Ep 880: Gemini Notebook: 7 New Updates and What They Unlock (Replay)

How long has it been since we featured Notebook LM on this show? Apparently so long. Well, it isn't even called Notebook LM. It has a new name. Yes, the tool once known as Notebook LM is no more. But don't worry, one of my favorite AI tools ever didn't go to the Google graveyard. It actually just graduated and got upgraded into the official Gemini lineup and is now called Gemini Notebooks. But the name isn't the only thing that's new. That's because over the last month or so, the team at Google has rolled out a bunch of new features that legit supercharge Gemini Notebooks and make it agentic by default. Which sounds really cool and all, but, well, why haven't we been talking about it? Because many of those upgrades were only for users on the expensive Ultra plan. But this past week, that changed. Now the updates that have completely changed the face of Gemini Notebook are available to all paid users. And that's why we're going to be showcasing the 7 most important updates today as we put AI to work on Wednesdays. All right, so on today's show, here's what you're going to learn. You're going to know why Notebook LM is no more and why Gemini Notebook might actually make more sense. You're going to know the one important grounded change that you really need to be aware of. And you're going to understand the real reason Gemini Notebook going agentic Matters more than ever. Alright, let's get into it. Welcome to Everyday AI. My name is Jordan Wilson, and this thing's for you. It's your daily unedited, unscripted livestream podcast and free daily newsletter, helping business leaders like you and me keep up with the nonstop avalanche of AI updates. I tell you what matters, what doesn't, how to use it, and you take that information to grow your company and career. So it starts here. But please make sure, if you haven't already, please subscribe to the podcast on Spotify or Apple and make sure you go to our website at youreverydayai.com and sign up for the free daily newsletter. We're gonna recapping the highlights from today's show in case you miss anything, as well as all of the other AI news and developments that you need to know. All right, so let's talk Gemini Notebooks. Yeah, Notebook LM is no more. So sad, right? But it's not gone. It just got a facelift both on the exterior and a lot going on under the hood as well. And this is one of those things I think over the last couple of years, a lot of people were kind of nervous because Google has done this before where they've come out with great products, right? And this isn't just in the AI age, in the pre-AI phase, maybe even more. They've come out with very popular products that a lot of people love, almost built like a cult-like following, and then they end up killing those products. So a lot of people were very nervous over the last couple of years as NotebookLM almost seemed like this side quest that was legit amazing, right? No joke. It's— I think we named it like our AI tool of the year in 2024 and 2025. So a lot of people were worried like, hey, don't take this away. It's very It's very unique because there's no, you know, the other competitors, you know, no one from Anthropic, OpenAI, Microsoft, you know, X, Grok, Meta, like no one else has something like NotebookLM that really just grounds its outputs in your data only. So luckily they didn't get rid of it. And it seems like if anything, Gemini Notebook is going to be along for the long run. So, uh, let's quickly go over, not gonna make you wait. Here's these 7 new updates to Gemini Notebook. I'm gonna be going over them pretty quickly now, uh, and we're gonna be doing some live demos so you can see these 7 updates in action. And then when we're done with the live demos, I'll come back and talk about these things a little bit more. So number 1, new update, already talked about it. Gemini Notebook is the new name. That's number 1. Number 2, there is something called collections. That's a way for you to organize notebooks for easier sorting. Number 3, and this one's actually really cool, uh, it's now there's automatic Google Drive syncing. All right, number 4, bringing the Gemini and search access. So your personal notebooks can now be accessed inside of Gemini and soon inside of Google Search's AI mode, which will be really handy. Number 5, new output formats. So yeah, no longer constrained to only what NotebookLM offered. Now Gemini Notebook, you can create PDFs, PNGs, documents, spreadsheets, PowerPoints, markdown files, charts, and images. Number 6, technically each notebook has its own secure cloud compute. Computer, which is really cool. Number 7, just more agentic intelligence. And that's probably the biggest upgrade that we're going to see in Gemini Notebook, now powered by Gemini 3.5 reasoning and Google's Anti-Gravity Agentic Search. All right, so let's do it. Let's take a look live, shall we? What could go wrong? I'm going to share my screen. Livestream audience, do me a favor. Let me know if you can See, all right. So what I did here is I took all of my notes from yesterday's show, had a lot of them. So yesterday's show was kind of on RSI, recursive self-improvement. I have all of my notes together, and this is something I routinely do inside of NotebookLM. I still use NotebookLM almost every single day. Sometimes I'll technically spend like hours inside of NotebookLM, just depends on what I'm preparing for. But To make this a little faster, I essentially just created the same version of this notebook inside of Gemini Notebook. I'm going to probably call it the wrong thing like 5 times. And then I have some prompts ready to go. So for our podcast audience, FYI, our AI at Work on Wednesday show, it's always a hands-on demo. Sometimes it's a little more visual, sometimes it's not. But you can always watch the video version of this on our website at youreverydayai.com. And in the podcast show notes, we always leave a link to today's episode. All right, so here we go. Um, I'm going to go ahead and enter these prompts. We'll probably check on them a little bit as they go, talk a little bit more about these 7 new features, um, and then we'll come back at the end and take a look. All right, so I'm just going to get these started so everyone can see, yes, these are going in real time. All right, so there's my first one that I sent, uh, there's my second one, There's my third one, and there is my fourth one. All right, so a little bit about the content in here. It is long-form content. I think it was like probably 60 pages of notes that I had for the show on recursive self-improvement. And if you want to go listen to that, like I said, that's yesterday's episode, episode 833. All right, where we talked about RSI explained, when AI starts improving itself, and what it means. So essentially, I have these four different duplicate. Duplicate notebooks that I just started going, and I'll kind of read the different prompts. Some of them I'll read in full because it's actually going to show us, you know, some of these new features that we're really pushing Gemini Notebook on. Well, just number one, see if they work. The first one I think will be pretty relevant to a lot of us. And I want you to also think as I'm going through these demos, think of your use cases, right? And probably before I do this, I'm going to give the quick like 60-second overview of NotebookLM. Sorry if this is, you know, a rerun for you, right? But I think it's important to just set that. So the biggest thing is NotebookLM— or sorry, Gemini Notebook. It just doesn't sound right. All right, there's essentially 3 different panes. There's your source pane on the left-hand side, there's your chat pane in the middle, and then on the right-hand side is your studio pane where you can create different kind of artifacts and outputs. So for your sources, in my example, I just copied and pasted, but there's different ways you can add sources. You can upload files, You know, PDFs, images, docs, audios. You can put in websites and YouTube videos. You can auto-sync now your Google Drive files, which is really helpful. And then, like I said, copy, copy and paste text, or there's kind of like a quick, fast research and deep research at the top. So number one, you're going to get your sources. And this is where the big difference in NotebookLM comes because when you're chatting in the middle, everything that you ask is going to be grounded in that information. And that's the biggest difference. And I think what makes Gemini Notebook really special and much different because, you know, for all other large language models, right, they're going to use a combination of number one, the files you upload just like Gemini Notebook, but number two, training data, right, which could be a good or bad thing, right? Sometimes training data isn't always the best. Sometimes it is. And then last but not least, you know, large language models now by default all connect to the internet. So the big difference with Gemini Notebook is it's going to ground those chat responses in the middle in just your sources, which cuts down on hallucinations by an order of magnitude through the roof. All right, so that's it. And then you can chat in the middle with your sources. They're gonna output there. But then on the right-hand side, you can create a bunch of different multimedia assets in the studio, audio overview, slide decks, video overview, mind maps, reports, flashcards, quizzes, infographics, and data tables. And we've actually done dedicated shows on some of these before, on the videos, I think the infographics, right? But essentially, you have some of the best of Google's AI products baked into the studio right here. You know, as an example, the visuals are created by Nano Banana. You know, the audio overview is created by Google's audio model. Some of the video, you know, you can also customize these things so that there's like a cinematic video as an example, and that uses elements of Google's video model. So essentially, you know, without even really knowing it, you can just click one button and use a lot of different of Google's, you know, different models. All right. So let me just quickly read some of these prompts here just so we can understand exactly what's going on working with this, this long, chunky data on recursive self-improvement. And one thing I will let you know, it's already been 3 minutes, right? And my first prompt is still working, right? And it's not even nearly done. And that's important because, you know, one of the big steps here is this is now powered by Gemini 3.5 and the anti-gravity harness, right? So what that means is before, you know, NotebookLM, And I can say that because before it was NotebookLM, it wasn't truly agentic in nature, at least not in the way that it is now. Right? Some of those new updates that I talked about was it has a secure cloud computer. So what that means is that each notebook can write and execute code for research calculations and data analysis. And in this first prompt, that's exactly what I'm having it do. Aside from having it create custom outputs, it's also having to under the hood. Do a lot of that. So here's the first prompt. What I said, I said, using this notebook and its sources, create a personalized AI budget decision package for a non-technical CFO and CIO at a 5,000-person enterprise whose agent usage is growing rapidly. Do not ask me questions. Make reasonable assumptions, state them clearly, and complete the entire task in one run. I'm saying first analyze 3 strategies. Number 1, use frontier models for every workflow. Number 2, use the cheapest available model for every workflow. Number 3, use a tiered architecture in which A frontier model plans, a cheaper model executes, and an independent systems evaluates. Then I'm saying compare them using total cost, successful task rate, retries, human retry, human review time, latency, vendor risk, and adaptability to future price reductions. Use the Luna and Terra pricing evidence from the notebook, run the necessary calculations, and select one strategy. Explain why the alternatives lose. Identify the strongest counterargument and state what future evidence Oh, it just put me to the bottom and state what future— where did I lose? There we go. What future evidence would change the recommendation? Then generate these 5 outputs concurrently. Number 1, a customized audio overview presented as a CFO-CIO discussion that reaches a clear decision. Number 2, a mind map connecting price changes, agent usage, model routing evaluation, human review, and total costs. Number 3, an editable Excel workbook with assumptions, formulas, 3 scenarios, sensitivity analysis, and a recommendation dashboard. Number 4, a Nano Banana powered 16:9 infographic titled Cheaper AI, Bigger AI Budget? That explains the selected strategy visually. And number 5, a 2-page PDF executive decision brief containing the recommendation, supporting evidence, rejected alternatives, risks, And next 3 actions. All right, so you can see, yeah, kind of a beefy prompt there, but you can probably already start to imagine how something like this new Gemini Notebook would be valuable for your use case, right? One of the important things to keep in mind about Gemini Notebook is the Output Studio now is you can just use it with prompts, right? So now you can see how a single prompt can do a lot of heavy lifting versus having to go into, right, all of these Studio assets individually and trying to build them, which is great because what you could actually do, right, starting to, you know, bridge in other terminology and strategies from other, you know, LLMs as well, you can create skills, right? I actually have a Codex skill that does something like this for me, right? It will go and, you know, according to my, you know, needs and personalization and customization, it will just literally control my browser, it will go in and do all of these kind of things for me. And you can see how it's going to have a much higher success rate and just be way more token efficient for Codex, you know, using browser use or computer use to go in and do this when I can just throw that big prompt in the chat knowing now that Gemini Notebook runs agentically, it has its own cloud computer, whereas before to get this kind of personalized output, multiple you know, artifacts, it would have taken a lot of work, right? If I was to just hand this over to an agent. So now that you really just have a much stronger harness, right? This is Gemini 3.5 using the anti-gravity engine underneath, right? Now the capabilities are just much, much higher. All right. So technically it's still working. It's still creating one of the audio overviews. But the actual chat completion now is done. So I'm not going to go through this too much in length, but I do want to describe, especially for our podcast audience, kind of what went on here. So I should have timed this, but it looks like it took about 5 or 6 minutes. And, you know, one thing I actually like about Gemini Notebook that's better than other aspects of Gemini, it actually has a decent chain of thought, which is nice, right? Because that's one thing that I think Google and Gemini is really lacking in compared to OpenAI. And Anthropic is being able to see what's actually going on under the hood. So little cheat code for you there. Gemini Notebook is actually probably better than the default Gemini in terms of knowing and learning what's going on under the hood. So, you know, I'm not gonna read all of the chain of thought because it's a lot, but you can go try this yourself. You know, click thoughts and see what's going on there. But it's developing budgeting solutions. It's evaluating the AI strategies, analyzing the model costs. You know, it's retrieving some instructions, analyzing strategies, analyzing pricing. Right? So I can go through and see exactly what it's doing. And I'm looking at some of these things. I can see that it's creating, you know, creating different files. So it's running some code. So, you know, it's executing code in there in its dedicated sandbox. So this was not possible a week ago. So the output here, I'm not going to read it because it's fairly long, but I'll read the beginning. So it says executive AI budget decision package. This package is designed for the CFO and CIO of our 5,000-person enterprise to address rapidly growing AI agent usage. Below, we analyze 3 potential model procurement strategies and deliver 5 unified synchronized outputs to drive your strategic budgeting decision. Alright. So we'll see here. I'm not gonna look at this, but it is exhaustive. Right? And the good thing is all of this is sourced as well. So, you know, it's saying, you know, workflow tokens, you know, for planning, you know, using 20,000 input, 2,000 output. I can hover over that and see exactly where it's pulling that from in my notes. So, you know, when I'm looking over the different strategies here, it did a pretty good job putting together 3 different strategies. One Strategy 1 was using frontier models only. Strategy 2 was using cheap models only. And then strategy 3 was a tiered architecture using a combination of both. And then it went through and, you know, went through and did these kind of combined monthly system costs, right? And, you know, what's kind of cool in this scenario is the tiered architecture, right? Which is kind of using the stronger model as an orchestrator, at least according to its calculations, that was actually going to better, right, than just doing the cheap only. And the cheap only was actually going to be more expensive, presumably because it was going to go through a lot more tokens and more monthly human review rework costs. Right. So such a cool tool to put together. All right. But then let's look over at our right-hand side in our panel and see what was actually created. So we did ask specifically for 5 different deliverables. We asked for a customized 5-minute audio overview, which is almost done. Number 2, we asked for a mind map connecting pricing changes. We asked for an editable Excel workbook. We asked for a Nano Banana powered 16:9 infographic and a 2-page PDF. So, so many of these new— so many of these things again were not available last week. You know, you really had to stay in this kind of confined output of what NotebookLM offered. But now with Gemini Notebook, right, it's really flexible. And that is actually really helpful both from what you can do in the middle pane because you don't have to be as rigid, right? But in my use case, right, if you really want to have agents working for you, I think this makes it so much easier for a Claude desktop or a Codex desktop, an agent living on your desktop that can control your browser. This is going to allow it to iterate and create much more valuable outputs I always wanna encourage listeners of our show, even if you're a beginner, right? We have to get out of the prompting phase, right? We have to be able to give these desktop agents as much context and as much information about what we need, then let them go do the work. And I'll usually have another, you know, AI agent audit their work and make sure it, you know, corrects any mistakes that it sees. But in this scenario, right, if I were to do this manually, even inside NotebookLM, it would've taken me 3 to 5 times longer, but that's just because of the new harness. So let's just quickly take a look at all of our other sources. So here we have our AI budget model. All right, so downloaded an Excel file. I'm just going to screenshot this here. Let's do that versus having to unshare and reshare my screen. Let's see, it's uploading, it's uploading. Oh gosh, let's see. All right, that one, let me, uh, let me try that again. Uh, the actual Excel sheet looks, looks great. Um, we have our budget model. Uh, let's, let's see if we can get that again here. Uh, let's copy this, shall we? Um, all right, here we go. There we go. We got it here. Uh, so we got a working, uh, financial model, right? I just quickly checked the spreadsheet. Gave me an XLS, open it in Excel, everything's working, right? It gave me exactly what I wanted. It gave me my different 3 strategies. It mapped out the cost. I can go in and change the formulas. There's little graphics, a little dashboard. It created exactly what I wanted. It works. Is it the most beautiful spreadsheet ever? No. Is it something better than I could have done? Absolutely right. I love working in spreadsheets, but I'm not necessarily the best at creating formulas, right? So again, Just think of all of the unlocks that something like this brings. All right. So that was our AI budget model. Did a good job. Then we asked for a 2-page PDF that gives an executive decision brief. All right. So now here we go. Livestream audience can see, but we have a 2-page executive brief. This is the AI portfolio optimization, the implement— implementing a tiered model architecture for the 5,000-person enterprise company. There we go. This is, you know, just kind of a PDF version of what we created inside the middle pane chat. As well as why, you know, what the best strategy was and the counterarguments for those. So perfect. It created that. Let's look at the other ones. We have our mind map. Okay, so here we have our RSI here. Then we have the economic shifts in pricing, the tiered agentic architecture, governance and control, and monthly system cost scenarios. And then I can click those out and explore a little bit more. in each of those. There we go. All right, so, uh, mind map, great job. All right, and then let's look at our— this should be our infographic created by Nano Banana. All right, so looks good. You know, I'm kind of, uh, you know, quickly editing it and it looks good. The only thing that I see wrong is there's a Notebook LM watermark. All right, Google, we got to update that, right? If we're trying to get rid of Notebook LM, uh, we gotta update the, the watermark to Gemini Notebook. Uh, but the It, this actually looks pretty good. The 16 by 9, nano by nano graphic for the CFO, CIO decision maker. Great. Looking at the, you know, 3 different strategies, it breaks down the human cost, the model cost, the, the pros and the cons. Great charts and graphics. Did a really good job. And then last but not least, it did just finish. I'm just gonna listen here. All right, perfect. It created a 21-minute deep dive on self-improving AI/token prices. All right, so, okay, interesting. So it even did put in the custom prompt. So that's cool. It went in, did everything correctly, right? So hey, the demo actually worked. All right, I'm just going to quickly look at some of the other use cases just so we can see what we did. In the other notebook, I said create 3 quizzes of ascending difficulty focusing on a different topic from my sources. This is a good one. This was actually a Gemini notebook example prompt, which I thought was a really good idea. If you're trying to learn something, you know, they have this quiz module, which is really good, right? But, you know, sometimes you might really want to be intentional about how you learn things and break it off and start easier so you can go through the paces and relearn things as you go. So this one, It created an easy quiz. All right, let's— should I do one question? Watch, watch me get it wrong. All right. What is the core definition of recursive self-improvement in the context of AI? A, is it a method where humans manually rewrite every line of code to make AI faster? B, a process where an AI uses its own capabilities to design and build a more advanced version of itself? C, an AI system that is only capable of performing one specific task like playing chess? Or D, a way for AI to search the internet and summarize existing news articles better? So I know it's B. All right, so that's right. So it has 3 different quizzes. Did a great job doing that of ascending difficulty. All right, let's look at our next one. So for this one, I said create a 5-level RSI maturity ladder, and I wanted an infographic and a PDF. So did answer everything. This is really good. It answered everything in the chat. My PDF here, it broke down the 5 kind of ladders of RSI. Did a great job. And then let's look at the infographic. Cool. Yeah. Yeah, this infographic actually came out a little cleaner than I thought, considering that we didn't customize anything. So yeah, we have the level one AI assist researchers. Level two AI executes a human design method. Level three AI improves AI infrastructure. Level four AI helps train or improve another AI. And then level five AI controls the complete improvement loop. All right, let's go. Let's look at our next one. I said using the old Luna and Terra pricing in this notebook, create an edible Excel calculator for agentic workflows. I have my Excel sheet there. I'm opening it up in another tab. Yeah, looks good. I'm checking. There's formulas. Everything's working live. Perfect. All right. And then one more thing before we wrap up here that I wanted to showcase. All right. The middle is agentic now. All right. So even though it's still going to be grounded by default, I'm going to go into one of my things here and I'm going to type who won the 20 2016 World Series? All right. Now I'm going to ask that question. So by default, it should come back to me and say, hey, I don't know. However, I can go out and find that information. All right. So why am I bringing this up? Because that is the one important grounded change that you need to be aware of. All right. So what's interesting here, and I didn't know this, it looks like because it had already started pulling things agentically on the web, because in some of my other testing when I did this for the first time, You know, so I just opened up another chat to show you this. I said, who won the 2016 World Series? In this example, it says your sources do not contain information about this. Do you want me to research this on the web? I said yes. So it looks like maybe if web research had already been active, had already been activated, that it might just answer that by default. So that's an important thing to think about because previously NotebookLM in the middle chat was always 100% grounded in your information. So that's a big difference and a big change. So in some instances, it's a big upside, but you have to be aware of that, that, you know, you have to be able to go back and look at the chain of thought in a lot of these prompts. It went and it was doing a lot of web research in addition to the information that I gave it, because in the prompts, I was kind of pushing it past the limitation of the sources. So this does kind of, in my opinion, change Gemini Notebook. And maybe That's good that we are getting away from the notebook LM and inserting Gemini Notebook because there's pros and cons to that, right? Like you saw in this example here, because it looks like because I had already triggered the anti-gravity harness in that first prompt and then asked it information that was not in my sources, it just well it just responded. All right, but it did say it says I ha I haven't imported those sources about the 2016 World Series. Oh wait, I looked at it. Okay, I didn't read it. Didn't. All right, um, it just said that I hadn't looked at those. Uh, do you want me to look them up? I, I got ahead of myself. Uh, so yeah, I can say yes. All right, so all right, scratch what I just said the last 60 seconds that, um, I thought because I had already activated the anti-gravity, uh, harness and it was already doing, uh, tool calls that it was going to go out and respond. I literally just didn't read it. I just saw the response and I'm like, oh my gosh, it responded. But no, me, uh, not sleeping enough. It literally said I haven't imported those sources about the 2016 World Series into your notebook yet. Would you like me to import them first? And then I said yes. And now right now I can click and see that it's going out to grab those sources. All right. I stand corrected. So it's still important to know though that, you know, once you do give Gemini Notebook access to, it is going to bring in information that are outside of the sources. So it does change, I think, again, for the good and the bad, but at least it does caution you or make you kind of manually approve that saying, hey, You don't have this information in your sources. Do you want me to go grab that information for you? All right. So that is a wrap. But to quickly recap, what do you need to know? All right. So, uh, number one, what's the big difference here? Gemini Notebook represents a shift away from just having that, you know, NotebookLM, all those sources. And, you know, essentially it was a non-reasoning model that could use Google's different multimedia AI. And now it's different, right? And I think that's the big reason for the name change is because now with Gemini Notebook, it is agentic by default, right? Having that cloud computer, being able to write and execute code, being able to, when you give it the okay, search the web, it really changes what NotebookLM was into what Gemini Notebook will be in the future. You know, the important grounded change that I talked about right there, you just have to know. And I think ultimately that will be a good thing as long as you understand what's going on. And then the real reason I I think notebook Gemini notebook going agentic matters more than ever. Well, you saw just through my example, it makes it much easier to work with, right? Before, you know, you were kind of restrained, right? You almost had to think in your mind in terms of outputs, right? You had to prompt Notebook LM according to the outputs. Now those outputs are so flexible, right? Before, you couldn't create PDFs. Before, you couldn't create CSVs, right? You couldn't create Markdown files. So you were much more limited in PowerPoints, right? You can do all these things now. So Gemini Notebook has become much more of an agentic coworker versus what it was before. There's essentially a handful of artifacts or a handful of formats, and you really had to use NotebookLM according to those outputs. Now it is extremely flexible and even more powerful than ever. And I guess that's why NotebookLM is no more. But Gemini Notebook at least is here to stay, and it is really good. All right, that's a wrap for today's show. Now you know Gemini Notebook, the 7 new updates, and what they unlock. If this was helpful, do me a favor, tell someone about it. Share this on LinkedIn, you know, go and subscribe to the podcast on Apple or Spotify. Then make sure to go to youreverydayai.com, sign up for the free daily newsletter. We're going to be recapping the highlights from today's show as well as a lot more. Thanks for tuning in. Hope to see you back tomorrow and every day for more Everyday AI. Thanks, y'all. Bye.