# Claude Code Tutorial for Beginners (2026)

## Метаданные

- **Канал:** Dapp University
- **YouTube:** https://www.youtube.com/watch?v=wbTofK6WwtQ
- **Дата:** 09.04.2026
- **Длительность:** 37:22
- **Просмотры:** 4,288
- **Источник:** https://ekstraktznaniy.ru/video/51774

## Описание

Claude Code Docs:
https://code.claude.com/docs/en/quickstart

🎞️TABLE OF CONTENTS:
00:00 INTRO
00:31 Claude Code Overview
02:42 Setup
06:18 First App
08:32 Helpful tips
11:56 Finished Todo App
13:40 IDE Setup
17:05 Modes
19:21 KANBAN App Prompt
28:52 Claude.md
32:11 Git
34:51 Models

Subscribe to this channel: https://www.youtube.com/channel/UCY0xL8V6NzzFcwzHCgB8orQ?sub_confirmation=1

Instagram: https://instagram.com/dappuniversity/
Twitter: https://twitter.com/DappUniversity
Email: gregory@dappuniversity.com

## Транскрипт

### INTRO []

In this video, I'm going to show you how to get up to speed with Claude Code fast using the latest most up-to-date information. This tutorial is for anyone who wants to leverage the power of this technology for coding even if you're not a seasoned developer. So, if you're new around here, hey, I'm Gregory and on this channel I've trained thousands of developers on emerging technologies just like this and today I'm going to show you everything that you need to know step-by-step using some of the most recent features that I've just launched within the last couple of days. So, let's dive in. All right, so let's start off with the basics like what even is Claude Code in

### Claude Code Overview [0:31]

the first place? Well, Claude Code is a command-line agentic coding tool put out by Anthropic, you know, the creators of the Claude chatbot. So, what the heck does that even mean? So, if you've used AI already, maybe you've used a chatbot like ChatGPT or Claude itself or Gemini etc. Basically, what you do is you ask it questions and then it gives you back answers. Now, you can use these chatbots for coding purposes like if you have small coding problems you can say hey write this line of code and then you can copy and paste it and put it in your project. I mean, you can even build full applications this way but it's not very efficient. So, Claude Code takes this to the next level. It actually turns it into an agentic process meaning you're turning it into an AI agent. So, essentially, what does that mean? Well, it basically it means that it runs on your computer. So, it has access to your file system. So, whenever you say hey build me an app, it can actually produce the files for the application on your computer and you can see what it creates, you can use it instantly, you can verify it, test it and add more code. Now, this needs a little bit of clarification because when you hear AI agent, you might think of something else, all right? You might have think of something like Open Claw, which is very popular where essentially you have this AI that's running 24/7 on your computer that's constantly running tasks for you in the background with a chat interface. And that's not exactly what's going on with Claude Code, okay? So, Claude Code is an agent but something like Open Claw is an autonomous agent that's kind of just doing its thing all the time. Claude Code is more like you ask it to do something and it can do a lot of complex things on your behalf and it just waits for the next set of instructions. Now, let's kind of break that down to three categories. You have a chatbot which basically you give it a question or a prompt and it gives you back an answer and you can do something with that answer, maybe put it in a code editor or just use the information. Now, Claude Code is an agent basically is you give it instructions and then it performs those tasks and actually generates files on your computer as opposed to an autonomous agent which is kind of just working, you know, 24/7 in the background. Now, one important thing to understand is that Claude Code is continually being updated and it's actually getting more and more features to make it more autonomous. I'll call out some ways that is the case later in this video but just know for now the agentic aspect of Claude Code is we're giving instructions and it's doing things and it waits for the next thing. All right, so let's get into the setup

### Setup [2:42]

process. So, the first thing that you need if you don't have it already is an Anthropic account for Claude. So, if you go to claude. ai/login, you know, you can just continue with email and if you don't have an account already you can set it up. All right, now it's important to note that if you're going to use Claude Code itself, you actually have to have a paid plan. You currently to my knowledge can't do this on a free tier. So, you need to sign up for a Claude Pro, Max, Team or Enterprise account. For personal use, you can get away with like a $20 a month plan and that will get you set up to start running Claude Code on your machine. And so, then once you've got that, you've got a Claude account and then you also have a paid plan enabled, then you're ready to get started. Now, while you're on the website, if you don't have it already, I highly recommend going ahead and downloading the Claude desktop application for whatever operating system that you're using. You'll just see the link right there, click download, install, you're good to go. Once you've got that set up, you can use the Claude chatbot directly from the application just like this. And one reason that I like having this available on my machine is if I'm ever stuck trying to do something in Claude Code or maybe I want to have a separate chat session open where I'm not actively using Claude Code, I can just use the desktop application right here, okay? Now, another reason to go ahead and download this is if you want to use Claude Code directly inside of the application, you can do that. So, at the top, you can see the chat mode, co-work mode and then code mode. You can actually use Claude Code just directly inside the user interface here, all right? So, it's important to understand that this application actually has access to your file system. So, before I was talking about Claude is an agent is an agentic coding tool that primarily is using your command line but you can use it here. This application, you can tell it to write applications, you can describe the app that you want and it'll actually create that application on your computer. You can connect all types of plugins and other things inside of here. However, I do find that you can do much more powerful work inside the terminal which is what I'm going to focus on in this video but I do want to call to your attention that you can actually use Claude Code inside of the Claude app itself. All right, so with that being said, I'm going to use Claude Code now in my terminal to show you how that works. So, I want to make this more of a developer focused tutorial even if you're not a developer already, you should still be able to follow along step-by-step with what's happening in the terminal. So, the first step is going to be open a terminal on your computer. So, I'm going to be using a Mac so I'll show you the Mac steps for this. If you're on Windows, I highly recommend using Windows Subsystem for Linux and running your uh command line with PowerShell in administrator mode. If you're not sure how to do that, well, you can basically just ask Claude how do I do those things. But for me, I'm on a Mac so I'm going to use my terminal here, all right? I'll open a new terminal session and we can just follow the steps that are listed on the Claude Code documentation quick start guide here. So, I'll put a link to this down in the description below and you can just copy and paste these commands directly into your terminal for your given operating system. But for Mac, I'm going to download this with curl using this exact command. Now, I've already done this but whenever you finish, you should see a success message after this is done. Okay, so once that's finished, the first thing that you're going to do is actually run Claude for the first time and you can do this from any folder in your terminal. We'll talk about the project specific folders here in a minute but essentially you just type the command Claude and the first time you run this, it's probably going to prompt you to log in. I tried to log out to show you what that would look like on my machine but I'm just going to run the login command again. So, basically, I'm going to log in with the Claude account with subscription. And essentially, if you're already logged into Claude in your web browser, it'll just ask you to authorize it and now I am successfully logged in. Now, one thing I want to call your attention to is if I close this terminal, then it's basically going to terminate my Claude session. So, essentially, whenever you're running Claude, you need to start it manually every single time that you open up a terminal. All right, so now our setup is

### First App [6:18]

complete, let's go ahead and create something with Claude Code. But before we do that, I've made sure this video is completely ad-free with no sponsorships or affiliates. So, if you've gotten value out of this video, the easiest and free way to support this channel is to hit that subscribe button down below and click the like. It helps these videos out tremendously for the YouTube algorithm. So, thanks a bunch and back to the video. Okay, so like I was saying before, you know, Claude is a coding agent and we're going to be using it primarily in our terminal. Now, there's a few different ways to do that. We can do it with just a basic terminal that you see on screen here and that's really all that you need to actually use Claude Code outside the application. You can also use it inside of a code editor which will allow you to see the changes that are being made. We'll do that here in a minute but for now, I just want to show you the bare-bones setup to get started quickly, okay? So, inside your terminal, you're going to be running Claude Code with a slash Claude command like I showed you before. Now, what I want to mention is that whenever you run that command, it's going to be scoped to a specific folder. So, what do I mean by that? So, whatever folder you run that command in, Claude is going to have access to the files inside of that folder and everything below it but not above it. So, the best practice here is to not run Claude necessarily from the root directory of your computer but the specific coding project that you actually want to be creating something inside of. So, let me show you how I do it. So, for me, I keep all of my coding projects inside of my code folder. So, I'll show you how to navigate to that inside of your terminal. So, again, you probably already know this if you're a developer but if you're brand new, you just do CD which stands for change directory and I go to uh tilde which is the shift character next to your one which just stands for home. I go forward slash code and now I'm in the directory where all my coding projects are. Now, if I want to create a new folder, I do make directory which is mkdir. I do space and I'm just going to give it a arbitrary folder name. So, in this case, I'm going to say Claude_Code test_1. And now I can enter into that newly created directory again with CD which just stands for change directory. CD Claude code test one and now I'm inside this newly created folder where I can run my Claude session with forward slash Claude or excuse me, Claude. No forward slash. I was getting mixed up with the command you actually run inside of your Claude session. All right, so a couple quick

### Helpful tips [8:32]

things I want to call out after we've actually entered into our Claude session is a couple things. Number one, how do we get out of this? Because that's something that happens all the time. You start something in your terminal and like how do I turn this thing off? Well, you can always just close out of your terminal session like I was telling you before but if you want to keep your terminal session open, what you can do is just type forward slash which that's the thing that you need to run commands inside of Claude. That's why I accidentally did a minute ago and just type forward slash exit and that will end your Claude session and now I'm just back in my terminal, okay? Now, I can open Claude again with a brand new session by just typing Claude, again no forward slash and you're back to where you were. Now, also, you probably see some more information on your screen here than I do, okay? That's for a reason. So, basically, whenever Claude runs by default, it actually exposes like your account information, your email address which I think is kind of ridiculous. So, I actually asked Claude like how do I turn this off and it kept giving me back the wrong answers. So, I had to do some old-fashioned Googling and try to figure out how do I not expose my email address every single time I boot up Claude and here's what the answer that I found. So, this will be for your bash profile or your bashrc or zshrc, whatever you use to manage your environment variables on your terminal, okay? You want to put this inside of there. Export is demo equals one and that way it'll tell Claude it's in demo mode. Whenever it does that, it doesn't expose is email address. This may have other side effects that I haven't experienced yet, but for now I know that it definitely hides my email address if I don't want to show it on screen. So, if you're confused about what that is, you can ask Claude to explain that file to you and how you use it and how you modify it. But if you are developer and you're sick of your email address being displayed every single time you run Claude in your terminal, that's how to fix it at least as far as I can tell. All right, so let's go ahead and create an application. Now, the cool thing about Claude is basically you can just tell it what to do inside of here. So, basically here's your prompt where you can type stuff just like you would type a chat bot, and whenever you hit enter, it's basically going to do the thing that you asked it to do. Now, a really quick tip here is you've probably seen videos of people dictating to Claude, I mean typing to Claude is really slow. I mean, if it's for short stuff, it doesn't really matter that much. But I'm like, "Say, hey, create me a to-do application. " That doesn't take you that long to type, but if you're giving it long instructions, it's a lot faster just to speak to Claude. Now, you see a lot of people using external software on their computer like WhisperFlow or even a Whisper instance on their computer to be able to dictate to Claude, but what's cool is in just the last week or so, they actually introduced voice capability into Claude code itself. So, what you can do, you can have the forward slash command and type voice, hit enter, that's going to put you in voice mode, which I was actually already in, so it disabled voice mode. If I type it again, it'll enable voice mode, and all I have to do is hold the space bar and it'll actually capture what I want to say. So, what I'm going to do now is I'm going to go ahead and create our first application. We're going to create a canonical to-do list application, and I'm going to just going to give it a really bare-bones set of instructions, and it's going to create it for me. So, I'm going to hold the space bar and I'm going to give the instructions. Please create a basic to-do list application for me that runs in my web browser. All right, so it's processed my voice instructions. I'm going to hit enter. All right, so it's working, and now it's asking me if I want to create this HTML file that's going to create the to-do list. All right, and so it's asking me if I want to approve these changes. I'll show you how to just bypass all that and actually have Claude just do things on your behalf, but for now, we'll just click yes. All right, so it's done. It

### Finished Todo App [11:56]

created the to-do. html, which is just a single file for this application, and I can just copy this file path here to open up the HTML file in my browser, and boom, there we go. It created a to-do list. It took less than 1 minute for it to create on my end, and let's just test it out. New to-do, go to the store, make a new app, and I can delete to-dos, and I can accomplish them, and it actually checks them off the list. I can see any active to-do items and any completed ones. So, it had some really good intelligent, you know, ideas out of the box. Now, what I want to call out here is that I gave Claude very minimal instructions. All right, I basically just said, "Hey, create a basic to-do list that just runs in my web browser. " And it just inferred what I wanted in my to-do list and made it. Now, that's fine for just little simple toy applications like this. Now, this is not a production application, and when you're going down the route of creating a production application, you want to be more specific about how you actually create your applications. And really, the quality of what you get out of this, regardless of what other people are telling you online, is what you tell the AI that you want and defining the outcomes that you want very concretely, and then specking that out and prompting it so that it actually understands what you're looking for, because really, the quality of the result is going to depend on the quality of instructions. And that's what I want to kind of focus on now as we get into a little more advanced setup, okay? So, that was basically all you have to do to get Claude code up and running, you know, just with your terminal here. I'm not even verifying any of the code that was written. I didn't even look at any of the code. I just tested that it worked in my browser. Now, let's take it to the next step. Okay, so we talked about

### IDE Setup [13:40]

different ways to run Claude code. You can use it in the app, and it can actually modify files in your computer. You can do it directly through the terminal like I was just showing you before. All right, but that was kind of limited in some ways where we can't really verify the code that's being written and a lot of other stuff. Now, another way to do this is to actually run a code editor or an IDE, which is basically an environment that lets you write and edit code like you would by hand, and then you can actually run Claude code from a terminal or a plugin inside of that. So, there's lots of different options out there. A lot of people like Cursor. I still like Visual Studio Code, okay? And that's what I'm going to show you to use right now. You can use Claude code out of the box. You can use plugins, all that type of stuff. If you don't have Visual Studio Code already or you don't have an editor of choice, you can download VS Code just from code. visualstudio. com for your operating system, and then just go ahead and get started. All right, so I'm actually going to create a new app that's a little more advanced than just a basic to-do list application, and we're going to work inside of VS Code. So, I'm actually going to start in my terminal. This is my preferred workflow. So, I'm going to create a new folder. Again, you can do this in your file browser if it's on Finder on Mac or whatever you want to use on Windows. Just like right click, you know, create new file wherever you want it. I just happen to do things from my terminal. So, inside my coding folder, I'm going to do make directory, mkdir, and I'm going to say Claude code test underscore two. All right, I'm going to enter into that newly created directory. I'll say cd Claude code test two. And then I can open up my projects in VS Code with my code shortcut like this, code dot. All right. And boom, there we go. We are inside of our Visual Studio Code app with our new folder. Now, if you don't want to do everything from the terminal, again, you can do it in your file browser, and then you can just do file open inside of Visual Studio Code, find wherever you created that file, and then boom, you can create a new project just like that. Now, it's pretty easy to run Claude code inside of here. They have a plugin that you can use. You can actually see that I've got, you know, access to Claude code inside of here. In the top right-hand of your Visual Studio Code, you should be able to click this button that has the bottom uh toggle panel right here. All right, and that's actually going to open up the bottom panel, and in by default, it should be opening up your terminal in the current directory. All right, you can also click this side panel here, which will toggle the sidebar, which is going to be the code folders that Claude is actually going to generate for us. So, whenever we tell Claude to create files, you'll be able to actually see those files listed here in the sidebar. All right, so it's as easy as starting Claude from your terminal inside of here. I just type Claude, and it opens up a new Claude code session for me, okay? Now, one thing that I want to mention is you can actually run multiple Claude code sessions at the exact same time, basically using different terminal tabs. So, if I added a new terminal tab here, I could do a Claude, and it's going to run two different instances of Claude for me, okay? So, I could have Claude doing something here, and then while it's doing that task, I could actually have it do another task with Claude. So, for example, I could be working on the front end in one, on the back end on the other. Maybe they're in different folders, maybe in the same folder. I'm using Git worktrees, or you can just clone the project name and call it a different folder. That's a little more advanced than I'm going to get into in this video, but just know that that's actually a modern development workflow is to have concurrent Claude instances going so that you can ship things really

### Modes [17:05]

fast. Okay, before we start building out our application, let's actually look at a couple different things inside of Claude code that you want to understand while we're doing this. So, first of all, you can always, anytime you're inside of Claude code, you know, just follow the instructions that are there on the screen. So, right here, it says question mark for shortcuts. So, if I just type question mark inside of there, it's not going to put a question mark in the prompt box. It's actually going to show all of the available shortcuts for me like bash mode slash for commands, okay? So, if I do forward slash help, all right, that will actually show me all the available commands. They organize them in general commands and then custom commands, which you can just move around with your arrow keys here. All right, you can read through this, and you're going to find all different types of things you can do with Claude code. I won't be able to get into all those in the video because there are dozens, if not 100 plus things inside of there. All right, escape to cancel to get out of that screen. Now, if I do the question mark again, it's going to show me something also incredibly valuable inside of there, which is shift tab to auto accept edits. Now, I'm going to click escape again, and let me talk about that. If I do shift tab, all right, shift tab, you can see this question mark for shortcuts is going to change to accept edits on. If I do shift tab again, it's going to say plan mode on. And if I do it one more time, it's going to take me back to the original screen. So, if you notice, that was three different things. Now, those correspond to the three different modes that Claude code runs on currently at the time of recording this video. So, the first mode is the default mode that Claude is set to out of the box. Now, if you recall a minute ago when I created the application in my terminal, it was saying like, "Hey, do you want to create this new HTML file? " And I had to actually select yes, I want you to do that, okay? Now, if I change the mode to accept edits, it's not going to ask me that. It's just going to create the HTML file for me, and it's going to, you know, essentially make my life a lot easier. Now, I can also do shift tab one more time to do plan mode, which is going to be read-only mode. It's not actually going to create any files on my computer. It's just going to create a detailed plan of the steps that it's going to execute in order to actually write the code for me. Now, this mode is actually very helpful, and it's the one that we're going to start off with when we create our next application. Okay, so

### KANBAN App Prompt [19:21]

for this application, what we're going to do is we're going to create a Kanban board. So, previously we did a to-do list application, which is pretty easy to one-shot. It's like the canonical software development CRUD application. But what I want to do is take this to the next step and actually create a more project management type of board rather than a single user to-do list, okay? Now, before I talked about, you know, the quality of the code that you actually get depends upon the instructions that you give it. And if you are a software developer, this is where you can really use your architecture understanding and things like that to get a lot more out of these tools than what you know an everyday person would be able to accomplish, okay? So, this is where you're getting the AI to be more like a pair programmer with you rather than just like you know, somebody who doesn't really know a lot about coding actually being able to create useful application. Now, you can definitely do that if you don't know a lot about code, but there's always going to be a competitive edge to actually understanding architecture and what the tools are capable of. So, what I'm going to do right now is actually put in a prompt and show you like how you can be more specific about what you want. Okay, so I've just dropped my prompt inside of there. Now, one really quick thing that I should mention is like if you ever get a lot of text inside of your Claude code instead of having to backspace everything manually, okay, you can hit escape twice and it will just clear everything for you. All right, that's another little tip. But, here's the prompt. Okay, so let's create a new Kanban web application like Trello, okay? So, I gave it an app example to reference. So, inside the app I want to be able to create multiple projects. Inside of each project I want to create cards that will get moved in between each column or stage. I want to custom define what each stage is inside of each project. And in the project view, I should be able to see a collapsed view of the card with the title. Okay? When I open the card, it should appear in a modal with a title and a large area for the description. It should support comments inside the card with the expanded view. I should also be able to edit the title from the project view. I should be able to uh edit the title and the description from the expanded view. And here's where the software development principles help out. Inside, let's keep everything monolithic, meaning a single repository uh with Next. js and let's use better SQLite, you know, for the database. So, I'm going to hit that. We're going to be in plan mode. So, while this is um populating, basically Claude is going to just create a plan for us. All right, you can actually see some information here. Like it's talking about the token usage, all right? So, basically that is the cost of what I've just asked Claude to do. Now, with your plan you have a specific amount of tokens that you're able to use in a certain time window. And if you're not careful, you can quickly eat up those tokens and find yourself waiting for them to reset, all right? Now, the good news is if you have a credit card hooked up to Claude, it's not going to over bill you. It's just going to run out of tokens, all right? But, that's what it's doing. We can actually see the current cost uh and we can see the amount of time that it's you know, taking. Now, we don't really know what Claude's doing behind the scenes. It's just kind of waiting for this plan to get created. However, if you want to actually see what Claude is doing, you can do uh control O, all right? And it will show you what Claude is sort of doing behind the scenes, okay? So, instead of just this you know, kind of black box process, it'll actually kind of show you uh the logs of what it's actually thinking. All right, so Claude has actually completed my plan, all right? And what I can do is see you know, the result of this, all right? Basically, it's going to um it's telling me everything it's going to do. It's going to create the database with better SQLite like I talked about. Here's the text stack, Next. js, better SQLite, uh etc. etc., all right? He's giving it self-context, which he can come back to later. We'll actually talk about how to save that context here in a minute, okay? We can see all the dependencies that it needs for the project and it's actually going to give us a order that it's going to build things in. And then finally, the verification criteria based upon the prompt that we created. Now, at the bottom it says Claude has written up a plan and is ready to execute. So, how would I like to proceed? Now, this is going to talk about the mode that I talked about before. So, um basically, what I'm going to do is say yes, auto accept the edits. And that's actually going to switch back Claude out of plan mode into the auto accept edits mode. That's what I want because if I don't if I manually approve everything, I'm just going to have to click yes every single time that it creates a new file or makes, you know, an important change to my computer. So, I click uh enter and it's going to do its thing. And just to note, you can see that accept edits on, it's manually um or excuse me, automatically changed the mode. Now, there will be a couple things that I have to improve like if it wants to create new um dependencies that I like on my computer. All right, now I've had to you know, manually approve a couple things that it wants to do, but it's not like I'm having to say yes every time it creates a new file. But, what's cool is as I'm monitoring this process, there's one reason that planning mode before you start writing the code is beneficial is you can actually see that whenever you do the plan mode, it sort of sets up a checklist for itself and that's one way that you can sort of monitor what Claude's doing is like it's already completed the first task in the plan. So, scaffold Next. js project and install dependencies, boom, that's done. We can see it scratched off the list and we can see right now it's working on the second part of this task, which is uh setting up the database layer. All right, so the plan was successfully executed. Took about 10 minutes on my end, just full disclosure. It's a lot faster than I would have coded this by hand, but still had to wait a couple minutes for it to work out. So, um we're going to try to test this thing out, all right? It's going to require some approval from me. Do I want to proceed? Basically, it's just going to run uh some commands for us. And so, what's cool about this is Claude is actually checking its work uh before we actually open this up. And one of the reasons I wanted to do this in uh a file browser or excuse me, a code editor, an IDE, depending on whatever you want to call it, is that you can actually see the files that Claude has created while it's been executing this plan, okay? So, you can actually see the app directory here. We can see the node modules and with the package. json that it's created, all right? Uh we can actually see the SQLite database. It's not going to show it to me because, yeah, you don't want to look at that inside VS Code, but you can actually see um let's see the apps, the projects, the page, uh the CSS that it wrote for us, the layout, the page, etc. Okay, so according to Claude, uh the app is finished, at least you know, the initial instructions that we gave us. Uh it's running here at localhost port 3000. You can see what's included with the board view, all the details that we gave it, all right? So, let's go ahead and open it and see what it looks like. Okay, so here's the blank dashboard. Let's give it a test drive. So, again, I wanted to create a Kanban that can have an arbitrary number of projects and then in each side of those projects we'll create our own arbitrary workflows with whatever cards we want to. So, let's go ahead and create a new project for let's just call it YouTube. Like I want to manage my YouTube video pipeline, all right? So, up to YouTube and now we can add, you know, any sort of column that we want. So, the first stage is going to be uh let's just call it you know, to do. And then the next one we'll say uh scripting. And then, you know, we'll just say ready to film. And then we'll just say recorded. Uh this is probably have better names, but we can just you get the idea. Uh let's just say edited. And then say published, all right? So, I might have different actual names for a project, but this is the idea. So, let's call this like how to use Claude code. All right? And then we'll just click save. That'd be a video idea. Close that. And then let's just see here. How to use Claude code. Let's actually open it and see what it created for us. How to use Claude code, that was the description. Okay, so the description is here, the title's up here. That's what it was. So, we'll say make a video about how to use Claude code. All right? Save that. Let's do another video. We'll say how to use terminal as a developer. All right? And then I can say uh yeah, you get the idea. And so, I can also add comments, right? Like they didn't save it. Yeah, so we have to actually hit save, which makes sense. We can comment on the cards. And so, say like this should be very detailed. Blah, blah, blah. Other team members can comment. We can edit the description. We can also, believe delete the projects as well. Okay? So, anyways, I'm not going to belabor every single thing inside of here. I think you pretty much get the idea. Uh we've also got an issue there. We could get into the debugging with Claude code, but this is a really good first start, okay? So, that's how I would start a project like this. Basically, be very specific about what you want, all right? Um if you're developer and you understand the architecture decisions, like I was like, "Hey, we need to have a database here because, you know, some people may not even understand they need a database. " Like with the initial to-do list application, it was just a simple thing that runs in your web browser. As soon as you click the refresh on the page, all those to-do lists are going to be gone. But, you if you're developer, you understand persistence, the need for that. Obviously, we're using better SQLite for local development. For production, you'd want to swap out something a little more robust. Um but, you can work that all out with Claude with your development plan. Okay, so

### Claude.md [28:52]

that's an idea of how you can create an application with some more advanced prompting um with a little more developer workflow inside of your IDE with Claude code. Now, there's a couple other things that I want to call out here because let's say that we want to come back and actually develop this application later, okay? Um then there's a couple things we have to understand. Number one is context. So, you probably you may I don't know, maybe you may have heard about this, but a lot of people in AI and using these models for coding and for whatever for purposes uh talk about this concept of context. So, context is something that the AI really needs in order to function, okay? Now, you can open a Claude code instance and have it read your code base and it can start to gather its own context, but if you have sort of best practices that you want to implement for Claude every single time that it writes code, it needs that inside of its context. Now, there's a lot of different ways to manage context for Claude and Claude code, but one of the simplest ways is what's called the Claude. md file. Okay, so let me show you where this is. So, inside of your um app, you may not see this just yet. Well, you see Claude. md right here, okay? But, uh there's a better way to do this, which is by doing forward slash init. All right? What it's going to do is actually create this for me. Okay, so whenever I ran the init command, after I had already written the code, basically what Claude did was it said, "All right, take all the important context that I need to understand about this code base, okay, and write it to this Claude. md file, which you can actually see right here at the root of the project. " You know, up here on the file browser, click on Claude. md. That was blank a minute ago, but now you can actually see that whenever I ran the init command, that Claude just sort of took all the relevant context and, you know, added it to this Claude. md file. You can see it's not super long. You don't want your Claude. md files to be super duper long, because it actually bloats the context window, it makes the models uh less performant, and it actually makes them more expensive to run, because it has to, you know, ship all this context every single time it runs a single command. So, you generally want to keep your Claude. md uh files pretty lean, and give it the most essential context that it needs. But, I find them very important, because, let's say I close down um this Claude session, or, you know, it wanted to run a command that said, "Hey, do you want to clear the context window before running this command? " The Claude. md file gives you a place where that context is saved to have the essential information that you need anytime that you want to do that. So, if you're adding other things like skills, or you're adding things like special GitHub commit information, like you might want to put that inside of there, best practices about what you want Claude to do every single time that you run a session, that's the place for it. Now, there's different places you can actually keep the Claude. md file. You can have a. claude folder, okay, here, which is a place where you can store your skills and lots of other things. I'm not going to get into all that in this video. The simplest way to do it is to just have a Claude. md file in the root of your project, but if you want to get more advanced and look into the. claude folder, um let me know down in the comment section below. I'll create some more videos on that. Okay

### Git [32:11]

so the next important thing that you want to use with Claude code is version control. So, I use Git for this, all right? So, what is Git? It's a version control protocol for your computer. So, what does that mean? Well, you can see with this Claude code session, I just created dozens of files with a single prompt, okay? Now, whenever Claude goes and makes more changes, if we add more features later, I actually want to keep track of the work that Claude does, because essentially the work that it's doing on my computer is destructive. And if I ever want to go back and, you know, revert to a previous version, that's what version control is all about. So, that's what Git allows you to do, and it in addition to taking the code and taking it off of your computer and putting it somewhere else, like on a on the cloud, like on GitHub, or on a remote server, if you actually want to deploy your application so that other people can use it. So, I'm not going to walk through how to set up Git in this video. Um you can, you know, check out other videos on that, or you can just ask Claude how to do it, all right? It'll tell you how to set up Git on your computer. But, I want to talk about how I actually use Git when I'm running, you know, Claude code on my computer. You'll see some people out there who are like, "I want to teach Claude how to use Git so that it just automatically commits stuff for me, and I don't have to think about it. It can just auto-deploy every single time that it makes a change, and maybe even, you know, put it on the server, whatever, right? " I don't actually like doing that for a couple reasons. Number one, writing a commit message is trivially easy, in my opinion. Like, AI, to me, is a lever, okay? And uh I want to use a lever to get maximum leverage. So, again, in about 10 minutes, it created an app that would probably take me at least a few hours to create, probably more than that, honestly, based on the complexity that was there, okay? Now, a commit message takes me 20 seconds to write. Now, if Claude writes a commit message when I don't want it to, then I have at least a pain in the butt of maybe a minute or two, you know, reversing that activity and making sure, with a sanity check, that I was actually at the state that I wanted to be in the software. You know, also, if Claude is just writing commit messages whenever it makes a change and pushing it up to GitHub, well, if I'm working with team members and I want to manage my branches properly, then like it might be pushing to the wrong branch, and yeah, you might have safeguards in there, but it's just a pain in the butt. Like, I don't see a big win with that. So, anytime that Claude makes changes, I want to be the human in the process that's like saying, "Hey, this is the point that I want to save that. This is where I want to make sure I can roll back changes to. " And so, like for example, I just say get status, uh get add dot, get commit dash am, and I'll just say, you know, initial commit. It's as easy as that. I can connect to my GitHub, get push, you know, blah blah blah. Okay, so one of the last

### Models [34:51]

things I want to talk about before we wrap up this video is how to change the models inside of Claude code. Now, again, you can use slash commands. Actually, let me go back into Claude code. You can use slash commands, you know, to do different commands inside Claude. So, if I do model, all right, it's going to allow me to switch between the different models that Claude is going to use. Now, why is this important? So, the default model is going to be uh the Sonnet model, 4. 6. Um you can do uh Opus, which is uh also 4. 6, the most complex type stuff. Um Or Haiku, okay? Now, I was actually using Sonnet for this, okay? Opus is actually the most powerful one. That might be your default of what you've actually got your setup. I can't remember if that's what came out of the box with mine, or if I've just been happening to use Sonnet for this type of work. Um i- if you're using Opus, I'll put it this way, it is the most capable model, okay? But, it's actually going to burn up your tokens pretty fast, okay? So, I mean, it's great, but I've been using uh Sonnet for most of the stuff in this video. It was perfectly capable, it did most of what I needed. Uh Haiku's going to be the quickest, and probably not the best for most of your coding tasks. It might be fit for some things, but um I find that Sonnet's pretty economical, but Opus is definitely going to be the one that's going to give you the best reasoning and the fastest um computation, but it is going to burn your tokens quickly. You can simply hit enter to change the model, okay? Hit forward slash model again uh to downgrade um to Sonnet, if you choose to do that. So, again, not sure what your default's going to be out of the box. Hopefully, you got a lot out of this video. If you did, again, I'm trying to make this video completely ad-free, no sponsorships, no affiliates. So, if you did get value out of it, the easiest, free way to support this channel is to just subscribe to this channel and smash that like button down below. There's a lot of other concepts that I could cover in this video that I just don't frankly have the time to do, and I want to be making more videos about those in the future. So, if you like this, that'll give me, you know, all the energy that I need to go in there, start talking about other things, like how to do these things in parallel, how to do sub agents, how to do orchestration, and how to manage your context MCP servers, all these other topics that you need to know if you want to be an advanced agentic developer with Claude code. So, hope you like this video, and I'll catch you in the next one.