Claude Code is the most insane AI coding tool ever (full guide)

Some of the things I'm going to show you in this video will change how you use AI forever. Claude Code is the most powerful AI coding tool I've ever used. And in this video, I'm going to show you my workflow that has 10xed my productivity with AI coding tools. I'm going to show you why Claude Code inside a cursor is probably the most revolutionary AI coding workflow ever. And I'm going to show you how you can set it up for yourself so you can get the same productivity boost I got when I started using these tools. If you stick around with me to the end of the video, I promise you, you'll be building apps faster than you've ever built them before. Let's get into it. So, what is Claude Code? Claude Code is the Agentic AI coding tool that lives inside your command line. So, it's not Cursor or Windsurf where you have an entire development environment or app just for it. It just works straight out of the box inside your terminal and command line. Here's the difference, though. It can take a tremendous amount of context from your entire application and work autonomously for hours on its own. There's no more babysitting needed with AI agents. Typically, when you're working with like a cursor or wind surf, you can see their AI agent on the right, you're constantly babysitting, right? Every step you do with it, you have to hold its hand to make sure it's doing the right thing. You can't just give it commands and walk away. You're walking with it, checking the code it's building, then going back and giving it new micro commands. Here's the difference with clawed code. It's smart enough and it has the context window where you can give it very largecale commands and then it'll go to work and you can work on other things. It truly feels like the firstofits-kind autonomous AI agent that codes for you. But here's the thing, there's a lot of people talking about clawed code versus cursor. Which one's better? I'm going to show you a workflow where you combine the two to unlock productivity gains you never thought was possible. I'm going to show you how you can combine clawed code with background agents, the new feature in cursor 1. 0 that let you spin up multiple agents at once so that you can build apps faster than you possibly can imagine, all without writing a single line of code. What I'm about to show you is going to feel like you have an army of AI agents working for you autonomously. It's going to be insane. So, here's the plan for what we're going to do in this video. I'm going to build an entirely new application out using this new workflow I'm going to show you. I'm going to show you step by step how we're going to get this set up. I'm going to show you why this is so different from any other way you've ever used cursor before. So, what we're going to do is we're going to start off. I'm going to build a notion clone with this new workflow. I'm going to show you how we do this. We're going to open up cursor. a new project. We have a completely blank slate here. What you see to the right is the built-in AI agent into Cursor, the one you're probably used to. But what I'm going to do is this. Now, we're going to set up Cloud Code. I'm going to show you how to set up Claude Code and I'm going to show you how this all will interact with each other. So if you're

on Mac, you're going to hit control tillday. That will open up the terminal down here. Once you're in here, you're going to paste in the command to install claude code. That's going to be npm install-gropic. And I'm going to put this down below so you can just copy and paste. But feel free to pause here, take the command from the description, and paste that into your terminal here. Once you have that installed, all you need to do is type in claude, hit enter, and it starts up Claude Code. So Claude Code's all set up. You can see in the terminal here, what I'm going to do is I'm actually going to move this terminal over to the right hand side. And we're going to use this basically as our agent. So if you rightclick and then do panel position, right, we can move our terminal and claude code over to the right hand side. And this is where we're going to work out of. Just makes the UI a little bit cleaner. So what I'm going to show you in a second here is how we're going to combine clawed code, which I showed you here, with the new background agents. For those who don't know, background agents is brand new in cursor 1. 0. When you go into the AI agent over here and click this cloud, it shows you this window where you can spin up new AI agents. And these are going to be agents that work in the background autonomously on small little tasks. And so what I'm going to show you here is how I use clawed code for big tasks that I can walk away does its own thing. I don't have to monitor it. and then set up background agents that do little tiny tasks that I don't have to worry about either. I'm going to show you how all this can work together to complete unbelievably advanced workflows. And all of this combined will unlock unbelievable productivity. But let's start with Claude Code and get the app set up first and then I'll show you how this all ties together. What most people have been doing with Clawed Code is they've been using Clawed Code just in their terminal by itself. So, first thing we're going to do is select our model. Now, you have different options here. If you're on the $20 Claude plan, you'll only have access to Sonnet 4, which is fine. Sonnet 4 is incredible. But if you're on the $100 plan or the $200 plan, we're going to have access to Opus, which is their best model, and it is a little bit better. It's not necessary. You're fine on the $20 plan, but if you're on the $100 plan or the $200 plan, you could unlock max productivity with Opus. So, let's set that up. First thing you're going to do is do slashmodel and then I'm going to type in opus. This is only if you're on the $100 or $200 plan. Hit enter on that. But feel free to continue with me. Even if you're only using sonnet, totally fine. So we set the model. Next thing we're going to do is actually set up the app. But before that, we're going to have Claude Code do some planning for us. So this is something people don't really talk about when it comes to Claude Code, but Claude Code is probably the best high-level thinker and planner of any AI tool I've ever used. So what we're going to do is we're going to use clawed code before we build out the application to not only do planning on how to build out the app, not only do the project management of how to build the app, but also do the agent management. And what I mean by that is figure out a plan for what we're going to do with background agents while the app is built out. Again, this is what is so uniquely amazing about this workflow I'm about to show you is we're not only using clawed code to build incredible applications, but I'm showing you how to use other AI agents simultaneously to get a lot of things done at the same time. So, we're about to build out a prompt that will manage all of that as well. So, there's two things I want to set up with you first before we run our first prompt. We're going to create two files over here on the left. The first file is going to be called claude md. So this is a markdown file that's going to have rules for claude code. I'm going to show you those rules in a second. The second file we're going to make is project plan. md.

And this is where the project plan cla's going to build. That's going to manage all our AI agents and manage the application is going to live as well. So you want to set these two up first. And before we run our first prompt, let's just put in our Claude rules here. These are going to be rules that are included as context to every single prompt that we give to Claude automatically. We don't even have to tag it. So, here are the rules. I'm putting them down below. Feel free to copy and paste them. Pause here if you need to. If you're working along with me, put it into your own claw. md. But first, think through the problem. Read the codebase for relevant files and write a plan to project plan. md, which is our new file here. The plan should have a list of to-do items that you can check off as you complete them. Before you begin working, check in with me and I will verify the plan. Then begin working on the to-do items, marking them complete as you go. Please, every step of the way, just give me a highlevel explanation of the changes you made. Make every task and code change you do as simple as possible. We want to avoid making any massive or complex changes. Every change should impact as little code as possible. Everything is about simplicity. I always give these commands to AI. You want to make sure everything is as simple as possible for the AI. The more you break things down into simple steps, the more powerful the AI performs. And then finally, add a review section to the project plan file with a summary of the changes you made and any other relevant information. Basically, we're setting the guardrails around how the AI agent will work with us and keep us updated and give us information. It's good to always track everything the agent's doing. These this will set up good guard rails for the agent. Also, shout out McKay Wriggley. A lot of these rules are inspired by things he's done with his AI agents. Very helpful. Okay, so here's the prompt we are going to give Claude code. This is an in-depth one, but this is the key

to unlocking the productivity we're going to get with all our agents combined. So, I'm looking to build a simple notion clone. And by the way, this is down below as well. Feel free to pause and copy and paste this. I'm looking to build a simple notion clone, an advanced notetaking app with file and folder system that allows you to create many different types of notes and files. I'd like it to be built with Nex. js. This is the tech that's going to power our app. Before we start building this out though, I want to do a little planning with you. I first want you to make a project plan for this. Inside project plan. down, please build an in-depth plan for the app. Have highlevel checkpoints for each major step and feature. Then in each checkpoint, have a broken down list of small tasks you'll need to complete the checkpoint. So basically, we're having it built like a project management document with major checkpoints and then small tasks it needs to complete for each of those checkpoints. Also in this plan include instructions we will give to a marketing background agent, a researcher agent that will research user needs and a feature planning agent that will build the road map for us. We will then review this plan together. So that's another key here. We're having it include instructions for three other AI agents that we're going to loop into this workflow that I show you in a second. This is where the magic will soon start coming together. So let's hit enter on this. Feel free again to copy and paste this from down below. Put it into your own cloud code and we'll build this together. I'm going to hit enter. By the way, as we're going, if you've learned anything so far, make sure to leave us a down below. Leave a like. All I do is make amazing videos on how to build with AI. So, what's going to start happening now is Claude Code is going to go and start building this plan for us. So, it's going to start thinking and designing what it wants to do, and then it's going to start putting the plan into this project plan. Markdown file. Once we agree to a plan together, then we can have it go and start building us out the application for us completely autonomously. But this is one of the really big strengths of Claude Code is the planning it does and just the highlevel thinking is unmatched by any other AI tool I've seen on the planet. Especially if you're using Claude Opus with this, you're going to see incredible results. And yes, it does take a little bit longer, but the results you get are so much better. You know, there really is two modes to use with Claude Code. planning and execution. And you really want to go back and forth between the two. Do planning and then execute with it. What we're doing right now is the planning section. Okay. And just as you saw as we were going there, it completed the entire plan. So let's take a look at what it did over here. Notion clone project plan, a modern note-taking application built with Nex. js. Okay, it has the highlevel architecture of everything that's going to go into this. And then it has the development checkpoint. So setting up the project and then it has every single step for setting up the project checkpoint to authentication system and every single step for setting up authentication core data models right and so what we're going to have it do is actually go down this list checkpoint by checkpoint and complete all the task by itself. This is something that the AI agents built inside a cursor and windsurf have never been able to do before. Those AI agents are really good for doing each step one at a time, each little micro step and babysitting it. But what we're what's really amazing about Claude Code and what I'm going to show you is how it can tackle all of this all at once and go down step by step. Let's see here. Does it have the plan for the other AI agents? Boom. Here it is. Marketing background agent instru instructions. So, it has the instructions for the marketing research agent. This is what we're going to give it. And this agent's going to work in the background for us and develop a marketing plan. It has the instructions for the research agent that's going to work for us again in the background doing a whole bunch of work for us. And then of course the feature planning agent that's going to build our road map. And what you're going to see us do is we're going to have four AI agents spun up. Claude code and our three other background agents that will all be working simultaneously to build this app for us while we just sit and watch all the magic happen. This is the productivity unlock. Right? Before in cursor, you were just working on one task at a time and babysitting the AI every step of the way. What I'm showing you here with this new plan and this new model is we have kind of our master developer AI agent in clawed code building the application out and then we have these new background agents which is this new feature in cursor where we're going to have employees working for us in the background doing all the planning for us. And there's a lot more to it. I'm going to show you how to set it all up and get it working. So, do you want to make this edit? We're going to hit yes on this. And now the document is in there and good to go and we can start making our own edits inside of here. What we're going to do now is I'm going to set off cloud code to start building this application and then I'm going to show you how we combine this with the background agents in cursor to do incredible things. So let's start this first. We'll set off cloud code. And again, what makes Cloud Code really, really amazing is it is able to take these really large workloads like what I show you here with major checkpoints and a bunch of different tasks under it and handle it completely on its own autonomously. It's actually so good that a lot of the time when I'm using cursor now, I can work on two apps at the same time because clawed code is so independent. So, I'll on one side of my screen have an instance of cursor up where I'm working on Creator Buddy, my AI content tool that I launched a few months ago, just adding updates autonomously. On the other side of the screen, I'll have another instance of cursor up where I'm building my next app that I'm going to release. And what's great is I can just go back and forth between the two. Once Claude Code finishes task on one side, I talk to that one. I set that off, then I go to the other cursor. It really has for the first time ever unlocked the ability for me to build out multiple applications at the same time and do it really easily. So let's do this. I'll set up clawed code now to start executing on this project plan and then I'm going to show you how to go over to the background agent and get that set up. Okay. So here's what I'm going to say next. I edited the plan down a little bit to just four checkpoints. Feel free to start working on those checkpoints now. Let's review and test the application after each full checkpoint. So basically what I'm saying is this. Start working on each microtask under every checkpoint. Then once all the tasks are done, we'll meet up again. We'll do a little code review between me and you, the AI agent, and then we'll see how it turned out before we move on to the next checkpoint. It's literally like having your own employee. I was a manager of a technical consulting team before I started working on my application full-time, and this is exactly how you'd work with them. You're literally treating it like an employee. So, I'm going to hit enter on that, and it is going to start going to work on its own, which is really, really amazing. So, as this goes, before we

spin up our other background AI agents, which really unlocks the next level productivity, we're just going to get this base app set up. We're going to commit it to GitHub because the background agents require the code on GitHub, which show you how to do, very easy to do. We're just going to let it build out this initial part of the application first. So, as you see as it goes, it's just going down the checklist and setting up every part of the application, right? It has its own internal checklist here of everything it needs to install for the application, and it just goes through. That's where the autonomous part comes in. With the cursor agent, you really had to babysit it every step of the way. Here, it's able to think on its own, come up with its own step-by-step checklist, and go down those tasks by itself to get things set up, which, as you can see, it's setting the entire application up as we speak. All right. It looks like it finished setting up the initial version of the app. We're going to test this and then we're going to set up our background agents. So, let's test this out. We're going to hit yes on npm rundev, which will run the application. Okay, looks like the base tech of the app is all set up. Let's see the project plan real quick to see if they changed anything here. Okay, boom. Look, they checked off the project plan as I was going. Again, this is one of the really big strengths of Claude Code. Breaks it down to task and complete the task one by one. and it updated this checklist as it went concurrently as it was building the app. So now we're going to set up what makes this workflow really unique is the background agent. So I'm going to start setting up the background agents. Now one thing that's required for the background agents is the code being on GitHub. And so we got to make sure that's set up too. So one last thing we do before we set up the background agents and really hypercharge this workflows. We just got to get this code on GitHub. For the background agents and cursor to work, you need the code to be hosted on GitHub because that's where the background agents are going to commit the code to. So, I'm going to go on github. com. I'm going to click new repository. And we're going to name the repository notion C for clone. I'm going to make this private and we are going to create the repository. And then I'm going to copy this link. And we're going to say let's get this code into GitHub before we continue. please commit it to and then that link. And now what's going to happen here is Claude Code is going to commit this code to GitHub so that we can start working the background agents in the background. And so the way the background agents work is they take the task we give them, they start doing work and they just commit the code straight to GitHub. So they're going to be creating these plans for us in the background, uploading it to GitHub while Claude Code works here locally and builds out all our code for us. So, I'm going to say yes so that I can commit all the code to GitHub. And then we're going to get the background agents working, start doing some really, really cool things here. Okay, looks like the code is now on GitHub. It just went automated by itself. Claude code went took all the code, committed it for us, pushed it for us. Now, we can set up the background agents and start doing some cool things. This is where our workflow is going to be hypercharged because this is where we're now going to be setting up basically multiple employees that will be working for us autonomously. So the first one we're going to set up is going to be our marketing agent. This is going to be the one that's building our marketing plan for us. So let's take the instructions that Claude Code originally gave us for the marketing agent. Let's

click the background agents button up here in the top right. And this is going to set up our background agents for us. And what I'm going to say is this. You are a marketing research agent that will plan the marketing for our notion clone we are building. It is an advanced note-taking tool. Please create a new file for the marketing plan. Here are your instructions below. And then I pasted in the lines from the project plan that described what this marketing agent should be doing. So I'm going to hit start on this background agent with the plan inside of it. We're going to connect it to GitHub. You're just going to authenticate here. And boom. Just like that, the marketing agent is going. So now let's set up our other agents, right? So we have one agent working for us right now. Let's take the research agent. Let's copy the instructions for the research agent. We'll say you are a research agent for our notion clone. Please take the instructions below and do research for us in a new file. Boom. Give them that instructions. Now we have the next agent going. Now we have two agents going for us right now. Let's set up a third agent. So, not only do we have our Claude code going, building out the application for us as we speak, but now we have three research agents going in the background and building us out the road map, doing user research, and figuring out what we're going to work on next. So, let's take now the feature planning agent instructions. You are a feature planning agent. Please take the instructions below and make a new file with your research. Boom. I paste that in. Now we have three agents going for us at the same time doing research. One other thing here too is like these don't have to be research agents. I'm now just making them research agents because we're at the beginning of our project. But as we build this out and Claude Code builds out the plan for us, we can have these go out and having have them build out actual features for us. Right? So maybe we have claude code working on the main portion of our app and then we have each agent going out and building out a different feature for us. This is the real unlock for us. What we were doing before in cursor with just the AI agent on the right was we were doing one task at a time and we were babysitting the agent every step of the way right took up a ton of time was definitely faster than manually coding. But now what we're doing instead is we are not only having the most powerful AI agent in the world claude code handling our main application completely autonomously by itself. We are also spinning up new agents in with our background agents in cursor to do research to build out micro features and to basically work for us autonomously as well. Right now we have four employees at our little company working for us to build out this application. And the best part is I don't have to do any babysitting. The app is just running as we speak. So just to show you what's going on right now in the application, I'll pull open our local host. So right now it has the template for the application done. So welcome to notion clone, a modern note-taking application. This is what Claude Code has built so far for us. But now it's going in and it's working on checkpoint 2. I'm working on checkpoint 2 file and folder system. So it has the step by step of all the tasks it needs. Right now, it's working on the first task in the checkpoint, which is create the sidebar navigation. And it's just working on that autonomously. If we want to make it go in work by itself, I can just say yes and don't ask me again and just keep going, which I so far I've been building out multiple applications with cloud code. I've had no problem with it working autonomously. I don't need to babysit it. So, I just give it full permission to do whatever it needs. while it's building out those new components and it's building out that sidebar. If I click on the cloud again, you can see our three background agents and they actually each completed their task. So they all completed their research. So let's look at the feature planning research first. So here we go. A new file was created with a comprehensive feature planning research document. It has the MVP framework, the road map, the feature spec specification best practices, implementation recommendations. Let's see what it did here. So again, this was all in the background while we've been building the app with claw code. So it has all the core features, the engagement features, has the user onboarding, search functionality, has retention features detailed out priority. So it has the it prioritizes all the features in here, which is amazing. It has priority levels. Look at this. So it has the road map for months one to three, four to six, seven to nine. It has an entire year-long road map built out. It has all its personas. So secondary personas, primary personas. As you can see here, this is a pretty indepth feature planning report it built here. So let's move on to our other agent. What did our other agent do while Claude code's building out the rest of our application? You can see the code just running here. The agent is running and building out the components. It really is amazing. It feels like I have my senior developer. I like to think about it like this. Claude Code is kind of my senior developer, maybe even VP of technology, my CTO. And then I have my agents, which are like my interns that are going out doing research. Maybe they're a development intern. They're going out building a little feature. I have an entire team of AI here working for me. So, let's look at the marketing plan it built for us. So, it has an entire marketing plan. So, it has competitor analysis of Notion, of Obsidian, of Evernote, of Microsoft OneNote. Uh, it has marketing gaps we'll be tackling. It does pricing research, knowledge professional. It has all these personas it went after. This is pretty amazing. And now what we'll be able to do after Claude Code has gone out and built out the application, we can just feed it the research from the other agents and have it build out the improvements. This is the incredible productivity unlock, right? Because clawed code is so powerful and able to handle huge tremendous tasks by itself, multi-step layered tasks without any oversight at all. We can spend our time working with simpler background agents, which is the newest feature in cursor to give them microtasks to go out and do that might be unrelated to what claude code is doing. If I wanted to do this before in cursor, here's what I'd have to do. First, I'd have to babysit this AI agent on the right. I have to say something like, "Build out the navigation sidebar. " Right? One feature at a time. Break it down to I'd hit send. I'd sit here and watch it and babysit it to make sure it built out that navigation sidebar. Then, if I have a little break in between, I would then go over to chat GPT. And here's where I do research. Now, I would sit here and engage with Chad GBT and say, "Hey, please do research on a notion clone, right? And then I'd hit send on that and I'd sit here and I'd have to babysit this chat while I go back and forth and it asks me questions and it does searches all while I'm doing this. Nothing's taking place in my app, right? There's nothing going on with the agent inside cursor because I'm not babysitting that. Everything is very engaged. " And while this workflow was revolutionary 6 months ago, it's not revolutionary anymore. Now we have way more agentic abilities with clawed code where we can give tasks and walk away. Then we can spend time with our other agentic abilities with these background agents. Give tasks and walk away. Now instead of handholding interns, this is more like managing an entire company of AI agents, which is an entirely new paradigm of building applications. All while I was talking to you right there. Claude code was building and now boom. Look at that. The sidebar with the folder structure and the files are good to go. That was all done while I was talking to you. I didn't do anything any babysitting, no coding while we were talking. That is all good to go. And that's kind of the power of this new workflow is not only one taking advantage of the power of cloud code and the fact that it can do so many things by itself, but two the fact that now we're layering on top of that the background AI agents and cursor to do research and do planning while claude code is going. This is the workflow that has unlocked so much for me. Cuz now imagine this right now. Imagine what I've currently been doing over the last couple weeks of using this workflow is I can now manage having multiple cursors open and have it building out my app creator buddy, having it built out my next app work buddy, and have multiple apps going at the same time in cursor. And the entire time while I'm having multiple apps built out, I'm able to walk away and do whatever I want while I not only have Claude code going and building out my app, but background agents building out micro features, building out research, doing marketing, planning for me, and all that. Now, if I wanted to have a background agent going, I can even say, "Write me announcement tweets for the launch. " I have that agent going now. So, I have these four agents now working for me. I'm going to have this build out my announcement tweets for the launch. I come back in here and I go, okay, continue on to checkpoint three. And now it's going building out the next section apps. So this is the shift from babysitting an agent to build it step by step and shifting more to a managerial role where we're now managing agents to build out the application for us which one saves us time so I don't have to sit here and babysit an agent writing code and two allows us to 10x our productivity because claude code's doing way more work for us and then we have background agents doing microtasks as we go along as well. This is by far the most powerful workflow I've ever seen with an AI code builder. Period. It's better than anything I've done with cursor. Windsurf. It's better than anything I've done with just straight up clawed code by itself. If you can combine clawed code with background agents and cursor like I just showed you, you're an absolute animal and you are going to get incredible things done. And the best part is if you take advantage of this right now, you're going to be so far ahead of your competition. Your competition is just hearing about Claude Code. You're moving on to the next step, which is using Claude Code and adding background agents on top of it. If you do that, you're in the top 1% of the top 1%. If you learned anything in this video, leave a subscribe, leave a like. All I do is make amazing videos on how to build with AI. I'm going to make way more videos on this workflow. I think there's so many different things you can do here. and I'll see you in the next video.