Introduction

[00:00:00]
>> Scott Moss: Welcome everybody to building AI agents from scratch. I'm your instructor here from the master. Scott Moss, super excited to be here to teach in this course. AI is pretty much all I've been consumed with for the last year plus, but specifically the last year running an AI startup and doing all the AI things there.

[00:00:18]
So it's pretty much, I feel like I just had to teach a course on it, because this is all I'm doing. But also, because, as you all know, there's just so much relevant things happening with AI, it's part of everyone's tech stack. And I think this is a very interesting but required skill that I think folks need to learn going forward whether you use AI as a tool to help you write code, or whether you're writing, or whether you're working on the tool using AI that writes the code.

[00:00:48]
Either way, you need to understand how it works. So, this course is just diving beyond the hello world, LLM, things that you see online and building something more functional, something more powerful. So here is the repo. It is going to be at my GitHub
>> Scott Moss: Username, which is Hendrixer, after my son, and it's just agent-from-scratch.

[00:01:14]
So go there, you need that code. When you pull it down, you'll be on the main branch. That branch is the complete version, so don't start on that branch. There's another branch called Step /one. Check out to that branch, that's where you wanna start. If you wanna see what it looks like when it's done, you can go to the main branch that's there for you to reference as you go along.

[00:01:35]
If you wanna see how did I complete it? How did I get it done? It'll be there. But again, if you also took my course, we're just gonna build this up from scratch, and I might do it slightly different than what the main branch is. I'll try to keep it the same.

[00:01:47]
But yeah, that's there for you. And then onto the notes. The notes are here and the read me, we'll go check those out. Just a page that I hosted on Notion, just super simple to just to add notes in here. Most of these notes here are just going to be for two purposes.

[00:02:02]
One, for me just to stay on track, really, because otherwise I'll just get lost and detoured because I have severe ADHD. And then two, a lot of the code that we'll be doing in that lesson is just in here. So you can just like, maybe he was typing too fast or I'm behind, I want to get caught up.

[00:02:20]
You can just go into one of these sections, copy the code, and paste it in if you want. But I also will be doing branches every single lesson that I push up to GithlHub that has the code that we just worked on. So you can just check out to that at the end of the lesson.

[00:02:34]
Let's just dive into the intro and kinda go over the prerequisites that you'll need, what you'll be learning, how you'll be learning, what you'll be building all the good things. Like always, interrupt me if you have any questions, but I'll try to be as thorough as I can and try to get to live coding.

[00:02:48]
So prerequisites, what do you need to take advantage of this course of creating AI agents from scratch? Well, you really only need two things. You need to know JavaScript because we're using JavaScript. Well, technically I'm using TypeScript, but you. You don't need to know typescripts. If JavaScript, you know TypeScript, it's totally fine.

[00:03:04]
If you've never seen TypeScript before, just write JavaScript. It'll work exactly the same. It's literally the same thing with optional types and then having some understanding of like, how APIs work. So if you've not to the point where you build your own APIs, but if you've ever interacted with an API before, whether it was like through an SDK.

[00:03:25]
Or using some type of fetching mechanism, like fetch or Ajax, whatever it is, as long as you know how to interact with an API and like how to make a request and get a response, that's it. Those are the only two things you need to know to get the best of this course.

[00:03:40]
And even if you don't know JavaScript, but you're really good with another language, a modern language like Python or Ruby and you're taking this course and you're worried that you won't build anything from it, you'll probably still be fine. So I wouldn't worry about it. We're not gonna be doing any JavaScript trickery, or there's no frameworks in here.

[00:03:59]
There's literally almost zero dependencies, other than some simple tools to help us with AI. So it's very simple on the programming side, which is crazy, because AI is super advanced, and it's so weird, cool. So formats, basically two things I'm gonna give a lecture, kind of like I said.

[00:04:20]
Introduce a topic and a section, kind of go over it from the definition of it, then I'll probably talk about my definition of it. And then I'll do some examples, and then we'll hop right into live coding, and then we'll take a break, and we'll do that on repeat until we finish the app that we're making.

[00:04:38]
But in the course, you will have a AI agent that you can play around with, build upon, and show off to your friends. And then, as far as like technical setup, obviously you need Git. If you don't have Git, get Git, that should be the first thing you do ever as an engineer.

[00:04:53]
The second thing is just make sure you have the latest version of node. I think 20 plus works fine. I don't even know what version note is on right now, but 20 plus works just fine I tested it with that. I also use bun, so if you do use bun, this repo has like a bun lock and all that stuff.

[00:05:09]
So if you'd like to use bun, you can use bun if you want. But I'm gonna use node and npm for the sake of just like making sure this works with everybody. And then the last thing is you need to open API account. I put a link in here to sign up for it, AI is not free.

[00:05:26]
Those GPUs go bur so you got to pay for them. And I there's a way for me to teach this course and have everything run on everyone's computer, but everybody has different computers. So some people might wait an hour to get a response some people might wait five seconds.

[00:05:43]
It really just depends on what version of Apple M1 you have or gaming laptop or whatever. So I decided not to do that. And we went with OpenAI, which I think probably is the most simplest one that we could use. There's other ones that we'll talk about, but we're gonna use OpenAI.

[00:06:01]
So make an account. I do think you have to add a card, and if you're worried about cost, if you sat here and ran the agent that we're gonna build over, and over, and over for the next two hours, you might pay a dollar, maybe. That's if you just ran it on a while loop or something, so there really isn't an issue with cost here.

[00:06:22]
You'll probably spend a couple pennies today on this. So, but you need it because, again, GPUs aren't cheap.