INTERACTIVE CLAUDE FULL COURSE 4 HOURS: Build & Sell (2026)

Welcome to the most comprehensive course about using Claude designed to take someone who's never even touched it before and then turn them into an expert who's running their own AI operating system with a full AI workforce. And we'll do it through 10 levels. Each one builds on the last. And as you graduate through them, you'll go from typing your first prompt to having your full team of AI agents replacing the work you or your team hates to do every day. And even if you're using Claude right now, most people are using only about 10% of what it can actually do, while others are literally running their entire business operations with them. They're landing clients and building income streams they didn't have before. And that's the gap we're closing today.

>> I will say I was so happy I found you. >> But I think the a lot of people are leaving the beginners. And so how you broke it down, that's why I kind of joined you was simply because you made it so simple for the everyday person to be able to understand it and how to be able to function and use it. That's why I'm where I'm at literally is simply because of what you've what your video.

>> These are the exact systems that I use to automate my business, eliminate all the busy work, and grow my income using Claude without having to hire a large team. And I'll be showing you not just how to use them for your own work, but also how others and you can build, package, and sell these solutions, too.

And how this is going to work is we're going to be learning through the story of Alex. Alex starts exactly where you might be right now. Too much to do, not enough hours in the day, and feeling like there might be a better way. Level by level, you'll watch him figure it all out. And by the end, he has automated most of the work for his job, built a side business, selling these systems to other companies, and has a full AI operating system with agents working around the clock for him. So, every concept becomes real through his story. And everything Alex builds, we're going to be building together step by step. I've actually spent weeks designing this course and broke it down into 10 levels. And by the time you finish this, Claude won't be something you open up when you need help writing an email. You'll have multiple agents running their own system controllable from your phone – and go from I use Claude sometimes to Claude runs my operations for me. And if you know some of this already, use the chapters below to jump around. But I'd still start from the beginning since there's still a lot of skills, hacks, and Easter eggs that I've hidden along the way. That's going to compound as you go and become priceless in the right hands. But the biggest thing is I highly encourage you build alongside me. That's how you're going to get the actual results. So, with that out of the way, let's get into it. So, most people actually open Claude, they type a message, they get an answer, and close the tab. That's it. That's their whole relationship with the most powerful AI on the planet. And that's maybe about 1% of what Claude is good for. But even if you use it this way, there are some fundamental things you should know before we get into the really fun stuff. So, let me walk you through the basics really quickly. So, most people actually use Claude through the web with claw.ai. So, when you log in and normally make an account, this is what you're going to see. And this is the web version. It's free to start. It works. This is where everyone begins. But what I want you to focus on is look at this text box in the bottom right here. This is where the fundamentals are. And I want you to understand before we move any further. And that's the model selector. So, if I drop it down, you can see there's three different models. And for these three different models, think of it like three different brains built for different jobs. First of all, there's Opus. And Opus is the specialist. It's more capable. It's good for deep reasoning and hard problems. So stuff like coding, building things, thinking through big plans. And this uses a lot more tokens and a lot more expensive because it's also significantly slower than the rest. Okay. The next is sonnet. And sonnet is basically your daily driver. This is where you'll probably be doing most of your tasks. And 90% of the things run here. it it's normally the most efficient model when it comes to speed versus intelligence. And lastly, there's Haiku. We don't use this that often, but for any task that where you need quick answers or a lot of bulk operations, you can do with Haiku super cheaply and super fast, right? This is not a heavy thinker, but is just for snappy judgments. Okay. And overall, my default rule is use sonnet, escalate to opus when sonnet answers feel a little shallow, and drop to haiku if you want to do quick stuff or something in bulk. And you see this last toggle right here, extended thinking. Flip this on when you want clot to actually think step by step through any decisions you're trying to make or build something out. It's going to pause. It's going to ruminate on each answer it comes up with. And it is also a lot more expensive and takes a lot more time. If you don't have the max plans, don't use this liberally, but for for normal everyday tasks, Sonnet should be good enough. That's about it for these three models. Now, let's move on to the next part.

All right, so the next thing I want to show you that's really going to speed up what you're going to be doing if you're just getting started is this box and this button right here. And what this is is this is the voice mode. And the voice mode is really great because what it allows you to do is just speak to Claude and Claude will be recording what I'm saying. And I use voice all the time because it's significantly faster than me having to type it out and I can think a lot more clear on paper. And one thing you're going to see me do throughout this whole course is use this app called Whisper Flow. It basically lives in the background on my computer. I hold a key and I talk and it drops whatever I say in here. And that helps me, you know, write prompts a lot faster and get it to do things. And it doesn't matter where I am, like if I'm in a dock, uh if I'm in the terminal, if I'm using cloud or here, voice overflow or these voice tools, I 100% recommend using these because using voice is significantly better than actually typing everything out. All right? And before we start becoming super users of Claude, you need to understand how the plans and how Claude builds you. And if you go to claude.ai/upgrade, then you're going to see Claude will have these two plans. All right. So, they have the pro plan and then the max plan. How this works is basically Claude charges you a subscription and it's not pay. You pay a flat monthly rate and you get unlimited usage as long as you pace it out. So, how this works is basically there's a rolling limit that resets every few hours or every week. So, you can't blast through a month's worth of tokens in one Sunday afternoon. But, if you space your usage out, you're basically can use Claude unlimited or just for a flat fee. For me, if you upgrade here to the $100, they give you another tier with the 200. That's where I'm at. With the $200 plan, if you pace yourself, you're pulling in from um what I've been doing, the API equivalent of, let's say, $3 to $5,000 a month on what I'm just paying Claude $200 a month for. That's a really good deal for me. So, I take that. And quick thing on that, if you actually go to platform.claude.com/usage, Claude, you can see your APIs usage. This is pay as you go, token by token. You don't need a subscription for this, but I'll save you the trouble. This gets expensive super fast. So, for normal everyday use of Claude, the subscription is almost the right move. The API is for developers wiring Claude into software. So, for right now, maybe start with the pro version and move up when you feel like you're hitting the ceiling. Pace out your usage. Don't cram everything. And don't pay for a tier that you're not going to be using yet. All right.

All right. So, the next most important thing you're going to do is make sure we have the Claw desktop app downloaded. Okay? And to do that, in the bottom left, you see there's a little download icon. If I hit that and then hit download for Mac OS, we're going to hit this button. All right? And then save this. Quick note as you do that, make sure your computer Mac or Windows is updated to the latest OS version. Cloud desktop app moves fast and the newest features only show up if your machine is current. If you're on something from five years ago that hasn't been updated in a while, update it now. You don't need a brand new computer, just a recentish one that's up to date. Right. And right here, I'm going to open this up on Mac. And to open this up, I just hit what got downloaded. And now it wants me to drag into my applications. And I'm just going to replace it just so you guys understand. And 80% of this course is going to run through this desktop app, not the browser. All right. The desktop app is where co-work lives. It's where cloud code lives. It's where all the power tools that show up in level two and beyond are going to live. The browser is fine to start, but you're graduating right now. Congrats. Download this, install it, open it, and log in with the same account. All right.

So once you download the desktop app, this is what you're going to see. And Claude actually changed around their UI a bit. So I'm going to tell you what to look out for. All right. So on the top left right here, if you click this and expand this, you'll see a couple different things. So right here on the left, this is now where we switch from claude to cloud co-work to cloud code. Okay, we'll be diving into this much later in the course, but I just make sure you know. And some things in the new new UI they changed is I want to make sure you know where the settings are. And the settings on is on the bottom left. If you hit your name and then hit settings and then I want you to go in here. First thing I'm going to show you is if you hit usage, if this like for example, if this blue bar is filled out all the way, you can't use Claude. They're going to put you on a timeout, right? So make sure you have the right plan and see where you're hitting these limits. Number two is if we go to capabilities, I want to make sure you turn on memory right here, these two. And what this is going to do is basically let Claude remember things about you across all your different conversations you have. So your preferences, your ongoing projects, and how you like things done overall, which means you stop repeating yourself each time you open a new chat. And we're going to be leaning on this hard. All right. And then if you scroll down a bit more, make sure you check on artifacts. All of these. And artifacts are what Claude uses to build out documents, decks, diagrams, interactive tools on a side panel instead of dumping text on. Okay, now you have both of those done by hitting new chat. All right, so we're back here and we're set up. Awesome.

All right, our initial setup is done. Now let's go through and learn what Alex has to do. So let's say Alex runs marketing at a SAS company. It's a small team, so Alex has to do a lot of different things in the company. And the problem is that the CEO just told Alex, "Hey, we're spending too much on paid ads. Pull together the profit and loss and marketing spend and come back tomorrow morning with a presentation on where to allocate the money and also find out how we can make more money." All right, so tomorrow morning is not a lot of time and there's data all over the place. There's credit card statements, there's platform receipts, there's revenue coming in from five different places, subscriptions, annual plans, one-off setups, all these. And the CEO expects a deck to match the brand guidelines. Probably Alex needs to pull all of this together, find where the money is going, figure out what's actually working, and build a presentation by tomorrow morning. The end goal is clear, a clear profit and loss breakdown, probably like a pie chart or something to show where the money is going and build a presentation and a deck to recommend what to do next. Cool.

Now, Alex is going to do this the right way. And to do that, what he's going to do is set up a project and then load in the context and then start working from there. So, let's start doing that. So, now to set up project, what I'm going to do is I'm going to open the sidebar and then hit project right here. And then I'm going to hit new project. Cool. Let's call this PNL for boss. And let's say I'm trying to create a P&L and see where we should allocate our budget for let's say the next quarter. And simply we're going to create the project. So there's a couple of things I want to show you here so you understand what's happening. All right. And first of all, if you look here, there is an instructions field. So if I open that up right now, what we're going to do is include the right instructions. So Claude has the context, right? And every project has a custom instructions field. This is what Claude calls a system prompt. And a quick explanation is a system prompt is a set of instructions that gets inserted before every single message we're going to be sending to Claude. Every time we type something into Claude inside the project, this system prompt will go first before Claude even sees what we're sending it. All right, this basically is there to set the tone, the rules, and the role before anything happens. So, when Alex types a question, Claude actually sees the system prompt first, then Alex's message. This is why this is so powerful. You write it once and it shapes every conversation automatically. You don't have to keep saying be direct or use bullet points. It's already there. So for Alex, let's write one. Hey, I'm the marketing manager at a B2B SAS company. I need sharp data first answers. Lead with num the number of fightings and then the reasoning in your responses. Use bullet points, not long paragraphs. And when you give me recommendations, give me the one you actually believe and do not hedge. Make visualizations as much as you can to explain what you need to match our brand voice. Be direct, confident, and don't add any fluff. So, this was just a basic one that I added. Now, every conversation Alex starts inside this project, Claude is going to see this and have this context loaded and it'll know how I like my responses. So, I encourage you to go try it out for yourself and then come back. Okay?

To make this a bit easier for the early levels, what I'm going to show you is let's imagine Alex has compiled all these documents that he may or may not need to compile his report. So, I'm going to take all these documents and drag it here. And you can see over time all these different types of files are being ingested by Claude. Great. So, Claude now has access to use these files when I'm going to be talking to it. Okay. So now that the project is created and the instructions have been loaded and the files have been uploaded and our memory has also been turned on, Claude now knows our company, the voice, the numbers, and what we probably want. Now Alex would be ready to build a thing like that P&L document. So let's try and do that. Hey, I just uploaded all of our credit card statements, ad platform receipts, revenue data. What I want you to do is break down the profit and loss for me. Please show me two pie charts maybe one for revenue by source and one for expenses by our channels. Please and one thing I want to show you is you see right here it's selected as opus but I know that I don't need opus right now. So I'm going to hit it. I'm going to hit go and then I will wait. And you can see clot just working. It's reading the ad spend data reading the meta ads all this stuff. And I'll just let it do its thing.

Okay, look at that. Now I can see two interactive pie charts. This is really cool. And this is what a new feature Claude just introduced in their chat function where it can actually build these visualizations in for you. Right? So anything you need to visualize or understand. Claude chat is a really good place that you can be using and I use these in my YouTube videos a lot. Great. So now just visualizing this data we can see that most of our money is being spent on meta ads and most of our money is being made from this monthly subscriptions. Cool. So now I'd assume the next logical question is to see if hey are these meta ads really worth it like the money we're spending and what are we getting per lead. So if I ask it, hey, based on all this data, where are we right now? What's actually working and what's not? I want to know which channels per marketing spend is bringing in the most leads so we can double down on that channel. Can you build me some visualizations for that? Great. So I just asked it to do that and I'll let it load. And what I want you to understand is to get in this mode of thinking when you're just in level one or two of how you can just interact with the chat, make it quickly inside the chat, make visualizations for you, read a lot of data, synthesize it, and give you insights to save you a bunch of time. So, we just spit it back out and I can very clearly see now how we're spending so much on meta ads 27k but getting back 1,200 whereas on stuff like YouTube we're spending 1350 and getting back 9600 and blog and Instagram all these and which means essentially it's telling me what to do to double down on Instagram organic YouTube blogs and what to cut. This is really interesting. So now I think because AI has access to analyze all that information, what I can do is make sure now AI can take all this data and then make me a presentation. Okay. So now with all this data, can you build me a presentation and an artifact based on our brand guidelines? You can look it up from bookend.ai. cover all the profit and loss, all the key findings, the recommendations, and where to shift our budget to, make it clean. This is going to be presented to the CEO tomorrow. All right, great. And I'm going to hit enter.

Awesome. So, it just got done with these slides, and I really like the color. It matched my brand. And we can see it made a really nice agenda where we're seeing a snapshot of all the stuff, which ones are working, key findings, and it's really made a very nice summary and a story we can tell our boss. Great. So now, Alex doesn't have to stress this just got done really quickly. And let's say he would want to present it. If I just hit this Google Docs button, then I can see right here, if I pull it here, it just pulls up my browser. I can make edits to it if I want to. And it already made and put this slideshow in Google Slides. So, it's really, really simple for Alex to present. So, this is one good thing that this claude chat does, which is create these type of artifacts. And these artifacts are really really useful for building stuff like dashboards, charts, diagrams, any interactive stuff. So for example, let's say we want to like kick it up a notch. What we can do is build maybe an interactive tool that we can play with and present rather than just having a slideshow. And I'll show you how to do that. So can you build me an interactive budget real allocation tool that I can share with my CEO? Maybe it's going to show our Q1 spending by channel versus the recommended reallocation. Let the user drag sliders to adjust the budget split between um paid ads and organic and show in real time the changes that are projected to bring in leads, conversions, and any cost per acquisition based on our quarter 1 performance data. Make it match our brand guidelines. and the CEO should be able to play with these numbers on the widget. Okay, great. So one more thing I want to say is sonnet is good for this but right now this is one big thing I need to make you understand and this is a limitation of the chat feature where now let's say I wanted to go to opus or haiku I can't do that because I'm talking with this model directly but to do that we're going to be moving and showing you in level two how that becomes really useful. So, I'm going to for this task because it's a big task and it needs to code and do all of that stuff. I'm going to turn on extended thinking and then I'm going to hit go and then wait for this to get done.

All right. So, now I see it built this interactive tool with all these sliders and what we can potentially project. And if I want to share this with let's say my CEO, I'm going to go up here and I'm going to save as artifact. Okay. And it just says save this visual earlier as an artifact. And it runs a new prompt. So I'll wait for that. Awesome. So it's made it as an artifact. Now I can play with it and see it right here. But what I want to show you is if I want to share this with the boss or anyone else, I'm going to go to the top right. I'm going to hit publish artifact. And then you're going to see right here you can publish to the web or just share with your team. And this makes it really easy for you to essentially build out really cool things in Claude and then actually share this with people without having to understand or know anything. So these are just small widgets and tools. And right here I just copied that and I pasted it. And right here you can see there is a clot artifact that anyone can play with. You can see you can adjust these sliders and it's all good. So, this is an really cool tool and this is kind of 90% of what the claude chat function does. And to show you a cool use case that I've been using it personally a lot for. So, in my claude club, if I go to classroom and one of these, you'll see right here, I'll have a bunch of these guides and which are essentially cla artifacts that I've made for my community members so they can follow an interactive plan and do things in a really cool way.

Okay, so with that out of the way, now Alex, let's say presents this to his boss and this is what happens next. So Alex presented the deck to the CEO and he loved it. But now the ask is, "Hey, can you do this for me every week? I want a weekly marketing performance report on my desk every Monday." The thing is, Alex knows how to do it. Level one, prove that. But think about what Alex has to do every single week. The first problem is you have to go find your credit card statements, download them, and then go find the ad platforms. all this data that you have to gather from places. You have to drag that all into Claude. That took like 20 to 30 minutes and it's going to take him a lot of time every week. Problem two is it's the same prompt he has to run every time like break down the P&L, cross reference the CRM data, and then build me a branded deck. All these things over and over. We need to fix both these problems. And to do that, we're going to be graduating into something called Claude Co-work. And congrats, you're moving on to level two. Great. So, let's get into that. Quick note before we dive in. I want to make sure you know that co-work will require the claw desktop app. Okay? If you're all if you're still on the web at claw.ai, you won't see this right here. You need the desktop app. And your Mac and your Windows need to be on a recent version. So, if you followed level one and download the desktop app, you should be good. Also, co-work is only available for the pro and max plans. All right, if you're on the free plan, this is where you'll want to upgrade to at least the $20 a month pro plan, and you'll probably need it from here on out.

All right, in level one, you use chat. Chat is great for quick stuff like asking a question, getting an answer, building one-off artifacts like that. But for real recurring work, like Alex's weekly report, that chat kind of hits a ceiling. So, for example, like in chat, we were managing everything like we were like do this and then we wait and then do that and then we wait. Co-work is the next level from chat. It's fundamentally different too where claude in chat it's responding with these artifacts inside embedded into the chat. Co-work doesn't do that. But what it does is let's say you describe an end result or a whole workflow. Co-work will plan every step, execute them in a sequence and use multiple models. So not just like it's not just this sonnet right here. It can use multiple models and can deliver the finished output. And one major difference of co-work is that it also runs on a virtual machine which is a sandboxed mini computer inside your computer. So cloud co-work will use that virtual machine to get its tasks done which means it also has a temporary file system where it can store some stuff and and in a limited capacity it can also interact with your own desktop computer. It's a real powerhouse for professional level work because it's pretty safe. It's wellguarded and gives claw the ability to actually interact with external services and the internet with any app you can really think of. And to get to co-work all you have to do is if you look up here there's chat co-work. You just need to hit that and you should be good to go. And you can see you can also if you click this button right here you can hit project create a new project and you can import your projects from before. So remember PNL boss we just worked on. we can create that project and it'll have the memory and everything preconfigured.

Okay, now that you're in co-work, I want to teach you a framework so that anything you do with co-work becomes incredibly effective. And I call it the GCPS framework. And every piece of real work that's going to be going through Claude will follow these four steps. And we're going to use this framework to solve both of the problems I just described for Alex. So first, we need to gather all our context. So, we want to make sure we use connectors to gather stuff we did in level one. So, what Alex did by hand, like downloading the files, dragging them in, and we're going to fix this to make sure Claude knows what to pull on its own. Then, we want to make sure we contextualize. And the way we do that is with projects. And project is going to give Claude the framing and your system prompt as well. It's going to give it the memory and all the instructions you need. We already did this in level one. And right now we're just going to use the same project. The same idea we did in the cloud chat but just now in co-work. Okay. Then we move on to P which is the process. Now this is where Claude actually does the actual work. It's going to be analyzing all the docs. It's going to be building and comparing. It's going to be building the pie charts the decks all those things. So in the chat where we managed every steps co-work is going to be running the entire whole process autonomously. And after we make a really good process, then we want to make sure we standardize that. That means what we're going to do is we're going to make a skill and package that output so it's in that format and in the right place and repeatable. Then we're going to be setting it on a schedule and when the skills come in, the reports are always going to be looking the same and it's going to be done without Alex have to reexplain the format. It's going to just happen on its own. All right. And then we can also control it with our phone with this thing called dispatch. Awesome. So we're going to be learning all of that. And that's GCPS, gather, contextualize, process, and standardize. We're going to be doing all those steps all together.

So let's start with G. The first step G, gather starts with the concept of connectors. And if you don't know what connectors are, connectors are useful for this exact problem in the gather step. So remember all those files we initially downloaded and Alex had to drag in and compile together like credit card statements, emails, um everything lived in different places like Google Drive, Gmail, HubSpot or your CRM, all those different places. The concept of connectors fixes that because Claude before this could not access your Gmail, your drive, etc. What connectors do is it gives Claude the permission and a key to go enter your Gmail direct and go inside your apps and once you're connected Claude can reach in and pull what it needs and they're built on something called the MCP protocol. That's model context protocol and that is something you'll be hearing interchangeably as you kind of learn more and more about connectors and MCPS etc. So you technically don't need to know how this works under the hood at all. Just know that this is a secure standard process that Claude uses to talk to external applications which means your existing permissions like read and write are respected and Claude can only see what you want it to see. And setting this up takes 30 seconds max. So let me show you how that's done.

Okay. So, imagine I'm Alex and this is what his team's Google Drive looks like. Let's say there's a marketing folder, finances folder, everything there. And each week the team dumps their receipts, invoices into a random folder or whatnot. It doesn't have to be organized. Then, let's say there's website data, something like in Post Hog. And of course, there's a CRM. Let's say we're using HubSpot, right? This is storing every lead that comes in, where they came from, where they are in the funnel, and if they converted or not. Getting the data from all these sources gets really, really messy. So, that's really where connectors change the game. And instead of Alex having to go and dig through the folders, the files, all these things, what we're going to do is we're going to connect a connector. And let me show you how. So, first what you want to do is make sure you're in the right project, which for us we were using PNL for boss and hit that. Okay. Then what we want to do is make sure we hit this plus, go to connectors, hit manage connectors, and then right here on the top, hit this plus button and browse connectors. So, we have to look for our files in a couple different places. So you can see right here you can scroll and there's hundreds of connectors and these are the official Anthropic partners. And I'll show you how to connect the external really really cool other connectors later in this video. But right now what I'm going to show you is let's say let's connect our Google Drive. So all we're going to do is hit this plus button and then Claude will pop out with a window and I'm just going to select my account and authenticate that way. Great. So you see now it's connected to Google Drive. That's number one. And then ideally I want to connect to Post Hog. So right here I'm going to hit the same thing and do it again. So, Postthog and right now I'm just clicking through and making sure and I'm just again authenticating with Post Hog. Great. So now Post Hog is configured. Now, the last one we need to configure is if I just go back to my Claude app. Now, I need to go to HubSpot right here. I'm going to hit plus and I'm going to connect my account for Booked In. So, remember, Alex works on Bookedin and then connect my app. Great. So, all of that is done. And now if you go back to this new session button and right here if we select my project again just to make it easier what I'm going to do is hit this plus go to connectors and you see right now HubSpot is turned on and I can turn on Google Drive so I can search my Google Drive and also Post Hog is turned on. Great. So to gather all our folders what we're going to do now is make sure we give Claude the right context. All right. So remember, we're back in co-work and we have the project selected. So I'm just going to be speaking to it right now. So hey, every week my team dumps their expenses, uh, expense receipts and invoices into our finance folder in Google Drive. Our ad platform reports and are in marketing and in Post Hog that has all our website analytics. Also everything you need is in HubSpot. So for our CRM data, what I need is like the presentation we just made and we built all those graphs and whatnot. I need to build the same P&L charts, the presentations like we did last time, the revenues versus the expenses and then the channel performance, the whole breakdown, right? Can you pull everything you need to do that for me? Great. So I just entered this. Now I'm going to hit let's go. And now I'll wait.

And so you see co-work will look a little different where co-work will have three different things where first it has this progress and what it does is claude will literally make a plan into what you do. Number two is it also has working folders. So you can see it working with this folder. There's some assets right here and it's reading all the summary whatnot. And it also can take apart all this context and use those files to overall make decisions over time. So this is a lot more powerful than just the cloud chat. And I recommend as you're working on anything professional related, use claude co-work. All right. So I'm going to let this load. It's going to take a little bit and I'll come back to you after it's done. And one more cool thing about cloud co-work I want to show you is you can't do this in chat but let's say in the middle of while it's processing what you can do is actually steer claude as you see it working right so you could just be like hey only worked for last week folder something like that and you can ceue up these messages so that's something really cool you can do in co-work as well all right so I see it got done and you can see after it gets done this is very important important is you can actually see the files it's been working with. So it has these decks. So right here, if we open this up, for example, it has this deck that's been working with. That's really cool. And then you see to follow the guidelines, it has, you know, all our colors and all these things. And, you know, we can know that it's been working with Google Drive, etc. Okay.

So based on all of this stuff now we have gathered all the right context. Okay. Now we know that from all of these documents we have given it the right context and it's also contextualized into what it needs to be doing and what data it's been trying to make. So what I all I wanted it to do was basically have the insight to essentially build those graphics. And right now we can see all these things got pulled. Now next step is let's say what we want is for it to make a deck. Right? Okay. What I want you to do now is can you make me a presentation deck? And this is for next week. So I should have some data in my drive and whatnot on all these sources that you've just used for the next week. I want you to build me a performance deck. And right here, this is very important. Uh right here, I'm going to take this. Remember, uh the first deck it generated. I'm going to take that and I'm going to go back to Claude and I'm going to paste this link in. This is the process I want you to follow. Make sure the deck always looks like this. And can you make me a new deck for the next or the last week? Right. So now that we have gathered all the data and we've given in the context of what we wanted to do, now what we're trying to do is get it to process all this context that's gathered and then make us the correct assets. All right. So I'll let this work for a bit and then I'll come back to you once this is done.

So one thing I want to show you and this is the reality of just working with AI. your first prompt isn't going to give you something perfect. So, if you see right here, uh what it gave me is not something that's the best. It's not, you know, uh following all the standards of what I would like, not like the old presentation. So, what I did was right now I just added another prompt. I said, "Hey, here I attached my slides that I really liked." And then I said make sure the output is follows the same structure has the agenda and all the same similar slides. The reason I did that and I wanted to show you is that it's okay to go back and forth until you get a good output that you like. All right, don't be afraid to do that. But this is the first time we want to make sure we do that end to end and then I'm going to show you how to make that into a standardized process. Okay, so this just got done and let me open it up right now. Uh, let me just reset the window and okay, this is pretty solid. Great. It has everything that we did and it's followed a very very nice structure, right? And we can see how Claude pulled the data from drive, Post Hog, uh, HubSpot, analyzed all of this stuff and built the pie charts, the B, uh, the bar graphs, all these. Okay, this is awesome. The deck looks pretty good. But here's the problem. Next week, we have to do this all over again. and the week after that and you know as it goes. Are we going to keep reexplaining to Claude every time and waste our time? No. We're going to make a skill so that we can get consistent output over and over again. But what is a skill? Right? So, let me show you. If you head over to customize right here and right here, you see where it says skill. Claude actually comes pre-built with a couple different skills. So the first one is the skill creator and you can see in the skill creator it is basically a text file and you can see there's also like some code in it and you don't have to understand what any of this does because the skill is a package of instructions. It's telling to use these Python files this code so we can get consistent deterministic output. All right. And that's what a skill is. It's a written down process. It's literally like a recipe. to like this one. The markdown file is a recipe card, the steps, the orders, all the measurements. Imagine like a kitchen. You can also reference different tools and whatnot. So, the connectors we just used, right? Okay. So, hopefully that makes sense on what a skill is.

Now, let's think about what we kind of did and break it down into a skill. And the way we need to do that is first go back to our GCPS process. So first of all to make this output what we needed to do is make sure we connected drive HubSpot post hoc all these right those are kind of the ingredients where claude is going to be pulling from we need to tell it to do that each time because there's going to be new reports and keep updating rather than having old stale data. Okay. And what we did is then we contextualized and processed that data that it just gathered. So we want to pull the latest financial data and then the ad data and also the CRM data. Then what we did is we processed that and made it into like for example we made it into some bar charts and some pie charts etc. And we want this exact output each and every time right so for example right here there is executive summary an agenda and then the P&L snapshot. So all these slides then there's also the brand colors. You see these are pretty consistent and we like that. All right. And because th that's the process we want. What we now do is just this. Now we claw to use that skill creator to make a skill. And how we do that is very simple. So and notice how I have opus selected because opus is a really good task for this. Okay. We just spent some time building a deck. What I want you to do is turn this into a repeatable skill that I can run every week with one command. And here's what the skill should do. First, we want to gather. So, we want to connect to our Google Drive, pull the latest weekly receipts from our last week, and then the credit card statements, uh the finance, the ad platforms, the post hog, all the stuff you just did. Make sure you gather that. Then what I want you to do is categorize all these expenses by channels. So where they came from. So paid ads, um, organic, all these things, same as what we just did. Then create the bar bar graphs and pie charts, whatever we need. And then like follow the same structure for the slides. And that is the output that I'm looking for. So look at all the slides how there's an executive um summary, there's an agenda, the P&L snapshot. The same way what I want you to do is follow this and create a skill out of this. So every week if I say something like uh weekly marketing report, you will generate this. All right? And you can see that was like a pretty beefy prompt. And of course you have to go back and forth with it, but this is just my first attempt. What I'm going to do is one more thing. Hey, if I tr maybe say slashweekly marketing report, I want you to run this thing autonomously the whole process. Okay, now go make me the skill. Great. That's what I just said. Now, that was a beefy prompt I entered with Opus because this is a very logical process and it's a hard thing to do. That's why I didn't use sonnet and I'm using opus to make the skill so I don't have to do it every time. Now you'll see after this is done, I'll demo how to use the skill.

Okay, so it just got done and you can see if you really want to, we can view the skills file and you can see there's a bunch of stuff. Um, and we can just save a skill. Now, if we let's say we're Alex and it's next week. If I just do /weekly marketing report and hit enter, what it's going to do is literally create a new presentation without us having to do all that back and forth. That's really cool, right? We just packaged everything up. And I'll show you the output once it's done. And one more thing is while it runs, I don't actually have to write doly marketing report. That's pretty cool. I can just be like, "Make me a weekly marketing report." And Claude is smart enough to go search my skills database and then run it on its own. All right, so let's check the output.

Okay, awesome. So we just ran one command and Claude gathered all the resources from spread out from different apps, processed everything and then built out all the charts that we were after. This is the same quality we got 5 minutes

ago when we walked through that entire manual process. Now it's just one command, right? That's the last S in GCPS.

Okay, so this is awesome because now we have what we want standardized and now we can share with our teammates and it our job to create these decks every week just became a whole lot easier. But we haven't even put it on a schedule yet. That's level three and let's start there.

Congrats, you did it. You built with the GCPS framework. You connected claw to your tools and even built a skill that creates branded weekly reports with one slash command. But the thing is, you're still the trigger. Every Monday, if you were Alex, you'd have to open up your laptop, open co-work, and type /w weekly marketing report and then wait. The work is automated, but starting the work isn't. And level three is going to fix that. And by the end of this level, Alex won't ever need to have to be at the desk again for work to get done.

So, let's get into that. So, let's say Alex is at a restaurant grabbing lunch and his phone buzzes. it's the head of sales and says, "Hey, we got a call with Acme Corp in 2 hours. Can you pull up their engagement data that we had with them and pull together a quick brief, please?" But, you know, the problem is Alex isn't at his desk and the laptop is open at the office. The connectors are on that laptop and the files are on that laptop. Everything is tied to that one machine.

This is why we need to be using Dispatch. Dispatch connects your phone to your desktop cla and what you can do is you can send it a task from the claude mobile app and it runs on your desktop using all the files, the connectors, skills and projects. So you can come back to finished work. You use your phone to send the instructions and your desktop does all the heavy lifting.

Now let's set it up. All right. And to set it up, what we're going to need to do is first get your phone and make sure you have the claw app and then go to dispatch. And you see it says dispatch from anywhere. dispatch all the stuff. So, let's get started. And then what you need to do is make sure you scan this. Okay. So, I'm going to scan this on my phone. Okay.

So, after you scan it, make sure you check on keep this computer awake and also let it give the ability to control your computer and in that way it can basically do or take any actions across your computer. All right. So I'm giving it all these access and then using my password to make sure now I can quit and reopen claude. And just to make it a little smoother, what you want to do is come here and then say bypass permissions so it doesn't ask you for permissions. Okay. So right here I got connected. Now you see it's the same message. And now if I just type hi, you can see the hi appears right there. And it has access to the same cloud sessions that I've been working on and we can actually start doing things.

So right here, I'm going to make sure I check allow all browser accents so it can control my browser. And like let's say I am Alex and I can do something like slash weekly review. Oh, whoops. My camera died. But now what I can do is, you see, I just asked, can you run skills? And it has access to all these skills. Now, uh, if I'm feeling a little funny, what I can do is just drop a voice note. Like, can you run the weekly marketing report, please? And you see, I'm just running. Oh, it says drop a voice note. But yeah, that's about it. Now, I can also send voice memos to my phone and it'll run these marketing reports. And that's as simple as that.

Now, if Alex is, for example, let's say somewhere else completely and he needs to run this weekly marketing report, he can just talk to his phone, make sure he has his computer on in his work computer at least. If it's a Mac Mini, that's why it's really, really useful because it's always on. And then it's going to keep running and when it's done, it's just going to send me uh a report. Great. So, right now, it's even giving me updates on what it's doing. And uh it's also, you know, just ruminating. So I'll come back once that's done.

All right. So you see it got done. And this is kind of cool because what I can do is I can literally access these files from my phone. Uh, if I was Alex, this is so much easier than me having to actually run my weekly commands, right? But you can take it one step further. Because Claude already has access to all that data, we can make it do cooler stuff with the data. So, for example, I could be like, "Hey, uh, Claude, can you pull all the engagement data for Acme Corp from HubSpot in the last 90 days? Pull the three emails we've had with their, I don't know, VP and put together a one pager with the key talking points and send that in an email to Dan. But let me review the asset here first. Okay. So, I just sent that over. I'll wait for it to finish. But, you know, now it has access to all the data and I can control my computer and do anything I want from my phone. And here is the one pager that it made me. Cool. So, now it can just email this over to Dan because it has access to my emails. And it's really easy to just do work from your phone while you're on the go with dispatch. This one pager maybe would have taken me 30 40 minutes if I were to do it alone without having access to all this. Now on the phone I can do it with 15 in 15 seconds.

And this patch is super powerful. But the thing is Alex is still the trigger. Like you see I have to do this stuff. Alex would have to pull out his phone and tell Claw to do the thing. But what if Alex forgot or didn't see his phone or on vacation for a week, right? And scheduled tasks is the next layer that removes the last human in the loop where we get to pick a task at a time and Claude will run this stuff automatically. So let's get into that.

So to do schedule tasks, there's three ways we can do it. And I'm going to show you number one. The number one easiest way to understand where it is is in the schedule section right here. So, if you click the schedule section, then you can see there's a section for schedule tasks. And make sure you know this that these scheduled tasks only run when your computer is awake. That's why a lot of people are buying Mac minis and just putting it on and turning it on all the time. But what you can do is you can hit new task and then let's say uh run marketing briefing every week and then uh of course you should be a bit more descriptive for the CEO that likes this every week. And then notice here in the prompt this is basically the prompt that's going to be run for clot. So I can do something like I can do slash and you can see it already sees my skills and I can be like run this report and then send an email to the CEO. His uh email is siminbookin.ai. Okay great. So because it has already access to my Google uh my Gmail all the stuff through my chats that's why I can do all of these. And then frequency you can pick one like hourly daily weekday weekly all this stuff. And you can pick a day, more options if you need to, whatever. And also, this is a big one, make sure you can put all your schedule task and you match it with a folder. So, if you have a lot of projects or clients that you're working on, you can have Agentic Workflows and these scheduled tasks for every one of them. All right? And then you can just hit save and then it's just going to run every Monday at 9:00 a.m. Cool. This is method number one.

Method number two I'm going to show you is if we let's say go back to this dispatch chat we were having what you can do is from inside co-work you can do slash you can do slash schedule and you can do the same thing you can be like hey can you make it so that you're running the weekly marketing report skill every week Monday at I don't know 100 p.m. EST and then you can do that and then hit go. And what's going to happen is rather than me having to type it in after this is done. I'll show you. And notice how this time I didn't even do the /weekly marketing report. I just told it to do that. And it Claude should be smart enough to go to my skills right here. You can see these skills right here. And what it'll do is run that skill and reference that skill and run it each time. So right now it's creating this schedule task without me having to do much. So this is just a little easier and let's see this in action. All right. So we can see it being done and let me go back to the schedule. So it should say like it you know it made this new one next run tomorrow at 107. But you know what you can do is you can open it up. You can see the instructions where it's stored all the folders. Make edits to this whatever you want. All right. And it becomes really simple to see and manage all your scheduled tasks and all your skills. So you know where everything is all in one place. Your connectors, all of these are in one place. All right. The weekly report is on a schedule. Now it's running every Monday automatically and Alex doesn't have to think about it anymore.

But then something happens. One of Bookton's agency customers that runs a travel agency reaches out and they know that Alex has been doing a lot of AI stuff and says, "Hey, Alex, you're the AI guy and I've got a problem. My team spends the first few hours every morning checking hotel prices across five different destinations for our clients. They check stuff like Booking.com, Hotels, Expedia, all these, literally refreshing every tab and writing down numbers into a spreadsheet every day. And these prices change constantly. And our clients expect us to book at the right time. Is there a way we can automate this?" And Alex thinks about this. And that's kind of the same pattern as the weekly report. It gathers data, it processes it, logs it, and alerts if there's any changes. Except instead of pulling from Google Drive, we're just pulling from other hotel websites. One problem though, those hotel websites don't have connectors. There's no Booking.com toggle on Claude's settings. And if Claude tries to visit those sites directly, it gets taken down by captures, spot detectors, etc.

Now, Alex knows how to fix this. And this is where things get interesting because what Alex is about to learn doesn't just solve Marcus' problem. It unlocks a whole new category of what Claude can do. So, I'm going to show you two ways on how we can tackle that. And the first way is we're going to get claw to control our browser. And the way we're going to do that is in Google, we're just going to do Chrome web store. Make sure you have Chrome downloaded and go here. Um, and then type in Claude. Make sure you download the Claude extension. Okay. So, I added the Claude extension. And then I'm going to log in with Claude. Cool. So, I just downloaded it. I'm going to pin it right here. And right here, if I click this, you see Claude said, "I understand." Next. Next. Let's go. And right here on the bottom, what I want to do is ask without asking. Act without asking. Then I'm going to hit this button right here. Enable with fast mode. Cool. All right.

Now, CL Claude should be able to control our browser. So, I could be like, and what I'll show you is there's two ways you can use this. One of them is just through this chat box right here. Just I can be like, "Hey, can you open up Google Flights andbooking.com on my browser, please? This is pretty simple. Uh, nothing crazy." And you see right now it just opened up Google Flights and Booking.com. So, it can control your browser. It can take actions on it, etc. Okay. The second way you can control this is if I go back to my cloud app and right here uh let's say go to co-work and let's make a new task right here. Um, what we want to do is hit this plus button and then go to connectors and this is also a connector and that's called claw in chrome. So, if you turn that on, you can also control it like this. Be like, "Hey, can you open Chrome?" And in Chrome, can you go to Booking.com and search for, let's say, the nearest hotels near the LA Convention Center, check in at April 22nd, and check out at April 24th for two adults. Sort by the lowest price, and take a screenshot of the results so I can see what's available. Pull the top 10 hotels names, the ratings, the room types, and the nightly prices. pull it into a spreadsheet and save it to a travel price tracking folder in my uh Google Drive or my Finder. So, right here, I'm just going to give it access to another folder just because I want to be neat. And right here, let me call it uh travel create, open, and then allow. And then let's go. Cool. Okay, great. So, you see right now uh it's getting started on this right away. And I'll come back once this is done. Wanted to show you a quick peek while it's working. So you see it's making uh progress right here. It set out its plan. It has instructions. And in this connector, if I click this, I should be able to see the browser pop up on the right in a second. And you see right here, we can see what's going on, what Cloud is doing, all the screenshots it's taking, all the actions. It's kind of just cool to see, but um normally you can just step away, make it, um do your thing.

Awesome. So it just got done, I can see. And cool. Claude just popped up with a new update that I'll do it in a sec. But uh right now you see it spat out this um hotels nightly price. Wow, this is way too low. Interesting. Um, yeah, and I can see in my Chrome that these prices are pretty accurate. Uh, yeah, so that's one way to do it. I personally don't like this the most because although this works, it takes a long time. One, two, it consumes a lot of tokens and it's not as reliable.

So, number two, let me show you how to actually make this super reliable using an external service which is actually meant to do all the scraping. And you can also take advantage of that to do any type of scraping on the internet that you would like. All right? Cuz right now, if you wanted to give Claude browser access, you'd basically be handing it the keys to your actual browser, the one you use every day. So your Gmail, your banking, your passwords, everything. And sure, Claude has good intentions, but do you really want an AI agent roaming freely through your personal browser with access to everything on your machine? One wrong move, and it's potentially touching stuff you'd never want it near. So, we're going to be using Fire Crawl to solve this by giving Claude its own dedicated browser completely separate from yours. So, Claude gets a sandboxed environment that's isolated, private, and controlled. And my personal favorite part about this is that this cloud browser stays logged in. So once Claude authenticates to a service you need, that session is saved and persisted. So next time you need it, firecrawl lets it remember. So you can control whatever access you give Claude. That's the difference between giving your AI agent your car versus giving it its own cars. And that's how we should stay safe and use tech like this.

Okay, so let's set this up. And to set it up, we're going to use the service called Fire Crawl that I really like. And I'm just looking for firecrawl right here. And um I recommend for most of you guys honestly you'll just can get away with the free plan. Uh, I already have a plan cuz I use it so much. So I'm going to go in here and what you want to do is now we need to give our claude access to our firecrawl. Right? And this is so without firecrawl if cla would try to visit a website it might get blocked or it might fail. With Firecrol, if I ask Claude to visit any website, even Amazon, Firecrawl will handle all that anti-bot stuff. And it's like giving Claude a browser that it knows how to operate a lot more efficiently than how humans are meant to do it, right? And it's actually really simple to install it. All we need to do is if you just Google fire after you have an account, firecrawl mcp, and right here, you see right here, I'm going to take this and then I'm going to go back to my cloud. And then what I'm going to do is I'm going to go to customize. And what we're going to do is we're adding an external connector. And this is why I was showing you this because I want you to know this is how we add external connectors that may not be in the cloud app store and how you can in uh install MCPS that you may not have access to or you want to. All right. So I'm going to hit this plus button right here and then I'm going to hit add custom connector and then I'm going to say firecrawl wl. And then I'm going to enter this in. And then what I'm going to do is paste in my API key. So I'm deleting this right here. And if I go back to Chrome right here, I'm going to take this key. I'm going to go back to Claude and paste this here and hit add. Great. And now you see Firecrawl has been added. And I'm going to make sure it always allowed. So all of the stuff fire crawl you can we can do it very simply. Okay.

Now it becomes super easy to use. I'm just going to hit new task. And right here you'll see if I say something like make sure I choose the right project. So I can be like hey uh use firecrawl to escape uh scrape hotel listings near the LA convention center on Booking.com for uh April 20th to the 24th. pull the names for the hotel, the star ratings, the room types, the prices, and any discounts or deals, and then organize it into a clean table, and then give me that table. So, imagine now the scraping is useful once, but let's say Marcus' team would need to know how the prices change day by day. I can just be like, hey, now I want you to also get this and track this over time. Every day, scrape the same hotels and return the log of results in a spreadsheet. So, I want to see the log over time. Each row should have today's date, the hotel name, the room type, and the nightly price. Compare today's price to yesterday's. If anything drops more than $30, flag it, right? And uh remember, we just did scheduling. So, I'm like combining all of these at once. So, again, okay, now I'm going to switch to Opus actually because I realized uh this is a very compound request. So, hey, can you schedule this to run every morning at 6:00 a.m.? Uh, scrape all the hotels and maybe call the file la hotel tracker.csv and also store it in my drive later. Okay. And this is a pretty hefty request. You see, I've been doing it a lot of things, but make sure you know because I've connected to Firecrawl. Now, I said, "Hey, use Firecrawl." I want to make sure in connectors right now I see firecrawl is on. So I should be good to go now. I'm just going to hit let's go. And it should work and it'll be a lot better than me having to actually go to my browser and I can use my Chrome without having my cloud, you know, hogging up my computer's resources because now my computer works faster. I get to scrape a lot more reliably. It has its own browser and I'm protected. So Claude doesn't touch any of my uh secret keys or tools or websites what I might have. So I'll come back once this is done and we'll go over that.

So this thing just got done. Let's open up these things just to see it. And nice. So you see it made me a really nice table and this was significantly faster than what I had been doing before with the Claude having to screenshot everything. So highly recommend using something like this. And look, there's a really nice table I can see. If I go back to my schedule tasks, it set this up. The hotel price tracker. This is what Agentic workflows are. If you think about it, Alex just built this for Marcus as a favor. Maybe it took an hour or so, but Marcus is probably not the only agency dealing with this kind of problem. Every travel agency, maybe hotels, has the same problem. Every e-commerce company that would track supplier prices would have the same problem, too. And even SAS companies that may monitor competitor pricing. You see, this is a repeatable workflow. Alex can literally package this up as something like Price Tracker and offer it to other customers or even sell it as his own service. Like he could say something like, "Hey, I'll set up automated price monitoring for your business for $500 onetime setup. Maybe you pay me a couple hundred to help maintain it." That seed that Alex can now build automations for other people's businesses is going to grow. We'll come back to this in level five when Alex starts building really, really cool agentic workflows for clients. But notice it now. The skills you're learning just aren't for your own work. they're actually really valuable and sellable.

All right. Now, Alex has built a lot of automation for Marcus. He did that favor and for his team. Alex now has the weekly reporting running on a schedule every Monday. But then the CEO says, "This is great. The marketing report is exactly what I needed. But now the sales teams want something similar. They want a weekly pipeline report. And the customer success teams, they wants a churn risk report. And if you think about it, all of these kind of look the same. It's the same format, the same brand, same quality. And the CEO asks if Alex can make that happen.

So, Alex decides to look into plugins. And what a plug-in really is is a bundle of multiple skills into one installable package. Think of it like apps on your phone. A skill is a single app like a calculator, the weather app. It has one job. A plug-in is a folder of applications bundled together for a role. So, let's say like an IT department, if you start a new job, they set up your email, your Slack, your Google Drive, a VPN, an HR portal, everything at once, right? You install that once and then it's set up for your job. That's what a plug-in is for Claude. It's basically a whole toolkit for a role.

So, let me show you how plugins work. And if you click into customize, you can browse plugins right here. And this is really cool. Anthropic actually already gives you a bunch of plugins. So, let's say for marketing, sales, legal, these are pretty generic, but if I add them right here, you can customize these plugins that Anthropic has already made for you. If I go back and you see personal plugins, marketing and it comes with a bunch of skills where it can review your brand, do a competitive brief, create content, all these little things and all that's all that is. It's just a bundle and you can install more if you'd like just by creating browse plugins and installing what you need. But the real value of plugins is the fact that you can build your own really, really easily, bundle it up, and then give it to other people. So let me show you how to do that.

Okay. So, let's say Alex wants to build the plugins for his team, right? If Alex wants to do that, then what we want to do is honestly just talk to Claude. What we're going to do is let's switch our folders again and go back to our P&L for boss because this is what we've been working through and say something like, "Hey, I want you to build me three plugins from the skills we've already created." And here's how we're going to break them up. So, the first plugin is the core toolkit. This is the shared base that every team may get. This includes the branded deck uh for the brand we've been using. This generates the branded powerpoints and maybe it also has the skill for pulling data and then also organizing and getting the right data from drive, HubSpot and Post Hog that we've done before from the weekly report skill. Now bake in our brand guidelines and our colors. Make sure it has the require connectors too. Okay. plug-in two, make one for marketing ops and inherit everything from the weekly marketing report using the full GCPS uh framework, but also make a competitive brief and a content performance skill for that as well. And then plugin three, have a sales ops toolkit and add some skills like pipeline report. This is basically what we want to do is get the data for weekly sales pipeline and make a deal uh deck based on that and then also a deal say stage analyzer so we can identify where deals may be stuck and recommend the next actions. One more thing for the sales can we also make a skill for follow-up drafter to follow up and uh make personalized emails based on any CRM activity in HubSpot that you see. For each plug-in, create the folder structure with the skill MD files for every skill and include the brand guidelines and templates and list the whole required connectors. Make each skill actually work and let's go.

Okay, so this was a really beefy prompt. I want to make sure I uncheck travel because we're not using that anymore. But right here, you see I word dumped everything I needed. I thought through it. I'm like, "Okay, marketing, sales, and uh everyone needs a core toolkit." So, right now, again, this is just the first attempt. What you want to do is you get your ideas on paper and then you iterate. But ultimately, I'm let's say Alex in marketing. Then I already have my skill. Now, I'm just making different tweaks and versions for what different data the marketing team, sorry, what data the sales team may need and what data the customer service team may need. And I'm telling Claw to just build it out based on that and save it to me. So after this is done, I'll show you the plugin and then I'll show you how we can share this.

Okay. So So it got done building my plugins. So right here you can see there's three plugins. And if I hit save on all of these, what's going to happen is if I go to customize right now, you can see the plugins appear here. And let's say you are the teammate. If you hit this plus button right here and then hit create plugin and then upload plugin right here, you can browse the files and whatever plug-in file you send your teammate, you can essentially make them install it. And if you're curious on where to find these plugins, then right here, you see how we have PNL for boss? That is a folder in your computer. So, so in your finder, if you just search PNL L L4 boss, you see if you open that up, you can see all the plugins and all the assets that we've been generating inside that folder. So, that's just a quick little hack if you are curious where everything gets saved when you're working in Cloud Co-work. Right. So, you can just take one of these files, give it to your teammate, and then tell them to go to customize, and then hit this plus button, hit create plugin, upload plugin, and browse files, and drop in your file, and they should be good to go and start using your plug-in. And there's a second method, which is using GitHub to share plugins, which we'll be diving in into later uh in some of the later levels because GitHub requires a little bit more technical understanding and we're going to be using GitHub a lot. So just just so you have a preview, if you hit create plug-in, and it says right here, add a marketplace, you can get it from a git repo as well. All right. So remember this for later, but right now there are two ways to share plugins, which is one, you can just share them with your team if you're on one of the teams plans. Number two, you can zip it up, send them the file, and they can upload it. And number three is with GitHub.

I want you to notice something. everything we've just done just by speaking to Claude. That is an agent executing a multi-step workflow on your behalf. This is the same as if you'd had trained someone to do it exactly how you like and they never forgot. And that's really what an agentic workflow is and you've been building them this whole time without calling it that. We'll go way deeper on this in let's say level five, but right now I want you to start seeing it for what they are. These are AI agents doing work on your behalf. But, you know, as Alex learns more and more about Claude, he now doesn't have to run any skills or reports anymore manually. The reports are running on a schedule. He can use dispatch on the go and his team has their own shared plugins and skills. This means he went from I use Claude to my department runs on Claude.

But remember that recommendation from the weekly report? The top line was basically just shift budget from paid to organic content. Now, Alex needs to build out that content. So let's say he starts to open up co-work to build Instagram carousels. The problem becomes his co-work is limited in that regard. Because it's a sandboxed environment, it also takes up a lot of resources from the computer. And this is where co-work really hits a wall. And that's where something new begins. So let's get into that.

Congrats. This is the next level. And in this level, we're going to be using cloud code. and we're going to break it down step by step so it's easy for you to follow and build all the cool things we're going to be doing. But what is cloud code? Well, cloud code is basically cloud running inside your operating system. So unlike co-work, it doesn't use a virtual machine or a sandbox. It uses your actual files in your system. And here's how to think about it. Chat is like texting, let's say, your smart friend. It's really good for quick advice. Co-work is like hiring a contractor who works in their own office with their own computer. They hand you deliverables through a window here and there. It does good work, but everything stays in their world until they pass it to you. Clawed Code, on the other hand, is hiring someone who sits at your desk on your computer with your files open and then they can see everything and then even work very long hours and everything they build also lives inside your machine.

Great. So, quickly just for some use cases, we use chat when we want to see some illustrations, make some artifacts, do some quick brainstorming or have some quick questions. co-work when we want some let's say reports research or deliverables anything that's like professional level work and then cloud code is the final frontier where we should be doing coding creating assets like images videos graphics and working with existing project files that may be pretty large in size basically anything that needs to persist on your machine and may need the power of your machine we should do in cloud code and the thing about cloud code and why it's different is because cloud runs in multiple places inside your computer so it can run in the terminal. It can run in VS Code. It can run in your own AI operating system, the cloud desktop app, and a bunch of others. But for this level, we're just going to be using the desktop app to slowly just get situated with it. And in the later levels, I'll show you when to use the other ones and why they may be beneficial sometimes.

All right. So, to switch to cloud code, what you want to do is hit this button right here. And on the very right, it's going to say cloud code. And you'll notice they've actually changed up the UI a lot. And I just want to give you a quick tour just so you know where everything is. So the first thing I want to mention is the old projects we've been working with. You can't access them anymore as easily. So for example, if I wanted to access one of the older folders we've been working with, let's say in cloud co-work, I would have to find PNL boss. So I would have to search for that file and then drag that in and then I can work on it. You see? So the folders are right here in the new cloud code. Okay. And then notice how you can like add different folders to work on as well. So, they've made that little change and you can go there. And on the bottom hand, you'll notice right here, so you don't get tripped up, they have some different model selectors. So, normally I'm using the Opus 1 million model because I'm on the $200 plan. Don't use this if you're not on it. And then you can select low, medium, high, max, etc. I just choose max right now because I have a lot of tokens. But please use the right model for the right task. Okay, one more thing I want to show you. Now in cloud code, what they did is we had the schedule section right here inside co-work. But now they've just renamed that whenever you go to cloud code into routines and it looks a little bit different. So you notice how we have one example routine made right here. And you can also see in the calendar. So which routines are running when and a routine is essentially the same thing as a schedule. It's basically let's say something externally can trigger cloud code to run an automation on your behalf or it can run on a specific time. These are kind of what we'll get into later called agentic workflows. But right now I'm just showing you the UI. One more thing I want to show you is let's say in the sessions right here you can open up multiple sessions and right here you can see you can move these around. So these are all your different projects and your folders etc. So you can go around and be very organized. Okay, you can also move them around like that. And even let's say you like working on one thing over and over, you can pin this. Okay, so that was really cool. Make sure we're going to be using dispatch later. So keep note of that. And a really nice thing is they also in customize is we can still keep using the same skills and connectors that we have been connecting to with our cloud co-work.

Couple more things I want to make sure you know is that this thing right here bypass permissions. This is going to save you a bunch of headache. And the way you enable that is by going to settings with by clicking your name and then going to settings. And then if you go to cloud code and scroll down, I want to make sure you check allow bypass permissions mode. This will make sure that you work a lot faster and you love your life a lot more. And one more thing is make sure cloud code on the web check these on as well. And then after you did that for cloud code, you should be good for the settings. Okay. And one more small thing we're going to be using throughout the course, which may look slightly different, is this plus button right here. So if you hit this and hit connectors, this is where we're going to see all our connectors, the ones that are enabled. All right. And if you want to manage your ad, it's right here. So it's more or less the same thing. I just want to make sure you know some of the little tweaks Enthropic has made to their Cloud Code app. All right. And lastly, just to make sure you know, we don't need this too much, but what Enthropic has enabled in their brand new UI is the fact that now I can open different things in split view right here. There's a lot of things you can do. And on the very right, you can open up a terminal, you can open up tasks, you can make a plan. We're not going to be needing 99% of the features, especially if you're just starting out, but I just wanted to make sure you know because it's more or less the same thing. It just looks slightly different. The schedules are now just called routines only for cloud code and they share different workspaces and everything else is good to go. And let's get into the course.

Great. So, let's get into building something. And first I want to start with what is it that we're going to build. First I want to start with what is it that we are going to build right now because there's going to be a couple moving parts because cloud code can do so much. All right. So remember how so remember how way back in the video we discovered that for the company that we were spending way too much money on meta ads and we need to reallocate into organic content. So what we're going to do is we're going to build an entire automation and pipeline that generates content and posts it all all automatically. All right. And ideally we get AI to do this and Alex so Alex doesn't have to do anything except maybe approve here and there. What I kind of want to also show you is that this automation we're about to build is actually super useful. And you can see this is my Instagram and I've actually been doing a lot of these. So these are all you see there's like a little image there's me all of the stuff these carousels are made and posted all automatically. So it does a lot of research on its own about the topics that I post about and then it creates these graphics and also makes it into my style and posts it automatically. And you can see right here here's the service that I can see my content calendar. And we're going to be setting all that up. All right.

But first, what I need to explain to you is to do this, there's a lot of little steps. We're going to be relying a lot on open- source technology. And what open source technology really is, it's kind of like this. It's hosted on the internet in this service called GitHub, which is basically a public library for code. And anyone can go there. They can store their own books, their tools. You can share it with others, etc. We're going to be using GitHub a lot. Okay. So, what you want to do first is go to github.com. And then from github.com, I want to make sure you sign up and I'm just going to be signing up with my Google account. Okay, so I just logged on to github.com and now I have an account. Right.

Next, what I'm going to do is I'm going to go to my community and right here inside classroom inside skills vault, I basically have this simple Instagram carousel link. So, this is a skill I've built out and you can see I've worked on the scripts and all the stuff and I try to make it onto my own. So, it uses my head shot. Uh, it uses my language. It knows all of that. So, all I have to do to get Cloud Code to work with this is I need to copy this URL. I'm going to go back to Claude. I'm going to go to code. And then right here, I'm in Piano for Boss. So, right here, let's get a little bit more organized. And I'm going to choose a different folder and let's say call it carousel maker. Okay. So I'm making a new project called carouselmaker. Then what I'm going to do is I'm going to paste this in. So cloud code can actually go to this code, pull it down and then start using it and actually produce out output from it. Right? So, uh, I'm going to also change this to opus. And I'm going to ask, hey, can you use this skill I just pasted it in? And then what I want you to do is build out a carousel. The topic of the carousel should be about, I don't know, a very long in-depth claude course about how claude skills can be stored in GitHub and how you can find more stuff. Simple, right? Okay. Then I'm going to hit go and you'll see it'll start working on its own. It's the same as using co-work chat pretty much. It's just a different interface, but the nice part is that it can use the resources for your computer and it lives in a lot of other different places that I'll show you in a bit. Now, I'll wait for this to load and after this is done, we'll go and review it.

Great. So, it just got done and now you see it's saying it saved it right here. But with cloud code, um, I can be very lazy. I can just be like, open it in my finder because I don't know where that is and I don't want to open it up. I can just tell it to do that. And it'll very clearly control my computer and start opening stuff up. And if you see, if I uh highlight all of these, open it up, you can see what it looks like. Great. Okay. So, notice some little things. Um, copy looks fine. uh design looks fine. Small some small things is it does not have my picture, right? I ideally wanted it to have my picture. So, what I'm going to do right now is I'm going to go to YouTube and copy my picture. And this is now what we're doing and what I'm showing you is how we iterate. Okay, the idea is, hey, we just got a skill out from, let's say, YouTube or whatnot or some community. And then now what I'm doing is I'm going to go to Finder. Okay. So, I just got my profile picture right here. So, I'm going to copy this image right here and I'm going to drop it here. And just for safe measure, and I'm going to show you how to do this. This is right here. So, I'll drop it here as well. But what I'm going to say is like, hey, hey, edit this skill so that it always uses my image. I just attached the image to you. And can you remake the carousel and open it up in my Finder again? Great. So that's all I had to do. I just p my pasted in my image because my old ones, the old carousel I generated did not have my image. And let's say it's good enough. It's just a matter of iterating. So let's say you if you didn't like the copy, you wanted a darker background, you could just say all that and it'll change it. But you see this is a template that I can have after this is done. And what I want to show you is the fact that hey, we can download skills from the internet. We can change it up and use it as we want. Okay, so this just got done. Um, if I open back my folder up right here, you see it edited all these. So, let's open this up now. Oh, nice. You see, now it has my picture. Simple. Now, I can just be like, "Hey, hey, make sure you edit the skill to always use this picture." And then, can you save the skill inside this folder? Great. So, what we did is we took a skill from the internet. And guys, if you're already a part of my community, I have a lot of different skills and I'm just I just keep adding new ones and making them. Uh, so go feel free to dig through and then have fun with it. But overall what you can do with this is now we have just made a skill where we can build carousels really easily. Okay, so that part is done. The next part is when we want to make sure is one we have the ability to do research and then we want to make sure we can take that research make a carousel out of it. Then the last one we will make sure it gets auto posted.

So I'll show you how to do all of those. All right.

So now let's say we have made these carousels. Now we want to post them on automatic. So we want to go to Chrome and then look for blot.com. And I'll leave all the links and stuff in the description below or in a resource file for the entire course so you can access all of that.

But right here, what you want to do is first you want to log in. Then on the bottom left, you want to go to settings. Okay? After you go to settings, make sure you connect your accounts. So all that is is basically you just hit one of these buttons like login with YouTube and then you select your accounts and then you log in. All right?

So after you've done that, go to API and then it becomes really easy because then you need to do is you copy this. Okay, right here you see cloud desktop cloud co-work. We're going to copy this and then go back to claude. Then we go to customize. Then just like firecrawl, we're going to add a new connector. So we're going to hit connector. We're going to hit the plus. We're going to add custom connector. We're going to say blot and then insert this MCP and then hit add. Okay.

Then you'll see right here's right here. And then you're going to hit connect. This makes it so uh really easy. This makes connecting it really easy. And you know I just hit approve and boom blot is connected. Now I'm going to make sure I put always allow. Great. Okay. That's good to go.

Now I need to just go back to my cloud code chat. So if I go back This was the build build carousels I've been working on. And right just to check it out in more options and connectors, I want to see if I have Blato enabled. Cool.

Now, what I could be like is like, hey, what accounts do you have connected in my Blot? Oops. I And you see it got my wording wrong. So, I'm going to say blowto and then correct it. So that's a really nice part about cloud code as well. You can steer it. So if you make mistakes in a prompt and you like hastily put it in, which you shouldn't do, but um you can always correct it. All right. And you see right now it's just going through my accounts and it knows it has access to four of my platforms. Cool.

All right. Now what I can do is I can just ask it to post. So I can be like or it doesn't matter what kind of schedule you have. So, let's say you have a folder where you make um 10 different posts. You could be like, "Hey, take each one and post them every day." You can just tell Claude and it'll do that.

So, for example, maybe I'll show you with a post and I can say something like, "Hey, uh, just like the carousel we generated, I want you to use firecrawl to do research on what new advancements and fun things are happening in the world of AI and claude. I want you to make a carousel and then I want you to post that carousel with Blot. Um, just to test it out, can you do one right now and set a schedule that I want you to be doing this every day at um 12 a.m.?" So, right now, uh, I want to see that on the schedule as well. And this is just like me talking to an employee. I'm like, "Hey, okay, I taught you how to do these carousels. Go make them and then make a new one. Make a new one every day and then post it."

If you want, you could not have it automatically post. You could ask it to just like review. And it's as simple as that. If you want, you don't have to make it auto post. You can just hold it so it just, you know, outputs all of them. You can go in, you can review it, you can make your tweaks, and then post it as well. So, you could have that step in between if you want to check. But what I just showed you right now, it will literally just go through, make and post, make and post, and it puts your content creation on autopilot.

Really, it's just a matter of you iterating and tweaking what exactly what you need. All right? And you know, you can do so much things just with the concepts that I've taught you. Like right now, we just did it for content, but you can actually make it into a trading bot with the schedule feature and these external APIs.

So, let me show you how I did that. So, I'm going to go back into my browser and I'm going to search up Alpaca. Uh, and you can see Alpaca trading platform. And I'm going to go here and I'm going to sign up. All right.

So, after you enter all your information and you go through, you're going to see you have this dashboard right here. And in this dashboard, if I go to the top left and I open this up, and you see I have an individual trading account, but everything we're going to be doing is not with real money. It's with paper money. So, what we're going to be using is this paper trading account. Okay? And you can, you know, if you want to make a new one, I'm going to say, um, trading claude. And let's say we can give ourselves, I don't know, $50,000. Okay. And save. And you can see right there, we're going to have a trading clot account. And I'm going to click this. So now I have $50,000 to essentially play with in the stock market. All right.

Now, I'm going to show you how to get this connected with Claude. To get this connected with Claude, what we're going to do is I'm going to scroll down right here. And you see right here where it says API keys, what we want to do is hit generate new API keys. Okay? And you can see there's three things. Endpoint, key, and secret. We're going to need all three of these. All right.

So, let's get cloud set up. So, I'm going to go back to my cloud app. And here, what I want you to do right now is hit this thing that says code. Okay. It's the same thing. We're just going to be talking to it, but all it is is a little bit of a different interface. Okay. So, cool. You see this little guy here, and that means we're good to go.

And I'm going to do is right here, I want you to hit this button right here and hit a new folder. Okay? Because we want to keep everything we're doing organized. I'm going to go to documents and I'm going to hit new folder and I'm going to call it trading. Okay? So now everything we do is going to be saved into this trading folder. All right. Okay.

I'm going to move this to the side. Now, what I'm going to do is I'm going to be pasting in right here this and then go back to claude and then say end point. I'm going to say key. Oops. Claude key. And then I'm going to say copy this one and then go to claude and then say secret. Okay.

Now, let's see if we can get our claude to trade a simple stock just to make sure our connection is solid. And after our connection is solid, we'll build layers on top of that so we can actually get it to do some really cool stuff. All right. All right.

So, what I'm going to be telling it is this. Hey, um, what I want you to do, I just gave you the documentation and my keys to connect to my Alpaca trading account. I'm just testing the connection right now. Can you please buy one share of Apple? I want to see it inside my account. All right, great.

So, all this is good. Uh, right now, just to make it a little faster, I'm going to right here click sonnet because this is a very easy command. And I'm going to hit go. All right. So, it's saying the order is placed successfully. Here's the summary. Now, if I go back here and if I reload. Okay, awesome. Uh, great. So, you see I already bought a share of Apple.

Now, you can trade by just talking to it. Be like, hey, can you sell that share of Apple and then buy a share of Tesla? Great. Trust workspace. And here, this is what I want you to do as well. Hey, can you make sure in this folder you save these credentials so I don't have to keep giving it to you when we want to trade. We're going to be using this account and in this folder we're going to be doing a lot of trades. Okay.

So, what we're going to do is because right here we gave it access to all this stuff. I want to make sure I don't have to keep giving it access. I'm just getting it to save these credentials inside a file. Right now, Claude can trade but it's trading blind. It doesn't know what to buy, when to buy, or why. It's missing the most important piece, which is strategy.

So, let's get into that. In this level, we're going to turn Claude into something that runs on its own. So, basically, a bot that watches the market, makes its own decisions, and based on the rules you set. And the first strategy we're going to make, you can think of it like a smart thermostat. Once you set the rules, if the room temperature, let's say, drops below 68° FE, you're going to turn on the heat. If it goes above 75° F, you kick on the AC. That means you don't have to sit there fiddling with your phones messing with the temperature all day. The thermostat does that all for you. It checks, it acts, and adjusts. And the bot is going to work exactly the same way. That means we're going to be setting some rules.

And let me show you how this works by actually explaining the trailing stop strategy, which is a really good strategy used by many different traders. Let me show you how the trailing stop strategy actually works. So, say you buy a stock at $100. Well, let's say we tell Claude, if this drops to $95, sell it. That's your floor. $5 is the most you're willing to risk on this trade. Now, if the stock starts climbing up and let's say hits 110, if it didn't go down, but your floor is still sitting at that 95, then you're exposed to a $15 fall. That makes no sense. So, what we want is Claude to move up our floor to 105, which means if the stock dips and hits the new floor, you're still up $5 in profit and you were protected the entire time. But what if the stock drops right out the gate, right? Well, that means it would just hit $95 and you'd sell it and you're out $5 and that's your loss.

Now, here's the important part. You're not stuck. You've still got that capital. Claude can now take that money and goes looking for the next setup. Maybe it's the same stock a week later at a better price. Maybe it's a completely different opportunity. The point is you lost a small portion that you're okay to lose and risk. Basically, you live to trade another day. And ideally, we find trades like this. So, let's say the stock keeps climbing and every time it does, Claude is going to drag the floor higher and higher. You can never fall back to where you started. It's basically now going to sell at 110. So, you get that $10 in profit, which is a 10% gain. That's a trailing stop. You set the rules once and Claude follows them every second of the day. You're always protected and you're always locking in the gains. So when a trade doesn't work out, you move on fast with the most amount of money still in your pocket.

Now, why am I telling you all of this? Because the worst thing you can really do is hand your AI a pile of money and be like, "Go figure it out." The rules aren't the limitation. That's the whole point. This is how you take what you know, your instincts, your risk tolerance, your read on the market, and you encode it. The same rules are true here. Claude can execute thousands of decisions faster than any human. It's executing your decisions, just running at the speed and discipline that you never could on your own. That's what makes this different from just letting AI loose. And I'll show you how to set this all up in the next portion.

And to set this up, it's actually really simple. So, I'm just back in my chat where I have been speaking with my Claude and I'm just going to be speaking to it, right? And just to make it a little easier, if you guys wanted this in my classroom, uh, if you guys are already watching this, if you go to skills, you'll see I have a bunch of skills here. So, this is the prompt we're essentially going to be speaking to our claude. All right, so let me just go to Claude. And right now, don't worry about it. If you want those docs, it's there, but you can follow along with me in this video. So, I'm just going to speak it to it. All right.

So, I want your help to actually schedule a trailing stop strategy on let's say Tesla, right? I want you to buy Tesla using Alpaca paper trading account. Buy, I don't know, like 10 shares at the market price right now and set up these rules. The floor, if the stock drops, let's say by 10%, sell everything. That's my stop loss. I don't want to lose more than that on this trade. the trailing floor. If the stock goes up 10% from what I paid, move my stop loss up, maybe move it up 5% below the current price. Every time it climbs, move another 5% up the floor again. So the floor only goes up, never down. And then I want you to also ladder in. If the stock drops, let's say by over 20, 30%. Buy a bunch more shares, let's say 10 more shares. if it drops by, let's say, 20%, buy 20 shares. This way, I'm getting better prices on the way down instead of just losing money. And after you set this up, show me a summary of every order. And right after you place it, so I can confirm this looks right. Okay, so I word vomited some stuff there. And um let this finish and I'll show you what this looks like.

So, I see it just got done and it says this is the current price. And if I go back to my Alpaca dashboard, I can see it set some price. It bought some Tesla. And then it even set a stop loss right here. Now, one more thing we need to do is make sure we go back here and be like, "Hey, can you set up during market hours every day that you're checking consistently when we need to move our floor up or need to make new stop- losses or re-enter, use the slash schedule to make sure we have that going and set your own schedules. So, this just executed one trade. Now, what I want is clock to basically be alive and keep looping and checking on a schedule like, hey, are these trades looking right? Do I need to re-enter? Do I need to move my floor up? all of that stuff. Normally, we'd have to do it, but right now with the command I just gave it, it should be able to do it on its own. Okay, so it just got done and it says literally Tesla trailing stop monitor and it's going to be doing that every 5 minutes from Monday to Friday 9:00 a.m. And it's going to keep checking. And just so you're clear on the left hand side, if you hit see this clock, if you hit this clock, it says scheduled. And if you open it up, these local tasks are already going to be running right here. You can see it already set the schedule for me. I didn't have to touch this. And if my computer is on, it's going to be running this on its own.

Now, we have basically made a trading bot. Now, let's go through a couple different scenarios. Hey, so just briefly and really quickly, can you tell me what would happen if let's say Tesla shoots up to $500 randomly? What would you do? And right now I'm roleplaying it just so you understand and I encourage you to do this as well so you understand what would happen in different scenarios and what would it actually do and if you do really need to change its thinking and thought processes around you could just be like hey change this around. Okay. And right here it just done it says if Tesla shoots to 500 on the way up the trailing stops kick in trailing activates your floor moves up up etc. And that's about it. Yeah, this is a really cool way to do ladder buys. Those don't trigger at all. You see, right now, I don't have any of them. So, I'm going to be like, can you think about what are good ladder buys and make sure you set that up in the schedule and update the strategy inside our whole strategy of what we're trading right now. So, as the price goes up, it buys in gradually. So, we're always at somewhat of a profit and we're pretty safe in our buys. And just so you see, it even gave me the latter levels of if it's negative 15%, it's going to buy 10. If it's 50%, it's going to buy 50. You can always change this around, but this is really cool now that me without me having to do any of this stuff, Claude can just action it for me.

Now, this is all great. But the thing is, we're still picking stocks ourselves. We choose Tesla because we like it, but that's a gut feeling. The biggest traders and the best traders in the world don't trade on gut feelings. They trade on information. So, in the next level, we're going to be showing Claude where the smart money is so Claude can actually make informed decisions. But what is smart money? Well, on Wall Street, there are people who move millions of dollars in a single trade. When someone puts, let's say, $50 million into a stock, they didn't really just do that off a gut feeling or flipping a coin. They have a lot of research teams and a lot of private data. They know something. These people are called whales. And when these whales make a big move, it leaves a trail like massive options orders, unusual volumes on a stock out of nowhere. And a group that does this a lot is surprisingly US politicians. But the nice part is that the members of Congress are required by law to report their stock trades. The data shows is that many of them consistently beat the market. They sit on committees. They regulate entire industries. They get briefed on policy changes before the public even hears about them. They know which companies are about to get a government contract and which ones may get investigated, but you can't see what they bought and follow them. And most people have no idea this data exists, but it's there and Claude can read it.

But the thing is, we can't just tell Claw to go scour the internet for these hedge funds trades and congressional filings and all that. There's millions of data points hitting the market every single day. No AI can really browse its way through all of that. But there are services that do this full time. These companies are scraping and organizing all this data around the clock. And the data is sitting there waiting to be tracked and updated every day. And what we need to do is plug Claude into that. And that's what this new skill and MCP does. Think of it like a power outlet. The electricity, the insider information is running through those walls. And we basically plug our claude into that so we can use it. These services have data flowing through their systems and the MCP is the plug that connects Claude to their data. And once it's plugged in, Claude can pull from at any time. it becomes a new skill and Claude goes from being smart but working with nothing to being smart and seeing what the big players in the market see. Next, what we're going to do is give our Claude the power to access these services and it's going to be super easy.

So, let's get into that. And just to emphasize why I'm so excited about this strategy I'm about to show you. This is me just backdating. Let's say a year ago, you were copying the same thing and the same strategy by copying the smart money. You can see if you were following the S&P 500, you were just buying and holding it, you would have made around starting at a 50k account, you would have made around 57750, which is around 15%. So, you would have made 15% more money. But if you copied McCall, then you would have made 67,400 by the end of the year, which is a 34.8% return. You can see how these politicians have access to this insider information and they can out compete everybody. And that's what I'm going to show you how to do.

And just to make sure our strategies are organized, what I'm going to do is go to another paper trading account. So right here, I'm going to use this summon account and I'm going to do the same thing. Okay. So I'm going to hit generate keys. I'm going to take all of the stuff where I can just copy and paste this. And then I can go back to my cloud. And then what I'm going to do now is make a new session. So going to hit new session and I'm going to paste this in. And now what I'm going to do is the service we're going to use to track these politicians. If you go to Google, I'm going to do Capital Trades. And Capital Trades is a free service that you can be using to see which politician is trading what and what's going on. So I'm going to copy this URL. I'm going to go back to my Claude and go right here. Okay.

So now uh by the way the prompts are going to be in the description below but right now hey so um I'm trying to set up a copy trading bot when I've also given you access to my alpaca account. I want you to make a new folder inside this uh folder so we can have this running and have the schedules done. But what I want you to do is number one, find a politician who has been having really good success in the stock market currently and is actively trading. And then what I want you to do is copy their trades. Okay? So set up your crown jobs and your schedules so you're always looking and using capital trades to see what they're up to and copy those trades. Make sure you buy, sell, buy the same options that they knew. Whatever you need to just copy them and then we're going to be tracking that over time. All right. And you're going to be using the Alpaca account that I just posted the image of. Great. I'm going to hit bypass permissions, enter bypass mode, and then hit enter. All right, that was super easy. Let me wait for this to finish and we can see what's going on.

Okay, so it just got done and I see in my schedule, it's already set a schedule to see what these politicians are doing. And then I actually also asked it, who are you copy trading and why? And it chose Michael Macau. All right, very interesting. And why is basically okay why the algorithm picked him over the others? Because it seems like he is very active right now and he's the top trader. interesting. The trades may be slightly stale, but that's the nature of how they disclose, we still get a lot of gains because mostly Congress can buy like 2 years out or something like that. So, they don't do day trading and they're not allowed to do it. That's why even if we're a little late because we know about it later, we can still make those gains. And just to reiterate, you can see this is a simulation for the year. If we started one year ago and did this exact same thing, if we copied this exact same strategy, this bot would have made an extra $9,650 in comparison to the S&P, which should have only made around $7,000. So that's a 2.2x return over the S&P 500. And that's how these guys just stay so rich. It's kind of crazy.

So, with that out of the way, now I'm going to show you the third strategy, which is a strategy that really advanced people use, which means you can get paid no matter which direction the stock moves. And to understand that, I'm going to explain how options work. Okay, so what is an option? Well, an option is basically a contract when two people agree to do a deal. And let me show you this with something you already may know. Okay, so let's say you pay your car insurance company $100 a month. In exchange, you get the right to file a claim if something goes wrong. And if you crash, the insurance covers it. You're protected. And if nothing happens, they keep your 100 bucks. They're basically getting paid to take on the risk.

Now, look at this exact same structure. What happens is you pay a premium to another trader where you get the right to buy or sell a stock at a locked price. Okay? And if the stock moves your way, you can exercise that contract and profit. If the stock doesn't move your way, the contract will expire. They keep your premium. It's the same deal. You basically paid for protection. They got paid for the risk. And here it is next to each other rowby row. It's identical. The premium, the right to act, the expiration date, and if nothing happens, the other side keeps your money. That's all an option is insurance on a stock. Okay?

And to dive in, there are only two types of options. A call option gives you the right to buy a stock at a locked in price. Say Apple is at $200. You can buy a call option with a strike price of $210. That means if Apple goes to, let's say, 230 before your contract expires, you can buy it at 210 and pocket that $20 difference. If Apple stays below 210, your contract expires and you lose the premium you paid. That's your max loss. Think of it like putting a deposit on an apartment. You'll pay $500 to lock in the rent for $2,000 for 30 days. If the rent price jumps to $2,500, you got a deal. If they don't, you lost that $500 deposit. And on the other side is a put option. The put option gives you the right to sell a stock at a locked in price. Say you own Apple at $200. You buy a put option with a strike price of 190. If Apple drops to 170, you can sell it at 190 instead of 170. The put basically protected you from the drop. That's the insurance side. You paid a premium to protect yourself against the price drop.

Why this is interesting is because with stocks and options, you can actually flip the table and become the insurance as well. So that is called selling options. So when you sell an option, you basically become the insurance company. you're the one that's collecting the premiums. So, someone pays you for a contract and most of the time that contract expires without anything happening and you keep the money. And the cool part is the insurance companies make billions doing this. They collect premiums after premiums from millions of people and pay out on a small percentage of claims. The math basically works in their favor over time. And selling options basically work the same way where you collect a premium upfront and if the stock doesn't hit the strike price by the expiration, your contract dies and you keep everything you were paid. And the wheel strategy is built on selling options. You are the insurance company. You collect premiums at every stage. And let me show you how that works.

Okay, so let's walk through the wheel strategy with a real example with let's say Tesla. So Tesla is trading at $250 and let's say you really like the stock, but you're not paying $250 for it. You'd rather get it at let's say $230. So if you sell a put at the 230 strike and for making that promise for saying I'll buy Tesla at 230 if the price drops there, someone is going to hand you $5 per share. That's $500 in your pocket right now just for agreeing to do a deal you already wanted. Now, two things can happen. Tesla can stay above $230, which means the contract might expire, nobody makes you buy anything, and you get to keep the $500 and you can do it again the next week. The other scenario is Tesla drops below 230 and you have to buy it. But remember, you collected $500 before already. So your real cost isn't 230, it's 225. You got Tesla cheaper than anyone else, and you wanted it anyways.

Okay, so stage two, let's say Tesla does drop and you get assigned, which means you have to buy the Tesla shares. Now you own 100 shares at an effective cost of 225. Most people would be upset, but you're not because you set the price. You got paid to wait and you bought it at a discount. Right? That's not a loss. That's a your plan working. Now, it's time for stage number two. Okay. Now that we own the shares and they're sitting in our account, what we're going to do is we're going to sell a covered call, which means we will pick a 260 strike price. That means someone is going to pay us another $5 per share, which is another $500 for the right to buy your Tesla shares for $ 260. If Tesla stays below 260, that contract expires, you keep the shares and you keep the $500 and you can sell that call again the next week. This is cool because you're generating income from a stock that isn't even moving because if Tesla does go above 260, your shares get sold. But look at the math. So let's say you bought the Tesla shares at 225. You sold it at 260. That's a $35 per share in stock gains plus the $5 you collected selling that put plus the $5 you collected selling the call. That's $45 per share in total profit on 100 shares. That's $4,500. And here's the best part. Once your shares get called away, you go right back to stage one. You sell another put, collect another premium, and the wheel just keeps spinning. every single rotation you get paid. The stock can go up, down, sideways, it doesn't matter. You're collecting income at every stage.

The wheel strategy sounds great until you try to run it for yourself. You're picking strike prices based on your market conditions and all these other things every single week. Most people who learn about this wheel give up after a few weeks because management is what buries them. And the cool part about Claude and I'm what I'm about to show you is that Claude handles all of it. Which means Claude itself can monitor the positions, pick expirations, and roll contracts when needed. Which means you just collect the premiums on a schedule and you'll check in, let's say, once a day.

So with that out of the way, understanding why Claude makes this so effective, let's set one up. Okay. So, for this we're going to be using the first trading Claude account. And I'm going to go back to Claude. And you see I actually have a prompt right here that you can copy paste. And just so you can follow along, I'll read it out to you. Okay. So, what this is is basically I want you to run a wheel strategy on let me see Tesla using my alpaca trading account. Stage one, you sell the puts. So, you sell a cash secured put on Tesla with a strike price around 10% below the current price. Pick an expiration 2 to 4 weeks out, collect the premiums. If the put expires worthless, sell another one and keep collecting premiums. If I get assigned, I have to buy the stock, move to stage two. Stage two is selling the calls. Once I own the shares, sell covered calls with the strike price around 10% of above what I paid. Pick an expiration, maybe 2 to 4 weeks out, collect the premium. If the call expires worthless, sell another. Keep collecting premiums. If my shares get called away like sold, go back to stage one and start again. And the rules are never sell a put unless I have enough cash to buy the shares if assigned and never sell a call below my cost basis, what I actually paid and track my premium across all cycles and check my positions every 15 minutes during market hours. Please set the schedule. If a contract hits 50% profit before expiration, close it early and give me a daily summary at Marcus close. All right, run this during market hours. Do nothing outside market hours. Great. That's what that prompt was. Now, I'm going to hit enter and wait. But that's about it. You're going to see after this is done. I'll show you in a bit. It's going to set the things in a schedule. It's going to keep running this. And I encourage you to maybe set up a paper trading account and then have a go at it. See how it will do.

All right. So, we built all of this inside the desktop app. And for 60% of the stuff we build, that's probably where we'll be doing most of our work. But I'm going to take this a step further because where we're about to get into, like managing multiple agents, building a second brain, running an entire AI operation, the desktop app alone isn't going to cut it for all of that. And to understand why I need to show you how this whole thing actually works at a high level and think of it like a house and inside the house there are rooms and each room has its own purpose. So in some rooms like a kitchen you cook in the office you work on project also your house has all your tools all your keys and let's say your connectors as well. What we're going to do is build an AI operating system which at a high level is kind of like your house. there is one project folder and everything else lives under there like you, your memories, your active projects, things from the past. Your AI operating system should be able to access the correct context that's live and happening over time and access it whenever it needs to. And the nice part about cloud code and the house, it works kind of the same way on how both of them have a lot of doors you can access it through. You know, like the house has a front door, a back door, the garage, you get into the same house and there's all the same content. And in the same way, you can use cloud code through different doors or different interfaces. So the desktop app is just one of them. You can do it through the terminal. You can do it through VS Code, Warp, your phone, a canvas, and AIOS. And those are all different doors to the same house. You'll walk through, let's say, the terminal door if you want something quick. And it's the same projects you're working on with the same Claude MDs. So Claude will know what to do exactly. And the reason we do that it's easier to cook in the kitchen than to cook somewhere else in your room, right? is the same way. It's easier to work on, let's say, content on a canvas with Claude and manage agents from a dashboard. Basically, different interfaces have different advantages and we're going to go over a bunch of them so you can work super effectively.

So, let me show you how to set each one up. Okay, so the first thing we want to do is make sure we have Cloud set up in our terminal. I know we initially downloaded the Claude desktop app, but that doesn't come with the Claude app inside the terminal pre-baked. So, I want to show you how to do that. And it's actually really easy. You just go to Cloud Desktop and ask, "Hey, can you install the Cloud Code CLI tool for me, please?" So, I'm going to run this. I want you to run it as well. There's other easier ways to do this as well where if you want to go to Chrome and then do claude code install, I'm just showing you two different ways. This is the second one. And right here, for example, what you can do is if you're on Mac, Windows, Homebrew, whatever, you can copy this. And then I'm searching my terminal up and then paste this in. You should be good to install it as well. And because this clot already has access to your terminal, it can do it for you if you ask, right? So you can just be like, do it for me. And you see it says installation complete. So these are just some easy ways to just download clot code uh in your terminal. The reason we're doing this is because if you have it inside your terminal then you can use clot code from anywhere including from your phone from the terminal. All right. Okay.

And to verify how that works is first in your terminal zoom in on the terminal a bit more. Okay. So in the terminal if you just type in claude you should be able to see this. And after you see this little guy, you're off to the races because now it's the same thing as using the cloud code on the desktop where we can just be like hello and you can use it um from here. And just a quick little tip. So let's say you're here but you want to you you working on a terminal session doesn't matter where it is and you want to use it back in the desktop. So if you do slash remote control and do enable remote control from the session look what's going to happen if you yeah if I reload this here in a half a second you'll see hi this is literally the thing we just talked about hello terminal it's kind of like they're all both of these are synced you see so that's just a quick hack I wanted to show you now let's move on to the more powerful ways we can use Cloud.

The second way a lot of people use Cloud and you'll probably be seeing this across YouTube um if they use it through VS Code. So, if you just download VS Code right here and right here, I'm going to open up VS Code. One second. And it's really simple. It's free. It's a free piece of software that every developer uses. So, it get makes it a lot easier. I personally use this sometimes, not all the time. But if you if you want to do this, what you want to do is first you want to go here. You want to search up claude right here in the extensions. So this is extensions. After you install this claude uh by hitting install here, what you're going to see is this little icon right here. And if you type it, then you can start using claude like hi. And the reason people do that is because they have access to their finder right here, which means they can open any folder that they'd like. And it's the same as really using it from the desktop. But I think the desktop is better than VS Code at this very moment. Well, the reason I don't like uh VS Code as much for someone that's starting new is imagine this. So let's say I open this random folder up. Okay, I can see all my files here. And right here, if I click any file, then if you click this orange thing, then you'll see Claude. By the way, you probably have to just sign in. Um, what you can do is right here. The problem is, let's say I say hi and I close this. It's really, really hard for me to get back that session and you might lose it forever. Okay, that's why using the desktop is a little better. So, I recommend that. But you'll probably seeing people using cloud code in this way where they just type stuff and then interact with these files right here. That's another way of people use cla and it's kind of the same thing. It's just through another window. Okay, people have different preferences.

So this way of working is arguably my favorite and I actually wrote this whole course with this tool and it's this thing I call clawed canvas or open canvas. I'm actively developing it and I'm giving it to my members. But we're going to be developing it actually live from scratch in this course later in the video with multiple agents and I'll show you how. But I'm just opening this application up. But you can see what this is is basically a canvas where I can like put in my notes. I can essentially make it light, make it dark. Uh I have a little file explorer on the side. I can go in, I can see what's going on. And this is like my script for um and you know the beats that I want to hit inside this section and this level. So there's a lot of stuff I can do here. Uh I can you know work on different projects have different canvases all these ways. The biggest thing I like here is right here. If you see uh this is cloud code and what I can do is I can open up multiple other cloud code sessions as well. So if I just hit shift t right here you see automatically it starts up cloud code. That's one of them. And it's the same thing as in the terminal. So this is cloud code that's in the terminal. And you know personally I like using codeex as well. So I can use codeex here as well if I really wanted to. But you know here I can just say hi something like that. This is claude. And the nice part about this is like for example can you um make a story of a cat and then insert it on the canvas as a markdown file. So it has context of everything that's happening inside this entire uh file and canvas. And what I can do is I can interact with it in that way. And you know, I like using these things visually. And I like using a lot of keyboard shortcuts. So this is what I'm trying to get at. It's like, you know, after you learn how to use Claude, what you can do is you can literally build your own tools. This is me making it dance around. Uh so it's just fun for me and I like building my own tools so I can work a lot faster. I can see all my automations right here. I can like, you know, draw around if I really really want to like, hey, uh, because I can bring in my tablet. You can make like little boxes and diagrams and arrows like this one. I wanted to do one first, etc. And even right here on the left, you see I even built in a little file explorer. So I can, you know, so you can, for example, uh, look through all your files. I can search what's going on. I can look into my, for example, agents.mmd. What's going on here? I can even drag it inside like for example right here and I can work with this really simply because essentially I can like take this it already knows but just for fun I added this little uh dragger icon where now if I connected this like claude will only work with this file you know so little things like that I like working with this thing called Poppy AI so I'm like hey I want to use my own cloud code subscription rather than pay for tokens so that's the reason I kind of built this and We'll be building this later in the course as well.

Okay. So, you can see right now, oh, let me make this bigger. Okay. So, light right now, it's telling me it wrote it. So, if I zoom out, oh, right here. It inserted it right here. Cat who owned the moon. It's it it's silly, but like, you know, this makes it really easy for me to like have multiple versions. I can be like, hey, spin up four different versions of this story or this hook, and I can judge them all, and I be like, "Okay, I'm going to work with this one." And then dive down from there. So, it really helps me making content. So the canvas is really cool because it's local and I can use it to visualize a bunch of things. But recently Claude actually just made their own agent managers. And how you can get started is first you have to go to platform.cloud.com and it lives here not in your cloud desktop app. The unfortunate thing about using this is the fact that you can't use your cloud plan and that's one of the biggest reasons I don't like using it because it gets incredibly expensive. Okay. And how you can get started is just by hitting this quick start button. And in the quick start, what you want to do is get its

Help to create your agent. But just to take you over a couple concepts. So there's an environment. And the environment is basically a computer that Claude will spin up and manage on its own. And that's where the agents you create will be working. Then there is obviously the agent, which has access to MCPS, skills, it has its prompts, its model, and it does, you know, whatever you need it to, and it does it by sessions. So you can imagine sessions like, you know, these are like different sessions. So different cloud chats, they're different sessions. Okay.

And how you can get started really easily is honestly just like this. You hit quick start, and then you can create any new agent. Let's say I want to make an agent that writes my YouTube scripts. So I can just be like, "Hey, I want you to create an agent that creates YouTube scripts for me using this skill." And the skill is just so I can give it some access. Let's say I give it access to this Ralph Wigum copywriter. Really simple. And then you just hit go. And the nice part about this thing here is right here, if you see. Okay.

So while this works on the background, the nice part about this agent manager is right here, if you see there is some skills. So it already has access to some skills where I can get it to make me powerpoints, excels, PDFs, docs. It already stores and all its files in one place, and you know, it already has some pre-existing context. So it'll remember stuff over time, and that's about it. It created the agent, and now if I go to sessions, right, right now I already have a session running, but if I can make a new session and say test agent, I'm going to create the session, and here, let me say hi, and you can see the reason people use this is because what happens is you can keep a record of everything your agent does in the cloud, and that means you don't need your computer to be on at any time, right? And cool. This is about it. It's like any other uh thing you're doing with Claude with the schedule tasks, all of that, but just running in the cloud with some more visibility.

But I want to take you over some things that are really important to understand, because because that is really how you understand when you use this and why you use this. So, I just want to quickly take you over the pros and cons of when you would use this. And the first pro is the fact that you can do multiple things at once with multiple agents that are running in the cloud. That means your computer doesn't have to turn on. The second biggest thing on why it's so useful is because it kind of behaves like OpenClaw, where it has built-in tools. There's session tracking, so you have a lot of visibility, and it's made and built especially for enterprises, really. And the fact that it remembers context over time is really huge. But the cons are really what kill it for me personally or any small business, because it gets expensive incredibly fast. The $200 I spent on my cloud subscription that I can't even put into this, that with that $200 subscription I would get, let's say, $5,000 of use. But with this current one, I would have to spend that $5,000. So, and one other big problem is that Claude will lock you into their ecosystem. That means you can only use Claude. That means if, let's say, Google or OpenAI just gets significantly better the other day, cuz you know, there's always a new best model, I cannot use those models, and that's a big problem, right?

So I'll show you an efficient solution that I'll cover later in the video, where you know, a lot of people have been buying these Mac minis, and these Mac minis come and become super useful, because that kind of becomes your virtual server in the cloud, and you can use cloud code and everything else to run an agent there, and you don't have to pay any of these costs, and everything works locally. So you have, so you have all the control of everything together, and saves you so much money in the long run, and gives you a lot more control. So let's get into that.

Just like the control center we just saw that Enthropic has in the cloud, I just made one that lets me control all the stuff locally. So this runs on my Mac Mini, and this is basically clock code that runs all the time that I can text from my phone. And here's some cool things I do with it. It's called AI agents control center, where if I, let's say, for my carousels, this Instagram thing, it's been developing and running my carousels on its own. This is how I manage multiple of my agents. I can like go to each project, see which agents are being active there. I have a little dashboard where I can like, um, you know, go through uh all the agents. If I want to be a little cute, I want to see all these guys right here. There's a conbon board where I can, you know, visualize what task is going on. There's a little vault where it's storing my keys. I need to hide this. Um, and you know, I also connected it to my Obsidian, which we're going to be doing in a little bit.

But essentially, I have my projects, everything that I've been working on, all my skills, uh all in one place, the plugins where I can see, this is really cool, is I can have a little calendar for scheduled tasks. So, I can essentially see when my automations and my agentic workflows are running. And here I have a couple different things. I can look at my usage, my memory, where, you know, my files, which ones they're connected to. So each time I text from my phone, because this also works from my phone, where uh right here in settings, if you just go to Telegram, set this up, uh you can start texting this thing. But, you know, I made a whole system where that I can just like work from anywhere. And for a little bit of fun, here's like, you know, I can talk to my agents through Pokémon. And here's like a cloud code session. And I can see some of the files, memories, etc. Uh, it's just me having fun building a lot of things. And right here, if you need help, you go to uh my cloud club, etc. But essentially, this is what I used and a bunch of agents right here through this dashboard to build that. Uh, I'll show you if you guys want this file. It's also again inside my Chrome right here. Install your operating system. I have all the links there.

So, I'm giving that away and helping people build these. So, if you guys want something like that, you can check it out inside the community below. You should get the link there. But honestly, you don't need any of that, because what I'm going to teach you right now to do is probably the most important thing you can do, which is setting up your AIOS such that you have a second brain that always remembers you. And you can take that, and we're going to be building that using Obsidian and Terminal. Okay.

So, what is an AIOS really? So, an AIOS is really just a layer that you wrap around your work that has actual context on your life, evolve with you over time, and can actually do things for you without you having to ask. Basically, it knows your projects, your goals. Let's say you take a meeting, there's some action items, and it'll action it on your behalf, and it knows the way you think, right? And why you'd really want one is because we all have a lot going on in our lives. Like, for example, in this project, I have a bunch of stuff that I'm working on. I have a YouTube channel. I have a community I'm running. I have a SAS business. There's personal health, relationships, all these things, right? And even like a whole bunch of research. It's basically all my knowledge lives in one place, and the AI has full context. So I actually interact with my one through Discord, Telegram, and iMessage, like each time I'm feeling something different. And the nice part is I always have access to my project, and I can drop in little ideas here and there, and it'll be stored and tucked away inside the right folder, and it manages itself.

But the thing is, how do you actually know it's providing some meaningful value to your life, right? Well, my friend Liam Oley actually laid out three really intelligent KPIs to judge if an AIOS is really helping you out. And I took this framing and built an entire system on top of it. And this is how you can really accurately measure if you're getting some meaningful return out of an AIOS. So number one, it's essentially the time away from your desk. How much of your day can you really run your business without you having to sit on a computer? Number two would be the percentage of tasks that are heavily augmented or fully automated by AI. And number three is revenue per employee. Basically, as more people in your team get access to these tools and AI, they should be able to produce way more output. So ideally, we need less people. So if those numbers are moving in the right direction, that means the AI operating system is providing meaningful return.

And for this to work, the AI actually has to know you, right? It needs context on what you're working on. And I've built this entire system on the folder structure and all the stuff that I'll share with you later so we can help you build it. And what we do is we build it around your northstar and your goals. And basically, it will automatically get populated with, for example, your meeting transcripts and stuff that happens in your day-to-day. And that's why I call this thing an evolving brain. It's like, literally, it's it's literally like an encyclopedia of me. You can see here, this is a YouTube video I just made. And from here I can see the transcripts. Let's say I just, um, there was a new one. I can see what links to each other. For example, it knows the people in my life. Uh, if I released a new video like the trading one, um, here it has all of that. It just pulls it in. It has access to all my email and whatnot. It's a living encyclopedia of me and every interaction that kind of compounds over time. So, it remembers the most recent things that are important to me and my goals.

And you know, there's a lot of ways you can visualize this. Hopefully, you saw this one earlier in the video that I showed that I use sometimes, but this one I've really been liking as well. And this one is built on top of Obsidian. It's free and it's open source. And it's honestly one of my favorite things. So, let's get into how we built this. So you can set one up for yourself and then even control your AIOS from your phone, right? Okay.

So to first install the AIOS, what you, what I want you to do is first go to obsidian.md and download this application. All right. So when you open up Obsidian, this is what you'll see. And what I want you to do is create a new vault from a folder. So the second option right here. So wherever you want to, let's make a new folder. Let's call it, I don't know, evolving brain. And open that. Now, what's going to happen is you're going to be met with all of this stuff, right? And the fun part now is we're going to need Claude to go mess with this Obsidian. Okay.

So, now what I'm going to do is I'm going to go back to my Chrome and right here, uh, just to guide you, in my classroom inside Vault inside Obsidian Second Brain, I have this link right here. So, evolving brain template. What I want you to do is copy this. Go to, go back to Claude right here. Select the folder and then let's choose the same folder we just made, and then just ask, hey, can you set this up for me? All right. So this is the first part. Ideally, it's pretty easy. And let me just give you a quick tour of this evolving brain template. So even if you want, you can build it your own. So this, what this is, this compromises of the whole structure that I've built out for you guys. So, it makes it a lot easier to organize your life, your projects, all these things, right? Plus, it has a couple Obsidian plugins that you might need. So, there's two specific plugins that you may need. One of them is this. It's Obsidian Terminal. And what that does is basically it gives you a terminal inside Obsidian. And what that, and what that does is gives you a terminal inside Obsidian. So, you can start running cloud code really easily. The second one, the second one is this one called agent files. And you don't need to know anything about this. I'm just showing you what it lets you do is, let's say you have a lot of these claw chats. And if you're working with, let's say, 10 different agents in the future, you'll never lose them. And you'll have one place to manage your skills, your agents, your commands, all of this stuff. All right.

So, what I gave you is just saving you a bunch of time and making it so that you don't have to do a bunch of stuff. So, this is saying it's done, which means if I go back to Obsidian, you will see I have all these different folders. Right now that we have that, what I want you to do, this is very important, is quit Obsidian and then relaunch it, because to set it up, it made it so that I gave you guys some plugins. So, now it'll have some new abilities. So in Obsidian, you can see right now just a couple of small things you can make like, um, Excaladraw drawings, and there's one that's really, really important where on the left hand side you see this button, it says open terminal. If I hit this button right here and if I do claude, if I just say claude, claude starts up, and I can start acting with it. Right? Before that we didn't have access to claude, and now Claude has access to everything. So, that's one more thing.

The last thing I want to show you is this button right here that says agent files. And agent files is basically a collection where you can manage your AI agents, your skills, all your commands, everything through anything you might be using. So, so for example, you see I have agentic workflows, all my projects right here. Any tools I have, I can connect it and I have visibility here. Any skills that, you know, we were just talking with and using throughout, and any skills I also I have, and most importantly, conversations. So for example, right here, uh, you see I was, this is the Claude chat I was dealing with inside Claude right here, and if I go back, if I hit copy here and if I hit this and then paste it here, what's going to happen is now I can continue that conversation, and it just makes it really easy for me to manage any conversations, whether it's from the desktop, whether it's in, I've have it in Obsidian, in terminal, whatever. This you don't need to use as much because it gets a little more complex. But all you need to know for now is right here, just learn to do claude, and you will have claude up and running here. Now if I hit command G, you see it already has a small little brain. There's not much in this. This is basically like, uh, all the little things that I've installed for you, like the fact that it's going to be dreaming it. You can process your meeting notes, all those. So this is just a template brain that we are going to be building on top of. All right.

So let's get into the next step. We're, we're going to be preparing our data so we can actually make a brain and a wiki for ourselves. All right. So to prepare our data, there, our, our brain or our digital footprint basically lives around a lot of different places. Maybe it's in Slack, it's in Discord, Gmail, Calendar meetings, your Notion, Obsidian, whatnot, right? And even your ChatGPT chats or your cloud chats. So, I'm going to be showing you how we can step by step go through it. And just to make it a little easier, uh, this is of the board that I'm going to be putting inside the community below just for your help. And a couple different things I just want to go through with you before we start is number one, I built in some identity files. So if you just try to answer these questions, there's a bunch of questions that I know that work really well, and then in Obsidian put it inside, um, your identity, then what happens is it's going to perform a lot better because it will already know you and the most up-to-date version of you. Number two is what we should get for context is all of this stuff. So for example, anything from our chat and text messages, iMessage, and I'm going to be showing you how to do a bunch of this stuff, right? Any AI conversations, any notes or documents. So, for example, if you like, let's say, store your life inside Apple Notes, you want to export all of that, right? Or your emails, your calendars, all of these. And it's really easy to do so. So, I'm going to show you how, you know, we can get AI to like dig through everything and then build out your own life, right? The first thing I actually already built for you. All right.

So, the first thing I want to show you is this is just the easiest thing. If you want to go to ChatGPT, if you used to use it in the past, and I'm assuming most people did, what you want to do is you want to go here, hit this button on the top right on the bottom left, if you hit settings, and then right here, if you go to data controls and then hit export data and confirm export, what's going to happen is in your email that you signed up for ChatGPT, you're going to be seeing this email from you where they're going to export, import all the data and everything it knows about you. Okay? Then you're going to get this email where you can download and export your data. What I want you to do is let's say go to Finder. Okay? And then let's make a new folder. Call it "my brain dump." So basically, we're dumping all the context. Okay? So what you want to do is you want to pull that inside this folder. And we're going to be pointing our second brain in this folder. So it can go and it'll know where this information is. Claude will take it and organize it. Okay, that's the first thing.

The second thing I want you to do is go to this website called Composio, and it's free. So if you go to the Composio, and it's free, and I'm signing up, and this is what you'll be met with. So what we're going to do is integrate Composio with Claude, and this one I'm going to skip for now, and you know, you can connect your Gmail, your Slack, all the stuff. So right now I'm going to continue. But what I wanted to show you is this. If you go to connect apps, okay, right now let's say your data is in Gmail. All you're going to do is like connect your Gmail, and then this is where all our data is. So for example, let's say I can connect my Slack, my Google Calendar, if I'm doing anything on Twitter, my Google Drive. Uh, and then there's also if you want to get onto finances, so it knows about your business, you can connect it to your QuickBooks or your Stripe. Uh, so what I found is Composio has the most amount of things, and after you connected with all the things that or connectors that you need. So for example, right here I just connected with my Gmail. What you're going to do is you're going to go to install. Okay. And install, I'm going to copy this command. And then what I'm going to do is go back to my cloud. So I just went back to my Claude, and then I'm going to hit customize. Then I'm going to hit connectors. And then what you want to do is you want to add connector. Add custom connector. Let's say composio Alex. Okay. So after you hit add, you're going to see something like this. And in here, what you want to do is just authenticate the same way we just did before. And then make sure you allow everything. Oh, I just want to make sure I allow this. Okay, great.

So after you do this, your composio should have access to all your data. So for example, uh, right here, let's say I choose my evolving brain, and then I can ask, and then I can make sure, hey, go to connectors and see, okay, I have composio, now I can just ask things about like, hey, do you have access to my YouTube data, for example? Hey, do you have access to my YouTube data, what was my last video and how did it perform? Um, and I'm going to use Sonnet because it's so fast, so what I wanted to show you here is in my connected apps, because I connected my YouTube, it can pull data from my YouTube or any app that I need. Right? So, let me go back to my cloud, use composio. See how it asks with browser or composio. That's how we can start gathering all the data. Then what I'm going to show you is I'm just, we're just going to be entering a prompt to our cloud, because and then it's going to use all these tools, connect all the data, and then essentially make our brain, right? Okay, so you see right now, uh, I literally just released, I have those, uh, my shorts auto posting, so it just detected that the views, the likes, all this stuff, uh, it knows it's a short form content, all these, it's been live for 2 hours, so it has live access data. And the reason I'm telling you to use this thing, uh, again. Oh, maybe I glossed over this, but let's say you're recording your meetings. So, let's say Google Meet, uh, or Fathom or Fireflies right here. You can connect all of these, and then you can just ask and put it on a cron job like, "Hey, pull in my meetings for the day." And then your brain will continually evolve with your meeting. Okay, that's how we essentially make it so that brain is constantly updating with our latest information. Okay.

So once we have those connectors set up, now all you want to do is be like, hey, can you use composio and in this evolving brain make a bunch of markdown files such that you collect all the information from my YouTube, maybe get the transcripts, put it there, and essentially what I want is to build my second brain. So go through all the apps that I'm control, uh, connected in composio, and then make me my brain. Okay, so that's what I want you to do. So just go through each one of your apps, and then just make sure you have a folder that's populated with your data. Okay, the other easier way you can do that is just honestly just dumping it in. Like think about it, if you have any of these files, just dump it in, and you should be fine. But the most important part is that right now what we do is we have essentially Claude, it has connected to all the apps you use. So it can be Slack, it can, uh, be Google, it can be your email, whatnot, everything, you're right, also your meetings, which means it can evolve over time. So, okay, so just to make sure I walked you guys through this one step at a time.

So imagine I run that command and I made a lot of mock files for Alex. So about his business, about his life, his personal relationships or his health, uh, little projects he's done like the carousel we just did before, all these things. And I'm going to use this as my brain to build our second brain. All right. So what I'm going to do right now is, let's say, I'm going to take, so I'm going to take this file and I'm going to go back to Obsidian. And then right here, I'm going to paste that. And then, hey, can you use this to build out my second brain, ingest all the data, and then make me my wiki as well. And then right now, it's just going to be loading for a while. So, what it's going to do is going to like use your finder. So, let's say all your files are going to go, everything. It's going to build out your all of these things based on what you know, and it's going to make a lot of guesses unless you want it to. And then what it's going to do is going to have and expand this brain by a lot more. So it has ultimate context of what you need, and I'll show you after this how we can actually get this evolving over time as well. All right, great.

So this is just finished running, and you can see the brain is a lot more populated and there's way more connections because I gave it all that data. And if I actually go out to my Chrome, what it does, it opens up this wiki for me. And you can see it's a Wikipedia styled index where for Alex it has a bunch of different things. So for example, you know how he's working on all these, uh, little things, and it has the direction for all the people, his personal health, his side hustle consulting, that all he's doing, his marketing operations, his AIOS migration, so all that data is there. It's really cool, it's all linked together, it's graphed, and it's very easy to search, and the even the people in his life, how he has, uh, connected to who, right? Okay, so this is just a quick demo, demo of the second brain.

Now, how do we actually use this over time, make sure we can access it from a lot of different places, and whenever we have, let's say, a quick idea across any device, we want to make sure this captures it, right? We also want it to be evolving over time such that, you know, we have a new meeting, let's say, with our boss, and we want that stuff to go in this brain. How do we do all, do all that? So, let's set that up next. And just to, uh, show you a quick couple fun things, it even like, you know, took out little things and made just random goals up for Alex. So, this is really cool. I highly recommend you to try this out, because it, it's really, it's really a lot of fun to see AI map out your own life, your aspirations, all these things without you even having to know. So, some things for me is like I've literally seen it find out stuff about myself that I didn't know, and I was like, "Oh man, I actually do that." So let's go to the next part.

And the next part is where it gets a lot better and evolves over time with you. So the way we're going to make sure this is self-evolving is with a few concepts that I want to show you, which we've already actually already done, right? So all it is is we're going to be setting up schedule tasks in our cloud desktop or any application that we want to make sure that we just say, hey, use our composio or our data and our connectors to pull what we need to from the latest thing, and then process it into our second brain. Right? So it's a really easy prompt. And for example, what I need to do now is, let's say, go back to Chrome, and I'm going to go to Composio. And right now, let's say I want my meetings to be always, uh, every week to be ingested, right? So I'm going to look for Fathom, because I use Fathom to record my meetings. So I'm going to connect Fathom. And you know, you can do this if Claude has its own, uh, MCPs, just just use those. So for example, inside right here in customize, again in connectors here, uh, if Claude has their own, just look for these first. I think it has Fathom. No, it doesn't have Fathom. That's why I used Fathom in Composio. Uh, that's why I remembered. And after you have basically, uh, Fathom connected to Composio, basically Composio will connect to Claude, you know, like, let's say your Gmail, uh, I'm going to call this, e, all of these, let's say Claude doesn't have access to email, you can put all of these in Composio and then connect them. Okay, so that's the goal.

Then all we're going to do is remember, because we're in the same folder, right? We're inside this evolving brain folder, whether we're here, whether we are in Obsidian right here. Evolving brain. The nice part is it, it doesn't matter where you're using Claude from. The fact is you can use the same Claude and the same connectors in this folder, and it's going to keep happening. So what I'm going to say is, let's say I'm going to just, just to make it easier for you guys. Hey, can you set up a schedule task so that every week, at let's say Saturdays at 6 p.m., I want you to pull down all the meeting transcripts using Composio and Fathom. Pull down all my meeting transcripts, and then I want you to pull down the data, save the transcripts in here, and then process it so that it lives in my second brain and it all gets indexed. Okay? Use all the skills that you need to. That's kind of the prompt that you want to run, and then you know it's going to keep going, and after that's done, you do this for all the apps that you need to and set up these syncs. And after you set up these syncs, what's going to happen is you're going to have a self-evolving brain that you don't even have to touch. Okay. And you don't have to touch. And it's going to keep getting better passively. And you're going to see this just getting better and better and better.

Make sure though, this is one thing I want to make sure you know, that remember, these scheduled tasks can be remote or local. Okay? Ideally, what you want to do for these tasks, if your computer's not on, on, you do a remote task. All right? So, a remote task for example, this is very important. When you do these scheduled remote tasks, let's say call it, um, second brain sync. This is another way I can do it. And you can also install a repository. What you can do right here is remember that you, I put my second brain, my evolving brain, inside GitHub. So I can sync it there. So all you would have to do is, let's say, install GitHub app, and then I'm configuring my GitHub, and then now it has cloud, should have access to all your repos. Now if I go back now, uh, let's select a repo. So then you want to accept the permissions. Okay. So it should be good here, and this is just my organizations. Then I'm going to accept the permissions, and it should be good here. So after you install that Cloud app, which you could be like, is evolving brain sync, and then type the prompt here. So it's like, hey, I want you to go through, use the Composio MCP. The reason I said that is because I wanted to use Composio and get my Fathom meetings, and then sync it in my GitHub to my evolving brain. Make sure what you do is you pull down the transcripts, and then what you want to do is process them and put them in the right places with brain process. Okay. And then all you need to do is just hit create, and you'll be good to go. And it's really as simple as that. So you'll just have tasks that are running in the cloud. And if you do it through here, it becomes a lot easier because you don't even have to honestly keep your computer on, and your life just becomes significantly easier where these prompts are running and it keeps updating your brain with any app you need over time. All right, so this is cool.

Now, let's move on to the next feature, where let's say we want a quick idea, we want to jot down, we just want to send a voice memo where we're just making notes, all of the stuff. How can we access our cloud and work with our project from anywhere we want? Right? That's the next step. All right. So, to connect the channels, I actually wrote a small little guide that I'm going to open up in my community for you. All right. So, we have the guide opened up on the left, and this is the terminal, and this is in Obsidian right here. Right here, I have Claude open. And if you hit this plus and then just do Claude, if you want to open up a new session, that's all you need to do. But let's follow this guide.

So, first what we want to do is make sure we have bun installed. And this ideally we do in the terminal because it says it right here. All right. So right here I have, do I have bun? And I'm just copying the next one. It's saying I'm good to go. All right. So I'm good there too. And of course I have Claude. So I'm going to skip part two. And then what I'm going to do is it's asking me to start a new cloud server. So when you do claude, what happens is it's going to start up Claude. And then we want to make sure we have this plugin. So I'm just going to copy paste this. And it says successfully added. Then I'm going to copy paste this. And this one is added, too. Great. So now we're ready to start our Telegram linking. So we go to Telegram. Then we'll see. First what we need to do is make a new bot and start BotFather. So I'm going to open up my Telegram app and I want you to search for this BotFather. All right. And then I'm going to hit start. I'm going to do slash newbot, right? And then we need to give it a name. This part is honestly sometimes kind of annoying because so many bot names are taken, but I'm going to say Alex, um, demo bot 1 2 3. Uh, you need a bot name at the end. Demo 1 2 3 1 2 3 unders and then bot. Oh, I guess that didn't work either. But you should do underscorebot. Let me side hide this. Okay. So this, this crazy name worked. Um, great. So, so this part is done.

Now let's move on to this one. So in cloud code, let's run this thing. Plugin install for what? And install for you in the user scope. Cool. So we just install Telegram. Now I'm going to reload these plugins. Okay. Now, what I'm going to do is paste this in and then I'm going to paste in the token, right? And to get the token, I'm going to go back to Telegram and I'm going to copy this thing and I'm going to paste it here and then hit enter. And I'm going to hit yes and always continue to install. And I just hit yes again. All right. So, next what we need to do is now we need to exit our cloud code. Uh, so I'm going to do /exit. Now, what we're going to do is relaunch our Claude with this command. I'm going to hit that and then let's go. Great. Now, this should be linked to our Telegram. What we need to do now is open up our bot. So, let's open up our Telegram again. And then, let me make this bigger so it's a little easier. So, now uh so I just expanded this just so it's easy, a bit easier for you to understand. But right here, what I'm going to do is in the search, I'm going to look for my bot. So, I found my bot. I'm going to hit start. And then what I'm going to do is right here it's saying to pair, DM me anything. You'll get a six pair of code. So, I'm going to say hi. And then it gives me this code. And then it says run in cloud code. So, I'm going to run this in cloud code. And I'm going to let it allow, allow. I'm going to allow it to do that. Great. So now Telegram should be paired. So if I open up my phone, I can see the same thing right here. And I can just say something like, "Hi." And then I can be like, "What do you know about Alex?" Now, just to show you that this is connected to the AIOS, I asked, "What do you know about Alex?" And you can see it working in Obsidian right here. And remember, because we set up our AIOS such that right now, it's connected to this brain right here. It's going to be pulling from this brain. And we have our self-evolving brain. So, if you want to essentially right here dump any memories that we want, we can control our cloud code. And now we have a portable system that we can use from anywhere that self updates over time. Okay.

So, and just if you wanted to see the response, this is, this is really funny actually if I go back to my Telegram. So, it says there's two Alex's in the brain. Alex Riviera, that's you, that's the vault owner, marketing manager and booked. So, that's what we've been working through all the, uh, time and Alex Chen. So, it's, it's so cool because it knows everything about us and we can just keep going from there. So, let's pair the second channel now. All right. Okay.

Now, to pair Discord, I'm going to open up a new one. And right here, what we want to do is first go to Discord. Then, first I'm going to open up my Chrome browser and I'll put it here. And then I'm going to go to discord.com/developers/applications. Then let me log in. Great. So I just logged in. Now I'm going to create a new application and let's name it Alexbot and hit this thing. And I am a human for now. Okay. Now we have all this stuff. Now we have to, we just did this part. Now we have to go to bot tab. Bot tab. So we go to bot tab and then reset token. So and then here's my token. Then I'm going to copy it. And then what we want to do is we want to scroll down and turn on message content intent. All right. So right here I want to make sure I give it admin privileges. That's number one. And number two is members intent. We're going to give it that message content intents. And I'm going to give it presence intent. So I gave it all of that. And then I'm going to hit save changes. All right. Okay, great. We're good there.

Now, what we want to do is we want to install the Discord plugin inside Claude Code. So, if I open back up my Obsidian, this is a new terminal. Remember, we have our, uh, Telegram running here. So, we can make a new one. So, I'm just going to do claude and I can do plugin install Discord and let me install that. Great. Now, let me reload the plugin. Okay. Now, I'm going to copy this. And then I'm going to delete this token and put my token in that it initially gave me. Now, I just hit allow all. All right. So, that's done. Now, it wants us to exit. So, let's do that. Let's exit. And then we copy this. And then we paste it. And we start it up. Great.

Now, let's go to Discord, make a channel. So, we're going to go to Discord. Okay. So, I just, all right. So, I just opened Discord. Then, I'm going to make a new channel by scrolling all the way down and making my own server for me and my friends. And then, Alex bot. Great. So, I have this. Now, I need to invite my bot in here. Okay. Okay, so once we have our server, if I go back to Chrome and right here, if I go to installation and right here, if I copy this link, if I make a new tab and then paste it right here, you'll see you can invite your bot into your server that you just made. All right, so that's when you want to hit authorize. Cool. And then let's open the app and go right here. Then what we want to do is go back to our Obsidian. So, I'm opening my Obsidian right here. Uh, let me open this thing back up as well. And just to make sure I have this again, I'm just going to run it. I'm going to add, make sure I have Discord. Okay, I should have it now. I'm going to reload the plugin. And now I'm going to make sure I configure my token, discord/configure. And I'm going to give in my bot token again. Whoops, I messed up here. So I'm going to copy this again. I'm going to paste this right now and then paste this again just so it makes sure it does the right thing. And yes, always allow. Okay, so cool. You can see all the stuff happening.

Now, if I go back to my Discord. Okay, so once Alex is inside your server, what you want to do is you want to click on Alex and then, sorry, you want to right click on Alex and then message. And then right here, uh, I'm going to blur this, but it's going to send you the pairing code. And then what you want to do is you want to just copy the pairing code and then go back to Claude. So Claude, and then paste this pairing code. All right. And just to make sure, you can also do both at the same time. And let's run this and then we can test it. Okay. So let me go back and let me try testing it. This should be good to go. By the way, way guys, uh, one more thing I want to make sure, uh, you know, is right here in user settings, right? Make sure you have develop. You go to developer and then have this turned on, because if you don't have this, some things may not work or you might not see something. So, I want to make sure you know that. Okay. So, Alex just got added to the server. And now, if I say, "Hi, what, what was the last carousel I made about?" And now I'm just getting it to search through its memories, you know. Um, I don't know if I actually fed that in. I just want to see what it says. That's how we make sure we connect our Claude to our Discord. Now I can talk with Claude through my Telegram and my Discord.

Just a disclaimer, wanted to let you know that this part we just did was actually really hard. And I want to make sure that you don't get stuck in the weeds of actually just improving your setup, which is basically like sharpening your axe constantly. I want to make sure we get into actually building stuff with it and doing the actual cutting with the axe. So, we'll get into that right after we're done. But, if you need any help at all whatsoever, I'm going to be inside the Claude Club. You can find that in school.com/cloud. And there, if you need any help, we'll help you out build whatever you need to with Claude and make sure you can get a lot of your work done with AI. Right?

And to give you a bird's eye overview of all the cool stuff we've done right now, check this out. So we initially developed our AI OS by giving it all the contexts from our old AI chats, our Gmails, our text messages, our notes, all the stuff. And we were able to put it into essentially folders and get Claude to do all of that heavy lifting. Right? Then what we did is we made sure we made an evolving context over time. So we connected to our meeting notes through the connectors and Composio and how we set up a schedule such that it evolves over time and gets better as we're working on newer and newer things. Right? So that's one big thing. And we even have our automations and everything placed in there. But something really cool is that now we also can do our projects. I spelled this wrong because inside the AIOS in the folder, all of these are also going to be meticulously kept, and we can work on it over time through anywhere we want, because overall how we set up all the interfaces to work essentially with our AIOS. So it always knows who we are and has context of what we're doing. So, and we've done that over things like Claude code through Obsidian, through let's say Claude the desktop app, where we just have choose the evolving brain project. We can

Work on it from co-work, uh, we can work from it with code and Telegram, iMessage, all this stuff. And we want to take this a step further. What we want to do is if you have a Mac mini and you build all we did on that, it's a server that's always going to be running. You have your own computer and cloud. And then if you use that with Jump Desktop, which is really easy, is just a screen sharing software, which I'll show you a short clip of a setup I have, such that I have access to a computer anywhere I go just with my phone, and what I am allowed to do is interact with this whole evolving system wherever I'd want to, and I have AI working for me through a lot of things like scheduled tasks and automations through these projects and workflows, and we'll take that even further. All right.

All right, guys. So, I wanted to show you something cool. That if you actually do have the AIO OS set up, what I do is I have my phone, you see, and my wallet and all the stuff. And what I can do essentially is, um, now I have a keyboard and a mouse that like folds up, and I'm ready to go and work with it from anywhere that I want to. And that's really cool. That I, I love like, you know, being able to be on the go and then just have my computer and my mouse and I have a little portable setup. And this is my Mac Mini. You see? So, for example, I can, I can use all the apps that I need to, everything there like a Mac, for example.

Congrats. You just built your first AI operating system. Let that sit for a second. Most people using Cloud are just typing one-off prompts into a chat window. You now have a living, breathing system that knows who you are, knows your business, and works for you even if you're at your desk or not. And let me hit on a few things about what that actually means. Because first, it's portable. Your AIOS lives in folders. It's synced across GitHub. And that means you can open it from your laptop, your phone. You can move it to your Mac Mini or a random server, anywhere. You're not locked in. Then it also has evolving context and memory. So every conversation you have with Claude, it's going to grow with you. And as you have your meetings, your emails, it learns from it over time. That means 2 weeks from now, Claude will know your patterns and your projects better than any employee you've ever onboarded. Third, it runs your automations based on what you know. So the AIOS doesn't just store information, right? It's acting on it. The channels you set up, the briefings, the weekly reviews way back in the other parts of the video, they're running on their own and they're going to pull from your brain and make it better as you go if you tell it to.

So, for example, like if you're an executive now, what you can do with the AIOS because you everything is connected, you can spin up dashboards on the fly and just ask something like, hey, what is my team's audit look like? What is our P&L? And it uses the right connectors to pull the data and builds up a dashboard that's live in front of you. This unlocks the three KPIs we've been talking about, where now you can work from anywhere since it's accessible from your devices or anything. You can make sure you can automate or augment any task with the agentic workflows we've been building out and just learning and building on top of those. And three, you can measure your revenue per employee because you can give them access to the same tech and just ask Claude, hey, what's our revenue? Who's the employee that's using the most? You can use all this to basically give your team an AI-powered OS as well. And you can have a much more efficient business just by doing this.

Now, we've built the skeleton. Now, let's start using it. And let's go back to Alex. You know, with Alex, we've been building all these automations for a while. And let's say now he realizes that his skills are sellable. So, Alex decides to start a side hustle, maybe building automations for other people, like content pipelines, operating systems. But to sell any of these services, you need a couple things. First, you probably just need a website so people can find you, understand what you offer, all of that. Second, you probably need a product to sell. So, any skills, automations, even a SaaS that we're going to be developing soon. Third, you need a way to collect payments. And fourth, you need to make sure people know about you. So you need to develop marketing assets and content and set up pipeline so AI is working for you. We're going to be building all of these from the AIOS using Cloud Code. So first, let's get on to building our website.

And to do that, what we want to do is we can use the desktop app to do this, because all we need to do is come here, choose the evolving brain. And this, just for reference, is using the same project here. You can see evolving brain, which means it doesn't matter if we're working from here or if we're working from here. It's the same project. All right. And one thing and one more thing I want to show you is this is where I think it really gets fun, because this is what like if you just prompt a website normally, what it looks like. It's kind of generic and it doesn't look very good. So just to show you some cool things, like this is Dribbble, and you can see a lot of cool designs here, and I'll show you how to replicate all of these. So I was scrolling Dribbble and I found this really, really cool website. Check this out. So, if I scroll, this dinosaur comes out and bites this thing. It's really cool. And I've been just scrolling this website, and you know, I love the design. It seems super professional. I would want to, let's say, replicate it. So, for Alex, let's replicate and create a website for, let's say, his agency or his side business. All right?

And it's much easier than you think. But just to show you, normally when people make websites, it kind of ends up looking like this where they just prompt, maybe they give it the website as a repo or an example. It's going to end up looking really basic. And because we don't really have a really good website skill. So in my community, I actually have this website cloner skill that you can use. And what we're going to do is we're going to use this cloner skill. And I'm going to show you how to do that to build a website that does stuff like this. And to do that, it's really simple. I'm going to go back to my cloud. Okay. So, I'm going to start a new session inside my evolving brain, paste this thing in, and first I'm going to install the skill. So, I'm just going to ask it, hey, can you please install the skill? What I'm doing right now is right now for Alex, I'm as Alex, I'm trying to start a side hustle where I'm going to be selling, let's say, my AI operating systems and content marketing automations that I'm building. So for that, I'm trying to build a website. So can you install the skill and then I'll give you an example that we can copy together? Okay. So this is the first prompt. Really simple. All I'm trying to tell it to do is install the skill.

And just to show you the skill and to show you what the skill can do, if I go back here, uh, if I click this thing, you can see like this is the original website they gave it. And then the clone becomes near perfect. And that is super impressive because most of the time, uh, I want to make sure you guys aren't making websites that look super vibe coded. Let me show you how we can actually build those really simply. Now if I go back to Claude. Okay. So now you see how it has Okay. Now you see this is really important. It installed the website. Now I'm going to give it context. And I'm going to do slashclone website. And I'm going to go back and then give it this website. And then I'm going to be like, "Hey, uh, what I want you to do is build me a website like this for me selling my AI services and maybe I want a cool headline like 'Killing Headcount' or something like that. And what I want you to do is build out that website in the same design so I can start sharing it with folks." Okay, really simple prompt. Um, of course, results will vary if you guys, you know, we'll probably get something out. Maybe we have to go back and forth with it a little bit, but overall at the end with the skill, you should be getting something really cool.

Okay, so it just finished and it came up with something really cool. Check this out. "Great AI, kill your headcount." But if you It has this little dinosaur as well. And if I scroll down, you can say, "Hey, I'm Alex." So because it has your context and it knows what you work with. So it knows about the content pipelines, all this data engineering, the rag systems, things you do, how you work. It's really cool how it designed it. Now the next parts just become, you know, changing some copy around, maybe some image, adding some images, and you're good to go. So this is number one. But number two is, let's say I want this website so I can share it with people. Right now you see this thing after I made this website locally right here, it says localhost, and localhost is basically this website is hosted on your computer. That means other people cannot access it, right? So to do that, what we need to do is deploy it. Right?

So there's two things we need to do to go to deploy it. I want to show you two things. Okay, the first thing we need to do is make sure we upload this to GitHub. So, hey, can you make this entire website project, make a project in GitHub and then upload it there, please? And if you're confused how to use GitHub, you should already have access to it when we did it before. Uh, remember right here, there's two ways. One easy way is also if you use Composio. So in Composio, if you go to connected apps and you come here and then connect your GitHub. So now it should have access to your GitHub. And remember, GitHub is just somewhere we store our code. We can pull it down. We can push new versions up. And then if I go back to my cloud and then I see right here I have access to Composio. We've done this before. And I'm going to make sure I give this always allow. Now what I can say right here, hey, can you for this website, can you, uh, put this website onto GitHub, please? Um, make a new repo so I can see it. Okay, so this got done. It literally says "Alex AI Service Website." Okay, great. So you see now it pushed this code up to GitHub, and what happens is now this code, that living version exists there, and each time I make a change, everything will be recorded, right? So now that it's on GitHub, what we can do is deploy this website, because initially it was on localhost, right? And it's really easy on how to deploy it. And the easiest service for most people is normally Vercel. So if you sign up for Vercel, see, you can actually already sign up for GitHub. And what you can do is you see I'm already deploying some of my sites on Vercel.

But what we want to do is I want to show you this here. Vercel has a really cool MCP. So we want to go to customize and we go to connectors and then we add a Vercel connector. Okay. Browse connectors. Vercel. And then I just allowed it and I authenticated with Vercel and great. So now I have Vercel and I'm going to just allow it to do whatever it needs to. Then I'll go back to my session and again, look, this is still in localhost. And if I go back and I'm like, hey, can you, uh, upload this site to Vercel so I can share it with other people, please? Pretty simple, but um, the concept is what's more important that I want you to learn. Like, hey, first we are putting this version and our code that we are satisfied with to GitHub, and each time we make changes that we like, we should be like, hey, push this to GitHub. That is basically like saving our progress. Okay, so we keep saving our progress. And then if we need to deploy it, we need to deploy this website to the cloud. Okay, cloud cannot do that on its own without another cloud service. So Vercel is normally the easiest. And if you just connected to a connector like Vercel and just ask it to do that and you have the connector, that's it. Now you can, you're actually very empowered to go out, build websites for a lot of people, start deploying them and charging them for it. That is as easy as that if you just learn this with Claude. All right, and I'll show you in a sec after this is done.

All right, so it just got done. If I just hit this and open its link, you can see this is up on the cloud and this little dinosaur thing is there too, which I love. Great. So that's about it. That's how you deploy websites to the cloud. And it's really simple once you know a couple concepts. And the cool part is that all this website was built without me having to give it all that context. It took it from my AIOS and built it on in itself. And just to take it a little bit further. So now we learned how we can have a website online. But you know, we were taking and selling all these really cool skills, right? We're building these skills. We're building these agentic workflows. We can sell it on this website as well. And it's so easy. So, I'm going to show you how to do that.

So, the first thing you want to do to build and sell is very simple. You probably, you will need Stripe. So, I'm going to go to Stripe. Make sure you sign up for Stripe. Okay? And after you sign up for Stripe, what you can do because Claude is so awesome, you can go to customize. You can go to connectors. You can add a connector which is browse connectors. Let's get Stripe. All right. So, I just offed with Stripe and I'm going to make sure I allow all and I will allow all. Great. So, now I have Stripe. What I can do right now is, let's say, um, and it can be any skill. Honestly, you can reskill, uh, resell any skill you would like. So, for example, let's say this cloner template. What you want, what you can actually do is you can come here. You can make a new folder. Let's call it, uh, clone product selling. And what I'm going to do right now is make sure I have Stripe. Uh, okay. I have Stripe and I have GitHub. So, hey, this is a skill I have. Can you make a fork of it and then make a product in Stripe such that when somebody buys this product, they will be sent and they'll be allowed to access this GitHub repo? Does Stripe and GitHub have a connection such that I can do that? And yeah, so it's honestly as simple as that. And then I'm going to wait for it. And you know, this can happen because Claude has access to your Stripe, it has access to your GitHub, it has access to your Vercel, and it has access to your skill. Now it can play with all of those different variables and it just does what you want it to.

So what it did, this is really cool. So it, you see right here, it forked that repo, that means it copied it, and then it made it private in your GitHub such that only people that buy will get access to that. Uh, all right. And it was that simple to set up. So now if I use this, it was that easy to do. Now, if I go right here, now if I take this Stripe link and I go back, then I go to where I made the website now and I just say, "Hey, can you make a change to this website?" So, there's like a on right after the dinosaur stuff, there's a big buy button where someone clicks it and they can buy this really cool skill. Say it's like a website cloner. Awesome. So, it just got finished. Let me go look at the site. And if I scroll down right here, "Clone any website in one command. Get the club." And then it comes to the Stripe page and I can pay for it and I'll get access to it. See, it's that easy to buy and sell skills and services. Now, with Claude, you can, as long as you understand the concepts, you can get it to do so many things. And this is just scratching the surface. So right now, let me show you how to actually even make this site better on its own and improve things with this really amazing concept that Andrew Karpathy came up with called auto research.

All right, so now we have a website, a product, a way to take payments. But here's the thing, like the carousel, the website, all of the stuff we generate these things kind of with a skill is the same, is the same every time. So every weekly report we're doing is the same, which is cool, but nothing really gets better on its own. And that's what we're going to be fixing in this level. And with this new idea from Andre Karpathy, who was one of the heads of OpenAI and Tesla, and we're going to be building a system that essentially measures and improves itself without you having to do anything. What it does, it runs its own experiments. It forms hypothesis and makes changes to see if it can measure and drive towards a meaningful outcome. Right? The concept is that Andre Carpathy wrote about this. The same pattern that AI researchers use to improve models can be applied to literally anything measurable. So let's say we're going to be giving Claude a metric and we're going to let Claude come up with and run its own experiments and it tries these things. It measures, keeps what works, and discards what doesn't all on its own. So, which means we're going to be like this graph shows right here. We're going to be waking up to better and better either code, loading speed, if you can imagine, all these different things. And they keep compounding because it's always linked to making new improvements. And this is revolutionary because auto research literally runs hundreds of cycles in the time we actually just do one. And the compounding is really wild because here's the part that really matters for us. If we can measure it, we can optimize it. So if you think about metrics like revenue per visitor, conversion rate, email open rate, page load speed, maybe ad cost per acquisition, churn, lead response time. All of these are numbers you can measure. And any number you can measure, we can run this auto research on.

So with Alex, we launched that consulting website. It looks great, but if we wanted to convert more visitors into, let's say, consulting calls or more sales, the site speed is very important, right? Apparently, every 100 milliseconds of like page load kills conversions, and a slow site is a leaky funnel. So, we're going to run an auto research loop on the website's loading speed and let Claude experiment overnight while we wake up to just a faster site. All right. And we're going to be doing this with this thing called EVO that I found, which was really good. It's a GitHub repo that just makes it a lot easier for you to run these kind of experiments. So, I'm going to copy this and I'm going to go to Claude and I'm just going to ask, hey, can you set this up for me? All right. So, you see it just started up. Now, all I'm going to say is like, hey, um, can you discover and just tell me what kind of experiments we can run for our website in this codebase to see what we can do? Use Evo discover. Okay. So now what we're just doing the first phase, it becomes really easy, uh, because all you do is there's two things you can do: it's just discover and optimize. So we discover, uh, like what kind of experiments we can make for our site, and then we just do optimize. And after we do that, we'll be able to see a dashboard with all the experiments EVO makes with auto research, and then essentially we can measure what's going on and have those changes done automatically. All right.

So, it came up with a bunch of stuff. So, it says it can perform Lighthouse score, load speed, bundle size, all these load times, etc. Improve the cost per visitor conversions. And these are all really solid experiments. So, just to make you see what we can do, let's run a bunch of experiments. Um, let's say just on the loading speed and show you what the dashboard kind of looks like. All right. So now that we have this, can you, uh, give me a dashboard of the EVO optimize for just let's say loading speed and that's about it. It's just like me understanding two things. It's like, hey, these are the metrics, and then I just say, hey, go optimize this, figure it out. So I'll come back when this is done because this is going to take a long time. All right, so that took a while. Let me see. Okay.

Okay, so I'm opening up the dashboard and this is what the EVO dashboard looks like. It's really cool. So now LTP is the Lighthouse core, I'm assuming. And you can see that it keeps improving these things over time. And we can see a lot of different sub agents ran. We can see what the experiments were that I don't know what most of the stuff is. But what I like about it is you can apply all these best practices. It knows what's going on. It's seeing, you know, it made some experiments, different, uh, agents, and it optimized our page load speed by so much by just doing this. And now we have a really nice dashboard to see it working. But you know, all of those, you know, this is cool, but the biggest thing is this: all of these improvements came from Claude. I didn't know what is being done. Every one of them is also documented right here that I can see and dive into, and if anything broke, like this thing, it reverted it automatically. The site now is much faster, which means the conversion rate is less leaky. And we never touched a single line of code. And what I want you to really internalize is if it has a number, it can be optimized. And if it doesn't, use Claude to theorize what number you can attach to it. Page load was just our first example because it's easy to measure and it's easy to see the improvements and it directly affects revenue. But this pattern works on everything, way more than code. You can put it on even a skill if you want. Um, you can test it on stuff like email open rates, ad CPA, landing page conversions, content engagement even, right? Like what gets you more engagement. And if we're making a product, we can test literally stuff like onboarding completion rates, different variations for the flow, what we do that may affect churn. This is so crazy. Like this is actually if you can just go to any agency and be like, "Hey, I will improve your X metric." you can very easily make a case for them to give you more money. The biggest thing is that every single one of this is a number, and everyone can, and everyone can just set this loop because all I did was just be like, "Hey, set up EVO and you go to sleep while Claude just keeps working." And this is what I want you to see. This final piece, this really should ideally change how you think about building anything going forward, because optimization should be the last thing you worry about because we can literally get AI to automatically overnight keep running and optimize this stuff for you and then even keep reporting for you, right? So, so you know, we did all this, but the really nice part about this which you may have noticed is this uses a lot of sub agents, right? And using sub agents is so useful because even when we're building, we can try out a lot of wacky experiments. So in the next level, I'm going to be showing you how we can actually take advantage of multiple sub agents to get to build our products and make really cool stuff you may not never have before. All right.

So in this level, we're going to be learning how to develop software products with multiple AI agents working for us overnight essentially. So we don't have to continually manage everything. And this is the real way. What we're going to do is we're going to have long-running teams of AI agents building a product for us while we leave it on overnight. And quick disclaimer, this is very important. We are going to need the max plans for CodeEx or Claude, whatever it takes, because we're going to be burning a lot of tokens because we're going to leave our AI agents running on overnight, right? We'll do this when we need to build something super big, like let's say a SaaS that you can sell, which normally takes a lot of development, right? But ideally, with Claude being so smart, we get to just, we get to just steer where it goes. And this whole chapter is essentially the basics on how we can get started with agentic engineering. All right? And in order to do this, we're going to make our AI agents behave like software engineering teams work in the real world. And let me give you a bit of background on how that works so you can understand at a high level and have more control over how to get this to work. All right.

All right. So to dive into this, I'm going to go back here. First, we need to answer the question: how the software teams work, right? And to answer this, there's a couple steps. The real reason software companies can ship features like every week, there's bugs that we have to deal with. We have to scale product and keep pushing. It's because of a process that we've developed over time. And it's a repeatable process that every good engineering team follows. And once you understand this process, you can replace it with AI agents. So let's walk through that right now.

So the first step of this process is essentially to make a PRD. Okay. Everything starts with this, and it's what a PRD is. It's basically a product requirement doc. All right. It's essentially a plan. This is where the team decides what they're building, why they're building it, who it's for, what features and needs, and how users will basically flow through it, and what edge cases we might need to actually consider. None of this is code. This is just thinking about what we want our product to do. And this step is incredibly important, especially with AI, because good engineering teams spend real time on this PRD. This part we shouldn't rush because every bad decision made here becomes a bug like a while later, and you'll see the AI getting confused and looping there. All right, but what I want to make sure you understand is the PRD is basically the foundation of our software product. So if I go right here and then make this. So PRD. Okay. So first we want to make our PRD. Then after we make a PRD, this is like a massive document, remember. And from that document, we go to step two, and step two is basically breaking it into tasks, break into task, right? So this PRD, what happens is (and don't worry, AI will do this, AI, I'm just going to add it so you're not scared), um, what it's going to do is just so because an AI cannot one-shot, like with one prompt, build the entire software project, right? So, what we're going to do, so what we're going to do is make sure from the PRD, we're going to make like task one, task two, and so on. And the thing is, uh, for example, this can be something like build, uh, signup page, small things. And then task two can actually just be like test signup page. And you know, this can go onto task, I don't know, like 200. The idea is you don't make this. We'll get AI to make this. So it knows Claude, what it's going to do is each of them will be their own separate thing that it can work within its context window. All right. Each ticket, so these tasks are called like tickets, are scoped. So let's say one Claude sub agent can pick this up, work on it and make sure it gets done, and hand it off to another agent to ensure it's all good. All right. Okay.

Okay. Step three, what we're going to do is so step three, what happens is agents create task agents action each task. So at a high level, what's going to happen is, let's say Claude, so you're going to have one master Claude agent, and that will spin up an agent, right? And that agent will be assigned to task one, and then it'll assign another agent and it'll be assigned to task two if they're not related. Okay. And it's just going to be happening on its own. And how you can visualize this is literally like this. So, so there's going to be a backlog. There's going to be in progress. Um, there's going to be a review and then done. Pardon my handwriting. Um, so task one is going to be like a Kanban board where the AI agents are just going to move it here while it's working on it, and then it's going to review it and then done. Okay. So after this is done, then it starts working on task two, and this happens over and over and over again. Okay, this, that's why we have to keep this running all night to make sure this is working. And then what's going to happen is, let's say these are going to be working. There's probably going to be a lot of different parts of code that all get merged into one codebase in GitHub. All right. So this is just a high-level overview. But overall, what I'm trying to tell you is how this works is remember: this PR into tickets, to code, to review, then we test and merge. Okay. So this is high level, and this whole thing is a loop. This keeps happening. Okay.

Now with that covered, what should we build? And the idea I had that what we should build is something kind of complex but also useful at the same time. What I wanted to do was help you build a desktop application that you can like play with on your desktop, and I wanted so that you can build it, you can sell it, you can use Cloud Code with it. All of these this, and building this application covers so many different concepts. So you should be able to learn everything moving forward. All right. And this is something that I use daily, especially to write my content. There's this, um, app that's similar to this and it's called Poppy AI. I think it's worth like $50 million. And you know, I've stopped using it after I had built this for myself. So, what I want you to get in the mind of is now you can build your own tools, especially with the power of these kind of agentic workflows. So, let's get into that. Uh, I think after you learn how to build this, it will be super useful, and you can even take it and sell it if you'd like later. All right.

To get started. I have this, uh, long-running AI agents link inside, and you can find this on the internet as well. I'm just like keeping everything here, but it's this software called AutoForge. And I've been using this to spin up MVPs and it's been working really, really well. So, if I go back to Cloud Code, what we need to do first is set up AutoForge. So, let's make sure we set this up. Let's choose a different folder and set up AutoForge. AutoForge. Hey, can you spin this up, please? And let me wait for it to spin it up. Well, wouldn't you know it, this is, um, insane timing, but this is why I want to make sure I help you guys set this up. Claude right now is down. That means my operating system, all of that stuff is also down, right? Uh, I, I was trying to log in and build this stuff for you guys, but then I realized it's not working. So maybe this is a better time than any on how we can set up our redundancies. And this is what I've been saying. If, like, for example, you set up managed agents with Claude and this thing happens, your system is down and you're like, oh no, what am I going to do? So now, um, let me help you guys set up a secondary service such that we're never only dependent on one thing and we can be nimble and always keep our systems active. All right.

All right. So to make sure we're never only just dependent on one thing, I'm going to spin up CodeEx. So to make sure we're always not dependent on one thing, the best alternative that I found is honestly CodeEx. So that's made by OpenAI. And if you already have a ChatGPT subscription, you can use it with that. So I'm just googling CodeEx. And right here, if you just download this for Mac OS, I'm going to spin up this app. Okay, so here is CodeEx. And CodeEx is fantastic. I've been really loving using CodeEx for a bunch of stuff, and I'm going to switch to GPT-4.5. And now because we set up our project and our AIOS, uh, locally, because it's not dependent on cloud, we can actually use this from here as well. So from here, if I go to my evolving brain and open that up, I can do everything that we've been doing from CodeEx. And here, let me show you something cool as well. Right now, Cloud isn't working, right? So after you install the app right here, I can just be like, "CodeEx," and then this works. So now I can just be like, uh, so then I can like paste in AutoForge and like, hey, can you set this up for me please? And then run that, and let me come back when that's good. But actually, for you guys, um, this is what I want you to know is that's why I like using this Obsidian, uh, because you know, I can use Claude, I can use CodeEx, and it's all with my files and my services, which means I'm not tied to one provider and I'm nimble and I can do whatever I need to. All right.

So, let me go through this process, and after it's spun up, let me come back to that. And you know, the nice part about these open source repos that we're using, it's not dependent on either of us using Cloud or building even with it. So, if Cloud is ever down, we're good. Um, we should never be dependent on one company ever because of instances like these, which obviously will happen a lot, and we want to make sure we are set up and secure because this is going to kill our business operations if we're not in a spot where we have other redundancies in place.

All right, so it's saying it's running. So if I say go back to my browser, it says, "Cool, welcome to AutoForge." And here's a little dark mode if you guys need, but okay, start project. Let's make a new project, and then let's call it, uh, project canvas. Okay. Oops. It needs hyphens. And now I'm going to allow it to make sure. By the way, just a quick little thing. It's okay. Great. So now it wants me to select the folder. So let me make a new folder and call it, uh, project canvas. And select this folder. So, and let's make a blank application. And then I'm hoping Claude is back up. If I go back here and I make a new terminal and I do "Claude." Hi. It looks like it's fine. Sometimes it just goes down. Okay, it looks like it's fine. So, I'm going to hit "Create with Claude" and wait for it. Okay, so this is what's come up.

And now what you want to do is spend some time going back and forth over it to come up with a really good plan on what you're going to build. All right. And just to give you a bit of background on how I like actually developing applications is this. So let's say you have an idea to build something. And most people, what they'll do is they'll go to Claude or anything and then just like say, "Build me this thing." Okay. I don't like that approach. I think that's the incorrect, wrong approach. I think what you should do (and I've just gotten way, way, way better results from doing this, and so did other people) is the reason being is because what you have to do is you get stuck in this loop where you ask Claude, "Hey, fix this." And then Claude does something and then you're like, "No, I don't like that. Fix it again." And then you get stuck in this, uh, so it makes something broken and then you get stuck in this loop over and over, and that wastes so much of your time. It's actually crazy because what's happening really is if you think about it, uh, if you want to build like an application, you want to have a good strong foundation, right? And if you do this approach or this approach, your foundation is like shaky. And if you expect to build something on top of a shaky foundation, you're never going to be able to do so, or it's just going to be really hard for you to build stuff overall. Right?

So what I like doing is, let's say I have an idea, then you make a plan. And I, I what I learned is like AutoForge helps you make the plan really, uh, in a right nice way. Okay. Then after I have the plan complete, then, then I make the MVP with AutoForge. It just uses Claude to make that. And then what I do is I will take over and refine. And you know, you can also just, also you, what you can do is continue doing this again with AutoForge. Uh, like, hey, you have an MVP, then you want to add a new feature or you want to fix something, you can just go through this loop over and over again. But what I like doing is I take over and refine. And you can do this from anywhere. Okay. And you'll see this in action and why it's so much better, because essentially what's going to happen is our plan is going to be broken down in sub-chunks, and those chunks, Claude, it's easier, uh, for Claude to tackle each one of them with sub agents rather than all at the same time. Okay, so let's go over this and make a plan.

So right now it's really simple. What do you want to call this project? What not? And right now, uh, the idea is for you honestly is just a word dump. So here again, this is why I like using voice tools. Hey, uh, I want to build like a spatial canvas workspace where I can use, let's say, Cloud Code or CodeEx, and think of it like a infinite zoomable whiteboard, but instead I can also use drawings. Um, and I want to make sure that it has tiles where I can actually do the work. I can see the markdown files. Uh, like if I tell Claude, hey, in this canvas, can you make a new markdown file with a story? It will do that. And then if I want to make sure there's also a file explorer that I can switch around. Okay, these blocks should be working in that way. And, um, the essentially the idea is to have something like Excalidraw or TL Draw like a whiteboard there, and I can have a tile where I can place my Cloud Code. I can type into that and it'll make new tiles that are essentially markdown files that I can edit and see and iterate over time. All right. Okay. So I just did that and then I'm going to hit go.

So now you're going to go through this process for quite a bit. So let's just go through this together. And this is just me doing a quick round with you. Of course, try to if you have an idea, flesh this out with Claude before you even get in here. Um, that's just a matter of you having a chat with an LLM and having a doc that you can just paste in. But now let's go through this.

Yeah. So these Claude tiles can basically just be terminals and the, uh, file explorer. Let's say there's a folder and I should be able to see the files. Maybe I should be able to drag those files inside that canvas, and in that way, uh, I can, you know, work with my Cloud Code and my text files very in a spatial canvas. All right. Uh, maybe call this project "Open Canvas." And right now this is a personal tool, but I want to set it up later so other people can use it. Uh, ideally I want this to be an Electron app that I can download on my Mac or, uh, yeah, MacBook or Windows, maybe. Okay. So, I'm going to go there. And an Electron app, just, uh, to make you guys aware, an Electron, like this, for example, uh, your Chrome browser, Muro, all these different apps. These are kind of like Electron apps. And Electron apps are the things you open up. It's not a web app in your browser. You can do either one. You can just ask for which one you'd need. Yeah, just make it quick. I don't want to be, uh, involved in any technical decisions right now.

And just to give you a quick little overview, this is just one framework that I found personally that works really well, uh, when it comes to development, but there are other really good ones as well. Like there is this thing called GSD framework, and it kind of works the same way where you make a plan and it works the same way as you take this link, you're like, "Hey, use this, install this plugin and help me set up an app." This is the GSD framework. That's one thing. And then there's Superpowers. And if you just search it for in GitHub, you'll find all of them. Superpowers is another one that works really well with your AI agents for coding. Um, GStack is pretty okay. You can experiment with all of these, see where it goes. But these are basically just, you know, again, other coding folders or projects that helps you make your projects a lot. I personally have found, uh, that I like using AutoForge a lot. GSD is probably my second favorite. So 100% recommend. Try out both. This one you would just install in Cloud Code and talk to it there. Uh, it doesn't have this kind of UI. What I like is having that little Kanban board that I'm going to show you in a bit. But, um, you know, feel free to use anything because at the end, because at the end of the day, all of them either use Cloud Code or CodeEx to do your coding out for you. This just helps me plan it out a bit.

Yeah, I would want separate whiteboards per project. Uh, I don't need the mini map, but it would be nice. Yes, we should have multiple terminals open. Yeah, I should have multiple terminals open. And, um, I wanted there's no preview mode. Just make it, make it like a TipTap like Notion where I can see it in rich, uh, way. And I should be able to edit it from there. It would be ideal for connecting tiles, but right now we don't need it. Yes, I want to freely be able to resize all the corners, the sizing, all the stuff. Uh, search would be very nice. The settings, preferences, maybe there's a, there should be a day mode and a light mode. Um, keyboard shortcuts is a big thing. Make sure you add in as many keyboard shortcuts as you can, because I love using that. Right. So, uh, yeah, I think it should automatically appear when I tell it to. So, that's the idea. Um, maybe I'll just pick a new folder in the system and it becomes a project in that way. It's kind of like an IDE. I would like Command Z and, and drag and drop. And you see how like there are so many little considerations that this thing is making and asking for my direction. That's why I like using this because rather me than me having to think.

about all the little nitty-gritty details. It's asking me on what it should behave like. When I open for the first time, maybe I make it a prompt so I can open a folder. Uh, and then in that folder, I can start making new canvases, uh, new cloud code terminals, and um, then also make new, uh, markdown files from there. And yeah, ideally, I should be able to duplicate it, but we can leave that for later. We don't need snap to grid right now. Maybe one consideration I want to make is that the tiles, as I write text, it gets bigger automatically. Uh, yeah, I should be able to minimize it. Closing a tile, it should just remove it from the canvas view. It should stay in the project. But in the file explorer, I should be able to right-click and maybe close it out. And yeah, ideally, it everything auto-saves.

Okay. So, notice it did a couple different things and it's still doing it. It created this appspec.txt, right? And this appspec.txt is essentially all the little tasks that it's going to be actioning. All right. So, it's giving you a couple different things. It made a lot of design choices. And now, let's continue to project. And this is what I was talking about. This is why I like this a little bit better because what happens, it creates essentially a Kanban board where I can see what's happening when and when I can stop it as well. And at the end of the day, this is just running on your AIOS and your um cloud code because it's basically like AIOS folder and then your project is underneath that as the project canvas. So we also have, so in the AIOS, we can always pick it up and then work from it whenever we want, right? So this will take a while, but after this is done, I'll come back and if you're curious, you can always peruse this thing where if you go to dashboard now, it's basically initializing all the features, etc.

Okay, so I waited a bit and you see it already created a bunch of different tasks and you can see this is kind of cool. It's like they already know which tasks are blocked because it needs a dependency, like you need to do something before it. So it's really nice that, you know, it builds out all these features and it's going to, after right now, it's really putting a lot of thought into what it's going to build and then slowly it's going to put each task in progress into done. And again, this will take a long time. Only do this after you're sure and you've already made a quick little prototype or so, maybe with just cloud code or codeex, and then dive into this to make sure your app starts from, starts out from a good foundation. I've personally already made a prototype and that was what I was showing you before and I've made that into the same app, right? This right now, it's just a matter of waiting. It's about keeping your computer, or if it's a Mac Mini, even better, just on and letting it spin and go throughout the night. And you can see right now, uh, it was 45 tasks and it's now built it into 91 tasks. The reason why more tasks are better is because we can have more sequential things happening at the same time and each feature will be tested.

Okay. All right. So this thing is active now. So finally, it's built out all the tasks and now you see they've spun up some agents that are actively working on these tasks. This is really cool to see and you can monitor it over time. And these are just for long-running agents. Again, this is for software development. This is not agents for other tasks. This is just if, let's say, you're going to bed at night, you leave this on and it's just going to be running and keep running and making sure you make a really good app. All right. So once this is done, I'll come back to you just to see the progress. I've been seeing, uh, it open up these kind of files here and there to test it out. Um, if I go back, I can see it's made progress on 14 of the tasks. Um, it's built a database, it's built the canvas, all these things. I can see a little graph of what is going on, what kind of services are dependent on what, uh, all the logs, the terminals, bunch of the assistants. So if I want to talk to the codebase about stuff like that. Oh, see, uh, they just popped out with a canvas. I, I can see the first glimpse of the, what's it called? The file explorer. It looks pretty good. And um, it's just a matter of iterating over time. But you see now how this is like an application. So I can like drag this around. This isn't independent on Chrome or something like that. It's right here on the bottom. If you can see Electron, it's doing all of this on its own by spinning up multiple sub-agents and using my cloud code subscription, right? So, it makes it really easy for me to develop a lot of apps and see it working over time. And you can see the ones that are active for coding. Uh, this guy is spinning up new agents on, um, on request, etc. You know, this is cool.

Now, we can see the first glimpses of our application. We can make new canvases. I can put in new, uh, I don't know, like terminals in here. Make little, you know, um, note, note cards. All these things are there. Uh, what is this like empty file line? Awesome. Yeah, I'm loving how this is going and that's what I want you guys to get in the habit of like, slowly let your AI agents figure it out. Ideally, to the way at which AI and software is moving, you just need to be able to steer it. But ideally, you understand the concepts of what's going on and how these things are honestly just being built out. This is just one concept that we just built from scratch. You see, we just built an Electron app with a canvas with a terminal, all the stuff that we can always use. And you know, slowly as we go. This is literally the exact same approach I use to build out this application that I use every day. And I encourage you to build out tools for yourself like this. And you know, the question becomes, how do you identify what do I work on? What kind of tools I need? And let me show you a really good framework that I use so you can essentially know what you have to automate and build tools and applications and agentic workflows for yourself so you can solve any issues you may be facing.

So for the next section, I want to make sure you avoid a very grave mistake when you learn how to use all these tools. And you know, this is the quote that I always get reminded of where it says, "Give me six hours to chop down a tree and I'll spend four sharpening my axe." Most people think this and think, "Yes, prepare. Good. Preparation is smart." But here's the problem. Preparation becomes its own addiction. And you know, I'll see people falling in this trap time and time again. It's like they'll see a shiny new tool, something will come out, maybe they play with it or they don't, and someone will say something like, "Oh no, you need local models running in your open claw to save the permanent underclass, stuff like that." And then they'll go do that and never actually get anything done. I want to make sure you avoid stuff like this and actually use the right technology. So let's actually get some output from what we have just learned and put what we learned into practice and get AI working for us, right? Let's not continue in this cycle where we're just inundated by new shiny objects and not actually getting anything of value out of AI and we're just spending more money there. It doesn't make any sense. And to do that, I actually developed a framework that I'm about to share with you. It's called the automation priority matrix. And once we go through that exercise, hopefully you'll know how you can start automating everything in your life and buy back a bunch of your time so you can put it in very high leverage tasks.

All right. All right. Now that we have our AIOS, we know how to build apps and automations. Now, the question is, what do you actually automate first? And the sad part is most people get this completely wrong because they jump into building automations for stuff that looks kind of cool and not the stuff that they're actually bleeding their time in. And then they'll see like three months go by and then they haven't saved anything or done anything real with AI. So, I put together this five-step framework to make sure you're doing everything correctly and I call it the SHIPS framework. Let me break down what each letter stands for. So, first, what we want to do, so the first letter is S, scan. We want to make sure we map out our workflows. And once we map out our workflows, this is something you might be able to see. So, what we want from scan is a chart like this. And I call this the automation priority matrix. And you can see right here, this is for Alex's workflow. And what he's done is he sees like he'll normally spend five hours a week on content production. And if he were to automate that, that would be a level five on his revenue impact. And I think it's very possible to do this. So that's a four on feasibility, which gives it a total point of 14. And if you go through this framework over and over for everything you do, then you're going to get this.

So, so how we do this is like this. So the first option is manual. Simply, you just write everything down that's recurring in your life every week. You dump that all in a spreadsheet and it takes about an hour. The second option, the one I like better, is software. So, what we want to do first is make sure we record our screen. And there's two softwares that I found that are pretty good to use. And let me take you over them. And the first one is Screenpipe. And if you want the links, they're again inside my community down below. But, you know, their code is all open source. So, if you copy this link and then you just say, "Hey Claude, set this up for me." It'll make sure it starts recording your entire screen and all your data stays local and never leaves your machine. So, this is basically constantly going to be recording your screen, seeing what you do and make sure you get a report. All right? And this happens for what's happening in your screen. It records your audio and stores it all in you. So, it stores it all inside your computer. All right. The other one, other one, if you want to use, is called RescueTime. And this one is a little easier to set up if you're not super technical and I think it's significantly easier to get started with. And I'd recommend maybe go for the $12 bundle where you get these time sheets. And what you want to do is you want to take those time sheets, right? You take those time sheets and then you put it into Claude, like right here, like we did in the very beginning. And then after you do that, I made a prompt for you. So right here, so this is a prompt for the priority matrix. After you run the prompt, after you dump in what you've been doing with your time, then you will get back basically a big sheet of this. This is an example output, but what this does is basically map out what you're doing day by day. So you know what is the most, what is the highest priority things you can be automated in your day-to-day life. All right.

So that is step one, number one on you scanning and auditing your entire workflow and what you do. And honestly, this is the best thing you can do for yourself because it's insane the amount of time you can buy back for yourself. Right? I would, I would suggest this is the single thing you want to get out and if you watch anything in this course, if you just get this report, then you know what to automate or even if you don't want to automate it, you know who to hire to take what over and then you just unlock a bunch of your time for yourself. All right.

So the next one, after we scan and map our workflows, is we want to make sure we hook everything up. So that part is really simple. We've been doing this for a long time. We've been doing that through the entire course. And hooking stuff up just means you go to customize, you go to skills and connectors, and then you make sure every tool you use, you hook up your Claude to that. And in that same way, we want to make sure we add all our context so that our AI knows everything that we do, all the people in our lives. So it has the right context to be taking action and providing a lot of output for us. All right.

So, this part's pretty easy. And then we build out our priority matrix. So, what I want you to do in the identifier portion is just simplify, which is you're just going to pick the first three. Forget about everything else. We're going to pick the first three and we're going to build automations to make sure this is the first stuff we action. All right? Forget about other tools. Forget about everything under it. Just, we're going to take the first three and maybe if you need my help, we can automate these together and you'll see how much time this really opens up for you. All right.

So after you identify and you pick your top three, then what we want to do is turn all of this stuff into a package. And you know, we've done this in the past as well where we've actually done a task, like for example, building that Instagram carousel and then we've actually turned that into a skill so that it's generating it on its own and even posting it so we don't have to do it ourselves. That is a scheduled skill. All right. So first, what we want to do is after we know what to identify, then we turn each one into a skill and then we make it run on a schedule. So if you remember what we've been doing in the past and inside, we had a bunch of scheduled skills like this evolving brain sync, that's what you need to get into. So you start having Claude do your tasks automatically, which is the highest priority for you.

And the last one is scale, which is really simple where we want to make sure Claude is set up to actually suggest new things it can automate for you. So, how we do that is again by setting up a cron job. What we're going to do is we're going to go to Claude and we're going to tell it, hey, I want you to every day or every week see what I've been doing because you've already have the time set up and then give me a report on things I can be automating that would be the highest ROI. That's it. That's really simple. You just get a priority report and then I'm going to show you an example of what I do in my day-to-day life. That was really useful. So, as an example, just so you know, I used to, you know, because I am a content creator, I have to do a lot of research. And what I did is I set and I realized Claude has been recommending, hey, maybe you don't have to do all this doom scrolling and research, let me just do it for you. And then I built this, uh, research scheduler and what it's doing is going to Twitter, finding all the newest cool stuff and then look, it's finding Karpathy's newest stuff. It's making, finding me all these SEO agents with all the links. And now if I want to actually dive in, I can. And it makes my life a lot easier because it's suggesting automations I can make and it's doing it for me. And that is the last part.

So with all that being said, hopefully you understand this theory. Now let's do this for Alex. So let's say Alex has made this priority matrix for himself. Now let's see how we can check off all these tasks. Okay. And to do the first one, content and video production. So, let me show you an example of some content and videos Alex potentially makes and then let's see how we can put that on autopilot. Right. Okay. Now, to make sure that we can actually take Alex's highest priority tasks and then make them into skills that Claude will do, we're going to start off with this. First, let's tackle the easy one, which is ad creation. Okay? And it's just, let's say, creating static image ads. So, um, just for reference, let's imagine, um, he has, this is the product that Alex has to create. So, let me take a picture of this. Okay.

Now, what I'm going to do is I'm going to get a skill. So, let me go back to my community. Okay. And inside classroom and the Claude skills vault, right here, I have two skills, a product ads and the memes ad skill. So right here, if I download this one and then go back to my Claude, I'm going to go to code and I'm going to make a new session and right here I'm going to attach. Okay, so I'm going to, I have this picture. I'm going to copy this image, paste it here. Okay, so I just pasted this image here and I'm going to be like, hey, can you use the e-com ad skill that I just put in here? And just for reference, this is the, you know, this skill, the e-com ads that I just pulled in from my community. I can also just pull this in here. And then I want you to for this product, create some image ads for me, please. And make sure I make, um, a new folder. New folder. Alex pink bottle. Okay. It's pretty simple. Mostly what I'm trying to show you here is the skill is going to be doing the heavy lifting and it's going to make some really cool ads. And this is how I want you to be get thinking about stuff. Find really, uh, find really cool skills online, take them, use them, and then manipulate them to how you would maybe want to use them. Right? So, I'm going to hit go and then I'll come back when this is done. This has got done from what I see. Let me see what it has, um, made. It made all four of these ads. Uh, this my other thing. Let me drag that in. So, huh. This is pretty good. You can see the, uh, bottle right here. Matches everything, especially for you. Pick your pink. This is really cool. It made all of this copy and everything up. I didn't have to do anything. Um, the skill I made it, what it does, it also grabs assets from the internet. It was hard to make, but I just wanted to show you kind of the extent you can take these skills into. And you can see this got these ice images and using Nano Banana Pro, it created this image of my cup with my screen. Uh, it's kind of funny to see this. Uh, put it inside this baby thing. And yeah, the I love these ads. Oh, this is really cool. It shows the contrast. So, I just wanted to show that. Now, what you can do is you could just be like, "Hey, put this on a schedule and make ads, I don't know, every hour." Okay. If you want to take a step further, you can go to Composeio and right here if you connect the apps and then I know they have a meta ads. Yeah. So if you can connect your meta ads account and then you can see which ads work the best. So you can make it essentially make new ads, connect it and then keep testing it and iterating on autopilot. All right. So, what that means is it literally gets rid of creating Alex's task of creating static ads. And it was as simple as that. See? So, it was as simple as that. We got rid of creating static image ads because you can just find a skill online, use it, and then make sure that you set up the schedule. So, it's automatically creating the skills. All right.

So, now let's move on to something a bit harder. And I actually really like this, which is content creation and video production. And this one is super interesting because now they released this thing called Seed Dance 2.0. And I feel like I've been seeing some crazy results. And I know this service called HeyGen right here has an MCP for Claude and Composio. So if I go back and go to home and connect apps, I should find HeyGen. Okay. Yeah. Yeah, I'm already connected. This is great. Okay. So I can reconnect to my HeyGen and then I would just need an API key. Fantastic. So let me sign up for HeyGen. And with HeyGen, what you want to do is you want to go here. You want to go to developers and then API dashboard on the bottom left. And then you want to create an API key. And then say, I don't know, for creating video ads. And then let's call it, it's for an agent. So they literally know who it's for. And then let's create this API key. I'm not going to show this to you. So I'm going to take that API key. I'm going to go back to Composio. I'm going to go to HeyGen. And then I'm going to hit connect. So this part I need to hide. But here I just connected my, uh, Composio. And then if I go back to my Claude in my customize, I should have Composio as a connector right here. And now what I can do is let's try something fun. Uh, maybe I'll make it here. Um, hey, can you use the Composio connector and through that use, uh, HeyGen and from there use the Seed Dance model and with the Seed Dance model, I want you to make an ad for Booked AI. And Booked AI is basically, you can go look it up what it does, but it's basically it helps you create voice agents for people like HVAC companies. All right. So, maybe make an ad about how an HVAC company will essentially, you know, they're missing their calls and then after they're missing their calls, they are not able to, you know, they're losing a lot of jobs, but maybe let's make it so that a friend, uh, shows them Booked AI and after the friend shows them Booked, they use it and they're like really happy because now AI is handling all their calls. So make a video of that and then send me the link or download the video. So really simple. Again, this is just the power of cloud code. You can just start iterating things super, super fast after you just learn the few concepts and after you learn about how you can get other tools to leverage and do things for you. So after this is done, I'll show you the results.

Okay, so it seems like this just got done. Uh, let's take a look. "Your AI agent. Always on, always booking." >> Uh, so, okay, it was pretty good. Some, uh, small mistakes. You notice how it says "Booked AI"? It's probably because when I spoke it out, it, uh, I, I didn't say "Booked in" and it didn't get it correctly. And then some other small things. It didn't use my logos. Uh, those I can just insert my logos and it should feed it that context and with Seed Dance and Omni Reference, it should be able to do it. Maybe see now the same thing, you can just put on a schedule, you can make more ads and start testing them. The only thing is you need to make sure you have essentially one way that you like keep, like a format or something like that so you can keep everything standard. So with that, let me try something fun. Maybe let's try to make a video avatar of myself becoming Batman or something. So, I'm going to go here and, you know, you can make an avatar of yourself. Uh, right here, you can hit avatar and then you can hit add a look and then you can see I already made one, but it's really easy to make. Is Okay, so if you want to make an avatar, you want to go to avatars right here and then you hit add a look and then you hit upload look. Then what you do is you have to record a video of yourself because it's so realistic. It's kind of weird, honestly. And then you hit record. Now you turn your camera and your mouse on. And then, oh, it, okay, u, and then, okay, I don't want my recording to turn off, but like you make that avatar and then after you're done with the avatar. Okay. So I'll make a new session and then I'll just use the same thing. I want to make sure I have the Composio API. Okay. Composio is connected and HeyGen should be connected. So I shouldn't be like, hey, can you use the Composio API and with that use HeyGen and with HeyGen, I want you to create a video using the Samin Yasar avatar and then maybe make a Seed Dance video of me of how I can, let's say I'm shooting a YouTube video and then I become Batman and I jump off or something like that. Okay, I'll try this out. I'll put it in Seed Dance and then, yeah, I'll wait. Uh, hopefully it understands my misspellings and whatnot. Okay, after this is done, I'll download the video and I'll show you what that looks like.

Okay, so this got done. This is kind of funny. "Hey guys, I have a secret routine. It's time you see it. Wait, duty calls. I am the hero this studio deserves and the one it needs right now. I have become something more. The night is just beginning and there is a great deal of work to be done in the shadows. The signal has been lit. I must go. Justice doesn't sleep and neither do I when the city calls for aid." Okay, this is kind of fun. Uh, but yeah, I like it. You see, now we've actually just built all these really cool ads and we can put these videos on autopilot. So now if we kind of just go back to, if you think about Alex's workflow, um, now what we can do is basically generating video ads becomes really easy because of how we just, um, made the video ads for Booked AI and then getting clips. And now I want to show you some little things using Cance. You can actually make really cool B-roll. That's one big thing. But if you want to make actually text and stuff like that, I want to show how you can do that with this thing called Hyperframes. And I thought that was really cool. So these are some other really sick integrations I'm going to show you.

All right. So the last part is I'm going to show you how to make clips and highlights from videos. And this is so cool because literally what you can do is you can edit your videos in natural language. And actually, this is kind of fun because this is actually the hack I used to, for example, make and edit my videos right here. So you see how I'm in the corner right here? I'm making like these motion graphics. So, I used to use Remotion before, but I found some stuff that's made it significantly better. And we can even do that to, like, let's say right now, if you've seen this, Hermosi has his highlights channel where he takes his long clips and then he like publishes these on autopilot. We can do that as well. And it's really easy because it's driven through this open-source software and it's called Vibe Frame. All right. Vibe Frame is free to use and it's really cool. So if I hit view on GitHub right here, what you can do is you can take this and they even have like a video to show you how to integrate it with cloud code. But, uh, what I like doing is what you want to do is you want to go to Claude, start a new session. Let's say make a new, uh, folder. I'm calling it frames. And then say, hey, can you set this up for me and install the MCP? So I'm going to hit go on here. I don't know what this opened, but okay. I'll let this happen and then come back. And just to show you, you can also give them this, uh, npm, this link right here. Um, what you will need though is a couple different keys that you will need to enter. So here are the environment variables. So you probably need an OpenAI API key, an Anthropic, a Google, and 11 Labs. Everything else you don't need. But after you have that, you should have Vibe Frames installed, right? And then after you have Vibe Frames installed, what you can do is you go back to Claude and then let's say you open this up. Make sure you install Vibe Frames and then, uh, let me go find a video in my computer. Okay, so I just downloaded this video. >> "This is the internet forever and it's all because of this." >> This is the video I released a while back. And now let's say this is actually how I edited the video. All right. So if you notice right here, there's these graphics here. I'm in the corner. Everything I did with what I'm about to show you. And right here, I'm going to essentially, I'm going to put this here. And let's make the clips just like how Hermosi does it, right? And it's really simple. And what I'm going to do is first I'm going to drag the video in. So now, um, Claude knows this is the video I'm talking about. Now what I'm going to say is something like, hey, I'm trying to make highlights for this video. So, a couple, and what I want is a couple of different clips that I can make a lot of videos that answer questions and it's clipped from this video. All right. So, for example, here's a channel that does this really well. He takes long form other videos. He clips them up and makes smaller videos. And I want you to do the same for this video that I attached. Make some clips, make a folder, and then show me. Okay. And right here, I'm going to go back. I'm going to take this, uh, and then I'm going to attach this here. And then I'm going to hit go. So let me wait for this to get done. This takes a bit.

All right. So this just got done. So it's saying six highlighted clips. Let me open it up. Show in finder. Okay. So let me open this up. Okay. Here it is. And let's take a quick look. You can see they're pretty small. Forever. >> It's opening up another screen. >> And you can see it's making these clips very simply. >> Okay. >> It's for a lot of businesses. And one of the best fits we found is real estate wholesalers. >> Yeah. And you can see this is like, you know, clipped up, actually my video into segments that now I can just automatically publish with Blot that I showed earlier in this video. Right. So now I have these clips. Let's say, um, I want to take one of these clips. I want to add some motion graphics. Maybe I want to do it for a long form, like something that goes on Instagram. Right? So for that, there's this other thing I found which is also open source and it's called Hyperframes. Right? I used to use Reotion a lot, but I've seen Hypermo, Hyperframes work significantly better for me. So, all you have to do is copy this, go back to Claude. Let me start a new session inside, uh, Vibe Frames and then maybe just to make it a little easier, I'm going to give it a clip. >> You can see. >> Okay. So, I'm going to take this clip. Let's say this is kind of complex to make it without burn. >> Okay. So, I'll take this clip. This is kind of complex because, uh, you see how there is right here, there's a screen share, there's me, all these things. It has to make sure it's kind of like one, uh, done in a way like that. Uh, where the clip is here. And then maybe I'm going to make it put some, uh, captions in the bottom and the top that are changing. All right. And to do that, what I'm going to do is I'm going to drag this here. And then I'm going to be like, "Hey, can you install Hyperframes?" And with Hyperframes, what I want you to do is take this clip and with this clip, make this for a vertical Instagram reel. And, uh, what I want you to do is put captions in the bottom. So you transcribe the video and put that there. And then on top, put something, I don't know. Okay. So let me wait for that.

All right. So, all right. So it just got done and just saying it's this thing called Hyperframe. So let's look at it. You can see "Fire crawl is" >> So here it is. >> Let me resize it. Okay, so here it is. You can see, okay, this did a good job where, you know, it knows that this is the Firecrawl video and right here, this is should be the captions. So, you know, let's play it just to see how it goes. >> "You can see Fire crawl is really quickly." >> Okay, not bad. And it's just a matter of iterating this over and over. And that's how I make my videos. Honestly, it's I have a skill and a process that takes my videos and let's say I have my, you know, talking head. I record this and I'll give it to this and then I'll be like, "Hey, for this section, I want you to make B-roll for it." And it'll make the B-roll. It'll place it in and then I'll just ask it to stitch all the videos together after I'm in a file. And that's how I kind of vibe code my videos. And with all of that being said, what you can do is now you can see is we have literally used AI to build and automate a bunch of this stuff, which means now Alex doesn't have to, you know, do any of this stuff anymore. And, you know, this is just me identifying use cases for myself where, okay, this is something where I'm spending a lot of time on. What I want to do is find a tool or a process that I can put on a schedule so I can essentially reclaim more of my time. And the more I do this, the more time I get back to do better and more high leverage stuff. And one thing I want to mention is hopefully you've noticed is to do all of this stuff, we were using a lot of open-source technology. And open-source technology is honestly where I see the kind of future of AI and cloud going. So I want to show you something really cool that I found that's an open source and that's where the whole shift is going really, where I think the future of AI is headed is essentially you can see all the clues inside open-source projects and, you know, we've been observing this and it's been true for some time and you can see every major frontier lab like Claude or OpenAI, everything they do was essentially inspired by something in open source, literally like OpenClaw. So if you guys have seen the the meteoric rise of OpenClaw like here, you can see how crazy the growth has been for OpenClaw in comparison to every other project in the past. But what happened was OpenClaw got really popular. This was open source. Someone with no team or funding, maybe they just love coding and they have a lot of coffee. They figure something out, they open source it, and then the frontier labs like OpenAI or Claude, they watch and then they replicate it and bake it into their product. And if you want to know what's going to be in Claude next year, you may want to look at where indie developers are and what they're shipping on GitHub. And you know, this is really cool because I'll show you this one thing that's been blowing up recently, and it kind of changes the game when it comes to actually just using a computer. And it's this thing called Clicky. And what Clicky really is, it's it's just a companion that follows you around in your cursor. And let's say you're confused about something, it's going to tell you what to do. Like for example, I'll set it up right now and I'll show you what that means. And the setup normally is really easy. If I'm inside this repo and I scroll down, it literally says get started with cloud code. And I would just have to copy this. Go to my Claude. Start a new session right here. Maybe I'll make a new folder. And let me start it up.

>> "Hey, my name is Farza and this is Clicky. Clicky is a little AI companion that follows you around your screen as you use your computer. And so how it works is all you have to do is press Control Option. Don't do it yet. It's not going to work. If you all you do is where you press Control Option and then you talk to Clicky and then in that moment Clicky can see your screen and can help you with whatever you're trying to do. So, let's say you're trying to, you're trying to write a better tweet. Let's say you're trying to learn how to color grade a video in After Effects and have no idea how. Let's say you're using Cloud Code and want to use skills and don't know how to properly use skills in your project. No matter what it is, uh, just hold it down, talk to Clicky. Clicky can see your screen and it can talk to you back. And it can also move around. So, it can also actually take that little blue cursor and point at things and do, do different things, uh, and show you different things. It can't click for you and it can't do anything like that, but it can show you things. Uh, Clicky only takes screenshots when you actually press that button. Uh, so right when you first start recording, it's gonna take a screenshot as context. Otherwise, it does not take any screenshots. So, yeah, let me know what you think about this. It's really early. I just made it in like two days or like in the last day. But yeah, let me know what you think. Bye."

And now look at this. So, if I go to Canva, for example, now, um, I have this little blue thing on top of me. Let's say I want to remove the background here or something like that and I don't know how to do that. Any software because this is connected to the internet. This thing knows what's going on in my screen and I can just be like, "Hey, I want to remove the background in this image. How do I do that?" >> "You're in Canva, so the easiest way is to use their built-in background remover. Click on the image first to select it. Then head up to edit photo in the toolbar at the top. And from there, you'll find the background remover option. Looks like you're on the max plan, so you've got full access to it." >> Okay, so you see it like literally points on where I should be looking at and then it tells me what to do. This is actually nuts because, you know, with this, what you can do is you technically understand every software because all the docs are on the internet and the LLM is following you around, helping you and guiding you do whatever you need to. So let's say something breaks, I can just be like, "Hey, how do I do this?" Um, like for example, hey, I just removed this background. How do I bring it back? >> "You can see the BG remover button is highlighted in the toolbar right now. Just click it again and it'll toggle the background back on." >> Oh, that was cool. Okay, so you see what I mean? This is so insane. For anyone that's actually new to a computer or they're trying to get learn something that they're trying to do, they always have a friend with them that they can just ask, "Hey, how do I do something?" And then it's up to them to go figure it out. And then they'll literally show you what to do. And then you also have control because it's not being able to click right now. It's just showing you, hey, this is how I do something. So this is what I mean. This is, and this is open-source software. So literally if I want, I can take this repo, all this code and I can build whatever I want on it. So for example, let's say someone is onboarding into my application. I can do that. There's so many like endless things that you can do where when you think about this and then you hook up your Claude or your LLM as it gets smarter, it kind of becomes insane on everything we get to do. And that's why, and you know, this is kind of funny because, um, when we think about it, Anthropic literally drops a new feature every day. Uh, this is the meme I'm always reminded of like, I wake up and then there's a new advancement. Like right now, as I was talking, they dropped this thing called Routines. And it's, you know, Routines, I was playing around with it and it's pretty much the same thing as right here inside schedule. If I make, let's say, a remote new task, now let's say I have a workflow, I can trigger it from any other application. I don't have to do it inside my Claude. And there's a lot of other little things you can do. They even dropped a new desktop app. And you know, the fact, fact of the matter is, now that like with this, uh, cursor thing, I don't have to learn it. What I would have to do is this can go out, it can know about this information and it can start helping me action what I'm trying to do. And that's what I'm getting at. All we need to do is understand how can we get our end result. And using the tools I've showed you, you should have everything regardless of what changes or what's evolving because you will always have the latest up-to-date information inside your cursor at, or at hand. So, I'm going to be playing with this. Maybe you'll see a couple YouTube videos from me about this because I think this technology is actually really revolutionary and you can, and it kind of like squashes your learning barrier of what you're trying to do. And that's the thing, right? Every tool, as we keep going, will change. Claude will get better. OpenAI will get better. HeyGen will get replaced. Composio might be called something else next year. It doesn't matter. But the skill I want you to develop is to look at a problem, identify clearly what you're trying to do, break it into agentic steps, wire up the right tool, and keep that loop running. That is the skill that you need to learn. And that's what compounds and makes it so that you never are intimidated by anything new coming out. Right?

So, the next level, I want you to actually sharpen your skills, but don't get stuck. All right? All of this stuff is brand new and it's not enough for most people just keeping up with it. So, I encourage you if you want to continue or take advantage of any of this tech, come join me in my Claude Club. Actually, here we're having an in-person meetup too. So, we can meet up in the community, which is pretty fun. But you know, if you've gotten any value from this course, the idea is to how can we use proximity to compress time and actually get you results with AI. And if you want to join, I'd be honored. I'd love to work with you. The link is going to be in the description. And you know, one more thing is if you got value from this, please hit the subscribe button. It generally helps me make cool stuff like this. And with that, I'll see you in the community. Have fun.