I Wish I Had This Tool When I Started Learning to Code....

Time is money. You know that. I know that. But when it comes to building something, learning something, a lot of us forget about research. And Notebook LM has quickly become my favorite tool when it comes to researching new topics. We're going to cover a lot of different use cases in this video. We're going to figure out how to reverse engineer an API. Get real nerdy there. We're going to go through different ways that you can basically download stuff to your brain faster. And this just so happens to be the core features. So, we're using Notebook LM to explain what Notebook LM does. This is the new mind mapping feature from Notebook LM. You can see the core features are yes, it's an AI research assistant. It has source integration, content generation, audio overviews. Really cool. Discovery of sources. been playing around with this one. It is great. So, if I click into one of these, you can quickly see all the details about this. And if I were to click on one of the nodes that does not have a little arrow next to it, then it'll actually fire off a query and it would minimize this, move this to the right, and then the chat would be in the center. So, this just so happens to be the first thing we're talking about because I figured it'd be a great way to explain it. So, you have the assistant that sits here in the center. You can ask it questions. It is a bit more ephemeral than most people are used to, meaning that it'll go away. So, you save things, you do one-off things and you save them. The flow of it goes, you add sources. You've probably used this before, but we're going to get into proumer use cases in a moment, but yeah, you add a source, you get the query thing in the center, the assistant, and then the right hand side, you have content generation, so you can have a deep dive conversation. You've probably seen this before, but have you customized it? That's awesome. You have the ability to go and create study guides, add notes to these, have a briefing dock, a timeline, FAQ, and then, of course, this new mind map thing. So, very cool stuff there. Now, how did I get all these sources? I didn't look up any of them. So, what I did when I went to create a new one, I have the pro version of this, is I hit create new. And what is new is this discover sources. So discover sources does what it sounds like or I can click into here and I can enter in a topic of what I'm trying to learn. So I'm going to take this route and hit I'm feeling curious and just see what it comes up with. It's like I'm feeling lucky. Some PM was like, "Yeah, we're bringing this back. We got this." So wisdom of the crowds, hard thinking, don't care at all. Next. So let's use a little bit more of an advanced use case. So now I want to not watch a two-hour podcast of George Hots. I just want to know what he talked about. So in the case where I added something, I just threw in the YouTube link. Boom. You see this stuff? But my responses may look different than yours if I were to add a prompt. So this is where you can define your conversational style. And what I've done is I wrote a little prompt that I'll just grab here. And per the founder of Google's advice, I made it threatening where he says you just get mad at it and it makes it better. So you can enter in the different custom prompt that you want there. And in this case, I just wanted bullets. And then I can select default, longer, shorter. I find if I combine my prompt of the bullet points with shorter, it's almost tur like I don't even I'm like now I'm missing out on some stuff. If I hit refresh, then it'll clear the chat history and I can then run a query and it'll respond with those custom instructions. You probably all know this. If you didn't, you learned something. That's great. I can customize audio as I said before. Now, let's get into an actual use case. So, I want to learn fast API. I'm a total noob. I give myself like a two out of 10 on the Python Fast API track. So instead of watching this 4 hour and 44 minute video where I'm sure the content's great. I want to just test the waters with it. I want to passively start learning stuff because I don't have the time. I don't have the four and a half four hour 44 minutes to sit here and watch this right now. Even if it's a 1.2, you know, it's going to take a while. So instead, I'll take that. I'll dump it in and I'll create a conversational podcast. Now, I customize this one to say that I want to have it be longer and more technical, but now I have a half hour thing and I'll play this while I'm doing some sort of trivial administrative work where I can I'm not fully getting it, but I'm learning more and then I'll get a little bit more amped to go and check it out. Then, of course, I can click on the mind map and see what's going on with this. What is the purpose of this? Do I ingest and store the data? How does that work? If I clicked on it, what are the key technologies discovered or discussed in here? Okay, fast API covers the web web framework, some of the details on it. Okay, cool. Easy URL routing, but just gives you that quick little again, if I watch a 4 hour and a half hour video, I'd be literally diving into the deep end of the pool. This is the way to do it. You have bite-sized chunks. You're using this to figure it out. I could say I don't really understand how time scale DBs work. So I can click on that and it'll fire off discuss what these sources say about time scale DB in the larger context of key technologies. So it's referenced the thing that it already knows and now it's given me this awesome thing and it has all of the references from where it was transcripted. transcribed. Basically, it has the reference from the video. And then if I pop this on. Absolutely. And making sense of it. That means needing a solid way to capture it, store it, and analyze it. Cool. Now, for the grand finale, actually, no. One last one. I saw some video about some crazy news and I was like, I want to see what George Hos. No, I did cover this one. Never mind. So, let's go to the deep end here. So, I wanted to learn how I could make this as a pipeline. Now, I'm going to not do this because it violates the terms and services of this product, but I'm curious because I found someone on Reddit that made this and I'm like, how did he do it? What did he do to make an API for Notebook LM? In my head, I'm like, okay, he's probably doing like some packet sniffing or something or a headless browser. I don't know. Never built this. Done like MITM proxy stuff and certified a signature for my iPhone to see whoop requests, but never done this. So, at first I tried to go in here and type in that box for discovery where I'm like, I want to reverse engineer this thing. D, but it wasn't that tight. Like, it didn't get me what I wanted. So, then I went in and I did an 03 request. So, I said, I want to create an API for a service that I have an account for, notebook lm, that doesn't have a public API. How can I do this? Below is a pragmatic road map for rolling your own API when the provider notebook LM hasn't published one. I split it into the key decision areas you'll face plus concrete next steps and code stubs you can borrow. Now my response will look different than yours because I think I have a setting in here. I do. So what do you do software engineer? And then what trait should you have? This is a really good prompt. So all of my responses come back this way. Structure your writing. Think and respond like a consultant. I know it's a tough word, but it keeps it concise. Segment your answer into categories when possible. Adopt a conversational tone. Use simple language. Incorporate real life examples. Blah blah blah blah blah. So, never use emojis, too. So, if you want that, comment below prompt for GPT and I'll reply. Then, check the guardrails first. Yes, it makes the mention of this. Yes, there might be some sort of enterprise loophole. Not doing that. Now it asks me the different approaches that I want to take. So I can do an XHR and GraphQL call. I can do what I was thinking which is like playright or puppeteer. It gives me a nice table. Gives me the pros and the cons of each. And then in practice most hackers start with A to map hidden endpoints then graduate to B for production. Cool. Didn't know that. And then it goes into depth on it of the actual things. Right. So to get back into notebook LM because now I want to learn more about this and I want to not just go and again like dive into the deep into the pool. I know enough about it where it's okay what would the topic you explained be or how would you explain this to your friend basically that's technical give me the keywords and it's unofficial API wrapping aka reverse engineering a private web API with headless browser automation. Boom. So that would be it. And then that's what I look for when I'm going to go generate a new set of resources. So then I go back in, I just paste in that thing. Most of the time you'll actually know what it is that you're going to try to research. In my case, I just found it and I was like, whoa, this could be cool because I want to just text a phone number and have it kick this thing off and then deliver me a podcast. I don't want to go do all this stuff, but whatever. Maybe they'll open it up. But then yeah, I get all these nice resources and I can click out to each one and check on it to see is this good, is this bad, right? So I would actually after this video in practice go and watch these and see and read a little bit, see what's going on. Also think this would be great for APIs. I tried with the X live search that just got released today. So maybe something's up with it. But if I went to what's an actual API that I need to do, it's not Twitter. If it's take a look at the API docs for image understanding. So maybe I want to learn how this thing works, but I'm curious if I can just pass it the docs because this could be interesting. Okay, now I have this. I'm not sure how this is going to work. This was not part of the plan. But let's first of all, I want to import that because I do want to learn this. So, let's spin up a new one. And I don't like that it always goes back to cards. Let's go into link. And I want to paste this. Can also do text, but let's see if it does anything beyond No, it's just that one. So I guess in this case you do a fuzz or something to print out all the URLs and then you can go and get them. But that's how I am using Notebook LM and I'm excited about it. I think it makes a lot of sense. I would be curious if you could do a codebase in the future just out of curiosity but I'm not sure. I think this is going to be a very good power tool for everyone that forgets the research part that is equally as important. We have all these people that are so focused and myself included on finding the best codegen tactics. But I think if you start to remember that for every hour, let's say, of development, maybe shave off a quarter of that for the research on the thing. Again, case by case classic, it depends answer. If you need to do research, go do research with Notebook LM. If you've learned anything in this video, make sure that you subscribe to the channel because I'm totally switching up the content. It's going to be fire. And like the video, if you want to be a part of our community of people that are basically just pushing one another to be better every day using AI. We have a lot of business owners that are learning to use AI in there. We have a lot of people that work at tech companies like Facebook. Sorry, not Facebook, but we got the equivalent ones. I don't want to name them. We got a lot of smart people in there that work in big tech. We got a lot of business owners. And then there are like the vibe coders as well that are just learning but they're serious about it. So you can check that out the video description. And that's all for today. I'll see you in the next one.