Transcription
Imagine a situation: you have a complex, huge project, technical documentation, presentations, slides, notes. In my case, this is Google Antigravity. Information about it is very fresh, and even the Gemini model itself doesn't know about this tool. And when working with the chat directly, I get almost only hallucinations.
Previously, to understand this and create good content, I had to spend hours sifting through a bunch of videos, create hundreds of bookmarks, and manually combine all the facts into one whole. But today, I will show you how to hack this system. I will show you how to create a bot that has read my entire archive in NotebookLM about Antigravity and now knows more about it than Google itself.
But I will not just upload files to the chat. I will connect NotebookLM's memory with Gemini's intelligence. Gems are bots within Gemini, if someone didn't know. Now I have an expert who sees all my notes in NotebookLM, remembers all my schemes, and helps me learn and write code in literally seconds.
Today, using my own example, I will show how to turn a file storage into your personal assistant and what the key difference is between connecting NotebookLM to a Gem bot and connecting it directly to the chat. Let's get down to understanding.
So, in my NotebookLM, I created this notebook called "Antigravity." I created it a few weeks ago, and it contains a lot of information about this Google tool, Antigravity. That is, in one of my videos, I showed how I did this. Why did I do this, you might ask? Well, the fact is that the NotebookLM platform is actually created for this. Yes, it is created for learning, for studying something new and some new materials. And in this case, Antigravity perfectly fit this situation.
So, that's how the stars aligned: a new tool, and I wanted to study it, and Gemini couldn't help me directly with this. Therefore, I came to the conclusion to gather all, all, all the necessary information. And as you understand, it is most convenient to do this in NotebookLM because you can connect up to three hundred sources there, absolutely different ones, and it will always store and keep them in one place, and then, if necessary, refer to this notebook.
So, returning to our notebook, here on the left side are my sources, as you can see. And everyone probably knows this. And here on the right is all the information that I extracted from there. During my long dialogue with the neural network, almost all these files that I have in Studio, I have practically translated all of them into sources. That is, well, from what I heard, when I access through the chat, it sees the sources, it doesn't see what's in Studio. And interestingly, it sees these sources, and if it wants to cite a source, it first gives numbers, and then if you enter these numbers, it will tell you what it's called. Well, very strange. But okay.
That is, if it took information from this source, it will say, "I took information from source number three." And if you ask it what number three is, it will tell me, "Here, here is such a video." That is, the ordinal number is third. That is, this is our data that is located here, it is constant, it does not change, well, unless some link breaks. This data is quite fresh. I can always add videos here through this extension that I have installed. It's very convenient. I click on any page, just this notebook. For example, being on YouTube, I can go to any video, here, it starts talking, I click like this and choose the notebook I need, "Antigravity," and add it to this notebook immediately, directly from the video page. And so I move through the video. If I like it, I see that there are good comments, it's a quality video, I add it to this same notebook.
So, let's move to Gemini and create our bot. Gem bots. Here, as you can see, I already have about ten of them, but these are mostly for my YouTube channel. That is, I search for news here, I create posts for Telegram, and I read videos. Yes, I send him the transcript of an English video here, he translates it for me, cleans up the code, and shows it. That is, I can simply read what is being said in the video. Well, and if we talk about bots in general, then this is essentially a chat, but with a written instruction, like a shell for what you will ask in the chat. But the general shell is written once, so-called instruction. Now I will show you everything.
So, let's click "Create Gem Bot." Name, let's write the name, let it be "Antigravity," short and clear. And in the description, let's write like this: "Expert on technical documentation Antigravity. Analyzes data based on a knowledge base." And in the instructions, I will insert this large instruction. Don't worry, I will provide it in my Telegram channel. By the way, about this instruction for this bot, it is universal. That is, you can replace it with, I don't know, "diploma work" or "how to plant trees," and it will all work. That is, go to Telegram, subscribe, support, and download this universal instruction for your Gem bot.
That is, essentially, as I said, here you need to give it general information that you don't want to repeat. That is, of course, when you later ask this Gem bot, you will ask for different information, but what will be repeated, that is, some data about you, in what format to give answers, you write it down once, and that's it. And this information is saved as an instruction in the Gem bot. Well, and the most important thing, of course, is the knowledge base. Here we need to add our notebook. Click the plus sign, "Add to NotebookLM," and select "Antigravity." Add. Click "Save." Well, and we can start chatting right away.
So, once again, in this case, the advantages are: the memory capacity, as I said. Plus, the fact that these files don't disappear anywhere and don't get blurred. Then, so-called grounding, as Gemini likes to use this word, that is, these files are constant. And when Gemini refers to them, it will constantly answer the same way, and not like it likes to make something up every now and then. Another advantage here is live synchronization, because if I just throw information into the chat, and then some new information appears, I will have to add it again. And possibly, it will be a different chat. But here it is collected in one place, and my Gem bot is always connected to this one notebook, and as soon as I add something there, it immediately sees it.
And the last, but not least, advantage, in my opinion, is simply the acceleration of work. I don't need to explain to the chat every time who it is, that it is a specialist in the Antigravity tool, and how I need the information to be presented. I wrote it down once in the bot, and that's it.
Well, and let's ask something right away. That is, the first thing that comes to mind, for example, I want to shoot the next video about this tool. By the way, write if you want to hear something like that from me. And I will write him this prompt. That is, as you understand, I also generate these prompts with the help of Gemini. Yes, why write simply and sparsely in two sentences, dictate, if you can ask, and he will make a large, beautiful prompt. This is called metaprompting. That is, I immediately ask him to analyze the loaded presentations, look at these capabilities, and then content requirements: summary, wow effects, visualization, response format, that is, everything, everything, everything that interests me. That is, as you understand, if I just send this same prompt to the chat, then, firstly, it will immediately ask me again, most likely I made a mistake by calling it Google Antigravity, there is no such tool yet. And even if it goes online and looks at what it is, it will not, it will not collect such content. Just on the first advertising page of this tool. That is, the first thing it will do is go and collect information simply from the official website of this tool. That's it.
So, let's click "Send." And here it answers me, yes. That is, it looked, analyzed the array of data, and compiled approximately this video plan. That is, it shows the time and starts here to talk. Here, about the foundation, that is, more deeply, of all this. In essence, you can only find this in the documentation for this tool. And it extracted all this from these sources, from these videos, from these articles that I saved there. And very detailed, cool, look, with file names, principles, in general, the quality of the answer is much cooler. That is, I asked it to answer specifically, and here it is specific, as it were, and without water, without any advertising brochures, but precisely based on the materials that are in this notebook.
Let's ask him in this chat if he sees infographics and presentations in my notebook. He sees even materials in Studio. That is, here is a slide. He noticed this one for some reason, but honestly, I don't like it. And I would like this one. I have ten examples of prompts here, different ones, like project ideation. That is, what kind of prompt can I give, and what will appear in the project? Yes, creating files in the project, and how to create them using which prompts. Let's nudge him towards this thought and, for example, specifically ask him: "Ten prompts." Let's return to our chat window. So, let's ask him.
And look how cool, yes. So, I have a presentation here in Antigrad. And here, look, based on this presentation. That is, firstly, we can look here, and if we just remember that it is here, we can also refer to a specific slide. Yes, so I just wanted to show that it sees visual data. Yes, in other words, if we have a presentation and we cannot copy the text from it, then we can do this cleverly. That is, it will go here and translate this visual context into text, where we can then copy this prompt.
And here I would also like to note that, as you know, those who have worked with NotebookLM, honestly, I don't like the appearance itself. This whole text that I see, plus these little links. Can't they have some kind of switch here so that I don't see these links? For example, I don't want to see them. I want to look. Okay, I'll switch and see where it got it from. Or, for example, like with OpenAI, in ChatGPT, I could highlight it like this and ask where it came from. That would be super. So, these little numbers, these incomprehensible dividers. In general, the interface itself, honestly, I don't really like it at all.
But we know that in this particular chat, in this Gem bot, our notebook is connected, from where it gets information. And this information will be constant and will never change. This is the whole advantage, the whole highlight of this NotebookLM connection to Gemini. I still lack a table here. Let's see how it will see a table. "Find the source about ten key prompts and make a table of three columns from it. The first column is the serial number, the second is the prompt itself, and the third is what will happen in the program window." Let's click "Generate."
So, the table is ready. Let's look at it. It looks like this. By the way, I have a video about tables, you can go and watch it. They are not quite, honestly. But okay, it doesn't matter. The point is not that. I'm just interested if it will rewrite it or try to create its own table. "Ten key prompt types." Table. Okay. Let's go back here and, for example, we know that we have this table. And let's ask him to show us this table right here in the chat. He answered me. But look, he, again, used the wrong file. Yes, probably I explained it not detailed enough. That is, I copied the name and said, "Find me such a table and show it to me."
Well, look, it referred to this file again. That is, I suspect that it does not see the table. As you can see, it says, "Yes, I see a table in your materials." And this is the key phrase. This is what I said at the beginning of the video. The fact is that this table should be here, in the materials. This can be done with ordinary text notes, and you can export it into tables, into documents. But the fact is that presentations and tables do not have such a menu item. I cannot use this table as a source, although it's very strange. Look, I can export it into tables, it will be saved in tables, and then I can add a source and add this source from Google Sheets. Why can't this be done immediately? I don't know.
Friends, let's summarize. Why did we go through all this, connect notebooks, set up Gem bots, write instructions? If you think globally, we live in an era of information noise. There is already so much information around us that it all gets lost in bookmarks. By the way, I have a video about that somewhere too. On hard drives, on old computers. In general, we spend much more time searching for information than creating it. My goal on this channel is to show you that artificial intelligence is not a scary future that will take away your job, but a tool that allows us engineers, creators, content makers to get out of what I call the "box." Use these technologies to free up time. Time for what's important. What I repeat in all my videos: for family, for yourself, for creativity. Everything I used in this video, I will post on my Telegram channel. So, come, take, copy, create, try. And at the same time, subscribe, support me there, and definitely subscribe to this channel. A lot of views without subscriptions. If this video was useful for you and motivated you to create such an intermediary between you and NotebookLM, such a Gem bot, then subscribe to the channel, give it a like. This will be the best gratitude for me. And if you want to thank me more than with a like, then there are all the tools for this under the video. This will help me develop and move forward immensely. My name is Ruslan. Build your systems and use AI wisely. Value yourself, value your experience. And goodbye.