Transcript: Build & Sell Claude Code Operating Systems (2+ Hour Course)

Source video ID: bCljOfCH8Ms

Transcript

• 0:00 — By the end of this video, you’re going to know exactly how to build your own AI operating system, even if you’ve never even opened up Cloud Code before or some sort of tool like that. I’m even going to be giving you guys for completely free the entire setup guide. So, literally all you have to do is download that, follow along with what I do in the video, and you’ll be up and running. If you guys don’t know who I am, my name is Nate. I’ve been deep into the AI game for almost 2 years now. I scaled my last AI automation agency to over $100,000 a month, and then I sold it. I’m currently now running one of the largest AI communities in the world. We just hit 350,000 members and this is a free
• 0:30 — community. I also run this YouTube channel and a lot of you guys have been asking me how I upload so many videos and how I stay consistent. It’s because of my AI operating system and I want to give you guys access to the exact same thing that I use. So, let’s not waste any time and just get straight into this one. I’m super pumped. Here we go. So, first of all, like what is an operating system? The first thing you might think of is like, okay, my Mac OS or my Windows iOS or my Apple iPhone iOS. It’s basically the layer between you and your computer or your phone. It’s where everything lives. It it has your files,
• 1:00 — your apps, your contacts, you know, it’s the screen that you’re looking at. And it’s not really something that you think about too much because it just works and it lives there all the time. But now we’re getting into this world where AI is taking over everything and it’s seeping into everything. And now that I have an operating system inside of cloud code, it is my OS, but now it has agents on top of it. It has intelligence on top of it. So, I’m really not exaggerating when I say that I could spend an entire workday with just cloud code open and I could still do everything I need to and I would still be more productive than people that are clicking around in all
• 1:30 — of the different apps. So, when we add intelligence on top of our operating system, we have an AI that can see all of our files, all of our communication, everything important going on in the business. And not only can it see it, but it can also interact with it. And not only can it interact with it, but it can also remember it better than you can because you’re a human and you use a brain and you forget things. But the AI not only has better memory, but it can pull things from the exact source and it can find it quicker than you can. Think about how much time you’ve spent on work about work, which is basically just the idea of searching for things. Your coworker drops you a file. You forget if
• 2:00 — that was in Slack or an email. You’re looking for an old Google sheet that you knew you were working on last month, but now you can’t find it because you worked on 15 more since then. But AI could find that instantly and grab it for you because it has all of the context, it has all of the connections, it has all of the capabilities, and it has the right cadence. So the combination of those four C’s is everything that you need really to manage your business and you can get it to the point where it truly runs in the background for you. Now you guys have heard me say Cloud Code and obviously there’s some other tools that we’re going to be talking about today. But what’s really important
• 2:31 — in this space is that you’re building things to be tool agnostic because the tools change every 6 months. Those of you guys that have been with me for a long time on this channel, you know that last year I was all in on Naden and I’ve made the full pivot to cla code because I’ve realized I’m just way more productive with cloud code. I really don’t think that I could have built this sort of AI operating system in nodn. I just don’t think I could have the way that I can do it in cloud code. Models would be replaced. API endpoints might be deprecated. SDKs might be deprecated. Things will just happen. But we’re going to be building this in a way that is
• 3:01 — very futurep proof because I’m going to change the way that you think. I’m going to change the way that you decide how to do things. I’m going to change the way that you build this AIOS so it can survive whatever happens. Just yesterday, I took my AIOS and I moved it over to Codeex just to make sure that everything was good. and it took Codex about 2 minutes to get adjusted. And now I have two different AIOS’s, Cloud Code and Codeex. And if I need to put it in anti-gravity or if I need to throw it into the the other new kit on the block, I can do so. So I’m going to be teaching you the durable layer that sits underneath all these tools and all these different buzzwords. Okay. So there’s
• 3:31 — two main frameworks that we’re going to be talking about today. The first one is more of a highlevel framework around the way that I think about AI, and it’s called the three M’s of AI. We’ve got the mindset, the method, and the machine. All right. So, here is my three M’s doc that I’m going to be reading over today. I’m not going to read every single word because that would take too long, but I will drop this doc in my free school community. The link is down in the description. I have all of the other resources that I’ve ever shared on YouTube in there as well for free. So, if you want to check it out in deeper, go there. But I wanted to spend some time here because the mindset and all of
• 4:02 — these frameworks I built out in a way that it’s really important for you to think about not only when you’re setting up your AIOS, but also as you use it every day and also as you scale it up and also as you start to, you know, bring it into other areas of your business because right now we’re focusing on you personally, but now that you’ve got this set up, your business is set up better to be more AI ready and your employees and your co-workers are as well. So it’s really important. So there’s three habits I wanted to touch on. The first one is the default shift. basically the idea of you know before you do any task just ask yourself how
• 4:32 — could AI do this or at least how could AI do 30% of it because the truth is AI is not going to do everything 100% it might be able to do 75 or 50% but that’s still a huge win and that is a productivity gain so here’s a real example the other day I needed to update over 300 YouTube video description tracking links and the old me would have just been like man this sucks but I just have to do it and I would have manually clicked on each video, changed the link, and then clicked in the next video and just done that. Probably would have taken me an hour. But the new me, my
• 5:03 — default is, okay, I don’t want to do that manually. Let me just brainstorm with cloud code and let me figure out how do I actually make that happen? Whether that’s an API or whether that’s an MCP or whether that is maybe even having to just do a browser automation script. But the point is, my default shift is so stubborn now where if anything sounds boring or repetitive, I’m not going to do it. I’m going to use my AIOS to do it. That’s habit number one. Habit number two is the function breakdown. Understanding that your role is a set of functions. Everything you do is broken into a tiny task and you can
• 5:34 — just automate one piece of that at a time. So an example, okay, if I had to think about how can I automate an entire YouTube video. That’s a very tall task. But break down the pieces that go into a YouTube video. Ideation is usually step one. And then you’ve got your scripting or you know your building or your your your slide creation. And then you’ve got packaging titles and thumbnails. And you’ve got descriptions. and you’ve got, you know, comment replies, that is a huge job. But when you break that into little things, oh, I can definitely automate ideiation or at least 95% of it. I can definitely automate scripting.
• 6:05 — I can definitely automate the slide deck creations for my videos. And when you have that mindset of just breaking it into little pieces, baby steps, it becomes way more achievable. So, in this example, if this is the full flow of automating a YouTube video, you just have all these little chunks. And as you slowly over time automate more and more chunks, you get to a place where you’re, you know, now you’ve automated like 80% of a YouTube video. And each of these chunks is repeatable. So maybe this is like the, you know, slide deck chunk. Okay. What if we have a different process? Let’s just say we’re now doing
• 6:35 — like, you know, we’re doing like meeting prep or something. Well, guess what? We can go ahead and grab this chunk and just move it over here and we can just fit this into this puzzle wherever it goes. So all of these little chunks and baby steps that you’re taking are reusable and that mindset shift is very important to have. And then habit number three, we’ve got the curiosity rule. Never accept AI output without asking why. Treat AI as a mentor, [clears throat] not a vending machine. There’s basically this idea of dark code because so many people nowadays are using AI to write code, which means they
• 7:06 — don’t truly understand what each line or each block of code is doing. And not that you have to be able to read Python or whatever it is, but it is very important to at least understand why did you build this? And if X happened, what would happen to the code? You know what I mean? So the whole idea of being curious is asking those sorts of questions. Why did you design it this way? What happens if someone, you know, submits an invoice that’s empty and this code is supposed to handle filled out invoices? And the other piece of curiosity is if you don’t know if something’s possible to be automated or automated with AI, then just ask. Okay,
• 7:38 — so moving on to method. This is really really important, but we’re going to come back to this a little bit later once we’ve kind of, you know, got everything set up. But this basically just talks about, okay, how do you actually decide what’s worth automating and how much of it you automate? So, there’s some more information here about that kind of stuff. And then the last M stands for machine, and this once again gets a little bit more technical. All of this you guys will be able to dive into later, but we will revisit some of the stuff later on in this course. So, those are the three M’s. Um, a couple of the key highlights that we already hit on. It’s never binary. The question is never
• 8:08 — will AI do this for me. The question is to what extent can I leverage AI here? 30% 60%, maybe it’s zero, but it’s very unlikely that it’s zero. Every task on your plate has a leverage percentage. You just have to find it. And mindset isn’t motivation. It’s the lens that finds the percentage. The other one is treat AI as a mentor, not a vending machine. Vending machines take a coin and they give you something. But mentors ask you questions. They push you back. They make you sharper. They encourage you. They give you ideas. And that is the relationship that you want to have with AI, not a vending machine. And
• 8:39 — then, of course, we just talked about how your job is essentially a tree of tasks. And thinking about it like that makes big, vague, overwhelming ideas more realistic. So, another thing that’s really important from the 3M is that productivity drops before it climbs. And I don’t want you to get overwhelmed and just give up because you think it’s too hard to build an AIOS. It is not. So, think about this. a solo operator and they are trying to build an AIOS and they’re doing a 30-day sprint. This bar basically represents like productivity,
• 9:10 — right? So baseline, they are not using AI hardly at all. They’re being maybe 60% productive. Day three, they’re slowing down. They’re noticeably getting out less output because they’re trying to adjust to this new way of living. And that’s just how things work. Whenever you make a change, you typically are going to expect a 20% decrease in productivity. And let me actually just draw this out because I think it’ll be easier to demonstrate. So if we just real quick go and draw a graph, then let me show you what this means.
• 9:42 — So let’s say your business is operating like this. [clears throat] This is a steady line. Now if you decide to make a change, you have to basically be okay with the fact that this change might decrease your productivity by about 20%. So this is the, you know, sort of like negative 20% gap you’re seeing. The question is, is this change worth it? Will this change ultimately result in more than a 20% dip? Because up here, what you might be getting with an AIOS is a 50% gain. So is the dip in 20%
• 10:13 — worth the 50% upside? Yes, because you’re getting an additional 30%. Now, on top of that, you also have the idea of a learning curve. So, if I can just go ahead and draw another little graph real quick. And why am I being so boring? Let’s actually use a color. So, the learning curve. People think that learning is linear like this. And it’s just really not. Learning is exponential. But what does that mean? It means that when you start to learn something new, you basically are kind of below the linear line that you would
• 10:43 — have been if you were just doing it the old way. But because of the exponentialness, the exponential nature of learning, you will pass that mark. It’s just that it’s going to take you a few days to get there. Maybe this is day one, maybe this is day two, and maybe on day three, you break even. But by day four and five, you’re way ahead of where you would have been if you just kept doing it the same way. So this gap is where people sort of tend to give up or get overwhelmed because they realize they’re being less productive. So those two things are really important for you guys to
• 11:13 — understand. Don’t give up. You get through it and then you start to see all the gains and now you are being 5x more productive, 10x more productive. I’ve seen that with myself. I’ve seen that with my team. It is an incredible thing to actually witness and by the time you’re 2 weeks in, you won’t even recognize how productive you’re being and you can’t even imagine if someone took away your AIOS and you had to go back to the old way. You would want to cry. Okay, so now let’s get on to the second sort of framework here. And this one is called the four C’s. This one is more specific to the way that we set up our operating system. And this is kind
• 11:44 — of the way that I’m calling the AISOS, the AI automation society operating system because we’ve basically just been getting everyone to do it in this way. And it is the AI automation society way. So what goes into that? Four C’s. We’ve got context. What AI knows about you, your team, your tools, your voice, your business, your money. And then we’ve got connections. So what data can it reach? Cloud code on its own can’t really do anything besides search the web. But your business data, your especially the important business data you need isn’t
• 12:14 — available publicly on the web. So you have to give it connections to all of your different tools and all of the different databases where your actual important data lives. And then we have capabilities. What can it produce? What can it do with that data? You know, it has to be useful in some way, not just a knowledge source. And then we have the final C, cadence, when it acts on its own while you sleep and while your laptop is closed. And all of these build on top of each other. You can’t have cadence without connections. You can’t have capability without context. And you have to go in this order. 1 2 3 4. And
• 12:44 — I’m going to show you guys your exact path to get from 1 to four today. The whole idea is that if you opened up a new Claude chat right now or a new Cloud Code session, how much would it know about you if you asked a question? So test it. Ask a question and see does this answer like a teammate or like an executive assistant or is this answering like a stranger who just met you 5 seconds ago. So here is the four C’s of an AIOS. And once again, the important part is that tools change every 6 months, maybe even quicker. But the platform that we’re doing right now and the foundation that we’re building is
• 13:15 — going to be able to move across whatever happens. And I think you guys understand what we’re getting at here. We’re setting up our context, which is basically the brain, making sure it knows our business. We’re setting up the connections to make sure that it can touch things, whether that is an MCP or an API or a CLI or, you know, whatever it is. We’re setting up the connections. The capabilities is that it actually knows how to do things. You know, you have a bunch of SOPs probably already in your business. Can you now give those SOPs to your operating system so that it can do it just as well and just as consistently as one of your employees does it manually right now? And then of
• 13:46 — course the cadence where it gets really really cool and it starts to feel more like a 24/7 AI employee for you. And those are the four C’s of our AI operating system. And we’re going to be talking about how you can actually test if you have those things in place. And we have a skill inside of the um template I’m going to give you guys where you can run an audit and just make sure that you’re good and you can find opportunities to add more connections or add more capabilities and things like that. But for example, if you have connections into your calendar and your tasks and your CRM and your inbox, it should be able to grab all that and help
• 14:17 — you out. You could ask like, “Hey, what’s on my plate today?” or “Help me plan my day.” And it would be able to read through everything, look at your calendar, and help you block off things based on your priorities and based on maybe previous conversations you’ve been having in that week. And with the capabilities instead of having to drop in like a full, you know, two paragraph brief for what you want, you could basically say, “Hey, we need a quarter three report and I need it by the end of the day and it will be able to go do that because it understands your SOP for that. It understands all the business data and there’s going to be less follow-up, less back and forth, less of
• 14:47 — you micromanaging it.” So, here’s a quick snapshot of the pillars, what you might want to do to test, what you would see, and kind of be like, “Okay, this is good. This was a pass.” and what you might see and be like, “Okay, this failed. I need to do some work on this pillar.” So, hopefully this is all starting to click. I’m not trying to hammer this home too hard, but context, connections, capabilities, and cadence. Okay, so now you have to actually start to get this into cloud code somehow. So, what I want you guys to do is think about these different buckets, right? Um, there’s actually kind of seven tier
• 15:18 — one buckets, but they kind of fit into these groups of ops, comms, data, and planning. So, what I want you to do right now is I want you to grab a piece of paper or a Google doc or come into an Excal or a mirrorboard or whatever you want to do and just kind of layer out these seven things. We’ve got revenue, customer, calendar, comms, tasks, meetings, and knowledge. And the cloud code template I’m going to give you guys is trained on these seven things. But before we get into that, I think it’s
• 15:48 — really important to just kind of sketch this out on paper. you know, do it in your head first before you jump into the whole onboarding of the AIOS or you might forget some stuff. And if you do forget some stuff, it’s absolutely no big deal. Like I said, I’ve been playing around with so many different AI operating systems over the past multiple months and I’ve changed it a lot. I’ve evolved it a lot. I’ve had to restart before and it’s kind of funny because, you know, I think the term AI operating system is very new and it’s kind of ambiguous and a lot of people probably have a different idea of what that means. Um, you know, I think first I
• 16:19 — built out an executive assistant and the executive assistant has kind of evolved into my AIOS, but a lot of people have also been throwing around the term like second brain. And I think a second brain is a very great start to building an AIOS. So there’s a lot of these things that you might have heard and this is just kind of my take on an AIOS, which is why I’m calling it the AIS OS. So anyways, you sit here and these are the tier one things that you should be thinking about as far as on your month-to-month, what are you looking at?
• 16:49 — What are you tracking? We’re tracking revenue. We’re, you know, communicating with customers. We’re making sure they’re happy. We’re obviously having things come through on our calendar. We’re communicating with people both internally or on meetings down here or externally. And we have project management. We have tasks. We have things to prioritize. and we have general knowledge. So when you start to think of each of these layers, it helps you figure out the connections you may need and the tools you may need. So that’s what this phase is for. That’s what this section is about. So revenue,
• 17:19 — let me just show you guys what I wrote here. I wrote school, Stripe, and QuickBooks. These are the places that I’m actually checking pretty frequently to see how many members we have, to see, you know, how much revenue we’re growing, to see our promotions, to see how those are working, to look at QuickBooks, to look at our P&L, to look at our expenses. These are the places where I’m frequently checking to make sure my business is doing well, that we have runway, and that we’re growing healthily. And then you have customer, you know, what does that look like for you? Do you have a SAS product? Do you have like a brickandmortar? Where do you have your CRM? Where do you have information about your customer, and
• 17:50 — where do you provide them value? So, for me, I thought of the two biggest ones that I’m checking, which are school and YouTube. And so, as you start to think about these different sources, these are where your connections are going to come from. And you want to start to think about how can I make sure that cloud code can talk to school, Stripe, and Quickbooks or School and YouTube. Every single one of these tools will have some way to connect the data to an AI system. We’re living in 2026. It’s possible. Some might have MCP servers, some might not. Some have API documentation, some
• 18:20 — may not. But there is a way. I promise you. Even if you have to spin up a browser automation, which I’ve made a video about before, you can get through in some way. Okay, so let’s keep going. Calendar. This one was pretty easy for me. Basically, just Google Workspace. My Google calendar, everything lives there. Even if people book in through Calendarly, even if someone, you know, does something on my ClickUp calendar, it all syncs to Google Workspace. So, all I need there is one simple tool. For comms, I have two main ones, I guess three, actually, that I thought of. So, obviously Google Workspace for my email.
• 18:50 — I’ve got ClickUp, which is where we do pretty much all of our internal communication. And then I’ve got Slack, which is where we communicate with a lot of our vendors because sometimes they throw us in their Slack space or, you know, whatever the case is. So I need to make sure that my AIOS has visibility into all of those sources so it can pull data from there if it needs to. And then we have tasks. So pretty much our project management mainly lives in ClickUp for our internal projects, our internal tasks, our deadlines, our prioritization. And then we also have notion because sometimes we’re working with vendors on different projects and things like that and maybe they’re using
• 19:21 — notion which is pretty common. And as you’re going through and filling this out, you don’t have to get it all perfect on the first try. You’re probably going to forget a few things, but that’s okay. You know, start small. Start with the most important core things that you use the most, and then you build up from there. So, meetings, I use Fireflies at the moment, and that’s really important for my cloud code to be able to look at those transcripts and see what meetings I’ve had because it can help me, like I said, plan things or respond to an email better or make action items better. And then finally, we have knowledge. And this one might
• 19:51 — seem a little bit more ambiguous, but just think about where else is important information that doesn’t live in these sources that you’ve already talked about. And so for me, I was thinking, okay, YouTube transcripts, because up here, I was mainly thinking of like, you know, my YouTube comments or my analytics, but right here, I’m thinking, okay, YouTube transcripts. I want all of my knowledge from YouTube videos in Cloud Code somehow. And then I thought about Google Workspace, of course, because I have tons of Google Docs, tons of Google Sheets, tons of just like things in my drive, even videos that I want Cloud Code to be able to access. And then local files, which is pretty
• 20:22 — easy. You don’t really have to do anything to connect that, but I’ve got a ton of local files that Cloud Code has helped me organize and move around and sometimes even send to people. Now, as you start to look at these different tools, think about this. You know, I joke around with people that if someone wanted to get a hold of me or if someone had a question for me, they would actually be better off asking my executive assistant because my executive assistant is also linked to all of these sources of data. So, it has all the knowledge I have, but it has a perfect memory and it never sleeps. So, if you
• 20:53 — can look at these sources and be like, “Okay, cool. If someone was able to talk to an AI agent that had all of this data and it could answer the majority of questions that my team or that you know you know some a customer might ask me then we’re probably in a good spot to go ahead and get started. And if there’s other things that come to mind when you sort of do that you know mental test then just chuck it down into the right category. And once you get to a point where you feel good about what these tools look like now what we need to do is we’re going to get started building our AIOS. So, the very first thing that I want you to do is go to my free school
• 21:24 — community. The link for that is down in the description and join this community. Your membership will get approved as quick as we can. And then once you’re in here, you’re going to go to the classroom and you’re going to go to all YouTube resources and you’re going to go in there and download the different resources for this video. You’re going to need a GitHub repo and you’re going to need, you know, those other docs that I talked about, but you need that to get started. And then we’re going to be doing this inside of Claude Code. And the way that I like to use Claude Code is in VS Code. You could also do the desktop app if you choose, but pretty much in all my YouTube videos, I like to
• 21:54 — use VS Code and that’s what we’re going to be using in today’s tutorial. It’s completely free to download. It is just an IDE and then we’re able to open up Claude Code inside of it. So, I’m using VS Code today. Google Visual Studio Code, download it, and then the next step is you need to be able to connect your Claude accounts. And you do have to have a paid Claude account in order to use Claude Code. You can do the 17 bucks a month plan or I guess it would be 20 on monthly, but you can start here. And if you’re hitting your limits really quick and you want to scale up, then scale up. It’s not a big deal. All right. So once you’ve done those two things, you’re going to open up Visual
• 22:25 — Studio Code. And it will look like this. Now, what you’re going to do first is you’re going to install the Cloud Code extension. So on this lefth hand side, go over to this button, click on extensions, and then you’re going to type in Claude Code. And once that pops up right here, you’re basically just going to install this extension. Once you install that, it’s going to prompt you to log in. And that is where you’ll log in with your paid Claude subscription. Once you get that set up, you’re going to see this button in the top right, which looks like Anthropics logo. And you’re going to go ahead and
• 22:55 — just click on that. And what that does is it opens up this little Claude code agent. And this is very similar to the way you would chat with Claude or Chatbt on the web. You know, you have a little chat box here. You can say hello. And this is going to call on Claude Opus 4.7 and it’s going to respond to you. It’s super simple. Now, what we have to do is we have to open up a project to work in. So, open up your file explorer, go to your desktop or your documents or wherever you want to have your AIOS main folder and just create a new folder. Just call it AIOS or, you know, whatever you want and create that folder on your
• 23:27 — file explorer. And then you’re going to go up to this top left hand explorer button and it says you have not yet opened a folder. What you have to do is just open up that folder. And that is basically our working directory or the project that we’re working in right now. So, you’re going to click on open folder and open up that one that you just created. So, I just opened up one called AIOS demo. Yours should look exactly like this because there’s nothing yet in that folder. So, all I’m going to do now is I’m going to close out of everything. I’m going to double click, open up Cloud Code, and now what we have is on the right hand side, we have our agent, and
• 23:58 — on the lefth hand side, we have our files. So, right now, obviously, there are no files in here. Now, what you’re going to do next is you’re going to go to my free school community, and you’re going to grab the link to the GitHub repo, and you’re going to grab the URL from it. And then you’re just going to message Claude and you’re going to say, “Hey, can you clone this GitHub repo into this current project? It’s going to go ahead and search the web for it. And then it is going to pull it in and you’re going to see a bunch of folders and files pop up on the lefth hand side right here. It’s going to ask you if it’s okay to allow this command. And you’re just going to go ahead and click yes, allow get clone.” And what you can
• 24:29 — see is now that it has been cloned, we have a bunch of these different files over here. So you can see that we have acloud folder, we have a archives folder, we have a context folder, decisions folder, a references folder, and then we have all of these other files right down here. So let me real quick take a second to familiarize with what we’re actually looking at. The dotcloud folder, this is kind of like your overall project guide, I guess, claude. What’s in here right now is we only have a folder inside of this called skills, which is why we have this little breakdown. And
• 24:59 — this is basically where all of our skills are going to go in the future. Now, what is a skill? A skill is basically just something that gives you six arms. No, I’m just kidding. But it’s basically a file inside of Cloud Code and it’s like a recipe. So, let’s say you had a skill for building a LinkedIn post or writing a LinkedIn post. What do you do every single time you want to write a LinkedIn post? Maybe you do research and then you generate a graphic and then you write the copy and then you review it and then you post it. So, how annoying would that be to explain every single time to Claude Code that you want to do those five steps when instead you
• 25:30 — could just package those into a skill and then say, “Hey, Claude Code, write me a LinkedIn post.” It would read your skill document and then it would go do everything. So, it gives you more predictable outputs, higher quality. So thinking of it like a recipe is a really good analogy because also let’s say you build this skill or recipe and it comes out not the way you want it and then you realize okay next time we do this I need to add like one more egg so that the you know brownies are cakeier or whatever then you just update the skill say hey instead of one egg do two eggs and then next time you say hey run the skill it’s
• 26:01 — better so the skills will ever be evolving if you want them to and it’s going to help out a lot so if we open up this folder you can see that I already have given you three skills in here we We have one called audit. We have one called level up. And we have one called onboard. So in just a sec, we’re going to run the onboarding skill to get you fully onboarded into this AIOS. But let’s take a look at what skills actually are. So inside of the level up folder, we have a skill.md. And this skill.md, as you can see, is a very simple markdown file that explains what
• 26:31 — the skill does. We have a name, we have a description, we have what the skill does, we have what the skill does not. And then we say like, okay, here’s what you do every time the user actually runs the skill. Phase one is the interview. Phase two is the method interview. Phase three is blah blah blah. It’s basically just an SOP. And what’s really cool about that is, let’s say you have an SOP for onboarding a client. You could give Claude Code that SOP and give it the right connections to onboard a client and say, “Hey, can you turn this SOP into a skill?” And then boom, you’ve essentially got an automation for
• 27:02 — onboarding a client. But anyways, I don’t want to get ahead of myself or start to confuse you guys. Let’s just actually keep going down real quick. What else we’ve got in this project? So, we’ve got an archives folder, which is where Claude will put old documents or things that you don’t need. We’ve got a context folder, which is where Claude will put different context files about your business, about you personally, about maybe the way that you like to communicate. Everything that it needs to know about you and your business, it can start to organize inside of this context folder. We’ve got decisions. It’s going to keep a log of important things that you guys have decided to do together.
• 27:32 — And then we’ve got references. And this one right here has a markdown file about the 3 M of AI. Now, this is a really good place to start. And I built this template so it’s kind of hackable. Over time, what’s going to happen is you’re going to build more and more files. There’s going to be more folders. There’s going to be different things. You might have projects. You might have quarterly folders. That can evolve and that is completely fine. The important thing is that Claude understands what the folders mean and what type of stuff goes in them. And that kind of stuff gets defined in a file called the claw.md. The claw.mmd is basically the
• 28:04 — master prompt for this project. So right here it says, okay, blank’s name operating system. You are Nate’s personal AIOS. Your job is to be their thought partner. Help them think, decide, and ship faster on this priority. And there’s a lot of placeholders here that are going to get filled out as you start to grow this project. Now you can see there’s a section here called your skills which is important because it has to understand when to invoke what skill and then we also have where things live. So context references connections and as you add
• 28:34 — more folders and files on the lefth hand side you’re going to just want to make sure that your claused file is getting updated as well so that claude actually understands where things live. So there’s more information in the cloudmd file. I’m not going to read it all right now. And this is another one of those files that evolves every day. My clawmd file in my project has probably been updated two times a day for the past multiple months because I keep adding new things or I keep taking things away. So none of this is set in stone. Don’t stress yourself out. As you learn more, your project is going to grow and learn more, too. And then what else we have
• 29:04 — down here are just a few other markdown files that are important. You know, we’ve got some connections here that are going to get filled out. We’ve got some expansions that can help you figure out, okay, once you finish Nate’s tutorial that you’re watching right now, what do you do from here? How do you make this better? So, those are just some other files that are in here, and I don’t want to read all of them out, but hopefully that little overview gave you at least a bit more comfort of what you’re looking at. So, now let’s actually go ahead and get onboarded. So, I’m just going to say in natural language, all right, I just cloned this repo. I want to set up my AI
• 29:35 — operating system. My name is Nate. Can you help me get onboarded into this project? And if you say anything like that, it should understand to invoke that onboard skill. And the cool thing about cloud code is you can basically see what it’s doing. So it said, “Okay, I’m going to read the project’s onboarding doc to see how I can set you up.” And we can see that it’s reading things. You know, we can see the exact files that it’s actually reading. And then it says, “Okay, welcome Nate. The repo comes with an onboard skill that does exactly this. It’s a seven question interview that scaffolds your day one file set. Since the skill was just
• 30:06 — cloned, let me read it and then I’ll walk you through it.” So this is where it starts that onboard questionnaire. The first question is, “Who are you? What do you sell and who do you sell it to? So, keep in mind because I’m just doing this as a demo for you guys, my answers are going to be pretty short. They’re going to be kind of vague. But for you, I would really spend some time here and give it context. You know, obviously it’s not the end of the world because you can you can continuously iterate, but I would definitely spend some time here and give it more than just a couple sentences. Give it a few paragraphs. And what makes it even easier is if you use a voice dictation
• 30:37 — tool. So, the one I use is linked in the description. It’s called Glido. You might see that right here. I’m actually using Whisper Flow, but I have fully switched over to Glido and I am an official member of the GLO team now. I just believe in that product much more. It’s faster, it’s private, it’s way more agentic. So, if you guys are looking for a voice tool and you want to support, then definitely check out Glido. Windows support is coming very soon. All right, so my name is Nate. I run a AI automation YouTube channel. I’ve got a free community and a paid community and we’re working on building out some other different types of programs. And um our
• 31:07 — offer is basically just we’re teaching people how to learn AI automation regardless of their background. We’re helping them figure out how they can get a career or we’re helping them figure out how they can get clients. And that is pretty much the type of person that we’re currently appealing to. Okay. So that is the short paragraph that I’m going with for now. Now it’s going to save that to the intake file, which is great. And that is actually this one right here. So as you answer, it’s going to save these answers right here. So as you can see, question one was who are you? and this is what I answered with pretty much right here. And then it’s
• 31:37 — going to keep going down the list and asking us these questions and it’s going to keep filling in our answers here so it can reference them later. So anyways, let’s keep going. Question two is to paste one or two things that you’ve written recently. Don’t edit them. Paste them verbatim. And the reason it’s doing this is because it wants to understand the way that you speak, the way that you talk. So I’m going to go grab one LinkedIn post and one school post and I’m going to drop them right here in the chat. All right. So I’m dropping in those two posts. Now, think about as far as like the way that you would want this to understand your writing styles. If
• 32:07 — you wanted it to just be LinkedIn, then you should say, “Hey, these are both LinkedIn posts.” Or maybe you wanted to give it, hey, these are two emails I sent to clients. These are two emails I sent to my team. I have different types of communication based on who I’m talking to. And that’s completely fine. You can give it that information. The more the marrier. Anyways, question three. What are your two to three biggest priorities for the next 90 days? So, I’m just going to make up a few different priorities here. you would obviously give whatever you’re really working on for Q3 or for Q2. What is your current sprint? What are some
• 32:37 — milestones that you’re currently working towards? So, I don’t think that you guys really want to watch me answer these next couple questions. So, go ahead and finish all seven and then I’ll meet you back here when you’re done with that. Okay. So, it now comes back and it says that day one is done. Your AIOS knows who you are, what you sell, what matters this quarter, and how you sound. So, today you could say, “What should I focus on this week?” Actually, let’s just try what should I focus on this week and we’ll see what it says. Tomorrow it says you can pick one tool from your connections and wire it up.
• 33:07 — And then on day seven, you can run your audit to see your four C’s score. So, if you really want to take it slow, you can do that. But, we’re going to keep moving forward and we’re going to start setting up some other things as well today. But, as I ask it, what should I focus on this week? Obviously, it knows because we just had this conversation. But what’s important is that it’s pulling from these files. It’s in the context folder right here. And we have these new three files that were added. We have about business, we have about me, and we have priorities. And it just read from those two together in order to see that we
• 33:37 — have AIS live on July 11th. We have our new program in development and team management and making sure that we can keep tasks on track and everything like that. And then it says, if I had to pick one thing for Monday, it would be the AIS live speaker lineup and where could the default shift apply here? to what extent could AI be leveraged on this task? So, as you guys see, like I said, I’ve trained this thing to be your mentor. Not just saying, okay, here’s what you should do or here’s what I found, but trying to train your mindset to actually shift. So, you’re starting
• 34:08 — to think about this exact type of question and multiple other questions that come with using an AIOS first. Remember, your goal should be, how can I do everything I need to do right in here, right in VS Code. I know I have my Chrome tab open right here. So, if I needed to go in here and click around, I could. But how could I try to be as productive as possible just in this interface? That is the default shift. It’s the whole cliche of, you know, like you only get out what you put in. So, um, yeah, if you don’t really commit to using this, then you’re not going to
• 34:38 — feel the ROI of the AIOS. Let’s take a quick peek inside of these. These are all just markdown files. So like about business, if I open this up, we can see that it has access to our offer, our ICP, it has our services, and it has our revenue model. This is obviously very very vague at the moment. It’s very minimal, but remember over time as you make decisions, as you launch new offers, as you, you know, make pivots, this will continue to evolve. Same thing with the about me doc. So right here, it just knows a little bit about me and about some of my top pains. But I could
• 35:09 — also give it some more actual context about me. I could tell it my age, where I went to college, you know, what I like to do in my free time. I can give it more data here. And then of course we have the priorities doc. And this one will always evolve. I think the best way to be doing this would be at the start of every quarter. You know, our business, we work in quarterly sprints. Not everyone does that. Maybe you work in two week sprints or maybe you work in, you know, yearly sprints, whatever it is. But find a cadence that you want to give to your AIOS so it knows the big milestones that you’re working towards. And it makes it really easy because for us inside of ClickUp is where we have
• 35:40 — all of our milestones for that quarter. So really the better way for me to tell my AIOS about my priorities would just be to say hey go to the ClickUp workspace called you know Q2 OTAAS and just read everything and those are our priorities for this quarter. And that’s where you guys are starting to understand the value of the connections because right now this thing only knows what we’ve told it because you know that’s the source of truth for it. But once we start to connect it to all of these things that we kind of sketched out earlier, once we connect it to these, it won’t have to keep asking us
• 36:11 — questions because the default will be instead of asking Nate for answers, I will just go look in school or I will just go pull the data from QuickBooks or I will just use my GWS tool. That is what it’s able to start to do. All right, so let’s say that’s where we’re at now. I’m going to go ahead and do a slashclear which basically just clears the conversation and I’m going to pretend it’s the next day. Okay. Hey, good morning. It’s day two and I now want to start to actually connect you to things. You know, what are the most important things that I could start to connect you to so you can get more context about my business and we can
• 36:41 — just start to grow your database of data that you get to access. Now, if you already know, okay, like ClickUp, Google, Fireflies, those are the three most important things, then you can start wiring that up right now. But you can also have a conversation with it. You can be curious. You can ask it what it thinks. and you can start to, you know, just kind of layer on top of each other the different connections. Now, when you think about an actual connection, if you’re not technical, don’t worry about it because all that happens is you say, “Hey, I have ClickUp. Go do research for me and
• 37:12 — figure out how to connect ClickUp.” And it just does it. The one thing you will probably have to do is you’re going to be the one that has to go grab the API key and you’re going to have to give Cloud Code the API key so it can actually access your stuff. It’s basically just a password. So what this came back and said, “Good morning, day two. This is the exact right move. I looked at your connections doc and I can see your priority order.” So here is the three things that you can do together. ClickUp for tasks, ownerships and deadlines and DMs and channels. Fireflies for all of your meetings and
• 37:42 — then Slack to get communication from other vendors. Now I actually disagree with this priority probably because I didn’t give it enough insight. I think ClickUp is definitely number one for us and then for me number two would be like my Google workspace. But anyways, let’s just start with ClickUp so I can show you guys how that works. It is a massive tool for us obviously. All right. So, one other thing that you can think about when it comes to ClickUp or you Gmail or Slack or whatever it is that you’re giving your agent access to is that you don’t have to give it access as you
• 38:13 — because your account probably has full permissions. It can see every space. It can download all the data. it can write whatever it wants. In some cases, you probably do want that, but in other cases, why not just create an account for your AIOS? What I did here is I created an account called Up AI. And now I give my uppai API key to cloud code rather than my own personal API key. And you can do this for all of your other tools as well. Maybe you want to get a separate API key inside of QuickBooks
• 38:44 — and you only want to give it read access. You know, you can do certain things like that to restrict the ability of the AI to make sure you don’t have a situation like I forget which company it was, but I just saw some news article where an AI deleted like a really big database and you just want to make sure that that doesn’t happen. So, per API key or per account, you can set different permissions. So, think about that. And what else that means is if you have a bunch of different AI agents, each of them can have their own API key. and if they’re spending money on that
• 39:14 — platform, you know, whether that’s a research platform or just an AI model, you can then see which automation is using how much money. So, think about the way that you separate out your permissions and your keys. But anyways, this is my uppi account of ClickUp. What I would do here is I would go to um up here in the settings. I would click on settings. I would go to the ClickUp API. And now I would just copy this token. And this is what I need to be able to give to Claude Code. And once it does that, it can access ClickUp. Now, obviously before that, it’s going to
• 39:45 — say, “Hey, can we connect the ClickUp MCP or, you know, how do you want to do this?” A lot of these AI models are going to want to default to using an MCP server. As you can see here, how would you prefer to wire it? MCP server. It’s the fastest if one exists for ClickUp. I don’t like using MCP servers because basically what happens when you do that is it gives you access to every single function and every single endpoint that’s possible, but you always don’t need all of them. And also having a bunch of MCP servers loaded into your project actually eats more tokens. So what I prefer to do is API endpoints. So
• 40:17 — here’s what I’m going to say and this is what you should probably say to all of the different integrations that you need to connect. I want to use ClickUp’s API because it’s more token efficient than having the MCP server. So what I want you to do is do research on ClickUp’s documentation about all the different API endpoints. it would probably be helpful for you to set up a reference guide, a markdown file inside of this project that has all of the endpoints stored so that later if you need to use a different one, you don’t have to go do research again. You can just reference
• 40:47 — that file. But anyways, go do research about that and then create me av file and I will give you my ClickUp API key in that file. So, let me just kind of like translate what I meant by that. It’s going to look up the ClickUp API documentation. So, if I search ClickUp API documentation, it’s basically going to find all of this and it’s going to look through it because there’s different endpoints to get a token or to create a task comment or to, you know, make a list or to remove things from a list. Every single function has a
• 41:17 — different endpoint. And every single time that Claude would need to do something in ClickUp, it would have to come research it. But instead, if we just let it research all of this once and then save it as a markdown file, markdown’s really easy for AIs to read. It’s also very cheap for them to read. So now we have our own sort of database of every single possible ClickUp function and we have that stored locally. So that is the research that Cloud Code is doing as we speak and it is going to save that for later. Now what is AENV file? The ENV file is basically just a secret file, right? So
• 41:50 — this is what it looks like. This is where you’re going to store all your secrets. So we will input a ClickUp API token and a team ID. Now the reason why we have a dot is because it’s aenv and it gets excluded from anytime we do a public push to a repo or anything like that. So it basically just means that we are protecting this API key. It’s it’s much more secure for you to paste in your API key into the ENV rather than you know giving it in the chat history. You know sometimes it might say hey you know just drop in your API key and I’ll
• 42:20 — set it up. Don’t do that. just tell it to create the env with a placeholder and then it’s so simple for you to go into ClickUp, copy this, and then you would literally just paste it right here, hit save, and now Cloud Code can use it. And if you ever get confused about where you find your API key or how any of this works, just ask Cloud Code to do the research, and it will figure it out for you 99% of the time. Also, guys, I realized that I switched over this mode to bypass permissions. So, there’s basically different modes in Cloud Code.
• 42:50 — What I like to do typically is start on plan mode if I’m going to build like a skill or I need help brainstorming. Um, but if you’re on edit automatically and you’re trying to do something like research like you just saw, it might stop constantly and say, “Hey, can I do this web fetch? Can I do this? Can I do this?” And if you get annoyed of that, you can switch to auto mode or you can also switch to bypass permissions. So auto mode uses a little bit more tokens because it basically uses an AI to analyze what am I about to do? Is this safe? If it’s safe, I’m just going to go ahead and do it. But if I’m doing something like a delete or if I’m doing
• 43:21 — something like a push or anything that might be a little bit risky, then I’m going to stop and ask Nate or I’m going to stop and ask the user. Bypass permissions just says, “Okay, you can go do everything.” I have never had an issue with bypass permissions. It’s it’s deleted things, sure, but only when I’ve asked it to. But you do run that risk of full autonomy. So, just wanted to at least let you guys know about that. The way that you can enable bypass permissions is if you go to the settings down here and you click on settings and you type in claude and then you can see
• 43:52 — right here allow dangerously skip permissions. If you turn that on, you should now be able to see this option. But you could always just use auto if you’re scared and then eventually if you feel more confident you can switch to bypass. But that’s why you might have seen that down there. Okay. So it says that ClickUp is wired. It says that we have a full V2 reference. So if I come into the references, we have a ClickUp API markdown file. And if I look in here, this is a very comprehensive doc. It is pretty long. So if cloud code does use this, it’s going to have to read all of these lines, but it has all the
• 44:23 — endpoints. And now it has basically full understanding of how to use ClickUp. So let’s go ahead and check it. It says, do you want me to run a test per assenee workload snapshot right now? It will see who’s holding what across all 17 people. I’m just going to go ahead and say sure. and we’re going to see if it’s able to use my API key that I gave it, use the team ID that I gave it, and use the right endpoints to find that information. And the cool thing is as you’re testing out these different endpoints or these different skills, if it hits a mistake, it’s going to fix itself. But what else is good about that
• 44:53 — is it found an edge case because every time that it fails, it learns and it can update something. So if this fails, I’m going to say, okay, update the API doc so that next time you do this, it never happens again. Or if you were if this was running a skill right now, like let’s say we ran a skill called team check-in, I would say, okay, can you go update the skill with the issue that you just ran into, how you fixed it, and to make sure that you never ever run into that issue ever again. So honestly, think of a failure as a good thing because it gave you more data about what not to ever do again. Another one of
• 45:24 — those super healthy Nate mindset shifts. Okay, so I’m not going to show you exactly what’s on the screen right now. I’m going to have like a blur over this or a box over this because I don’t want you guys to know what I’m up to. But all of this is legit. This is exactly what’s going on in our ClickUp. And at the end of this, you guys can’t see. I’ll show you this line, but it says the three things worth doing. It gives me a number one, which is definitely seems like I should probably check up with this person right now. We have number two and
• 45:54 — number three, which all make sense in the context of my business right now. So, that is proof that that’s working. Let’s do one more quick thing. I’m going to go ahead and just do a /cle. Actually, I’m just going to open up a new tab and I’m going to say shoot Nate a message in ClickUp. Actually, I don’t know why I’m not speaking. ClickUp and just say, “Hey, YouTube. Making sure that this is working.” Okay, so this will be interesting to see if it decides to use a channel or if it decides to just go ahead and DM me directly.
• 46:24 — But what is it doing? It’s reading the connections doc and then it’s reading the ClickUp reference dock and then it is going to shoot me off a message. Now, there’s one thing that I would say here that you might want to consider is if you have like a couple main API endpoints that you hit, you probably want to put those somewhere that makes a little bit more sense. So, for example, we don’t want it to read every single time that entire doc because that could just waste tokens. But what’s interesting is, you know, let’s take the example of the team check-in. Whenever we run the team check-in, we could say,
• 46:54 — “Okay, when you ran that, what endpoints did you use?” And it would say, “I used endpoints X, Y, and Z.” Okay, put endpoints X, Y, and Z in the skill file. So, you don’t have to look at the full API documentation whenever you want to do that specific skill. So, I hope you guys are starting to understand here why it’s a little bit difficult to give you like a step-by-step playbook for setting up your own AIOS because everyone runs their business differently. Everyone has different connections. But I’m trying to give you guys the way that I think about the, you know, the mindset of setting
• 47:24 — this kind of stuff up. So anyways, it decided to create me a task. So it didn’t completely understand what I wanted. So in my ClickUp, my real Nate ClickUp, not the UPAI version, I got a task here and the task says, “Hey YouTube, making sure this is working.” And it assigned the task to me. So it misunderstood my request a little bit. I was hoping it would send me an actual DM, but we can see here that it was able to create a task and assign it to me. So, at least it understands that. And really, what I was trying to prove to you guys there is that it was able to use a different endpoint as well. That
• 47:55 — is ClickUp. Now, before we set up another one and before we talk about some other stuff, I wanted to show you guys these skills that I have built in. So, the first one is called audit. So, I’m just going to go ahead and do / audit to show you guys what this skill actually does. This is basically going to judge your AIOS. It’s going to look at the four C’s. So, context, connections, capabilities, and cadence. And it’s going to tell you what you might be missing, and it’s going to tell you what you need to set up. So, right here, it says, “Okay, I’m running the four C’s audit. Let me scan the project
• 48:26 — structure.” And it’s going to come back with a grade. And this is really great because if you’re getting overwhelmed with what you should be doing, just run/ audit and see where the gaps are and kind of help plan out your next steps from there. Okay, so April 30th audit, we got 54.5 out of 100 and we’re currently, you know, we literally just got this set up. So, honestly, I think this is being pretty generous with context. I mean, 18 out of 25, it could definitely get a lot better. Connections, we’re at 16 out of 25. It could get a ton better. So, maybe we
• 48:56 — need to tweak this actual audit a little bit because we only have one connection and that’s ClickUp. But the point being, it’s going to start to go over your strengths and weaknesses. So, ClickUp got wired the same day. Okay, maybe it’s doing relative because it’s like, okay, on day one that’s not too bad. The connections.mmd is comprehensive. We have ClickUp, Fireflies, Gmail, Slack, and revenue. But here are the top three gaps and they’re ranked by leverage. So tier one domain is thin, meaning only one of seven tier one domains are reachable. So if we think back to this where you worked up, we had seven of
• 49:26 — these, right? Revenue, customer, calendar, comms, tasks, meetings, knowledge, and we’ve only hooked up ClickUp. So, it’s a little bit of comms, a little bit of tasks, but still not anywhere where we want to be. So, it says, “Why are fireflies next?” Then we have no recurring trigger or cadence, which is the fourth C. We’ll get into that later. And zero useruilt skills and zero agents. So, that’s really where we get a lot of the leverage is when you come into this AIOS and you build your own custom skills on top of everything. Now, speaking of custom skills, let’s go
• 49:56 — ahead and also run this other one. This is called level up. So, I’m going to go ahead and do a slash and I’m going to do the level up skill. And I actually didn’t realize that it suggested that it said, “Hey, these are structural gaps to explore capability gaps, which is what your AOS could do that it can’t do yet. Run level up after this audit.” And also, you can see it says that do you want me to save this? So, every single one of your audits can get saved so that you can see how you’re actually improving your AIOS over time. This level up is basically going to analyze
• 50:26 — your business, your connections, everything that you’re doing. And it might even ask you questions, but it’s going to show you some areas where you could just level up your stuff. So, it says pulling from your stack. You have a top pain of managing your team. You have quarter 2 priorities. You have things that are reachable now, and you have things that are reachable soon. So, here’s what you should do. Answer these questions however feels natural. One sentence is fine. Walk me through this past week. What did you do three plus times? Drudgery. Is there anything that felt manual, boring, or copy and paste? Smart intern test. Anything where you thought a smart intern could absolutely
• 50:57 — handle this, but you ended up just doing it yourself because explaining it would take longer. Constraint. If 500 new community members showed up on Monday, what would break first? And growth lever. What would give you 500 more clients tomorrow if it were running on autopilot? So, if you answer these five questions, there is absolutely no way that you’re stuck. There’s absolutely no way that you think, I don’t know what to add to my AIOS. I don’t know how to make this better. Every time that you answer these five questions, it will find an opportunity for you to automate something, to build a skill out of something, to connect something else.
• 51:28 — And that is why I built these two skills to go together, audit and level up, to make sure that you’re constantly improving and that you don’t ever feel alone and you don’t ever feel stuck with your AIOS. So, I’m not going to type that right now. I’m not going to answer that cuz we have a lot of work to do on our own. But, I want to show you guys something I did recently with Claude Code and Google Workspace. So, I live in Google Workspace, right? Um, everything that I do is pretty much inside of there besides ClickUp. And what happened was I realized that across my company, everyone had spun up so many different
• 51:58 — Google Docs and so many different Google Sheets and we had tracker sheets all over. And Google search is horrible. It’s absolutely horrible. So, I was able to connect Cloud Code to the GWS CLI. It is one tool and it gives you access to everything in the GWS environment. So, I’m going to show you guys exactly how to set that up in just a sec here, but I was able to do some really cool things. So, let me just show you guys a few of these cool things that I was able to do. If I go to my sheets, I have this video database that has all of my videos. It has the thumbnail, the title, a link to them, and a summary, and all of the
• 52:29 — resources that are associated with that video. And this is now my video database. And I don’t do this manually. Every time I post a video, the video gets pulled in, the resources get pulled in, and then a row gets updated on here. And I built this because it was able to search through my entire drive. It could find all my YouTube videos. That was a different connection. That was a YouTube data API. But anyways, it could find all that and then it could find all my resources and then it could just put everything in here. Super super powerful. Another example, we have this pretty massive doc about our AIS live
• 52:59 — event and there’s speaking slots and there’s, you know, different timings and there’s different outreach things and there’s a lot of data here. And I basically said, “Hey, Claude Code, go read this document.” I gave it the link to this. It opened up the doc in Google Drive. It read all of it and then I said, “Create me a database, a tracker sheet.” And it created this Google sheet because it analyzed everything that was going on in that doc. And it was able to label everyone. It it made all of these have drop downs with color coding. It assigned these tasks to everybody. It was able to just do everything for me.
• 53:29 — And this would have taken so long for me to do manually. Another thing I was able to do is I have like all my YouTube videos stored locally and it was able to search through that, organize everything, date everything, and then upload all of that to Drive. And then it deleted everything locally. And now what I could do is I could have it organize all of these into different folders based on the date. So it can just help you organize things, move things around, create things. And all of this living inside of my Google Drive, inside of a shared company, Google Drive is awesome because it can find things instantly.
• 53:59 — And everyone else on my team, I was like, “Hey guys, can you use the GW CLI real quick and can you look through everything you’ve ever built for the company and can you put it into the shared drive?” And now all of our AI agents are able to touch all of the things so much better. So the GWS CLI is a huge unlock. Unfortunately, if you use like Microsoft and that is your main environment, the GW CLI probably won’t be too useful for you. I’m sure there’s something different for Microsoft’s environment that you should use. But I’m going to play a quick clip for you guys about how to set up the GWS CLI and a
• 54:30 — little more information about it. So if that doesn’t appeal to you or not interesting to you or you already have it set up, then go ahead and skip past this next part. But I’m going to play this clip real quick. Google just dropped what some are already calling the most powerful workspace CLI on the internet. So if you’ve got a ton of stuff that lives in the Google environment just like I do, then you’re going to love this because now any of my cloud code projects can access everything. And all I had to do was install one simple thing. So here you can see I said what can you do with GWS which is Google Workspace CLI. So it can search, list, upload, download, move,
• 55:01 — copy, share anything in my Google Drive. It can do anything in my Gmail. It can do anything in my calendar. It can do anything with Google Docs. Same thing with Sheets. Same thing with Slides. And it also has multi-step workflow recipes. So these are basically like skills. These are chain command patterns for common tasks like creating docs from templates, reading sheet data, and creating a report doc, finding free time, and scheduling a meeting. And there are over a hundred of these that it actually has. So out of the box, when you give Claude Code the GWS CLI, you can do anything across any of the tools.
• 55:31 — And you also have access to over a 100 skills. So, I don’t know how many times you guys have tried to use something like Claude or Naden to build you a Google doc. And you do this over API and it ends up just looking like something like this. It literally just looks like raw markdown and it’s obviously horrible. And sometimes to go along with a YouTube video, I make resource guides that look like this, but obviously they have to be formatted. I’ve got like a header up here and I’ve got links and different things in this format. But now I can literally just take the link to a YouTube video. I can drop that into Cloud Code and say, “Create me a YouTube
• 56:01 — resource guide.” It’s going to go ahead and download that transcript from the video. And now what it’s doing is it’s creating the Google doc, not via API call, not via MCP, but via a bash command, meaning it’s literally running a terminal command in order to talk to Google and make this. So, it just actually created the doc. Here’s the ID. And now it’s going to populate it with what I need. And now it finished this up. It gave me the link. I’ll click into this. And we can see, boom, we have an actual resource guide. It’s got the image inserted up here as a header. It’s got a link that goes right back to my YouTube channel. It breaks down the market traditional automation. It goes
• 56:32 — through all this stuff and then even has my CTA at the bottom as you can see after all these horizontal lines to join the plus group. So that was obviously just one quick example, but there’s so many different benefits here using this workspace CLI. The first one is that you have one interface. So basically, like I said, it was one GWS CLI that cloud code now has access to and it can access my Gmail, my drive, docs, sheets, calendar, admin, and more. It’s also JSON first with structured responses. So our AI agent is really, really good at working with it. We have auto discovery, meaning the CLI is pretty much always going to stay up to date automatically. Pretty
• 57:02 — much zero maintenance because we authenticate and then we’re going to be good to go. It has built-in skills for triage, for prep, for generations. Like I said, there’s a hundred others. And it’s not much overhead because it’s basically just one tool. It’s not the same as like having all these different API endpoints or all of these different MCP configs and tools that would take up more context. But I know you’re probably wondering, what is a CLI? It stands for command line interface. And what we’re typically used to is a GUI or a graphical user interface where we can see buttons, we can see form fields, and we can click on things and that’s how we
• 57:32 — navigate, but computers are more navigating by text and by commands and by typing. So that’s really all that a CLI is. And this is an open- source Google Workspace product, and obviously it’s completely free. So I’ll leave a link to this GitHub repository down in the description if you want to check it out, read more about it. But I’m also going to walk through some of the key details right here. The first thing that I wanted to show you is if you go down here to the skills, this is where we can actually see all of the different kind of recipes they call them for pre-made multi-step workflows that it has. As you can see, creating events from sheets, creating presentations, creating meat
• 58:03 — space, label and archiving emails. There’s so many different patterns that you might use from this pre-built library. Now, if we keep scrolling down, what you’ll also notice is that right here it says this is not an officially supported Google product. Now, that doesn’t mean that it’s unsafe. This is an actual Google product, but the reason why it’s not officially supported is because right now it’s more of like an open- source beta. It’s kind of a developer playground rather than like an enterprisebacked software. And you can see right here that it also says, “This project is under active development. Expect breaking changes as we march towards V 1.0.” So, this thing is
• 58:33 — already really good out of the box and it’s only going to get better. And you can see, like I said, when Google Workspace adds an API endpoints or method, GWS picks it up automatically. So, you might as well chuck it into cloud code right now and start getting used to it. Okay, so I just uninstalled this so I can walk you guys through step by step how this actually works. It’s super easy. What I do is I basically copy the link to this GitHub repository as you can see. And I’m going to basically just give it to Cloud Code and say, “Hey, I want to install this GWS CLI, read through the documentation, and help me install everything that I need
• 59:04 — to install, and then we’re going to get set up.” So, this is basically going to do all the research for me, and then all I have to do is follow its instructions. So, it read the docs. It’s looking at what we already have installed. It basically saw that I already had some of the prerequisites. So if you don’t have those, you’ll have to install those. And then it told me that we needed to install the CLI. So it did that. And now we have two options. So the first one is to install G-Cloud CLI so that we have automatic setup and off. Or we could do it manually by creating our own project and whatnot. So let’s just go ahead and try option A. Okay. I thought this was going to be just like a simple command
• 59:35 — that it ran and then we were good. But it’s actually like some other thing to install. So let’s actually go back and try manual and I’ll just show you guys I guess the harder way. Okay. So I’m going to go to this link. go to our Google Cloud Console and make sure you’re signed in with the right account up in the top right. And I’m just going to go ahead and create a new project just to show you guys what this would look like. So, new project. I’m going to call this one Claude Code GWS. And we’re just going to go ahead and create this project. So, this is spinning up right now as you can see. And now that it has been created, I’m going to select it so we’re inside of it. And then I’m going [1:00:05] to go up here and type in APIs and services. Click on that. And we have to set up our OOTH consent screen. So, I’ll click on this. and it’s going to say get started. Click on that. We have to give our app a name. And then we have to choose an audience. So, I’m just going to do internal because I only need this right now for my own organization. If you want to do external, it’ll basically have you do testing or published. And if you do testing, just make sure that you add your email as a test user. And then all you have to do after you put in your contact information is hit I agree. And [1:00:36] then you go ahead and create that. Now, once that has been done, you’re going to go to create a client ID. So, I’m going to go back into APIs and services. I’m going to go to credentials and then I’m going to go ahead and do a create credential oath client ID. Now, in here, we’re going to choose a desktop app. I’m going to just call this GWS and go ahead and hit create. And now, we have our client ID and our client secret. And so, what you’re going to do is download this as a JSON file. Now, you can see here that it says to download that file and save it to your global.config/GWS. [1:01:06] So, basically, if you can’t find this, just say, “Hey, can you give that to me in a full path?” And then you can paste that into your finder or your file explorer and it will take you there. It will probably look something like this. And then you just drag in that credential thing. I called mine client secret. And cloud code will be able to look at this globally now. And so what you’ll notice is that we didn’t in this project yet enable these APIs. So let me just show you what happens without that. So it says the last step is to run GWS off login. So I just said, hey, I finished option B. The credentials are called client secret. And then I told it [1:01:36] to run the O login. So that should basically open up a tab for you, but if it doesn’t, then you can ask for it to give you that URL so that you can actually authenticate in. So you would basically choose your account that you want to use. And then you just have to basically confirm that it can access all of these different things as you can see. And then when you hit allow, you should be properly authenticated. After that, it’s going to come back and say, “Okay, cool. Let me see if everything works.” Now, this hasn’t been perfect on the first try every time, but if you just go back and forth a little bit, say, “Hey, that didn’t work. Hey, this is what I’m seeing.” It will be able to [1:02:06] get you there. It’s going to be your best friend for something like this because remember it can read all of the actual documentation. And now it says that the O is working, but we have to enable these APIs in our Google Cloud projects. So basically just clicking open these one at a time and all you have to do is hit enable. So it’s super simple. You just have to do this like I said for all of these different services that you actually want to be able to use. So that’s why I did this on a new project cuz I just wanted you guys to see that. But if you already have one that has all these enabled, then you can just use that project and generate that OOTH client ID. So there you go. You can see that this works. I said, “Can you [1:02:37] find my Google doc that I made in April of 2025?” And I went ahead and pulled links to all five of these because obviously that was a very vague request. And now we could take action pretty much anywhere in Google Workspace super simply with this CLI. But like I said, I just got this set up today and I’ve been playing around with it a ton in my executive assistant project and it’s been awesome. It can literally do anything. So here I’m asking it to grab my unread emails from today and based on what it knows about my business and my priorities, give them a score and if the priority score is below five, just mark it as unread automatically. All right. So, here you can see it said, “Got 30 [1:03:07] unrated emails. Here’s my priority score based on your business context.” And as I scroll down, you can see that it’s getting different ratings. And based on what I’m seeing right now, this actually looks pretty good. So, then I started playing around with Google Slides because I use Gamma right now. But at some point, I could imagine that if this gets good enough, then I wouldn’t need Gamma anymore. And this is a free option compared to Gamma subscription. So, I had it create me a slide deck and it was okay. I threw in my brand guidelines. I threw in my logo and I said, “Hey, can you see this? you created this using the Google slides and it’s okay but there’s some weird things that I need you to [1:03:37] fix. So then it came back and said I cannot see the slides I just know how to build them programmatically so that’s why there may be some errors with spacing and stuff. So then I basically just gave it access to ChromeDev tools so that it could open the page screenshot it look at it and then we built a plan to add visual validation to this Google slide creator skill. So now you can see as it’s going through it actually takes screenshots and then it can make fixes based on that. So then after it fixes everything, it says, “Okay, cool. Updated the skill. Take a look at it now.” So I’ll open up this link. Brings me to Google Slides where I have this slide deck. It has kind of my [1:04:08] brand colors. It’s got the logo up top right. And then as we go through, we can also see that the spacing is a little bit better. It’s still not perfect obviously, but we have custom images here that were generated with Nano Banana 2. And even the images are kind of on brand with the sort of orange and blue color scheme. As you can see, we’ve got this one with the WT framework. We’ve got this slide. And it even ends with a CTA for the free school community. So just to see what else happens, I’m going to say take a look at the slide deck and do another audit. How could you improve the skill in the future? So it’s going to go ahead open [1:04:38] up a tab as you guys just saw. It’s going to take images. It’s going to flick through the different slides and capture them. And as you can see over here, it now says take screenshot. And now it’s reading that screenshot right there. Now it just moved on to the next slide. And it’s going to go through and look at every single slide. And then it’s going to come back with a plan. And we could probably do a similar visual and validate flow with creating Google Docs as well. So now you can see it’s almost on to that last slide. And I hope it fixes this last slide because what you can see here is that the spacing is really off down here. So you can see it [1:05:09] came back with an audit. It came back with some future improvements. And one thing that I did notice is that because I made the window smaller, its screenshots were probably worse quality. So it said presentation mode screenshots would probably be better. But anyways, I just wanted to give you guys a little taste of how you can use the GWS CLI. but also use it with other tools to make the functionality even more powerful. So, just remember that this is very new. There’s a lot of people out there on Twitter right now saying that this is insanely overpowered. There’s also a lot of people that are saying that it just feels kind of finicky. So far, for me, it’s been pretty great. Everything that I’ve asked it to do or find or schedule, [1:05:41] whatever it is, it’s been doing that pretty much perfectly. But, there are some people saying that it’s asking them to reauthenticate multiple times. So, if that’s a little frustrating, I guess just keep in mind that it will only get better. And we’re not even to version one yet. So I would definitely recommend that you come to this GitHub, read about it, but more importantly get this thing installed in your cloud code setup and just start using it, using it, using it. Okay, so we’ve got the foundation of our AOS set up. We’ve pretty much linked in our context. We’ve pretty much linked in connections. At least we understand now that we’ve, you know, looked at all these things. We understand how we could [1:06:12] go look at the documentation. We understand how we could grab our API keys and stuff. And we understand how to hook everything up. And now we are at this place where we want to keep leveling up and keep moving up the chain of the four C’s. So next thing to think about is capabilities. Now when I say capabilities, what do I mean for the most part? I pretty much primarily mean building skills. Skills are those reusable recipes we talked about that help you do things more efficiently. And so everything that I do on some sort of cadence, you know, every time that I need to pull in my YouTube analytics, [1:06:43] every time that I need to check in with the team, every time that I need to build a slide deck, all of those things I’ve built skills for in my AIOS. If I need to create images, if I need to create a LinkedIn post, all of these things are skills that I’ve built over time because I’ve challenged myself to do everything to do as much as possible inside of my cloud code. So building skills, building custom skills for your specific SOPs, your specific workflows, your specific tasks. Remember how we talked about this thing earlier? Every single task that you have can be broken up into chunks. And all of these chunks and this overall larger thing can all be [1:07:14] turned into a skill. So, what I want you to do right now is I want you to grab a piece of paper and I want you to think about it’s a Monday morning. You wake up, you sit down at your desk, you’re ready to do work. What do you do? Write down what you do on a Monday. And then, unfortunately, pretend it’s Tuesday. You go to your desk. Write down what you do on Tuesday. And as you do this throughout the week, as you do this at the end of each day, look at what you’re doing. Circle things that are repeated. Circle things that you see pop up consistently. Circle things that you [1:07:44] hate doing. And all of those things you circled are things that you should come into Cloud Code with and say, “Hey, I have this process. You know, every Monday I have to organize all of the different tickets that we got over the weekend, and I have to tag them and classify them, and then I have to route them to the right team, and then I have to follow up with them throughout the week. I hate doing that. Can you help me turn this into a skill? Sure. Let me ask you five or six questions to make sure I understand what you want and let’s build a skill together. And that is exactly how you start adding more and more capabilities into your agent. So, in [1:08:15] order to get a better understanding of skills, I’m going to go ahead and play a video where I break down a bunch of the ones that I use, how they work, and how you think about building them and improving them over time. It is super cool. It is the biggest unlock inside of an AIOS. So, I’ll see you guys in that video. I have genuinely never been as productive as I am right now because of Claude’s skills. This image really does sum it up. I feel like I’m working on multiple different computers and multiple different tasks at the same time without sacrificing quality. It just comes down to one simple word and that is leverage. With Claude skills or [1:08:45] any agent skills for that matter, you have way more leverage than if you were doing this by yourself. So, in this video, I’m going to be breaking down what skills are, how they work, and how you can build really, really good ones, even if you’ve never heard of the concept or built a single skill ever before. So, let’s hop into a live demo real quick and get going. All right, so here we are in Cloud Code in my Herk 2 project, which is kind of just like my personal assistant. Now, if this stuff over here looks overwhelming, don’t worry about that right now. Just worry on what I’m actually asking the agent to do. So, I’ve got the skill that I called morning coffee that helps me plan my day every morning. So, it’s not morning [1:09:16] right now, but I’m going to run this so you can see how it works. So, as Cloud Code’s figuring that out, what I’m going to do is open up another agent. And in this one, I’m asking it to run a pulse check on all my projects and commitments to see how things are going. I’m opening up another one to create me an Excal diagram of the difference between local AI models and closed source models. And let’s just do one more that’s going to scrape the comments from my recent YouTube videos and give me an analysis on what I need to improve. So what you’re looking at right now is four different agents running in parallel doing things for me. And that took me probably 30 seconds to ask them to do that. And because I built all of these [1:09:46] skills, all of these agents have all of the context about my business, what’s going on with our projects, my YouTube channel. It has literally everything it needs. And now all of those agents are done. Here was my February 26th morning coffee. I had three things on the calendar. So, what it’s going to do is it’s going to look at my ClickUp. It’s going to see what else I’ve got this week and look at my tasks and then help me plan the rest of my day. So, this is the plan that it suggested. All I’d have to do is say, “Yep,” and it would block off everything for me. And for me, that’s huge because I don’t have decision fatigue anymore of what I have to work on. The second agent came back with a pulse check. 2 days until the end of the month. Here’s where everything [1:10:16] stands. So, obviously I’m going to blur all of this out, but it’s basically catching me up on all of the different main initiatives that we’re doing this month and this quarter and making sure that everything is on track. And right here, you can see there’s a few things that I need to follow up on manually. And this might have slipped through the cracks because I’m so busy making YouTube videos if I didn’t have this personal assistant to check up on me using the skill. The third agent came back and has finished the Excal diagram and I pasted it in and it looks like this. So, if I needed to make a video about this, I wouldn’t have had to take my own time to create this visualization. So, for the comment analysis, that came back and we can see [1:10:46] all of the comments, all the views, and things that I need to address either in future videos or in the comments. We got some confusion. We’ve got some cost things that I need to cover. We have to stop demoing toy examples for tool videos. Seems like you guys really want to see some anti-gravity stuff. I promise that’ll come soon. And then we’ve got top three priorities. So, I’ve been recording now this video for about 6 minutes. Just think about if I would have done all four of those things myself, how much context switching I would have done and how long that would have taken me. Okay, so now that you guys have seen an actual demo and hopefully you’re a little bit excited to learn about cloud skills if you haven’t [1:11:16] used them yet, what actually are they? So skills are reusable instructions. You write them once, you save them as a skill. You can trigger them anytime and you’re going to get way more consistent results because it’s going through that same process every time. So this visual right here was actually an AI generated image that I used with a skill. It was this one right here in my cloud code called Excalaw visuals. But sometimes with AI generated images, they don’t spell words right. As you can see here, this is all messed up. This stuff is messed up. So, I also have one, as you saw in the demo, to create Excal diagrams. And this one creates the [1:11:48] actual Excal that I could move and edit. And the words are always perfect because it’s actually just typing. And those two skills alone have saved me so much time. And what I’m going to do, by the way, for all you guys, is in my free school community, the link for this is down in the description, I’ve added a new classroom section called agent skills where I’m just going to be dropping a ton of these skills that you guys can go grab for completely free. So, before we dive into this today, just real quick, why should you care? And there’s three big reasons. You can be way more productive as a person because you can automate things like you just saw me do. And you can legitimately build a personal assistant that can do almost anything for you. The second one is team [1:12:19] leverage. So you can turn existing SOPs into automations really, really easy. And if you build something new, not just you can use it, but your entire organization can. So everyone as a group is getting way more productive, which will almost undoubtedly result in growth of the business. And also monetization. We’re entering this new world where skills are having a big moment and you’re able to capitalize on a lot of this stuff. Now, I’m not saying this is going to be a viable business model for a long, long time. So, you shouldn’t bank on it, but it is something to be aware of just like when people were selling end workflow templates and things like that. But once again, it [1:12:50] just comes down to one word, which is leverage. This isn’t just theory. This is something that we’re seeing with clients. This is something that we’re seeing internally in my own business. This speed of work that we’re able to achieve right now feels insane, but that is going to become normal. And if you can’t do that, you instantly become way too slow and way too expensive for the business and they might not keep you around. We’re actually making it a priority to make sure all of our employees are using cloud code. Because now I have all these different skills that I can just run with a simple slash command or a simple natural language prompt and get in one day a week’s worth of output. Because once again, one [1:13:20] person can figure out the best way to do something and turn it into a skill that the entire team can use. But they don’t just generate text. They’re basically automations. They can run scripts, they can call APIs, they can create things, they can have sub agents, and they can be called on from agents as well. So, this is truly AI automation. And just to really hammer it home, they’re basically SOPs for your AI agents. The same way where you would train a human employee by letting them read through an SOP to understand the process, and then they’d be able to do it. You just train an agent on it. You give them the skill, they read it, and then they do it. And the coolest part about it is the more [1:13:50] you use the skill, the better and better it gets. So, we’ve talked about a lot of these benefits, right? But what actually is a skill? Well, it’s just a folder and it lives somewhere in your project. The most common example is probably going to be in your.claude/skills/skll name and then you’ve got like a skillmd or a markdown file. So right here in my herk 2 project, you can see up top we’ve gotclaude. I open up.claude, we have agents, rules, and skills. Right now, we’re just talking about skills. If I open that up, we can see all of the different skills that I’ve created in. So let’s say for example, the excalagra [1:14:21] diagram skill. If I click into that, we have a skill.md. And when I open that up, you can see we have the name of the skill, the description, and then we have the actual workflow. Step one, understand the concept. Step two, plan the layout. Step three, generate elements. And this entire skill basically teaches my agent how to build these Excal diagrams for me. So, like I said, that is the anatomy of the skill. We’ve got front matter, which is kind of between these two dotted lines, and that is in something called YAML, which you don’t need to worry about what that means. This is just the way that it’s [1:14:51] kind of indicated, sort of like your markdown or your JSON or Python. Now, up here, we’ll have a name and a description, which tells Claude Code what the skill is called and what the skill does. So, as you can see, this one is called Excalraw- Diagram. And there’s a brief description about what it actually does or when to use it. And then we have the step-by-step rules, which are basically the instructions. And this is what Claude actually does once it decides that this is the right skill for the job. Now the interesting thing about skills is that sometimes you need way more data. So for example, let’s say we’re writing a LinkedIn post. [1:15:22] We have a skill for that, right? But what needs to go in the skill is other information sometimes like a company tone of voice or maybe your LinkedIn, you know, tone of voice, a target avatar, current priorities, a logo, maybe there are other things that you want to put into a skill besides just like the step-by-step instructions that will make it better. So the question is, where do these things go? Well, there’s typically two options, but essentially, as long as you’re pointing to the right path in the skill.md, you’re fine. So, let me explain what I mean by that, and then I’ll show you what I mean by that. So, first option is to have it [1:15:53] self-contained. So, in yourcloud/skills/skillame, you can have the skill.md, you can have your scripts right there, and you can have your references right there. Or option B is that they’re not directly nested right under that skill. So here we have do.claude/skill/infographic and the skillmd and we still have our scripts and our references in the same project but it’s just not nested directly under that skill. And so I know that might have made no sense. So let me show you exactly what I mean by that. Okay. So here we have a skill called idea mining. And so what happens in here [1:16:24] is basically used when someone asks for content ideas, video ideas, what to make next or to run idea mining. And so in here I gave it some context, right? My channel has this many subs. It’s about AI automation. My content pillars are naden, rag agents, cloud code, voice AI. And then what I gave it is a bunch of references. I gave it channel data which is YouTube channel.md. I also gave it the raw data which is a JSON file. I gave it a competitor list for me. And I gave it an actual script to run analysis on my YouTube channel. And so in this case, what you can see is that I went [1:16:55] for option B where I’m storing those reference files and those scripts, not directly nested in this skill. So basically what it could look like is within the skill itself, we could have a folder called, you know, references. We could also have a folder in here called scripts. And then within both of these subfolders, we could have more things like, you know, in the references, I could have channel data. And in the scripts, I could have YouTube-analysis.js or py, whatever it is. Basically, the idea is it doesn’t matter where those [1:17:26] actual reference files or scripts live as long as you point to the right spot in the MD file. So in my case, where these actually live is in a different folder. So here for the channel data, I would basically just go all the way down to references and then I could go down to right here, YouTube channel.md. So cloud code reads the skill and then it’s able to find this if it needs it. Same thing for the scripts. It would go down here to scripts and then it would find analyze YouTube. py and it would just pull this in if it needed it. So hopefully you guys are with me. However you want to set it up works. And I think [1:17:56] that’s the most overwhelming thing about cloud code right now is that everyone uses different kind of folder architecture. But don’t worry guys, I’m totally on top of this. I have a skill that I built out called skilluer. And this one I’ll be giving away for free once again in my free school community right here. And all you’d have to do is load in the skill builder and then it’ll help you build out everything you need. And I’ll be showing a live demo of that in a few minutes here. So the skill.md is the actual brain itself and the supporting files are the tools that it can use. That doesn’t mean every single time the skill is invoked that those [1:18:26] reference files will all be called. And just in case you guys were wondering if you’ve watched some of my previous Cloud Code videos where we’ve used the WAT framework to build automations, this is very very very similar. In that framework, the W the workflows were the markdown file SOPs. That’s basically the skill. The tools were the actual Python scripts and that’s basically just the scripts that you might write or the references that you would add in. So if you’ve already been building some WAT stuff, you will pick up skills super super quickly. The cool thing about skills is that you don’t have to build all of them. Obviously, as you’re working with cloud code and you’re finding that you’re doing things [1:18:56] repetitively, you can go ahead and build a skill for it. But there’s an official library from Anthropic of Skills. There’s a community of everyone that’s open sourcing their skills and giving them out. And there’s a marketplace where you can share and sell or you know download skills from people. And then you would take that skill or that essentially a prompt and you would add your own flavor to it. The one thing I would say is just be careful and make sure that no one’s trying to, you know, give you a skill that has any malicious intent in there. And all these skills can work across different products. So cursor, anti-gravity, codeex, because [1:19:27] it’s so based in markdown, and it’s essentially just a prompt, tons of different AI models can use them. Okay, so how does Claude know when to use a skill? Well, there are two ways to actually trigger them. The first one is you can be explicit, which basically means you can do a slash command and say the skill name, and it will just directly fire off that skill. Or it could just be natural language. So if I had a school post skill, I could say slashschool-post. Or I could natural language just say, “Hey, help me write a school post about X.” Cloud would find that skill and then invoke it. So when you ask Cloud to do something, it will first read through [1:19:58] the cloud. MMD file. It will analyze your request and it will search through the skills and see which one do I have that helps with this query. If it finds one, it will invoke it. But if it can’t find anything, then it will basically just use its general knowledge. So not every single request that you give to cloud code will invoke a skill. Now a really important part of that is understanding how skills stay lightweight. Because if you’ve been using cloud code, you know that context management is a huge deal. And if you had all of these skills to look through and all of these skills are, I don’t know, hundreds and hundreds of lines, then if cloud code was searching through [1:20:28] all of these every single time, that would surely eat up a ton of your tokens. So what’s used is something called progressive context loading, which basically means we have three levels. Level one is the initial search where cloud code only looks for the name and the description. So right here you can see let’s say we ask for an Excal diagram. It would basically search through all the skills but it would only read the YAML front matter. So it would read the name and the description. And typically this front matter is only going to be you know maybe roughly 100 tokens. So it stays very lightweight. And then moving down to level two let’s [1:20:58] say it identifies okay cool this is the right skill for the job. Then it would run the full skill.md and it would read through everything. And so that’s when it would start to actually understand what goes on in the skill. And that might be anywhere from a thousand to a couple thousand tokens. And then level three is once again a decision. Only load in the extra files when needed. So if I need to look at any scripts or references or templates or I need to pull in some brand assets or more context, I’m only going to do that if the specific request requires it. And so [1:21:28] hopefully now you’re starting to understand a little bit more about under the hood what’s actually going on when you ask Claude Code to do something for you. And you can always go to Cloud Code docs and go to the skills section and just read about how this stuff works. It’s really, really simple. On the doc itself, it will tell you just make sure to keep the skill.md under 500 lines. Move detailed reference material to separate files. And so, I know this may seem like it’s just a lot of information being thrown at you. So, let me just kind of contextualize this and slow it down and reassure you guys. You’re never ever ever going to write a perfect skill [1:21:59] the first try. The way that I build my skills is I have Claude Code do something with me. I walk it through the steps, you know, each time. And then when we’re done, if we’ve went from point A to point B, I say, “Cool. This is something I do once a day. Let’s turn this into a skill. Ask me more questions so we can make sure you have all the information you need.” And once again, I’m going to show you guys opening up a brand new project and setting up a skill from scratch so you understand the full process, but I just had to give you guys some context first. Now, we have this thing called the feedback cycle, which basically means you invoke the skill, you actually watch the agent work, you [1:22:30] give feedback, and then it fixes the skill, and then you do it again. And so, the first couple times you run a skill, you may feel like, eh, this feels very AI generated. But by the time you’ve run that skill 10, 20, 30 times, every single time it gets better. And so, that’s why it’s actually important to watch the agent work the first couple of times because that’s how you’re able to identify opportunities to speed it up and save tokens by doing things like this. So here’s an example of the pulse check skill that we actually ran earlier. Now this skill gets invoked when I ask for a pulse check or checking in on commitments. And what it does is it reads through some context of how [1:23:01] OTAAS work, which is important for it to understand every single time it reads the skill, which is why I put it here rather than a reference file. And what it has to do is it has to do a live lookup on my ClickUp to understand what’s going on. So what I did is I hardcoded in these list ids because when I was watching it, I realized every single time it was doing this, it was calling the ClickUp MCP and it was gathering all these lists and it was searching and parsing the results and then it would extract the ID and that just was taking so long and it was costing me a ton of tokens. So I realized that’s always going to be the same. Why don’t I just give it in the [1:23:32] skill document the list IDs and now it knows how to do that instantly every time and it doesn’t waste all those tokens. And on top of that, I know that searching through ClickUp can consume a lot of time and tokens. So, I built a specialized sub agent that in this skill, I say, “Hey, delegate to the ClickUp searcher agent with this query in order to do all of this searching so that you don’t blow your own context window.” All of that’s handled over there and then you only get the information that you need. So, there’s a lot of advanced things that you can do to manage your context. I’m not going to dive into all of that right now. We’re just focusing on skills, but just wanted [1:24:02] to give you a little taste of what’s possible in the skill.md files. So, another good example of needing a reference doc like that is in my skill builder skill. I obviously use this when I’m creating new skills, optimizing skills, auditing skill quality, things like that. And a lot of the inspiration I got from this was of course straight from Claude Code Docs itself about how to actually use and build and optimize skills. And so, when I was building this out, I I was watching the agent, you know, run the skill and I realized it’s searching every single time. It’s doing a web search and it’s crawling the [1:24:32] entire document, even if I just need a little piece of information. So, what I decided to do was I told it to basically scrape that whole thing. And then I gave it a reference.md, which is basically the documentation. So, I’ve got my skill.md. And what it does is it references that full file if it needs it. But really, the main idea that I’m trying to drive home here is that processing markdown files for your agent is so much quicker and cheaper than actually making API calls or HTTP requests, you know, executing functions and reading tons and tons of tokens. So, the goal is your skills will get to a [1:25:03] place where you can invoke them, focus on something else for 10, 15 minutes or whatever, and then come back and have a finished result that is really, really good. But the first couple times that you are testing out a skill, I think it’s a really good idea to just sit there and watch it and see what it’s doing. And a lot of people have asked me like when do you know when to build a skill? Well, basically just go about your work and if you ever realize that you’ve done something already or you’ve instructed something differently like I tell my claude to not use m dashes. Okay. Well, that’s probably a good idea to put that in the prompt, right? So, if you ever find yourself doing a process [1:25:34] or repeating prompts, then that’s probably a good use case to build a skill around it because skills don’t have to be complex. They could literally just be a 50line markdown file. All right, so we’re about to hop into a live build of a skill from scratch. But what I wanted to do real quick was go over the six-step skill building framework. So number one is the name and the trigger. What is it called and the natural language that would basically fire it off? Number two is the goal. So in one sentence, what will this skill accomplish by the end? What will be the output? Number three is the actual meat of it. That’s the step-by-step process. [1:26:06] If you had to do something manually, exactly what do you do in what order? What do you look at? And what decisions do you make? Number four is the reference files. What context do you need? Do you need images? Do you need understanding of current projects, current priorities? Do you need style guides? What do you need to do the job well? Number five is the rules. Think about what could go wrong and then the agent can help you building guard rails and constraints around that. And then number six is kind of like after you’ve built it, it’s just the self-improvement loop. And after the live build, I’m going to talk about actually testing and iterating and what you need to do to [1:26:36] make them really, really good. But for now, that’s the six-step skill building framework. Let’s hop into a live build. Okay, so here we are in Visual Studio Code, which is where I like to use Cloud Code. If you don’t have Visual Studio Code, just go ahead to a browser, type in VS Code, and then go ahead and download this. This is what it will look like. If it’s your first time using Cloud Code in here, you just have to go to extensions on this lefth hand side, type in Cloud Code, and then install this, and then log in with your paid Anthropic subscription. Now, after that, you’re going to click on this top left button, and it’s going to pull up this little thing that says you have not yet opened a folder. What you need to do is [1:27:07] open up a project to work in. So you could either open up one that you’re already working on or you could go ahead and create a new folder and then open that one up. For the sake of the demo, I just opened up a new blank folder called a bunch of skills. And I’m going to show you exactly what to do. So the first step is to go to my free school community link in the description. Go to the agent skills classroom and download the skill builder folder. Once you’ve got those files ready to go, first thing we want to do is just set up this workspace real quick. Initialize this project with a simplecloud/skills structure. Cool. So, as you can see, [1:27:38] that got set up. We have aclaude. We have a skills folder. And what I’m going to do is in this skill folder, I’m going to create a new folder called skill-builder and hit enter. And then I’m going to take those two files for my school community, the reference and the markdown, and I’m going to put that right in here. So now we have this skill builder set up with the reference file and the actual skill markdown. I’m asking it if it can see that new skill that I just added. It says, “Yes, I can see it.” And I’m basically just going to say, “Cool. Let’s run that skill to build a new one together. So now you can see what it did is it basically is [1:28:09] reading the skill right now. This is the instructions that we saw right in here. As you guys know, since that’s how skills work, it starts to read this. So here we go. I built this skill to actually ask you questions so that it’s way easier for you to communicate what you want. So the first thing is what problem are you trying to solve? What we want to do is content creation because in this skill, what I want to do is building branded infographics. What kind of content does the skill create? What’s the specific use case or workflow? And I’m actually just going to choose other for this. And I’m going to say educational infographics. Now it’s asking how we should trigger this skill. [1:28:39] So does it want to be natural language or do we want to just use slash commands? And I’m just going to say both is fine. And now we’re moving on to the step-by-step process, which is really important because at this point we haven’t told it what text stack we actually want to use or anything else about our business. So walk me through what should happen from trigger to output. And it has some good guesses, but what I’m going to do is do other and explain this the way that I want it built. I will tell you what I want an infographic about. You will create a concept. You will make a request to key.ai to use nanobanana to generate the [1:29:10] outline or sorry to generate the image. And you will also look at the brand guidelines that I give you so that everything that is created follows my brand colors and typography and stuff like that. The output format that I actually want is a PNG, not any of this stuff. Does this need to be conversational or fire and forget? I’m just going to go fire and forget. All right. All right. So, how does the key AI nanobanana integration work? Is it an API call? Yep. We’re just going to go with an API call. And in these options, you could literally say, I don’t know. Let’s try different things. You know, help me figure out what’s best. It’s [1:29:40] asking where those brand assets live, so I’ll put them in a folder. And where should the generated PNG infographic be saved? Yeah, sure. Let’s start a new folder called projects, and we’ll throw all of them there. So, it’s going to keep our project organized as well. So, now it’s asking about brand guidelines. I created this folder, and I put in our kind of color scheme as well as the actual AS logo. I have put in both our AIS brand guidelines and the AIS logo. I want to make sure that in the top left corner of every single infographic that’s created, the AIS logo appears exactly as I’ve given you. But I think [1:30:10] you guys get the point here. I’m going to answer a few more questions and then I’ll just show you when we have a result. And now that we’ve done that, what you can see is it is going to create the skill. It’s going to create the logo overlay. It’s going to create a supporting reference markdown file for all of the API details that it’s going to need. So that’s great. It’s going to register the skill in claw.md and it’s going to log its decisions. All right, so it fully built the skill. It created all those files for us. We just have to give it a key API key so it can actually run this. Okay, so I threw in my API key and then I said test it out with an [1:30:41] infographic about cloud skills. That’s it. No other context. It invoked the skill right here and we will see what happens. Okay, this is really interesting. So what it’s doing is it is generating the image and then it’s just going to overlay the logo. So, it’s going to be a lot more consistent than giving the AI image generator Nano Banana my logo. So, I didn’t even tell it to do that. Let’s see how it looks. Okay. Well, I don’t love this. We’re just going to go back and ask it to change some things. The logo on the top left doesn’t look great. I gave you a logo with a transparent background. So, it should just be overlaid on top and we [1:31:12] should be able to see the background behind it. The actual infographic itself is all right, but I actually want these to always be one by one aspect ratio. Okay, so I made some suggestions and it’s going to try again and it’s going to update its skill. So we’ll see if that’s better. All right, so second time we run the skill. Let’s see if it’s any better. All right, there we go. We’ve got the logo up top. We’ve got cloud code skills, custom AI workflow, command prompt, trigger, front matter, config triggers, AI agent delegation, document output. So just keep in mind all we said was build an infographic about cloud [1:31:42] skills. And this was run number two. Every single time that we do this cycle, remember we talked about the feedback. We would basically watch it again, give more feedback, and then keep going. And after we run this probably five or six more times, this would be really, really good. And then every time I ask for an infographic, it’s going to be consistent. And just to show you guys what was actually built, if we open up the infographic builder skill, we have the actual skill itself. So we have the front matter right here, the name, the description, we’ve got what the skill does, we’ve got context, so here’s where it links to the actual brand guidelines [1:32:12] and logos. We’ve got the step-by-step workflow right here. And we can see right here for full API reference and parameters just see the markdown file so that you don’t have to actually go search the web and search through a bunch of tokens. You can just read this markdown file. All right. So we’ve talked about a lot of stuff about skills today and we just built one live. So what I want to talk about now is really how do you bridge the gap from like a 90% good skill to making it pretty much 100%. So testing, iterating, and debugging. There’s different symptoms and there’s different fixes. So let’s just kind of go down this list one by one. The first symptom might be it does [1:32:43] the wrong steps or in the wrong order. Well, you would just tell it to edit the skill.md instructions. You could get missing tone, style, or context. In that case, you’re going to add reference files. And of course, those have to be pointed to correctly in the skill.md. You could get the same mistake happening over and over, then you’re going to add a rule. If it struggles with a tool or an MCP or it keeps searching for the same things, then create some sort of reference dock for it. If it works good, but it could get better, then that just means you have to brute force it. you have to just run it over and over and over and keep nitpicking at what it does [1:33:13] wrong or maybe not wrong but what it could improve on. If the skill isn’t triggering, then check the YAML and make sure it is specific enough. If the skill triggers too often, then maybe try disabling model invocation and that is something that you can see in the claw docs, which basically gives you control over if the skill can only be invoked by natural language or only be invoked by the slash command directly or both. So, like I said, if you want to look at some more advanced stuff, then definitely head over here to the actual doc. But at this point, we’ve covered almost everything about these skills. One thing that I would call your attention to is the actual front matter reference because we saw the name and the [1:33:44] description, which is what’s required every time. But there’s lots of other things that you can add in there. Here is the disable model invocation like we just saw. But you can also give it allowed tools. You can also give it an argument hint. You can give it a specific model to use. You can give it specific context. You can give it hooks. You can give it a specific agent. And so all of this lets you get really, really granular on the exact skill and how you want it to be used. But don’t get overwhelmed. You really only get to that point once you’ve ran the skill a ton of times. Now, another thing that I need to hit on real quick is where do skills actually live because what we’ve seen so [1:34:15] far is just building them right in ourcloud/skills folder. But when you’re doing this, they only exist in that specific project. So whether that’s my her two or my, you know, the one we just spun up, if I went to a different folder, that skill would no longer be able to be accessed by our cloud code. But you can also create skills that are actually global. And you do that by doing that in a different directory in your kind of overall home directory. And that’s basically indicated by the little tilda right here. And so that means every project you use in cloud code, no matter where you are, that skill would exist. So for example, I have a [1:34:46] front-end design skill that is installed globally. So that whenever I’m anywhere, if I need to do front-end design, it just is able to use it. And just in case you want to look at it in a different way, right now what we’re doing is we have our project, right? So, herk 2 and then we have dotcloud and then within dotcloud we have skills and then your skill and then your MD, your references, whatever and then maybe another skill. But if it was global, you might not actually see it in your project. It would just be within your overall home directory. So, the reason why you might want to do this is if there’s something [1:35:16] very specific about you, your business, your workflows that you want applied to every single project, no matter what, maybe your company context, your company projects, your tone of voice, whatever, then you can install that globally instead. All right, great. Right. So, now that you guys understand skills a little bit better and you’re excited to start building some, the last piece of the four C’s is cadence. And what does cadence mean? Cadence means a couple things in my mind. It means that because you can now turn off your laptop and things will still run, you have to figure out the different triggers. And you have to figure out what are things [1:35:47] that I should actually set up on some sort of cadence. Whether that means you want to have a skill running in the background or you want to have an actual deterministic automation running in the background or it means you want to have some sort of dashboard with live data across all of your different connections and you want to have one clean place to look at all of that. So let me address a lot of those things that I just mentioned and how you can start to work in a cadence around that stuff. So one of the very first things that you can do is you have this AIOS project, right? You should probably be putting this onto a GitHub repo. You can keep it [1:36:18] completely private. That’s 100% fine. But you should probably put that on GitHub. Now, if you’re asking yourself, what is GitHub? Let me just explain that to you real quick. So, basically, everything that you build in Cloud Code is a bunch of folders and files, right? So, in this example, we’ve got the AIOS with our, you know, our readme, our claude, we’ve got our references, we’ve got ourcloud with our skills. This is basically what your project is at the end of the day. And this is all being stored locally. Meaning if you have one laptop that you’re working on this with and then you open up your second laptop, [1:36:48] they will not sync. You will only have that on one laptop. So GitHub is basically a place for you to upload your folders and files and basically your codebase. So if you put this into a GitHub repo, short for repository, then you could basically just pull that repo in. So if you guys remember at the beginning of this video when you cloned my repo, that’s because I uploaded it there so anyone could go clone it. So you can upload one privately, meaning no one else can touch it, but you can because you have login access, right? So [1:37:19] now if you open up your other laptop, you can pull in that GitHub repo and just keep working. So you have your AIOS no matter wherever you are. But the second thing you can do on top of that is you can actually use that GitHub repo with Cloud Code on the cloud or an OpenClaw agent or a Hermes agent or whatever of these other agent frameworks that you want to use, a codeex even. you could plug that into any different AI harness. So, if you wanted to be able to use your AIOS whenever you’re, you know, out of town or if you are on the road [1:37:50] and you don’t have your laptop with you, you could connect it to like ClickUp, you could connect it to Telegram because you could put, you know, cloud code on a VPS or you could have it running on a Mac Mini or something. There’s different ways that you can have access to your AIOS 247. You know, you could use, like I said, a Hermes agent or an OpenClaw agent as well. But that is one thing that you can do so you always have your AIOS with you no matter where you are. Now, what else can you do? You can actually start to schedule these tasks. You can start to schedule all of your capabilities. So, all of your skills, [1:38:20] you can have them run every hour. You can have them run every Monday at 6 a.m. You can even have them triggered on some sort of, you know, web hook. So, those are the things that you can be looking at next. Now you guys remember earlier I mentioned that you could also use the desktop app of claude because in claude you have code, you have co-work and you also have claude chat and you can use claude code right here in the desktop app and you can run your AIOS just fine. I just prefer VS Code. But what’s cool in here is that you have routines and when you have routines they can either [1:38:50] be local or they can be remote. The difference here is that a local routine has to run with the desktop app open. But a cloud routine can run with your desktop app closed and with your computer off. As you can see, local routines only run while your computer’s awake. But on my plan, which is 200 bucks a month max plan, I can only have 15 remote cloud routines running per day. And you can even open up the calendar and see, okay, today’s Thursday. Here are all of the routines that I have running on a Thursday. On Friday, here are all the routines that I [1:39:20] have running on a Friday. And this is where you really start to set up some nice cadence of things that actually happen while you sleep, which is just absolutely awesome. And what’s cool about these is they actually run skills. So for example, this is my school wins engagement scheduled task or routine. What happens is the first instruction here is run the wins engagement skill. So a routine basically just injects this into a real claude chat. It doesn’t run some sort of, you know, Python script on the cloud. it runs a real clawed session [1:39:50] the same way that if I copied this exact prompt right here and I dropped this into my AIOS it would do the same thing. So let me just real quick show you today at 10 a.m. when this ran it did it like this. It dropped in this exact prompt. It ran the skill called wins engagement and then you can see down here cloud code basically ran the skill found all the results and then it executed everything for me in my school community. So that is kind of the highle overall really cool aspect of routines. I’m going to go ahead and play a clip right here just to dive into the cloud [1:40:20] routines a little bit more because I think that those are very important to understand. And there’s a few gotchas where if you don’t watch this video, you might end up getting stuck. So, I’m going to play this video real quick and then you guys will be right back here with me. Cloud Code has finally brought us routines, which basically means you can inject a prompt into Cloud Code, but it can be running on the web, so your laptop does not have to stay open. And I’m so excited about it. I’ve already been playing around with it. I’ve been migrating my automations over there, but there are a lot of little gotchas. So, I’m here to explain exactly how you can actually set up these automations so that they work. So, today, April 14th, Claude tweeted, “Now, in research [1:40:51] preview, routines in Cloud Code, you configure a routine once, which is basically like a prompt, and it can run on a schedule from an API call or in response to an event, and it runs on Anthropics web infrastructure. So, that’s awesome. So, you can call a routine from an API, you can have GitHub events trigger it, or they can be scheduled, which are like the scheduled automations that we already have, but now they run on the web. So, you really can create these from anywhere. You can do it right here as a scheduled trigger to run scheduled remote agents which is in the terminal. You could also go to clawai/code. So you could do it on the web. And right here you see I have three [1:41:22] web- based routines right here. Or what I’m going to be showing you guys today is just doing it in the desktop app. Because right here, if I go to my scheduled tasks, you can see that I’ve got some like these four that are local. And then I’ve got these four that are running inside of a GitHub repository. So these are the remote ones. If I go up here and click on new task, this is where we could set up a new local task or a new remote task. It’s very similar. You set up the name, you set up what Claude should do, and this is the actual prompt. So, I’ll talk more about that in a sec. But then you would configure your model, your repository, and your cloud [1:41:52] environment. You set the cadence, hourly, daily, weekdays. I think the minimum is once an hour, like you couldn’t go like every 10 minutes or something, but still not bad at all. This is where you could configure all of your connectors. So, if you need to connect Slack or Gmail or, you know, whatever it is, you can connect them right here. But you can also just do your regular API endpoints with your API keys. And then, of course, you’ve got your permissions. So, you can choose how Claude should be acting. Now, the one thing about these are these are meant to be a oneshot prompt. You’re not around. So, you probably want to make sure that [1:42:22] it doesn’t ever have to stop and ask you questions. Otherwise, what’s the point of the automation? So, like I said, there’s tons of things to dive into here, and I’m not going to try to bore you guys, but some of this is really important because when I first got this set up, my automations weren’t just migrating over and working. So, I’m going to tell you guys the issues that I ran into, and hopefully answer everything that you need to know so that you won’t have to go into the comments and ask these common questions. I can just answer them right here for you. So, let me just first of all, real quick, show you guys what I tested out. The first thing I wanted to test out is if I came in here and I created a new routine [1:42:53] for just shooting a message to my ClickUp. Obviously, that’s not any value, but I just wanted to see how it worked because what I wanted to do is see if I could do this without adding my connector of ClickUp. And I was able to actually get this to fire off, but it didn’t work right away. So, let me show you guys what I ran into. So, the way that this works is you need a GitHub repository to sync it to in order for this to actually run. So, it’s going to clone my Herku project right here in the web. It’s going to be able to read my cloud.mmd. It’s going to be able to read my scripts and my skills. And then after it finishes the job, it basically just destroys that little cloud GitHub clone. [1:43:25] But as you guys know, you don’t push your secrets into GitHub because if you see here my my Herk 2 project, this is myv file with all of my API keys. And this is listed in the git ignore, which basically says, hey, when you push to GitHub, you don’t include these files. So what that means is in here, if this is only looking at your GitHub repo, there’s nov. So how do you get your API keys into this routine that runs on the web? Well, what you do is inside of this scheduled task, you have a cloud environment. So if I click on this one, [1:43:55] you can see this one is called Nate Hertz cloud. So if I open up the settings, what do you see? You have the name of this cloud environment. You have the network access and you have environment variables. So right here is where I put in my YouTube API key, my ClickUp API key, any of the other API keys that I need to give this cloud environment access to. And then the other thing you have to do is you have to look at the access levels because right here you can see that this one is on full but by default this will be on trusted I believe and that means you can only download packages from verified sources from Anthropic. And when we talk [1:44:26] about this later I’ll have a link which you can go see all of them. You could even do custom if you wanted to allow specific domains that aren’t on that list. But in order for ClickUp to work in this case I had to go on full because when I went on trusted it said hey we can’t actually do that. But when I change this to full, it let me send a message to my ClickUp. And that is how I got this message right here that says, “Just testing that the remote tasks work and the credentials work.” So basically, when these run, whatever you have here as your instructions is what gets prompted. And that’s exactly the same way that the scheduled tasks locally work. So right here, you can see I say send a message in the internal ClickUp [1:44:56] channel. And right here, the actual thing that it says was send a message in the internal ClickUp channel. So, think of a scheduled task or a routine as you basically typing in a prompt and then someone coming in to your laptop and typing it in for you. So, it’s the exact same type of interaction as you talking to Cloud Code. But that’s why once again, you want to make sure it’s specific enough so that it can basically oneshot it. Okay. So, let’s take it a little deeper. Now, what I tried to do is I did another one which I wanted it to be able to use the YouTube data API [1:45:26] in order to grab some YouTube comments for me and give me a little analysis in, you know, ClickUp or whatever. So, this is the prompt I said, right? Analyze 50 of my most recent comments from YouTube and give me a quick bullet rundown. My YouTube API key is available as an environment variable. Use it directly from the environment. Don’t look for av because what happens is in your repo, right? So in this her 2 project um when I normally run this it grabs all my API keys from the env.mmd and realizes that’s where a lot of those [1:45:56] live. So by default it’s maybe going to try to look in thev and it’s not going to be smart enough to figure out and so for clickup it was fine. It figured it out but for some reason with this YouTube one it didn’t. So I had to explicitly tell it hey look in the environment variable rather than in thev. So you can see this first time I ran it 1241 I didn’t say that and it couldn’t do it. It said like, “Hey, I can’t find that. I’m getting an error.” And I even tried to tell it here, and it still didn’t work. But then on this most recent run, when I updated the prompt a little bit, it was able to fetch it right away using the API key. And now I [1:46:28] have a remote, you know, routine that would work. Obviously, I need to update this. I’m going to migrate over my other automations, but this was just for testing purposes. And then another one that I do is I have some automations here which basically opens up a browser using playright CLI and it does some stuff in my school community because there’s no publicly accessible API. We’ve kind of figured out a way to automate it without using browser. I’m not really going to dive into that right now. But what I wanted to tell you guys about that is I tried to basically move over this school wins engagement post or [1:46:58] sorry automation into a remote session. So, I copied the exact same prompt that was in my regular scheduled task and then I just added this little snippet at the end. But what happened is this wasn’t working because it basically said, hey, you know, like when you do this, it spins up a browser, but there’s no cookies because all of this is running remotely and all I have to look at is the GitHub repo. I can’t look at the local, you know, cookies that we’ve used in the last couple sessions of this automation. And so, it doesn’t seem like this would work because once again, it has no access to that stuff. So if I wanted to do an automation like this, I [1:47:29] would have to use um an endpoint that takes authentication in the form of like actual cookies or a header or you know like an API key because every single one of these runs is going to be stateless and after the run the GitHub clone just gets deleted. Now the exception of that is if the automation is changing something in your codebase or doing a review. If it does do that it will create a new branch for you or it will give you some sort of output and not just delete everything that it just did. But for an automation like this it would just delete it. But hopefully after you guys have seen those examples, you now [1:47:59] have the ability to come in and you know make some changes if you need in order to make sure that your automations are running. And what I mean by that is you understand this should be a very specific prompt. This is how you change the model. You have to have a GitHub repo. You can change the settings for your cloud environments right here. You set the schedule. You add any connectors you might need, which would honestly be a little easier if you added just like a Slack connector. And then you can set your permissions here. Now, the other thing to be aware of is you do have limits. So, if I come over here to my settings, you can see if I go to my [1:48:29] usage, we have our regular session limits, our model limits, but for additional features, we have daily included routine runs, and I haven’t run any yet on the actual schedule. I’ve just been testing them. Um, but we are at zero for 15. So, I could only have 15 automations running with routines per day because I’m on the max $200 a month plan. Your limits would be less if you’re on Pro. I think maybe three or maybe five. I’ll I have that information later on, but just something to keep in mind. All right, so let’s just dive into a little bit more of the details here that may answer some questions you guys [1:49:00] have. I think it’s pretty clear at this point what it is. Um, I’m going to give you guys this entire doc as well as anything else I’ve talked about in my free school community. The link for that is down in the description. So, some of the stuff I may not cover. If you want to read more about it, then just go ahead and grab that free resource. So, we know what it is. I think we know how it works, right? Like you define a routine, which is a prompt. You connect a GitHub repo. You could also trigger it by APIs or by a GitHub action and then you can connect your connectors and basically it acts as you talking to your own cloud code. Because of the fact that [1:49:31] this is working off of a cloned repo, it’s going to read the cloud.MD file automatically every time. So if you have a massive project like a Herk 2 project for example with tons of context and tons of stuff maybe you don’t want to put that repo into the cloud to be a routine run because there’s a lot of context in that cloudMD and in that whole GitHub repo that might not matter for this automation. So maybe you’re better off setting up a specific GitHub repo per scheduled routine. But of course, cloud.m best practices putting [1:50:02] in the information that’s important because this stuff is going to drain your cloud code session limits the exact same way as it would if you were open up in cloud code just talking to it. So once again, three trigger types, schedule API, which I think is really cool. You could have a different automation make a post request to some sort of routine. And then of course GitHub so you can have it automatically fire off kind of on a web hook based on new PRs, new pushes, new issues, new releases, things like that. So how does this compare to what already exists? We [1:50:32] have routines which is the new feature. We have desktop scheduled tasks and then we have something like just a /loop command. So routines run on Anthropics cloud and these other two run on your machine. Do you need the machine on? No, for routines that’s huge. But for desktop scheduled tasks and for loop, you need your machine on. Do you need a session open? No, that’s the same across all three. Do they survive across restarts? The first two do, but loop does not. That has to live within a specific session. Local file access, no for the routines because it works off of the GitHub repo. And for the next two, [1:51:03] yes, you have local file access. Permission prompts with routines, it’s fully autonomous. And for these two, they are configurable. And then the minimum interval routines is 1 hour. And these two are both could go every minute if you want. Okay, so let’s talk about the environments. Obviously, your env get ignored unless you push it into the GitHub repo. You know, ultimately, if you push it into a private repo, you’re probably okay, but you want to be really really careful because then, you know, there’s history there and if other people, you know, end up collaborating on it, you just don’t want to do that. [1:51:33] So, you want to put your API keys in the environment variable like I showed you guys earlier. You want to look at the network access, whether that is full or trusted or none or custom, and potentially some setup scripts. So, that’s not something I showed you guys yet. If you’re creating a new remote task, you can do a setup script, which is basically just a script that will run when this new session fires up before cloud code launches. So if you need to install any packages or anything like that. Okay, so what’s the difference between trusted and full? So trusted only reaches the known vetted services from Enthropic, which I thought I linked [1:52:03] right here, but I just linked it there. This basically shows you all of the different domains that are allowed. So right here you can see we’ve got enthropic services, we’ve got version control, we’ve also got some cloud platforms like Google, stuff like this right here. These are the ones that are kind of already verified. So what is the risk of going on full? Well, if Claude reads malicious content during a run, then it theoretically could be tricked into sending data to an external server and with trusted that outbound request would get blocked. Now practical risk for private repos where you control the inputs is very low, but I definitely [1:52:34] just wanted to at least acknowledge that. So connectors. This is different than just adding your API key. This is more of like the connectors you would add to your actual claude chat or like claude co-workth into like Slack or ClickUp or stuff like that. Here are some security details. I’m not going to go super deep into this. You could also do some more research and download this doc. But of course, there are some things to be thinking about like your API triggers or what’s going on with your GitHub repos and the branches because once again, everything is going to be running as you. So if you’re not testing out these routines before you [1:53:04] just kind of send them off every hour or something, you just have to be thinking about what could happen without permissions and you know stuff like that. Limits and quotas. So it looks like on pro you can have five runs a day. On max you can have 15 runs a day and on team and enterprise you can have 25 routines a day. If you hit the cap the orgs with extra usage enabled can exceed it on metered overage. And then we have the minimum scheduled interval which is one hour. And there are also resource limits. So every one of these routines in the cloud runs on four vCPUs, 16 gigs of RAM, and 30 gigs of [1:53:36] disk space. So once again, just be thinking about, are you putting an absolutely massive GitHub repo up into the cloud right now to run? That could just be wasting resources for no reason. So what persists versus what gets destroyed? The clawed branches gets pushed to your GitHub repo and the session also stays. So as you saw, if I came into here and I looked at all of these tasks, I could see all of the past runs and I could go look at them to see if something’s going wrong. but the actual cloud environment that gets cloned will be destroyed. Basically, the rule of thumb here is if something’s local or if cloud code can’t reach it in [1:54:08] your GitHub repo or via an API, then it won’t work. We already talked a little bit about writing good prompts, but you definitely want them to be more specific. For example, with my um scheduled automation here, this is much more specific, right? I have a skill that I wanted to run. I give it the order of operations, but something more like this YouTube comments one, this is not what you’d want to put in there unless you were defining a skill to just let it run because once again, this is supposed to be a oneshot prompt. So, you wanted to make sure it gets it right on the first try. Okay, so why is this so exciting and why does this beat normal [1:54:39] automation? Because we are actually keeping the agentic framework. If you if you know when I talk about the WAT framework where we have workflows and agents and tools when we actually push those automations to the cloud and it’s just a you know sort of a Python script we’re losing the agentic piece we’re only sending off really the tools and the workflow but in this case we’re keeping the WA and the T all running together because the agent is looking at the you know cloud MD it’s looking at its scripts and it’s figuring out what to do and if it runs into errors midrun [1:55:09] it will selforrect and if you configure it the right way it will be able to sort like leave a memory trail and it can leave like you know updates even though each run is stateless you can still have them kind of continuously get better and real quick let’s speedrun through these common questions do I need to know cron syntax nope you just can schedule in natural language super easy can it access my local files nope it only gets what’s in your GitHub repo or your APIs what model does it use can choose any of the models as you guys saw can you watch it work in real time yes you can hit run [1:55:39] now and then you can obviously watch it go right there same way you would in Claude You can even talk to it after it’s done or interrupt it and then continue going. Can it use my MCP service? Yes, that is what the connectors are. Can teammates use my routines? Nope. These belong to your individual account. You might be able to share those if you’re on a team plan, but I haven’t actually yet tested that myself. What’s the cost? It’s just your normal subscription usage. So, keep that in mind. What happens when a run fails? Every one of them will be stored in your history. So, you can go see why they failed. You could maybe even have it at [1:56:09] the end of every single routine, say, “Hey, if this does fail, just shoot me a Slack message to let me know.” Things like that. And can I test a run before going live? Yes, in fact, you should test it multiple times before it goes live. You just go into the routine, you hit run now, and then it will pop up as running. And then you just watch it, you know, watch it go through its order of operations, and you can inject, and you can help it correct itself. So that you have confidence that once it shoots off the prompt next time, you won’t have to get in the way at all. Cloud Code can now remind you to do [1:56:39] things, check on things proactively for you, and work for days straight without you ever touching it or needing to give any input. So, here you can see I just said, “Remind me at 10:23 a.m. to check on my project.” It goes ahead and uses a cron create tool to set this reminder. There we go. 10:23 just hit. I didn’t touch it and it just said, “Hey, Nate, this is the reminder to check on your project.” So, just shot off this one that says, “Every 10 minutes, check my ClickUp to see if there’s any new developments on our project.” It’s using the loop skill, as you can see, which is a new built-in skill. And it creates a [1:57:09] cron for every single 10 minutes. And now this would run for the next 3 days, every 10 minutes until I told it not to. And this doesn’t have to be every 10 minutes. It could be every hour. It could be every 5 minutes. It could be whatever interval that you want. And this is all thanks to the newly released feature or skill loop, which is a powerful new way to schedule recurring tasks for up to 3 days at a time. And this is so funny because less than 12 hours before this was announced, the scheduled tasks include code was also announced. So, right off the bat, those two features might seem like they’re the exact same thing, but they’re actually super different in how they work, and [1:57:39] they have different use cases. So, in today’s video, I’m going to break all of that down and tell you everything that you need to know about it. And by the way, if you haven’t watched my new scheduled tasks video, then check that out right up here and then hop back over to this one. All right, so as you guys just saw in the quick demo, we now have the ability to use loops, which means that we could say something like /loop every 5 minutes, check on the deploy, or we could just say that in natural language, which is awesome because it invokes the loop skill and then it creates that cron job right here in cloud code. And you’ll notice that this is in my VS code. So this is available in your terminal, in cloud code desktop app, in VS Code extensions, wherever. [1:58:11] This is just a core part of cloud code now. So if you’re not seeing this, just make sure you update your extension or you update cloud code. And this lets you set up loop intervals or reminders. So reminders, like you saw that first demo, I just said, “Hey, at this time, just tell me this.” And in that session, it will bump up a message without you triggering it. Or you could have them be intervals. So you could say every 2 hours. You could say every 30 minutes. Whatever you want that actual interval to be. And what’s cool about it is it does it all in the same session. So if I leave this session up, every 10 minutes, it would check everything right here, which means that it’s able to continuously read through what happened [1:58:41] in the past one, and it continuously sees what we’re doing. Now, obviously there are some pros and cons there, but just wanted to point that out. The major con there being your context, making sure that if something does go off every 10 minutes, you’re not going to get a huge report and then every 10 minutes you just more tokens, more tokens, and then context rot. It’s basically scheduling a prompt that you would be sending in here and then firing off, which means you can loop skills. So, if you want every 20 minutes, for example, run a skill called review PR, you could tell it to every 20 minutes run the skill. It would run it, it would wait 20 [1:59:12] minutes, and then it would do it again. And of course, you could use actual slash commands to invoke both the loop and the skill. Or you could just say, “Every 20 minutes, run my review PR skill.” And of course, the onetime reminder feature. So at 3 p.m. or in 45 minutes, remind me to do this or check in on that. And Claude will basically pin that time, it’ll create that cron, and then once it’s done, it’ll just delete itself. So whether that’s, hey, at 4:30, remind me I have to go do this, or every hour, remind me to just stand up and like look away from my screen for 5 minutes, it can do that. All right. All right. So, there’s a couple things [1:59:42] that I wanted you guys to understand about how this actually works. So, let’s just play around a little bit. Hey, at 10:40 a.m., can you please remind me to take out the garbage? Cool. So, what that’s going to do is it’s going to use the cron create tool, and it’s going to create that basically schedule to remind me take out the garbage. And what you can see here is the actual prompt. So, at this interval, which is just how cron works, it’s basically going to shoot a prompt into this window that says remind Nate to take out the garbage. You can see the recurring equals false. Now, of course, the key is if the session is [2:00:13] closed, then that cron is going to automatically be killed. So, now something interesting. I’m going to open up a new session and I’m going to say, “Hey, every hour, can you just remind me I need to stretch my neck?” And I’m going to shoot this one off. And we’ll see how this one is a little bit different because this once again creates a cron. We have a prompt. And you can see in this one, we don’t have the recurring equals false. We just know that this cron is going to go every hour. But these loop jobs or task jobs are per session. So these two tabs are two different sessions. So if I came into this session and said, “Can you [2:00:43] please tell me all of the scheduled loop tasks that we have today?” It’s going to use a tool called cron list, and it only can see the 10:40 a.m. take out the garbage. It cannot see the task that exists in this session because they’re independent and they’re separate. Now, one interesting thing to notice is that this session didn’t actually invoke the loop tool. The loop tool basically tells it how to set up cron jobs and how to use the cron create. So if you don’t see loop, don’t worry. It’s still actually doing this in a loop. It’s just kind of about the actual wording. So, if I was to open a new one, let’s see if I [2:01:13] actually call the loop tool right here. So, I do loop and then I just say, you know, um, check my ClickUp. This one is going to default, I believe, to 10 minutes if you don’t specify a time. And it might invoke the loop skill because we actually called it to, but looks like it didn’t because it knows exactly what it needs to do already. So, the point being, all that matters is that the cron is being created. It doesn’t always matter if it invokes the loop skill or not. And then if you wanted to cancel one of these jobs, all you’d have to do is either close out of the terminal or just say, “Actually, I don’t need this anymore. Go ahead and cancel it.” And [2:01:44] that one invokes a different cron skill called cron delete, it shoots over the job ID. And now that it’s canceled and one final thing to keep in mind is in VS Code, if you close out of a tab and then you just open up that conversation again, that still will kill those crons. So you guys just saw how pretty much all of this worked. We have cron create to schedule. We have cron list to list them and then we have cron delete to cancel them and all of those can be invoked with natural language which is awesome. So now let’s get into some of the limitations and then I’m going to compare them to the actual scheduled tasks feature. So the first big one is [2:02:15] that we have a 3-day loop expiry which is just basically for safety. It auto cleans things up if you forgot you had all of these loops running. So once you create a loop it basically has a 3-day timer on it. It can run for day one. It can run on day two and then on day three it can run up until that last fire and then it will autodee. And if you want anything longer than 3 days, then you would either just recreate that loop or that probably indicates that you should just turn this into a legitimate scheduled task. Now, the other thing that you can do is if you want to completely disable scheduling, so maybe in your natural prompting it’s [2:02:45] accidentally creating all these crrons, you could go into your environment variables and just disable that and it would probably be able to help you figure that out. So the other things here are that if you close the terminal, your tasks are gone. It doesn’t have catchup. So the scheduled tasks, if you, you know, opened up your desktop app and you missed a bunch, it would catch up automatically. This doesn’t do that. And there’s no persistence, meaning after your 3 days and you wanted to do that same loop again. It would be a fresh session. But obviously, there’s tons of things you can do here with context management and reading different files in order to kind of Frankenstein your [2:03:16] own fix there. So now that all the features have been explained and you’ve seen a demo, I think that probably you understand the difference between the loop and the schedule tasks a bit better now. But let’s just go over some of the key highlights. The loop has your 3-day expiry. It’s all done within one session and there’s no catch-up. It’s basically a help me now or help me on this project for today type of function. The schedule tasks are dis stored. They’re longived. They have catchup and these are like daily, weekly, monthly functions that can run indefinitely. Of course, with both of these though, the terminal or [2:03:46] you know the app has to be open and this one is only currently available in the desktop app. But I can imagine with how fast Enthropic is shipping things, maybe by the time you watch this video, scheduled tasks are already out for the terminal and extensions as well, the way that loop is available in cloud code everywhere. So basically, it’s one simple question. Do you need help right now on a project or do you need help with something every day or every week? And that’s how you decide if you use the new loop feature or if you use scheduled tasks. So I thought I’d end off real quick by giving a few maybe practical [2:04:17] ways that you could actually use loop rather than something scheduled. So maybe all day you’re waiting on a very urgent email. Just set up Cloud Code to check in on that email every 5 minutes and if it’s there, it can automatically let you know. Maybe you’re working on a deploy and you want to just pull that and check every, you know, hour or so if everything’s working okay. Maybe you’ve got a deadline due at the end of the week and you need a 3-day sprint to be constantly checking in on the team and checking in on progress. Maybe you’re testing and iterating. Maybe you’re watching logs. Maybe you’re tracking changes. There’s so many different use cases here. There’s so many different ways to use the loop to prompt an agent [2:04:48] to have different files, to use different skills, and it’s really, really cool the way that you could potentially set these things up. All right, so now you guys understand the routines, whether that is a local routine or a cloud routine, you understand how you can set them up. You’re basically turning a skill into a schedule. You notice the pattern, you identify the trigger, you map it to a routine, and then you start to have more of your cadence. And I’m going to talk about this in a few more slides, so just give me a sec. But there’s other things that you can start to do here as well. you can start to visualize your system a little bit. So, if I open up Obsidian [2:05:18] real quick, do you guys remember how I talked about putting all of my YouTube transcripts into my AIOS? So, it can answer questions about them and it knows what tools I’ve used and what videos I’ve made and all that kind of stuff. This is me just visualizing all of the markdown files, all of my transcripts. And I’ve used sort of like this Carpathy LLM wiki idea to be able to visualize this and to be able to see connections between every video, every tool, how they’re referenced, how they relate to each other. And it’s really, really cool for me to be able to visualize it. Now, [2:05:48] Obsidian giving me this visual layer doesn’t change anything fundamentally about how my AIOS actually uses the data and is able to look at it and talk to it and edit it. But sometimes it is nice to be able to have that visual layer on top of it. So, if you guys are interested in figuring out how you can turn all of your business context, all of your, you know, meeting transcripts, all of your YouTube video transcripts into a visual thing like this using the LLM wiki, then I’m going to head into a video right here where I explain exactly how to set this up. If you’re not interested, just [2:06:18] skip past this part of the video. But if you are, go ahead and give it a watch. What you’re looking at right here is 36 of my most recent YouTube videos organized into an actual knowledge system that makes sense. And in today’s video, I’m going to show you how you can set this up in 5 minutes. It’s super super easy. You can see here how we have these different nodes and different patterns emerging. And as we zoom in, we can see what each of these little dots represents. So, for example, this is one of my videos, $10,000 aentic workflows. We can see it’s got some tags, it’s got the video link, it’s got the raw file, and it gives an explanation of what this [2:06:49] video is about and what the takeaways are. And the coolest part is I can follow the back links to get where I want. There’s a backlink for the WAT framework. There’s a backlink for Claude Code. There’s a backlink for all these different tools I mentioned like perplexity, visual studio code, nano banana, nadn. It also has techniques like the wt framework or bypass permissions mode or human review checkpoint. So as this continues to fill up, we can start to see patterns and relationships between every tool or every skill or every MCP server that I might have talked about in a YouTube video and I can just query it in a [2:07:19] really efficient way now that we have this actual system set up. And the crazy part is I said, “Hey, Cloud Code, go grab the transcripts from my recent videos and organize everything.” I literally didn’t have to do any manual relationship building here. It just figured it all out on its own. And then right here, I have a much smaller one, but this is more of my personal brain. So, this is stuff going on in my personal life. This is stuff going on with, you know, Up AI or my YouTube channel or my different businesses and my employees and our quarter 2 initiatives and things like that. This is more of my own second brain. So, I’ve got one second brain here, and then I’ve [2:07:50] got one basically YouTube knowledge system, and I could combine these, or I could keep them separate, and I can just keep building more knowledge systems and plug them all into other AI agents that I need to have this context. It’s just super cool. So, Andre Carpathy just released this little post about LLM knowledge bases and explaining what he’s been doing with them. And in just a matter of few days, it got a ton of traction on X. So, let’s do a quick breakdown and then I’m going to show you guys how you can get this set up in basically 5 minutes. It’s way more simple than you may think. Something I’ve been finding very useful recently is using LLM to build personal knowledge [2:08:21] bases for various topics of research interest. So there’s different stages. The first part is data ingest. He puts in basically source documents. So he basically takes a PDF and puts it into cloud code and then cloud code does the rest. He uses Obsidian as the IDE. So this is nothing really too gamechanging. Obsidian just lets you visually see your markdown files. But for example, this Obsidian project right here with all this YouTube transcript stuff that actually lives right here. This is the exact same thing. and here are the raw YouTube transcripts and here’s that wiki that I showed you guys with the different um folders for what Cloud Code [2:08:52] did with my YouTube transcripts. And then there’s a Q&A phase where you basically can ask questions about YouTube or about the research and it can look through the entire wiki in a much more efficient way and it can give you answers that are super intelligent. He said here, “I thought that I had to reach for fancy rag, but the LLM has been pretty good about automaintaining index files and brief summaries of all documents, and it reads all the important related data fairly easily at this small scale. So, right now he’s doing about 100 articles and about half a million words. So, there’s a few other things that we’ll cover later, but the [2:09:22] TLDDR is you give raw data to cloud code. It compares it, it organizes it, and then it puts it into the right spots with relationships, and then you can query it about anything. And it can help you identify where there’s gaps in that node or in that you know relationship and it can go do research and fill in the gaps. All right. So why is this a big deal? Because normal AI chats are ephemeral meaning the knowledge disappears after the conversation. But this method using Karpathy’s LLM wiki makes knowledge compound like interest in a bank. People on X are calling it a game changanger because it finally makes AI feel like a tireless colleague who [2:09:53] actually remembers everything and it stays organized. It’s also super simple. It will take you 5 minutes to set up. I’ll show you guys. You don’t need a fancy vector database embeddings or complex infrastructure. It’s literally just a folder with markdown files. That’s it. You literally just have a vault up top. So in this example, it’s called my wiki. You’ve got a raw folder where you put all of the stuff. And then you’ve got a wiki folder, which is what the LLM takes from your raw and puts it into the wiki. So in here, you have all the wiki pages which it will create. But then you also have an index and you have a log. So for example, in my YouTube [2:10:23] transcripts vault, here is the index. You can see that I have all these different tools which I could obviously click on and it would take me right to that page or after that I have all the different techniques agent teams sub agents permission modes the WAT framework and then we’ve got different concepts MCP servers rag vibe coding we’ve got all these different sources which are you know the YouTube videos and then when I have people or when I have comparisons they will be put in here in the index and then we also have a log which is the operation history so in this case in the YouTube project the log isn’t huge cuz I only ran one huge [2:10:53] batch of the initial 36 YouTube videos, but now every time I have one, I say, “Hey, can you go ahead and ingest the new YouTube video into the wiki and then we’ll see every single time we update this.” And then, of course, you need your claw.md to explain how the project works and how to search through things and how to, you know, update things. It’s also a big deal from a cost perspective, token efficiency, and long-term value. One X user turned 383 scattered files and over a 100 meeting transcripts into a compact wiki and dropped token usage by 95% when querying [2:11:24] with Claude. And obviously token management and efficiency is a huge conversation right now and will always be. The other thing that’s really cool about this is there’s not really like a GitHub repo you go copy or there’s not a complicated setup. You literally just say hey cloud code read this idea from Andre Karpathy and implement it. And people on X are now talking about like this is how 2026 AI agentic software and products will be made. You just give it a highle idea and it goes and builds it out. And Karpathy even said, “Hey, you know, I left this prompt vague so that you guys can customize it.” And I’ll [2:11:54] show you the ways in my two different vaults right now that it changed things a little bit based on the context and understanding of what the project is actually for. Okay, so this was the original tweet I just showed you guys and then he followed up and said, “Hey, this one went viral. So here is the idea in a gist format.” So if you open this up, this is basically just another explanation of the core idea of how this works and why the architecture indexing all this kind of stuff. And by the way, this is the part where he says, “Hey, this is left vague so that you can hack it and customize it to your own project.” So we’re going to come right back to this in a sec, but the first prerec that we’re going to do, it’s not [2:12:24] necessary, but I like to have a nice little front end to see the relationships is we’re going to go to Obsidian and download it. So, if you just go to obsidian.mmd, you can see this is the completely free tool and you’re going to go ahead and download it. So, just for your operating system, download this and then open up the wizard and open up the app. So, when you open up the app, it’ll look like this. And what we’re going to do here is we’re going to create a new vault. So, down here, you can see I have Herkbrain and I have YouTube transcripts. I’ll just make it a little bigger. I’m going to go to manage vaults. I’m going to create a new [2:12:54] one. And now, we just have to give this a name. So, I’m just going to call this one demo vault. and you’re going to choose a location where you want to put this. So, I’m just throwing this on my desktop and I’m going to go ahead and create this vault. Then, what you’re going to do is go to wherever you like to run Cloud Code. So, in this case, I’m doing it in VS Code. And I open up that folder. So, demo vault. We get an Obsidian and then we get a welcome.md. So, I’m going to open up Claude. So, I’m going to do it in my terminal. I’m going to run Claude. And lately, I’ve been liking using my terminal better for Claude. I like to do it inside of VS Code, but the reason is just because I [2:13:25] like to see the status line and I have, you know, a little bit more functionality. So, anyways, now that we have Cloud Code open, here’s what we’re going to do. We’re going to go back over to the LLM wiki thing that we got from Andre Carpathy. We’re going to copy all of this and we’re going to go back into Cloud Code and then just paste it in there. So, that is the prompt from Carpathy that’s going to build out everything we need. And then before we send that off, we’re dropping this in which you guys can screenshot and then just throw into yours. But I’m saying you are now my LLM wiki agent. Implement [2:13:55] this exact idea file as my complete second brain. Guide me step by step. Create the cloudmd schema blah blah blah. So this is just telling it what it needs to do with this idea that we just got from Kapathy. So anyways, on the right we have this cloud code running and on the left we have our obsidian vault and you can see it just created those two folders. So it created the raw and it created the wiki as you can see. Now, by default, it threw in four folders. It threw in analysis, concepts, entities, and sources. Once we start to populate stuff, we can talk to it to see if that’s actually the way we want to do it or not. Because it’s interesting in [2:14:25] my personal kind of second brain, the wiki is literally just markdown files. There’s no structure to it. And in some cases, that’s good. Carpathy actually said, “Sometimes I like to keep it really simple and really flat, which means like no subfolders and not a bunch of over organizing.” But then you guys did see in my YouTube transcript one, there were different subfolders. And I think that in this case it actually makes more sense. But you can see that it went ahead and it created a claw.md. It created an index and a log and then a few different folders in our wiki. But now it’s saying, “Hey, let’s go ahead and try it out. Drop in your first source into the raw folder and tell me [2:14:56] to ingest it.” Okay, so I’m at this website called AI2027. If you guys haven’t read this before, it’s kind of an interesting read. So go check it out. And now let’s say I want to get this into my vault. What I could do is just copy the whole page, right? And it might just come through a little weird. or we can just get an Obsidian extension which lets us basically take articles right from the web and just put it right into our vault. Super easy. So search for this extension called Obsidian Web Clipper. You would go ahead and add this to Chrome. So then when you’re in the article that you want, you basically just click on your extensions, you open up Obsidian Web Clipper, and then you [2:15:26] can just chuck it into your vault. And then right here, you’re going to want to set this to RAW because this is the actual folder that it’s going to put it in. So you can go ahead and click add to Obsidian. Open Obsidian. And then now you can see in my raw section we have this AI 2027 source with the title the source and it’s not super super populated yet because the LLM in cloud code is going to do that. So here is our file. I’m going to open up cloud code and say awesome. I just threw in an article called AI 2027 into the raw. Can you please go ahead and ingest that? It [2:15:56] might ask you some questions. It might also be helpful to before you start ingesting stuff say hey by the way this project is specifically for my second brain. So, personal things, business things, whatever. Or this is just a research project. This is where I’m going to chuck you all the articles and all the things that I want to learn about and all the things that I know. So, there’s different ways that you can set up the project as you saw with mine. One for YouTube, one for just personal second brain. So, now what it’s doing is it’s going to read through this article and then it’s going to figure out where should I chuck everything into the wiki. It’s not just going to create one MD file for this. It might create five or [2:16:27] it might create 10. And there’s going to be relationships between each of the different sections that it creates. So, it’s kind of doing its own method of chunking. Now, one thing I want to call out real quick is with this extension, if you go here and you open up the options for it, you can see that you can actually change where by default the folders are dropped, which is in the location section. By default, it’ll be going to a place called clippings, but just go ahead and change that to raw. Okay. So, here it came back with all these questions, right? It said, “Here are my key takeaways from this article, blah blah blah.” And now it’ll ask you, “What do you want to emphasize from this [2:16:58] article? What’s your focus? How granular do you want to be? what’s your plan? So, I’m just going to say I want this to be extremely thorough. This is my passion looking at where AI is going to go. Um, and this whole project, by the way, that you’re setting up in this vault is basically just going to be my place to dump in research about AI. So, help me keep all that organized so that I can query it and that I can, you know, keep my thoughts related. So, that’s just a quick example of what it might look like for you to give it some more context to continuously build your project. So, I’m going to switch over over here to the [2:17:28] graph view because I think it it’ll be interesting to see as it is starting to go through and create those different wiki files. It’s going to go ahead and it’s going to create all those relationships and we’ll be able to watch it in real time. All right, so it’s creating all of the wiki pages now and you can see that it said it’s going to make about 25 because there’s so much stuff going on in the original AI 2027 article. Okay, so our first one just popped in here and there a second one just came through and now you can understand you’re starting to see where do you have hubs or where do you just have little individual nodes? So this is a major hub someone named Eli someone [2:17:59] named Thomas Daniel and you can see all the different relationships here with things like AI governance with things like open brain superhuman coder. Okay so that ingest took about 10 minutes. So sometimes you have to be a little patient with you know it reading through everything and organizing everything but it does a lot of heavy lifting of course. When I uploaded the 36 YouTube transcripts in batch it took about 14 minutes. So it kind of just depends but it created 23 wiki pages. We have the source. We have six people, five organizations and one AI systems page, [2:18:29] different concepts, so technical alignment and geopolitical and then an analysis and then it asks some questions about it so that it can help make the relationships and make the structure even better. Now let’s just open this one up a little bit deeper and see what it actually did in here with this stuff. So we have this is the source with all the main relationships. So as we start to add other articles, we will see other big kind of like nodes and maybe in some cases we’ll have relationships between like compute scaling with different articles that we upload as well. So let’s just see if I click into the main [2:18:59] source, we can see the tags that it got. We can see the authors and we can click around. So here’s a link to OpenAI. Okay, what’s OpenAI? Here’s references in AI 2027. Here’s some other connections with OpenAI like modelsp spec. Okay, we’re in model spec. Let’s take a look. We can see other things about modelsp spec. And we could also go to how the LLM psychology model works. So this is just super super cool all the relationships that we get. And once again, all of this stuff that we’re looking at was derived from one article and automatically organized and related. So the question now is like what do we do from here? Do we query it inside of [2:19:31] this environment? Do we query it from somewhere else? And that’s completely up to the way that you want to use this. So for example, with my YouTube project, I’m probably just going to keep this here. And whenever I want to ask questions about YouTube or if I want to turn this into like a website, I can just do that from here. Or if I need to, I can point a different project at this folder since everything’s here and it can crawl through the wiki, it can read the index and it knows how this stuff works because you can give it the clawmd so it understands the project as well. So for example in this one which is just my second brain where we have all of the different things about like I drop in my [2:20:01] meeting recordings, I drop in, you know, ClickUp channels, summaries and things like that. This is something that I want to use in my executive assistant. So what I did in my executive assistant here called Herk 2. If I go to this cloud.MD, MD. You can see that we have a wiki path. So, whenever you need to read things about me and my business that you don’t have already, you would basically go to my Herkb brain vault. You would go to that directory and then you would read through the wiki. You can read the hot cache, which I’ll explain in just a sec. You can read the index. You can read the domain subindex. And then you can also just search through everything [2:20:31] here. And I said don’t read from the wiki unless you actually need it. Here are some things that you might do that you don’t need to go read the wiki for. And all of this is my business knowledge. Now, if you guys remember, if you watched my video on setting up an executive assistant, I used to do this with context files inside of this project. And when I changed over to this method, I actually saw a reduction in tokens that I was actually calling in this project. So, the thing about the hot cache, right, I didn’t actually have this in my YouTube one. So, if I go to YouTube, you can see there’s no hot cache. But, if I go to the herk brain in [2:21:01] the wiki, you can see there’s a hot.md right here. And this is basically just a cache of like 500 words or 500 characters that it saves, which is like what is the most recent thing that Nate just gave me or that we talked about. In the context of my executive assistant, this is really helpful. You know, it might save me from having to crawl different wiki pages. But in something like the YouTube transcript project, I don’t really need a hot cache. So, another thing that I alluded to but didn’t really cover was the idea of linting. So, Karpathi says that he runs some LLM health checks over the wiki to find inconsistent data, impute missing [2:21:33] data with web searches, find interesting connections for new article candidates, things like that. So, it basically helps you run a lint, you know, every day, every week, whenever you want, which helps make sure that everything is scalable and structured in the right way. And it might even come back and say, “Hey, I don’t fully understand this. Can you give me some more info or can you grab some more articles that might help me out here?” So now the final question about this that I wanted to cover is like does this kill semantic search rag? And the answer is no, but kind of yes. And it all depends on the goal of the project and the goal of the [2:22:04] context, how much context you have. So here’s a really quick chart that I had my cloud code make. I was in my Herk brain where I dumped in a bunch of information about Karpathy’s LLM knowledge and I just said, “Hey, can you please explain Karpathy knowledge as simple as possible, keep it super concise and um compare it to typical semantic search rag.” So, it found Karpathy’s idea. Instead of a database, you just give the LM well organized markdown files and it compares it here to the actual semantic search rag. So, actually, I might as well just read it [2:22:34] off from here. So it finds it by reading indexes and follows links rather than using similarity search. So we’re getting a deeper understanding of relationships because they’re links rather than just saying, “Hey, these chunks seem similar.” As far as infrastructure, it is literally just markdown. So like I said, you don’t even need the obsidian. You just need these markdown files. Whereas with semantic search, you need an embedding model. You need a vector database and a chunking pipeline. The cost over here is basically free. Your only cost is going to be tokens. Whereas over here, you might have ongoing compute and storage. And for maintenance, you just run a [2:23:04] lint. You clean up things. You add more articles. You give it more context rather than having to re-mbed when things change. But right now, the weakness of course with the uh LLM knowledge wiki is that it doesn’t scale huge across enterprises, right? Because it’s just a bunch of files. Um and that is where the cost will probably get more and more expensive than going to something like standard semantic search or knowledger graph or light rag or whatever other tool is out there for that. So here you can see if you have hundreds of pages with good indexes, you’re fine with wiki graph. But if you [2:23:35] were getting up to the millions of documents, then you’re going to want to actually do more of a traditional rag pipeline. At least for now with how the current models are and everything we know right now in April 2026. Okay, awesome. So that was Obsidian stuff. Now let’s talk about what else you could do with clawed artifacts because you have all of these different connections, right? You’re at this point now where you have a bunch of stuff hooked up. So you could have, you know, your school dashboard hooked up, your Stripe information, your YouTube stuff, your ClickUp tasks, but what if you want to [2:24:05] be able to have a visual place to see that stuff? Because, you know, once again, we now have potentially a visualization of our data, or sorry, not our data, but like our knowledge. But what if we want a visualization of our data? Well, you could go ahead and custom build all of this. But the first step, I think, would just be to leverage cloud artifacts and see what you can already do. So once again, I’m in the desktop app. I’m going to go to co-work, which is the middle section, and I’m going to click on live artifacts. And you can see here that I’ve got three different dashboards here. I’ve got a Fireflies dashboard, I’ve got a weekly [2:24:36] commitments dashboard, and I’ve got a QuickBooks dashboard. So, for example, I’m going to click on the QuickBooks dashboard. I’m not going to show you guys this data obviously, but what this does is as soon as I open up this dashboard, it makes a call to QuickBooks and it starts to pull in this data and populate it with real analytics. And so now I have this really simple place in my clawed environment where I can actually see our revenue, our expenses, our net profit, our cash on hand. I can see the trends. I can even get an AI breakdown of our runway and it’s looking forward and it’s doing a bit of a, you [2:25:06] know, financial analysis. I can see all of this information right here. I could also do this for my weekly commitments. So I’m looking right now, this is pulling data from ClickUp and it’s organizing it to show me tasks, completion percentage, what’s at risk, who I might need to follow up with. So once again, it’s just another way for me to visualize all of this data. And you can set them up with artifacts super super quickly. You just come into Cloud Co-work. You would click on new artifact. And this basically just starts a new conversation with Claude and it says, “Okay, hey, let’s build you a dashboard. What do you want to see? What [2:25:36] MCP servers can we connect to? What other connectors do we have access to? Let’s build you a visualization.” Now, that is really nice, but it’s not perfect. And ultimately where you might want to go is using Cloud Code to build you an actual dashboard that not only just has like, okay, your QuickBooks is this one and your ClickUp is this one, but maybe you want a dashboard to see everything because if you have all these connectors, wouldn’t it be great to just have one dashboard for your business where you could look at all this stuff? Yeah, it would be great, but it requires a little bit more configuration because [2:26:07] obviously you have to sync all the data, but then you have to build some sort of refresh cadence. Maybe you want every single night all of that data to refresh. Maybe you want it to refresh every 25 minutes. You would have to build out that logic and you would have to build it up as a routine or you would have to put that on something like trigger.dev or modal. So there’s just a little bit more of, you know, complexity. There’s more moving pieces, but it’s definitely doable. And so the reason why I recommend starting with artifacts is because artifacts you can spin up in 5 minutes. And if you spin up an artifact and then you realize, “Wow, [2:26:38] I didn’t even check this all week.” Then why would you have went through all the effort to build your own custom dashboard? You know, maybe you build an artifact and you’re like, “Wow, I actually check this every, you know, I check this four times a day. I should probably build my own custom one.” Then you know, so another one of Nate’s mindset shifts is P. What does PC mean? Proof of concept. You want to build something that’s super easy and it’s lightweight enough that it proves yes or no. Don’t waste time. Don’t waste money [2:27:08] investing into something that might not be proven yet. So prove the concept first. Lightweight, cheap, build a cloud artifact. If you like it, if the concept is proven, then you can dedicate some more resources towards building a custom solution. Because honestly, what I’ve realized is I don’t really need that much data visualization, at least in in that sense, because of the fact that I can just message my assistant and I can get answers instantly. And if I ever want to generate trends and I want to see what’s going on, I just say, “Hey, can you pull data from these five sources from the past month and generate me a dashboard or generate me a report [2:27:40] because that way I’m able to get stuff when I need it on demand rather than having to spend time building and maintaining some sort of dashboard. All right, so now let’s say we’re at a really good spot. We’ve got a bunch of context. We’ve got a bunch of connections. We’ve got a ton of capabilities and cadence. What does it look like on the dayto-day and on the week to week? Well, first of all, obviously trying to do as much work as you can just in cloud code and not leaving and switching between different tabs and context switching. But on top of that, let’s talk about the daily loop. So, what I would start to do is every morning go into cloud code and say, “Hey, help me plan my day.” If it’s [2:28:10] helping you plan your day in a really good way, meaning it can look at your priorities and look through messages and look through your calendar, that’s really good. Keep doing that. But if it doesn’t, note down why not. What might it be missing? What context might it need in order to plan your day better? And then at the end of the day, look back at it. What skills did you use today? What did you have to correct your assistant on? What did you have to, you know, copy and paste in? And then tomorrow, you can sort of patch some of that stuff up and just make small iterations as you go. And then on the week to week, of course, you can use the [2:28:41] audit. Maybe every single Friday, you run an audit and you see what you need to do. Think about how many skills got used. Think about which ones got used daily. If they’re getting used so often, why not just automate them? And the other thing to think about is sometimes when you actually decide, I need to build an automation, it might be overkill to go with an agentic skill. What you might actually need is more of a boring is beautiful workflow automation because workflows, deterministic workflows, beat AI agents nine times out of 10. Most of the stuff that we were doing for businesses were just automations. We barely even used AI [2:29:12] sometimes. And hardly did we ever give them like a fully autonomous AI agent. It just truly wasn’t needed for most business processes, especially especially when you start to actually break them down by task. You know, automate one chunk at a time and you’ll realize how little AI and how little autonomy is actually needed for each individual chunk of a process. And if you do get to that point where you want to be able to build something that is more of a regular Python script or more of a deterministic automation then then basically just ask cloud code to help [2:29:42] you build that and then you can say hey I need to push this into trigger or modal so that it can actually run 24/7. It’s really really easy. It literally just takes that natural language in order to actually do. So let’s talk about some of the success criteria that I think about when I try to understand is my AIOS actually providing me value and is it working? And I wanted to call these KPIs, but I didn’t because KPIs usually indicate that there’s something objective. There’s an objective metric that you’re looking for. And we’re not looking for really objective metrics [2:30:12] here. We’re just looking for kind of subjective success criteria. So the first thing is that your team could reach out to the AIOS. Meaning ultimately it’d be great if they could just message your AI OS instead of you because if the AIOS knows more info than you or has better memory than you and can pull the exact source and it never sleeps, it would actually be more efficient for people to message that instead of you. And I started to realize when someone would ask me a question, I would take their question and I would just paste it into my AIOS and then paste the answer back to them because [2:30:43] once again, my AIOS had more data than I did. Number two is that you stop opening new tabs. You don’t have thousands of browser tabs open. You don’t have thousands of desktop apps open because you’re doing as much of your work as possible right from cloud code itself. Number three is that knowledge leaves your head. So you stop having to worry so much about all these little things you have to remember. You stop having a million post-it notes everywhere. You just have all of your knowledge inside of your system. It can send you reminders. It can make sure you’re not losing track of things. And it helps you [2:31:13] out in that way. And honestly, it can get rid of a lot of your stress. And if even just two out of three of these things are true within the first month, then it definitely took and you’re definitely starting to get that exponential growth from building an AIOS. And once again, just wanted to hit this one more time. This is personal. Like the whole point of me building this course was for your personal gain, for your personal productivity. It’s great to be able to think about, okay, my whole team needs to adopt this or I want to start teaching other people how to do this. That’s great. But you have to [2:31:43] start with yourself. You can’t scale a system if you haven’t lived in it yourself. You can’t help your team connect to all these different data sources if you haven’t connected to them yourself. And once you’ve set up your own personal AIOS, your whole company can start to build around your AIOS. A company where every operator runs a personal AIOS is a company that is truly AI ready and all the data is AI ready. Because think about it, if I didn’t realize that I needed to do everything through my Google Workspace, then I wouldn’t have realized the benefit of having a shared drive and I wouldn’t [2:32:13] have been able to communicate to everyone else to use GWS CLI to get all of their stuff into the shared drive and organized. And now all of us are benefiting from that. All right, so I’ve been standing here yapping at you guys for way too long now. Um, I really hope you guys enjoyed. I hope that this gave you some structure and it got you really excited about what you can actually do and where you can take this thing. If you don’t have a current voice to text tool, please go download Glyo. I actually see a future very soon where we have more of a voice OS where I could throw away my mouse and keyboard and I [2:32:43] could just talk to Claude Code inside of my AOS. So, I’ve got my voice OS feeding into my AIOS and I’m literally just sitting here, you know, I’m just laying in bed talking all day and work is getting done. So, if you want to support and if you need a voice tool, check out Glido. Link is in the description. But that is going to do it for today. I really hope you guys enjoyed. And now that you’re set up in an environment where you’re using cloud code way more hopefully, you’re going to need to really start to think about context management. So, I’m going to highly recommend you check out this video next where I talk about how to make sure you’re not hitting your session limit [2:33:13] super super quick. And I’ll see you guys over there. As always, thanks for making it to the end of the video and see you in the next one. Thanks everyone.