high-quality sources for you while you can just, you know, chill out for a couple of minutes and come back to it when it's ready.
All right, looks like our Deep Research is complete. And you can see this comprehensive report that it has created. So, I'm just going to press view. And what you're looking at right now is that this report is created pulling from multiple sources, and it has built this out. This isn't a paragraph summary or anything like that. It's a real structured document, and this is the moment where the pattern starts to kind of make sense. We're using AI to build the knowledge base that the agent is going to reason from. You didn't have to write anything manually. So, I'll add all these sources to my notebook. If you wanted to, you could vet all these sources, learn more about them, but for our purposes, I'm going to trust Notebook LM and just press import. And that's pretty much it. That's the whole knowledge layer done within a couple of minutes. All these sources are going to be embedded into this notebook that we can reference in our workflow later.
So now we'll work on the behavior layer, the gem. This is where we define how the agent thinks and responds. So what we're going to do is, if you click on these three icons over here, expand menu, we'll click on that and go into Gems. So once you're in here, what you'll see is that you'll see a gem made by Labs and also at the bottom, a Gem Manager and My Gems. We want to go over here. So click on this new gem button over here. And we have to give it a name. So let's call this an Investing Analyst.
So then in the instructions section, I have given it these instructions, which are: "You are a personal investing research assistant. When evaluating any stock, sector, or trend, always ground your analysis in the investing framework in your knowledge base. Apply the evaluation criteria. Flag any red flags defined in the framework. And always present both the bull and bear case. Never give a buy or sell recommendation. Your job is to structure the analysis so the user can make their own call." So, this is the instructions, which is kind of similar to our prompt, but you can copy the same prompt, change it around for your instruction section. It's up to you.
Now, the key step now is creating this knowledge base over here. If you didn't have Notebook LM, you would manually upload your own frameworks or files, whatever you have. But all we have to do now, which is going to save us a lot of time, is click on this plus icon. Click on Notebook LM. And then you have access to all the notebooks that you have created. So I'll attach a notebook, the one that we just created right now, which is the AI Investment Strategy and Risk Analysis one. I'll click on that and press add. This is the key step that connects the two layers. The gem now has the framework as its permanent knowledge base and is going to be referencing that whenever we use it. And once we have given it the knowledge base, all we have to do is click on save over here. So let's just click on this. So our Investing Analyst gem has been created.
Now we can start interacting with it. So let's interact with it. I'm going to give it a simple prompt like: "Run Nvidia through my investing framework, one of the biggest AI companies." Let's see what it does. So look at what it's doing. This isn't a generic Nvidia overview. It's running it against the specific evaluation criteria in the framework, flagging the red flag section, structuring the bull and bear case exactly the way the document defines. And that's the grounding in practice. The agent knows you better and is reasoning from your material, not improvising. This is exactly what I mean by a Notebook LM Gemini agent. It's not a chatbot you have to brief every time. It already knows the framework. It already knows how to apply it. You just bring it to the stock.
And if we take a look at the output, we have our scenario analysis, bull and bear case summary for us, showing that it meets our framework solvency criteria with the net cash position and leaves the foundational economic layer of AI. However, there is a projected decline in market share and the risk of an industry-wide "reality check" quote unquote AI bubble. Now, if I compare this to just asking Gemini, "Should I buy or wait?", you're going to know that, okay, the strategy is not going to be one-size-fits-all there. The case for buying: market leadership, high valuation, positive outlook. You can already see the difference in the output quality over here, right? It's pretty generic versus what we have over here is structured to the framework. And that framework is customizable. You can go into Notebook LM, change our framework, give it more restrictions, and play with it, right? So, we always have access to our framework. We don't have to summarize or redefine our framework every single time.
And just to show you guys that, I'm going to give it a new stock called Palantir and tell it to do analysis on that. So, it's the same framework, different stock. Nothing has changed. That's the whole point. You built the system once, and now every question you bring to it gets evaluated through your lens automatically. That's what separates this from just having a good conversation about investing in a regular chat.
And this applies way beyond investing. Think about whatever you do repetitively: content creation, research, studying, legal work, product decisions. The pattern is going to be identical. Your knowledge goes into Notebook LM. Your behavior logic goes into Gem. You connect them, and you've got a Notebook LM Gemini agent for that domain. A student could load every course material for the semester, syllabus, lecture notes, textbooks, into a notebook and build a single gem that already knows the full context of their coursework. A content creator could load their past scripts, style notes, and audience research so the agent drafts in their voice without being told to every single time. A product manager could load specs, user feedback, and competitive research. So, every question gets answered in the context of the actual product. Same two layers, different knowledge, different instructions. That's pretty much it.
And the last thing I want to show you guys is that your knowledge base is persistent. When you add new sources, your Notebook LM notebook gets updated, and that notebook is already referenced to your gem. So, you don't have to touch the gem at all. It already has the new knowledge. So, the agent grows as your notebook grows. So, you're not rebuilding anything. You're not reconfiguring anything. It's just going to compound over time. But that's the Notebook LM Gemini agent. Two layers: knowledge and behavior, connected and persistent. It gets more useful over time instead of resetting every single session. And this is one of the most underrated things you can do with Google's AI stack right now, and most people aren't doing it because the tools feel separate. Now you know how to connect them.
Links are in the description. Drop in the comments what you would build your Notebook LM Gemini agent for. I want to see what people come up with, and maybe I'll do a video on that too. Who knows? But that's it for today's video. Make sure you guys are subscribed to the channel. Follow our new newsletter as well at universeai.behive.com, as well as subscribe to the main channel World of AI, and support us on X by following the Universe of AIZ as well. Until then, I'll see you guys in the next video.