Transcription
Today's video, we're going to go over everything you need to know when it comes to building out an AI startup, or specifically, an artificial intelligence integrated software. This video will be valuable to people that want to build out an AI software from the ground up. Alternatively, if you don't want to code it and you want to actually hire a team to build out the software, I'm going to show you the infrastructure required and that is optimized for the context of AI software. Therefore, whether you're coding it or purchasing a team or hiring a team, you will have insight on what, what to use, how to use, and why to use. Let's jump in.
Welcome back, y'all. In this video, I am doing it because I've received a ton of comments in the context of building out your own AI software, the tools required. I even received comments on how do I hire a development team. I've done videos on this topic, but I'm creating this video as a way to basically organize all of that into one nice little video you can watch. There'll be a lot of annotation cards throughout this video that I'll reference if you want more details or information on the topics I will speak about today. Furthermore, if you're new to my channel and you're like, "who is this guy? How is he qualified to even talk about this topic?" Currently, I'm developing my own AI software called TubeStamp. I've been at it for the last 9 months. I know all the implications and everything I've learned through this development journey, you can learn with me. This is kind of where that playlist gets into at the end of this video. You can check out TubeStamp yourself, but in today's video, let's go ahead and see everything we use to build out a software so you can start building out your software or hiring a team to build out your software.
The way I structure today's video is going to be by blocks. So we got a couple of things to talk about when creating a software. It is not a simple one application to start developing. No, this is not no-code tools. We're not going to be talking about Bubble and everything above the board in that regards. No, we're talking about genuinely developing a software line by line with code. If you want to take this journey, Corbin, why would I want to take this journey? What's the advantageous benefits of doing it this way? Check out that video right there, which talks about the implications of using no-code comparative to actually coding it out. So, hiring a team, I give you options for both ways. The first major thing we're going to talk about in today's video is going to be building tools. How do we actually create said software? Then we're going to talk about front end, which you probably already know, but that's basically what you're watching right now on YouTube. It's like what the consumer or the user sees. Then we're going to talk about the backend. Then we're going to talk about the different AI providers we could potentially use within our software. And then finally, resources you can use and leverage within the creation of your software. Let's go ahead and start with building tools.
Side note, we'll give this stick figure his name Jeff. Jeff wants to learn how to build our AI software. Where do we begin, Jeff? Okay, here we go, Jeff. First thing we need to understand is how the heck do we even build it? Like what's the thing we're putting the code into? And we're going to go ah, and move this. We're going to use two major tools here. We're going to use VS Code and we're going to use GitHub. The description of this video is going to be very link heavy. I'm going to go ahead and leave links to basically everything I referenced in today's video so you can go and just scour through the description to kind of click off and find the ones that you are interested in. On top of that, I will give the pricing associated with each one of these softwares, so, tools to start off here, VS Code and GitHub, completely free. Therefore, to get started, go ahead and just download VS Code. With VS Code, you're going to be able to not only add extensions which are useful, but overall, this is where you're going to be building out your software locally on your machine, which we'll get into a little bit later. The second major one that we're actually need to actually build the software is going to be GitHub. If you're familiar with development or you've coded, you have 100% heard of GitHub. So this is probably a no-brainer for a lot of y'all. If you don't have an account yet, it's free. Proceed, go ahead and sign up. This is where we're going to be able to push our local branch to our development, so, staging, so, production branch that we use within our software. Corbin, what are branches? What are you saying here? Branches are our different versions of the software. So our local branch is when we run emulators on our local machine. This is where we can basically have it where we're safe from basically incurring a lot of costs when it comes to function calling, invocation calling. Think of it like a bubble where if you really screw up in the code and something goes crazy, you can just turn off the emulator, no issue there. You're staging or QA is where we're going to be producing a check before we push to our production, which is going to be the live facing website, web app, whatever you may say on your software. Knowing this, these are going to work hand in hand. You're going to want to use these as your building tools. Let's jump over to front end.
Coming over to front end, oh, the amazing front end, the thing that makes the UI look good or look like a 2004 HTML site that you're just like, "how has this not been updated yet?" Trust me, I know some sites out there. Here is what we're going to use. This is going to be up to debate. I already know people choose sides. Choose your side. Do what you want. Do whatever you're most comfortable with. I'm just going to be proceeding in a manner that is most comfortable with. I programmed in my past. Therefore, our front end here is going to be React. We're going to use calls within React using a Firebase backend to kind of reciprocate different functions and different actions within our software. Corbin, Corbin, why do you have Firebase in your front end? That's a backend thing. Does this guy even know what he's—chill out, chill out. We use Firebase within a front end when it calls upon different actions such as authentication, database. There is going to be certain invocation calls within the front end that we use with the Firebase. So I kind of left it here because it works seamlessly. I suggest you use React as my personal experience. React is you can get really, really solid-looking UIs within this library. On top of that, you're going to be looking at when you're coding in the context of front end with HTML, CSS, JavaScript, you're going to have different JSX, everything above the board when it comes to the front end. This is where you're going to want to go with when it comes to building out your software. That's what makes Firebase so attractive, as we're able to access a lot of fundamental stuff within software such as authentication. Firestore, this is going to be storing relevant value data that we have with our software. Firebase, I love it. And another reason I like to use it is because the UI is amazing. All right, so I love user interface when it comes to Google products. No, I'm not sponsored by Google. You can sponsor me, Google, if you want to. Let's jump in a call, but yes, Firebase is amazing in this context. On top of that, to start using and leveraging Firebase, it's actually free. What you'll notice is that to get access to the Spark plan, it's absolutely no cost, and then when you upgrade, it's a pay-as-you-go model rather than a fixed cost model, which is amazing in the context of software. If you don't necessarily know costs associated when it comes to software and basically why this is so advantageous, I encourage you to check out this video here where I show you a micro-niche example in the specific context of the software I'm building out that shows you how like crazy it gets to when it comes to scaling and making money. I believe in that example—don't quote me—it was either I could store around 80 million user videos when it comes to just metadata, and it would cost me like a buck a month or something like that. It's something crazy. Check out that video. Let's jump to the front or back end now.
Because we're using Firebase inherently, our back end is going to be with Google Cloud. Oh my gosh, why are we using Google Cloud? Use AWS if you want to. I'm not going to argue with you all in the comments. If you like it, you don't—if you don't, you don't. Choose whatever you're comfortable with, y'all. Side note, what I'm noticing in the AI space is like people get really toxic about like what AI model you use. Chill out. Just use the one you like. If you like Gemini, use it. If you like Perplexity, use it. I like ChatGPT. I'm going to use it. For a backend though, we're going to use Google Cloud Functions and we're going to use Python. As I've been developing this software with a partner of mine as well, I have learned that it is actually very advantageous to use Python. This has to do with a lot of major things. One big thing is that when you are making a software that's going to have an AI integration, you kind of want to use Python. It's not negotiable. So proceed in that manner. Google Cloud Functions amazing as well. You're going to be associated with your Google Cloud Functions project, your Firebase, like it's all going to be connected. You're going to have a really, really clear understanding of cost associated with each action. The name of the game when it comes to building out a back end is to basically use it as least as possible so that you can scale very easily and make a lot of money very easily. Therefore, your front end has to be optimized when it comes to handling certain data points and reference points within your software and within your workflow. Workflow, Corbin, what do you mean? Like what are you saying right now? Basically, your front end should be optimized to handle local storage, session storage, so you don't have to reference the back end as much. You want to limit the amount of times you call your back end as at scale. If you have a very optimized and functional backend, you're going to be able to make a lot more return on your underlying value you provide. Also, my voice is like really raspy right now, so keep that in mind. That's why I keep voice cracking a little bit. Therefore, on that topic, when it comes to creating an optimized version of your backend when providing value to a user, you can provide value with a very unoptimized way of your backend and a very optimized way of your backend. Typically, in the context of when you are proceeding with the value point you're providing, you're going to want to use functions that are invocation-based, so, pub/subs. You don't want to use, um, functions that are based on data changing within your logic or alternatively, or on a pub/sub time counter where basically every 15 minutes, no matter what, the function, uh, launches off. Right? There's ways to go about this and ways to structure your backend and database to ensure at scale you're not getting like thousands of invocations or thousands of database switches. Like keep it clean.
Next, we're going to look at AI providers. I did a whole video on AI providers when it comes to API right there, so you can check out a further in-depth of what I'm about to say. But to give you a brief overview, here's my opinion. First things first, last time I checked, Gemini, it didn't have access to a production-level API. That might be different by now, so let me know in the comments. Here is my personal opinion. This is an opinion. Proceed with the model that with whatever data point or value you're providing your consumer is most effective at handling the value. What do you mean by that? Supposedly Anthropic came out with a new model that is better than GPT-4 at an API level, supposedly. If you feel like it's better than GPT-4 in the context of how you're making your software, read with Anthropic. What I would encourage you to do though is use discretion. Go ahead and use 3.5, GPT-4 Cloud, whatever it may be. Check which one gives you the best value. But on top of that, if you truly are going to create a software that is going to basically give access to thousands of people value, you're going to want to choose a model that's very cost-effective. And high-level models like GPT-4 Cloud, 3, very, very cost heavy. So what I would encourage you to do is use a model it's very cheap. Take that how you will. Check out that video.
Finally, resources. Woohoo! We have our resources here and whoa, that's a cool effect that I don't think—I think that's a glitch. Basically, we're using ChatGPT and we use Python Package Index. The Python Package Index allows us to search community Python packages that could be valuable in the context of a software. I encourage you to check out different projects that you can import into your project that could be needed in the context of the value point you're trying to provide. Typically, in the context of handling this kind of information, you want to ensure—oh, trying to zoom in, y'all—you want to ensure basically the actual package itself is up to date. You want to ensure that there's a good community around it, and you want to ensure that there isn't a lot of errors around it. Take that as you will. Be in the description down below. Finally, why is ChatGPT here? ChatGPT here is because I think it's a great helper when it comes to creating code. Corbin, what about Copilot? No, I had a way better time at Copilot. What do you—you even talk? Okay, go with Copilot. I like ChatGPT more. Specifically, I like ChatGPT with custom instructions, not using the GPTs. No, no, no, no. Custom instructions. Check out that video right there if you want to know how to create custom instructions to effectively code or any use case for that matter. This is huge, y'all. This is big, y'all. The idea behind coding isn't necessarily I don't know how to code. The idea behind ChatGPT is when it can teach you how to code, of course, but alternatively, if you know how to code a block but you could just have ChatGPT code the block, that saves you a ton of time. It's not about not knowing the logic of the block; it's more in the sense of like actually typing out the block takes time. This is why this is extremely powerful because the issue here isn't bad code; the issue here is more specifically the user doesn't know the right code, therefore doesn't know how to procure it to give you the right code. There's a lot of implications behind using ChatGPT behind coding or whatever, everything above the board. What I can tell you from my experience is I'm staying with it. So that's it, y'all. That's all the building blocks it requires to make an AI software. Corbin, I don't know how to code. Corbin, I am going to university and I'm learning XYZ. Should I learn this lesson? Should I do this lesson? Stop. Just go. If you don't know how to code/you do know how to code and your next question is what video should I watch next, what XYZ, or course should I take? Stop. Open up ChatGPT, watch that custom instructions video, understand how to create effective custom instructions for the context of code, and begin. You now know the building blocks that's kind of behind me to build out a software and everything I basically said right there, except ChatGPT for $20 a month is free/maybe you have to pay a couple bucks because you're hitting the invocation calls and Google functions, everything. Therefore, the question isn't what video do I watch next unless you want to learn more specific context when it comes to concept and software, which you can watch at the end here, but the question is you just need to get going. You don't got time to learn how to create a backend by watching a 20-video playlist on YouTube that is specialized with Google functions. No, no, no, no, no. Start building it. Start using ChatGPT as your educator and proceed. The most powerful thing you can do in the next five years is learn how effectively learn through video and through text, more specifically text when it comes to ChatGPT and coding. These kind of videos I show you on this channel are very much like, "here are the tools, hit the hammer on the nail." The actual hammering on the nail takes time and takes learning. I do plan on doing an entire playlist dedicated to coding out an entire software live, but to be honest with y'all, I'm currently coding out a software, so it'd be like pretty heavy on my day to do like a video coding software and then coding software in the later in the half the day. Like, we'll keep it as it is. That is coming in the future, but biggest advice I can give to you is watch videos like this that give you context, but then proceed. So check out the video or check out the playlist at the end here if you want more context of software, income, revenue, very niche cases when it came to building out my software, everything about the board. I'll see you in the next video. This is the playlist I was referring to, from concept, the software. That's a random video, that's my face. I'll see you in the next video.