Category: Nate Herk Youtube

So right here you can see that I’ve got Jeff Bezos, Paul Graham, Alex Rosie, Gary Vee, Dan Martell, Gino Wickman, and Warren Buffett all discussing a business question that I had. You basically shoot off a question with different options and all of these different council members will kind of come together, roast your idea, and help you think about it in a different way because each one of these seats has one specific question that they’re focusing on. And now you can see what’s happening is they’re actually debating with each other. So Jeff Bezos was debating with Gary Vee. Same thing with Paul Graham and Gary Vee. Now I’ve got Alex Herozi going after Gary Vee too. And wow, everyone’s deciding to pick on Gary Vee in this roast room. This is crazy. There we go. Warren Buffett goes after Hermoszi. But the point is, like I said, everyone has their own question they focus on. So Warren Buffett’s question is, what does it do to cash risk and long-term value? Alex Herozy’s question is, will people pay? How much? And how do we close? And Gino Wickman’s question is, can the business run this without the founder holding it all together? Now, after all three of those rounds happen, we can see that we have a final verdict, which was option C. And then we can also download a PDF, which allows us to read this, and we can sort of look at this more in depth. And we could also share this with our team if we wanted to. Now, this is really just a visual layer on top of what actually goes on in Cloud Code, which is basically just running a bunch of sub aents based on these different council members that I have created. So, you don’t need this local host open in order to run and use this, but it’s just a nice visualization of what’s going on. And this is completely customizable, meaning you can switch out different council members and you can build other ones, too. In fact, inside of this repo that I’m going to give you guys, there already are a few other ones, too. Hey Claude, what are the other council members inside of our roast room? And so that’s Glido actually shooting off a prompt right now to Claude Code, as you can see. And then it’ll show us who the other council members are. Cloud code comes back and says there are 11 seats. We have the default seven that you’re seeing here. And then we also have Naval, Cody Sanchez, Sarah Blakeley, and Leila Hermosi. Of course, you can build as many other ones as you want. So, that’s a new feature of GLO where we’re able to actually orchestrate different chats between different coding agents. If you want to try it out, you can get a free month using the link in the description. But anyways, just wanted to show this off to you guys and let me break down how this actually works. So, this is a repo called the roast room as you guys can see. And basically, we have these seven different advisors as well as four extras. And I’m probably going to keep adding more in the mix as well. But what I wanted to show you is the advisers. What are these council members? What do they actually live? They’re actually just a bunch of sub aents. So if I go into the claude, actually let me zoom in a little bit more. You can see that we have agents and in here we have one markdown file for each of these sub aents. So Bezos, Blakeley, Buffett, Gary Vee, Graham and all the way through. And if we open one up so you can see what it is, it’s literally just a markdown file that holds information about the sub aent. So for Jeff Bezos, for example, how he thinks, what are the sources that he references, how he makes decisions, all that sort of stuff. So that when I actually shoot off a prompt to the roast room, it’s going to go through the different council members and it sort of adopts their personality of how they think and how they make decisions, what they, you know, and it’s able to pull from things like their actual conversations, their lectures, their books, their principles, things that they put out there publicly, articles, stuff like that. So what you’ll notice is they will all have their own agent file, which isn’t really that long. It’s some things like how you debate, how you talk, your lens, but the really interesting stuff here is the brain right here. So we can see all the sub aents, but if I actually go into the skills in this repo, we can see that there is a roast room folder that has brains. And once again, there’s a folder for every single council member that you want to spin up. So if we, for example, open up Leila Hermosy’s brain, we can see that she has hot books, frameworks, positions, principles, sources, stories, and voice. And it’s pretty much the same thing that all of these other council members have as well, but they all obviously look different. The content is different. So if we open up Ila’s hot, what happens is we can see her pillar is can your team and your leadership carry this and who has to change. It shows us what this was built from. So for example, Laya was built from 48 sources, about half a million words from 2021 to 2026. 14 interviews where she’s the main guest and 34 videos of her own on leadership and things like that. Every person here has eight rules. They decide by frameworks, decision loops, things like that. Obviously, any books that they’ve written, frameworks that they talk about frequently and that they quote, their positions on certain things, pricing, discounting, raising money, debt, their principles, different sources that it actually pulled from, specific stories that they might talk about in videos, interviews, things like that. And every single one of these brains for the individual people is built the exact same way. If we open up Dan Martell, we can obviously see that he’s got books in here as well. We can see that he’s got frameworks that he’s using all the time. You can see the buyback principle, the buyback rate, time and energy audit, the drip matrix. All of this stuff should sound familiar to you guys if you have followed, you know, Alexi or Dan Martell, value equation, grand slam offer, all of these things that they actually talk about all the time. A quick break to tell you about the sponsor of today’s video, Fish Audio. Now, anyone who’s tried AI voiceovers knows that the words come out right, but the delivery doesn’t always because it might read a joke or a warning and an apology in the exact same tone of voice. So, Fish Audio has a new model that is their most emotive yet, and it’s called drama 3. You direct it the way that you would direct an actual voice actor, which means that you can write the direction in plain English inside the text, like slowly realizing something’s wrong, and everything after that tag will follow that direction. So, one line can start calm and it can end in a panic. Thanks for calling. Your order’s on its way. Wait, that’s not the right address. Who approved this workflow? And the tags work in other languages, too. It also generates full conversations. You can give each speaker a voice. You can mark who’s talking. And every character keeps their own voice across the entire scene. Just run it one more time. Of course, that’ll be attempt 47. So, if you’re putting it on video, the streaming endpoint sends wordle timestamps back with the audio, so your captions line up exactly with the speech. Now, on the build side, you set the model to drama 3 in the fish audio API. They also have an MCP server too, so you can connect it to cloud code or codecs. And they also have an official NADN node. So if you’re building something that talks like a voice agent or a game character or just narrating your videos, then try drama 3 on the fish audio API with the link in the description. Huge thanks to Fish Audio for sponsoring this part of the video. Now, let’s get back to it. And then I added this new skill called New Roaster, which basically just lets you build new roers. So for example, if I came in here and I was like, “Hey, new roster.” and we wanted to do, you know, Elon Musk and we shot that off. This will basically create Elon Musk’s brain and his agent file and all of his sources. It will build him out and also build him a little cartoon character that will show up in the actual roast room itself exactly the same way that all of these other Roast Room council members have been built out. So, you can keep scaling this thing up. And obviously, I’m going to let this thing build out Elon and then I’m going to add that into the repo so you guys can all use Elon as well. So, while this is building out Elon, as you can see, let us do a new session. So, open up the roast room. That’s pretty much all we have to do is ask for it. And I’m just going to switch this back to Opus here. And it’s going to open up the roast room. We’ll open it up in the local browser inside of the desktop app. And then we will shoot off a prompt so you guys can see how this actually works. Now, once again, you don’t need this visual interface opened up. It will obviously run the sub aent roast room just the same, but it’s nice to see. But one thing that I will say about actually spinning up a roast room is typically you want to put in a decision. You want to put in options. So two or three options. You want to give it context so that they can actually reason through it and make a better call. Right? So let me just shoot off a quick example. We are looking at hosting another in-person event in the summer of 2027 and we’re trying to figure out where to do it. We’ve been, you know, analyzing different options like San Francisco, Salt Lake City, Austin, Chicago. We obviously want to be able to get good sponsors there, local incentives, get a lot of people there, and you know, so airports should be important. Um, public transportation, good restaurants, good activities to do outside of just having the event, things like that. So, help me figure out where this should be and do this on the roast room council. So this natural language request should be just fine. But you also couldn’t come in here and do a slash roast room. And then you could just invoke this like that. But we will go ahead and just get rid of that for now and shoot that off and it should be good enough. One thing that this will do is it will help you figure out what would be the best council members to have for this specific debate. As you can see, it’s going to load the skill. It’s going to gather event and business context. So it will ask you questions if it doesn’t feel like it yet has enough. And then it will actually go ahead and run the debate for you. So, you can see it is going to ask me a couple questions here. It has four questions that it’s asking me. And by the way, we already did book this event in Salt Lake City, June 16 through 20. So, if you guys want to attend, then definitely check it out. But for the sake of the example, I’m just going to pretend that none of this has happened yet. So, I’m just going to say not signed, open choice. It says, “What size and budget should the council plan around?” And I accidentally just skipped through those questions, but the last one says, “Who sits at the table?” And it gave me a recommendation. And it also lets you play with it if you wanted to put different people in. So that’s part of it is thinking about who should actually sit on the council for your specific debate. But now that’s going to be it. It’s basically going to run. It’s going to process this. Even if this was closed, like I said, we would have sub agents that would be going through and doing this no matter what. And the reason I decided to do this with sub agents, because when I actually started doing these roasts, I started doing this with agent teams. But the reason I’m doing sub agents now is because I found after doing lots of testing that agent teams was more expensive for a minimal amount of actual value ad. So doing sub agents works pretty well because they start to actually talk to each other and stuff like that either way. So it works just fine. Now in this case you can see that we took out Dan Martell because his is more about like protecting the founders’s time and we put in Naval and his question is does this build leverage and equity you own or does it just rent out your time? So you can see it was great that we brought in Nal for this specific type of roast room. And obviously last time you guys saw the PDF that it creates which is pretty high level but all of this stuff is stored locally and you’re able to dig in. You can ask more questions about it. You can inspect all that stuff if you want to look at the debate more deeply. So it’s not lost data of course. Show me where all of the debates are held later if I want to be able to dig in and ask more questions about it. So as they’re going through round one and writing their opening positions, I can just show you real quick. If I go to the files and I go down here to projects and then we scroll down to a project called founder council right here you can see that there is different sessions in here and these sessions will show us different roast rooms that we’ve run since creating this visual interface and we’re able to actually then dig in. So if I look at this one, we can see that there was a brief, there was a round one, round two, and a verdict. And this will hold all of the session logs from each of these people because as they start to write their positions, we’ll see that all of their reasoning comes through even when they start to debate. And we can view it right in this interface as well. So now you can see they’re starting to come through with their opening positions and we’ll see below their name what they voted for. So Gary came in with voting for A and gave it an 8 out of 10 score. Gino did the same. And you can see A is Salt Lake City and B San Francisco, Austin, Chicago. This one, Alex Heroszi, also voted for Salt Lake City. So, everyone voted for Salt Lake City except for it looks like Naval went for B, San Francisco. And then, like I said, we could click in and we could look at exactly why. We could see the risk. We could see what they would do. And then we can keep seeing what everyone’s going to do. And now, what’s going to happen is they’re going to read each other’s positions. And then they’ll start to go after each other and try to poke holes in each other’s opinions. In this case, they all are in pretty much good agreement, which is nice. But if they did all completely agree, then the chair or sort of like Claude code, the facilitator of this debate, would make one of them argue for a different side. So they kind of play devil’s advocate and we get to see more perspective as well. But in this case, we didn’t have a unanimous decision. So the debate will continue as expected. Now, what I think is really fun about this is we’re able to leverage the subject matter expertise of other people. And humans are really good at their domains, right? So like for example, I could not build a world-class copywriting agent. It would be tough for me because I don’t have a copywriting background. But what I could do is go find someone who I think is the best copywriter in the world. I could extract their frameworks, examples of their copy, how they think, how they decide what to write, and then I could put them into this council. And when I have questions about copy or, you know, how to market something, I could leverage them inside of this roast room. And it doesn’t mean it’s going to be perfect, right? It doesn’t mean that whatever this roast room decides I blindly take and accept, but just getting in this mindset of thinking about who has done something that you want to do, who has done something and proven it. Now, how can you go out there and study what they did and how they got there and basically just try to follow that path? I think it’s a lot better to do that than just trusting Claude’s gut or trusting your gut sometimes. Like following someone that’s actually walked the path you want to walk is a pretty good signal, right? So, doing these councils helps me just look at things through different perspectives and, you know, think about things that I might not have thought about because Paul Graham brought up this idea or because Naval brought up that idea. And so even outside of the bounds of this roast room for example, I think that just thinking like that when you’re building out other skills or you’re trying to automate different processes, I think thinking like that is a really good way to think. But anyways, you can see they’re actually coming through now. They’re deciding on final votes. And you can see that we’re probably going to win, you know, Salt Lake City. It’s 7.9 out of 10 right now. And that is probably the verdict that we’re going to get. But of course, we could go through and we could read everyone’s key points. As you can see, here is Hermosy’s key point. Here’s Wickman’s key point. You get to see here Bezos conceded to Gary Vee and Graham about them being right on certain things but also disagreed the most with Naval. So anyways, that’s how you can sort of start to read through some things in more depth. We can see that we have the PDF now which I’ll go ahead and download real quick. And now we’re able to look at the fact that Salt Lake City won which is actually pretty cool because that is where we decided to have this first inerson event. Now anyways that is done. We could obviously do more rounds if we wanted to. We could bring in other council members. And now going back to check on Elon, you can see that it created the actual portrait for him and it created these different views so that we can see what it will look like when he’s actually talking. And then if I go to the repo, we can see that we now have Elon added in here. If I go to the brain or not the brain, the agents, we can see right here we have Musk. And then of course in the actual brain because it’s supposed to do that through the new roster skill, we have a brain for Musk. Once again, it has hot books, frameworks, positions, principles, sources, stories, and a voice.md. So, we are now set up and we are able to use Elon Musk inside of this roam council. If you guys want to grab this GitHub repo and play around with it, then just go to my free school community. This will be linked in the description. Go to the classroom and then grab it and from the AI skills or from the all YouTube resources. All of my other free GitHub repos, skills, prompts, guides, everything is in the free school community. So, anyways guys, that is going to do it for this one. I hope that you enjoyed you learned something new. And if you did, please give it a like. It helps me out a ton. And as always, I appreciate you guys making it to the end of the video. And I’ll see you on the next one.

So, after testing over a 100 AI tools for my business, I noticed something, which is that most of them are just noise or a fancy solution to something pretty basic. But there are seven tools that have actually changed how I run my business every single day, even though there are so many new tools popping up all the time. Now, some of these may sound pretty simple or even boring, but they have had a bigger impact on my business than I expected when I first started using them. So, today I want to talk about what they are, how I use them, and why you should, too. So, the first tool is not going to shock anybody. It’s actually two tools. It’s Cloud Code and Codeex. They’re AI agents that work right on your computer. So, they can read your local files, they can write code, they can build things, and all you have to do is talk to them in natural language. And for me, they’re the number one biggest lever in my business. If I had to only use one of these tools on this list today, it would be either Cloud Code or Codeex. I can use them to make slide decks or spreadsheets and Google Docs. I can use them as a brainstorming partner, as a council, and basically my entire operating system that my whole business runs on. My second brain lives in there. All my projects, my skills, all my knowledge, all my meeting transcripts, all of that kind of stuff. It drives about 90% of my daily workflow. It’s also helped me cut out a lot of other tools that I used to pay for because now I can just build them myself. So, I used to pay for something like Gamma to make slide decks or something like Poppy to help with content and now I can just build my own little custom apps instead. I’ve even built my own version of Loom that I use because once you can just describe the tool and that you want, you know, you don’t really need to rent someone else’s tool. Now, it’s not me saying it’s easy to build software that you could scale to hundreds of thousands of users, but if you’re just using it or you’re using it for a small team, it’s not as risky. So, if you guys run a business, a good place to start is a tool that you’re paying for every month, but only using one feature of. Or you can point something like Cloud Code at SOPs and client docs, and your team can ask it questions instead of asking you. Now, people do always ask me which one of these two that I use, and the answer is it’s both. I pay for both, and I use both every single week. You know, sometimes one model is ahead. Sometimes my usage just lasts longer on one, so I’m bouncing back and forth a lot. But in general, you know, I think Codex is a lot better at computer use, browser use, and cloud code is a better thought partner when I’m brainstorming. And right now, the usage lasts a lot longer. But that might not be true in 2 months. So, it’s not a matter of which one is best. It’s which one of them is best for the task that I’m doing right now. And how do I build my whole like AIOS, all my files and my folders and projects in a way where I can switch between the two and it’s not a hassle or it’s not a pain at all. But let’s move on to number two, which is Opener. Open Auto gives you access to over 500 AI models from more than 80 different providers through just one account. So everything’s in one place. You’re not having to manage a bunch of different separate keys or different billing information or different passwords for things like OpenAI, Anthropic, Google, all the other AI model providers. So any automation that we’ve got running in production right now, whether it’s for my business or, you know, for one that we are putting in production for a client, it pretty much uses Open Router as the AI model provider powering it. you know, because I can also set up fallbacks so that if one model goes down or errors out, it just tries the next one, it falls back instead of the whole automation breaking. And I can create a separate API key per automation. So the analytics are showing me exactly which one is costing me how much money. So if you’re building automations for your own business, which I’m assuming all of you guys are, then you could have a cheap, fast model, sort your incoming leads, and then you could have a smarter model, write the actual replies or do different reasoning, and that can all happen in one dashboard from one provider. It’s a really, really simple tool, and that’s kind of why I like it. Everything is in one dashboard, like I said. And it can all be on one credit card. And switching models is basically just changing one line. Super easy. Real quick, guys, I’ve got this completely free kit that I want to give to you all on helping you build your own AIOS. Whether you need to build it or whether you need to scale it up, this is the full kit. It’s based on my actual framework I call the Herk Loop. Context, connections, capabilities, and cadence. This thing comes with six skills. It helps you set up a brain. It helps you set up your priorities in your business, and it interviews you. It is exactly how I built my own actual AI operating system, and it’s how I’m able to move so fast. So, if you guys want this, the link for this is down in the description. But anyways, let’s get back to the video. All right, so moving on to number three. Hostinger VPS. Now, Hostinger is sponsoring this part of the video, but I’d have it on the list either way because I actually do use it and I’ve been using it for a long, long time. A VPS is a virtual private server. So, basically, a computer in the cloud that stays on 24/7 and it keeps all your stuff on that private server running even if your laptop or your desktop is closed or turned off. Now, this one probably sounds the most boring on the list, but it’s where a lot of my stuff actually lives. I’ve got one running and an end. I’ve got one running, you know, Hermes agents and openclaw agents and all the other open source agents you might want to host later or even open source models you might want to host for yourself. You know, my Hermes agents can automate different parts of my business over here. Like I’ve got one that’s trading stocks for me and one that’s making Instagram reels and carousels for me 24/7. And then I use some just to test different cloud code setups. And the reason that I use Hostinger is because I can spin up a server really quickly with just one click of a button without worrying about the technical side like scaling or compute or you know running the different commands or whatever it may be because I don’t really come from a technical background or like a DevOps background. So because of all these one-click templates for trending tools like either Enidan or Hermes or OpenClaw or whatever comes next, it makes it so much easier and it also makes sure that I don’t have to worry about it when I’m trying to sleep at night that something’s going to shut down and I’m not going to be able to figure out what happened and why. And if you guys want to try this one out for yourself, you can go to hostinger.com/nateherkvps and that will be linked in the description. And one more quick thing on this guys, Hostinger told me that they’re having this Black Friday sale from pretty much now until about halfway through December. So there’s going to be lots of different sales and Cyber Monday and lots of different things going on. So I’ll let you guys go check it out and explore those deals. But anyways, let’s get back to the video. So number four is Grockbot. And I use this one a ton. Grockbot lets you set up bots that do work for you. And they all share a cloud computer with a browser, file system, and a terminal. So, they can basically do anything that you could do yourself on a real computer. And what I love about Grockbot is how fast I can set up an automation because it’s got connectors in there to all the apps that I already use. I have it on my phone and it syncs pretty much automatically. I can schedule routines super easily in there. So, basically automations that run on some sort of schedule or a web hook and they just keep going all the time in the background right from my pocket. So, for example, after every single Q&A call in my plus community, one of my Grockbots grabs the transcript, posts it in my community, and then it also puts it in the classroom so that members can find it later. Another one is posting in both of my communities every time I drop a YouTube video. So, it’s like these little repetitive things that I might forget or if I did forget, I’d have to like catch up on later. Grockpot can now just handle that stuff so I don’t have to be like perfectly consistent and there’s less room for human error for me to make a mistake. I’ve got another one set up that keeps me up to date on everything that’s going on in AI. It’s just pinging me little reminders and I’ve got all these little different Grockbot personalities set up and each one has a very specific job. The way I think about it is it’s like having a bunch of Open Claws in your pocket, but with way less technical setup. Because with OpenClaw, you’re kind of figuring out like the server side and the connection and figuring out the hosting and everything. But with Grockbot, you just describe what you want and it’s just an app that you would install on either your laptop or your phone. All right, so moving on to the next one. Number five, this is Glido. And full disclosure, I am a partner in Glideo. But that’s how you guys know that I actually do use it and trust the vision of it. So it’s a voice dictation app. So, you just hold a key, you talk, and then words get typed and pasted into whatever app that you’re actually in. So, I basically don’t type anymore. Prompts, messages, emails. I just speak into my mic and it’s there. But you can really customize all your settings in there. It’s also the fastest on the market and it’s completely private. But the reason why I wanted to put this one on the list is because of command mode inside of GLO. This means instead of just dictation, it can also take action for you. So, I can have it search the web, summarize a YouTube video, do deep research, find and open files and apps for me, or you can even look through all of the different things that I’ve said to Glido. And all of this is obviously stored locally. Like I said, it’s completely private. And what’s cool about the command mode is you can build custom tools for it. So, basically, you can build custom MCP servers on top of it. So, basically plugins. And you could have it like kick off threads into cloud code or codecs or sync to your notion or to your task app or whatever it is. Ultimately, the vision we have for GLIO is that you can control your whole computer with just your voice and get just as much done away from your desk as you would if you were sitting at it. And even if you only use the dictation feature and not the command mode, it still saves a ton of time when you’re writing emails or SOPs or whatever it is that you may be doing, just talking is faster than typing. Okay, moving on to number six. This one is key.ai and it’s basically open router but for images and videos. You get pay-per-use access to a bunch of image and video models in one spot like VO or Clling or Cance Nano Banana GBT image 2. Anytime that I’m making images or visuals for a carousel or for a thumbnail or for an Instagram reel or whatever it may be, I’m using Key.AI because you only pay for what you generate and it’s been the cheapest option I found when I’m making a lot of images and videos at once. So for you guys, that’s stuff like product photos, ad creatives, short videos for socials, you know, whatever it may be. And you don’t have to pay a monthly subscription. You’re just paying per usage. And the last one here, number seven, is trigger.dev. It hosts your TypeScript code and it runs on a schedule or whenever something kicks off like a web hook. So unlike a VPS, there’s no server that you have to manage. It’s just your code and it just gets executed. We use this all the time for all the automations in the business that are more deterministic, meaning they’re more predictable. They do the same steps every time, like processing documents or syncing data between two different tools. Those don’t really need a full cloud code agentic harness loop AI agent. and it would kind of be a waste to use something that powerful. It’d kind of be overkill for something that simple. So, we just build out a script using cloud code and then we put that script in trigger.dev for it to automatically process. And I don’t even have to open up trigger.dev. I don’t have to like it’s not very confusing because I just tell cloud code, hey, I want a TypeScript trigger.dev automation for this use case and it will build it. It will also deploy it. It will also test it and it can also update it for me. And because it’s just running code, it’s super super cheap as well. So, you basically just pay the second it runs. And the free plan comes with five bucks of credits every month, which will honestly last you quite a while. So, for your business or for your client’s business and you’re setting up a bunch of deterministic codebased automations, I would definitely recommend using trigger.dev. Okay, so now you have seen the seven tools that I use to run my business. And obviously, this is my list. You know, it kind of changes over time, but these are the seven that I keep coming back to and that I have been using consistently over, you know, a year and a half, two years now. Now, if your business is doing different things than mine, then this exact list might not be the best one for you, and that’s totally fine. But I know that learning some of them can feel pretty overwhelming at first, and that’s why I try to cover as many as I can on this channel and inside of my community. You can also find all the free resources I’ve ever given away inside of my communities, and you can ask questions about how other members of the community are using these tools or finding success. And if you are really serious about using AI tools to build a business or just earn some income on the side, then you can join my plus community with everything that you need for the business side of AI as well. So, both links are down in the description. But that is going to do it for this one. So, if you guys enjoyed, you learned something new, please give it a like. It helps me out a ton. And as always, I appreciate you guys making it to the end of the video. I’ll see you on the next one. Thanks everyone.

So there are over 100,000 GitHub repos built for Claude right now, and most of them are honestly a waste of time. But after trying and actually using more than 100 repos, I found 10 that I cannot work without. Some of these help me finish projects 10 times faster, some save you money on tools you’d otherwise pay for every month, and some have completely replaced me in some parts of my workflow. So in this video, I’ll show you what they are, what problem each one solves, and how to use them to build something that businesses will actually pay you for. So let’s not waste any time and just get straight into this one. Okay, so the first repo is Hyperframes, and this is from Hagen. It’s taken over a huge part of my video editing. You basically just tell Claude what you want in natural language, and it builds the video out of HTML and animations, and Hyperframes renders a real MP4, so a real video, right on your computer. You can get captions, motion graphics, title cards, product promos, all that kind of stuff. And then what I do is I build skills on top of Hyperframes to make things like YouTube video intros or Instagram reels, course videos, ads, so many things. Editing is no longer the bottleneck for me and my team. I’ll literally drop in a raw clip and say, “Hey, can you add captions and a title this?” Or “Can you turn this into an Instagram reel?” And a few minutes later, I’ve got a finished file. Now you might have heard of something like Remotion, which also does this, but I’ve just found way more success with Hyperframes. It’s not a sponsored video at all, and I’ve enjoyed the experience using Hyperframes a lot. It’s open source, it has no per render fees, so it’s completely free. And pretty much any business that posts videos will pay for you to edit videos for them, like a gym that wants 20 caption clips a month or something like that [music]. You can sell that as a monthly content package, and you don’t have to have any video editing experience to actually use these tools. Okay, so number two is one of the things that I use the most, and it’s the Google Workspace CLI. It lets Claude work with pretty much anything in your Google account, like your Drive, Gmail, Calendar, Docs, Sheets, and Slides, and it formats everything beautifully using the CLI. Now you could obviously use the built-in connector for Google inside of Claude, but those only work inside of the Claude desktop app. They don’t cover everything. But this is a CLI, a command-line tool, so Claude Code, CodeX, or any other agent that you build with can you the CLI. And it comes with over 40 skills, so Claude already knows exactly how to use it. Now, setting this up will just take a few minutes because you do need a Google Cloud project for the login, but then Claude can just walk you through it step-by-step. You just have to give access to the apps that you actually want to use through the CLI. And most small businesses run on Google, so weekly reports, invoices, follow-up emails, anything like that, you can help them automate that quicker. All right, repo number three is Apify, or it’s technically the Apify MCP server. There’s a ton of data out there that Claude just cannot naturally, or natively I should have said, reach and touch. Sites like Instagram and Reddit, they might either block it or just [music] need JavaScript to load it, so Claude basically gets back an empty page. So, Apify connects Claude to thousands of ready-made scrapers for Instagram, TikTok, Amazon, tons of different websites. So, I’ll basically just ask it, “Hey, can you find bakeries in Chicago on Instagram and give me their follower count and download the videos and show me the ones with, you know, no websites in their bio, things like that.” And then a couple minutes later, you’ve got a spreadsheet right there. The connector is open source, but the scrapers run on Apify’s cloud, so you might have to pay a few bucks depending on which scraper you’re using. The free plan gives you five bucks of credit a month, and then the other actors you might have to pay for, like I said, per result. There are also some other open source repos out there that help you scrape different things like Firecrawl, but Apify has been really easy to use as well. All right, so full disclosure, this next one I built, and it’s called Scrollcraft, and I’ll link it in the description. You’ve probably seen those Apple style websites where you scroll and the whole page moves like a movie and it’s super immersive, there’s different layers, it just feels cool. And if you ask Claude for one, you usually get a very generic page with a couple of different weird text fade-ins or something like that. So, Scrollcraft gives Claude a real scroll engine, plus the design rules that I’ve picked up by building a bunch of different websites, and it has a checker built in that screenshots the scrolls and looks at the positions, so Claude can fix things before you ever have to instruct it to fix those things. My AI Automation Society site, my personal site, our event sites, every site that we build uses this repo. And what you’ll notice is that not all of these sites have to be super, super scroll-driven and immersive. The repo understands things like taste, and depth, and layering, spacing, typography, color schemes. It understands just tons of great stuff for website design and for feel. Because ultimately, the point of a website is to feel good and to be able to actually convert the user to do what you want them to do. Nowadays with AI, a basic landing page is pretty much a commodity. But brands that sell online still pay a designer real money for, you know, a custom scroll experience or something that’s not as easy to vibe code. So trust me, this repo is going to come in handy when you’re building your next site. Okay, now repo number five. This is something that I found almost by accident, and it’s been on every product since. It’s called Humanizer, and it makes whatever Claude writes stop sounding like AI slop. It’s built on a guide that Wikipedia editors use to catch AI-generated text. So it knows all of those like AI tells, when you can just immediately look at it and it smells like AI. Things like not this, but that. All of those sort of setups or T-ups. Everything coming in groups of three. Over-the-top hype words. You basically just have to paste in website copy or the cold email that you’re about to send, or whatever it is, and you just tell it to humanize it using this repo. And it keeps what you were saying, but it drops the stuff that makes people instantly scroll past or lose trust. Because obviously, copy that reads like a person wrote it is a way easier pitch than AI-generated content. Okay, so those first five were really about landing a client. The rest of these here are more so about when a client asks you to build something real, like a cleaning company that wants customers to book online and pay a deposit. And you’ve probably had this happen where you’re maybe an hour or two into a build and Claude starts randomly giving you answers like it forgot everything. And that’s because long sessions get compacted or you maybe enter context rot, and details from earlier get summarized away. So repo six fixes that. It’s called Agent Skills from Addy Osmani. Addy spent 14 years at Google and now works on Claude code at Anthropic. So it gets the plan out of the chat and into files. You can run {slash} spec and it writes a real spec with you. You can run {slash} plan, which breaks it into small tasks. And then you can run {slash} build, which works through them one at a time with a check on each one. So when a session inevitably gets messy, you can just open up a fresh one and it can pick up right where you left off. The spec will also ask about things that you probably wouldn’t think to bring up, like double bookings or refunds or customers in a different time zone. All of these edge cases that’s really important to bake into your actual spec or your build. And it’s something you can hand the client before you build anything, so they know exactly what they’re actually getting. That way you guys stay aligned all the way through. And if you’ve been using something like Superpowers from my skills video back in May of this year, Superpowers is still really good, but Addy’s is easier to follow if you’re newer. And there’s also other skills like Matt Pocock’s which are really, really good, especially if you’re wanting to build like a developer, even if you’re not a developer. So I guess Matt Pocock’s would be a bonus GitHub repo inside of this one right here. But let’s move on to number seven, which is Expo’s official skills for mobile apps. You can basically go from an idea to a real app running on your own iPhone in an afternoon, even if you’re on Windows. And they include Expo’s own list of 20 things that make an app look AI built, plus the current rules for getting into the actual App [music] Store. So one quick tip, tell Claude to use the Expo skills by name because they don’t always load on their own. You will also need a $99 a year Apple developer account for a test flight or the App Store. And most shops already have something like Square, so sell them the one workflow their current software doesn’t fit. But trust me, if you guys want to get an app into the store and you want to build it in a way that actually makes sense, this repo’s going to save you a ton of time and tons of headaches. Okay, now number eight is on the back end of that app that you’re building or software that you’re building with Supabase and Stripe’s official skills. So if your app needs logins or a database or payments, which pretty much every app does need, Claude knows each company’s actual security rules, like making sure users only see their own data and checking that a payment really came from Stripe. So only install Stripe’s skill called Stripe best practices, you don’t need the whole plugin to start because the full plugin might try to route every vendor decision through Stripe. And then you can set the Supabase connector to read only and then log in as one test user before you launch and make sure you can’t see other users’data. But trust me, I know that the back end and the payment stuff can get very confusing. If you’re vibe coder, you might get worried about some security stuff, so use these skills. Now number nine is what you run before you hand anything over to a client or to the the internet. Because Claude will happily tell you, “Yeah, this works.” But, it might not have actually clicked those buttons and checked if it really did work. So, Playwright CLI from Microsoft opens a real browser. It clicks through your app like a customer would, and it writes tests that you can rerun over and over, and basically just verify everything. And since it’s a command line tool instead of an MCP, it doesn’t dump the whole page into Claude’s context. So, it’s a lot lighter on tokens as well, and it’s a lot faster, honestly. So, this helps you verify and test a ton of things and find those edge cases. And then, we also have Claude security, which is Anthropic’s security plugin. It sends a team of agents through your code looking for holes, and separate agents try to prove each other wrong, and they try to look for different findings. [music] And you basically get a plain English report of everything that survived, and then it can write all those patches and just execute on it for you. Now, clients don’t usually care about all these tests, but what they really care about is that the booking form works and that their customer data doesn’t leak. So, build this into every project. And last but not least, we have number [music] 10, OpenSEO. Because once the site or app is live, people have to find it. It plugs Claude into real keyword data, rankings, backlinks. So, instead of generic advice, it tells you the exact page and the exact keyword where you’re sitting [music] at. Agencies use an audit like this to win the monthly SEO retainer. So, just know the free trial credits are small, [music] and by default, your data sits on their server unless you host it yourself. Now, some of you guys might be wondering about some of the token saving repos like Caveman or RTK, Ponytail, things like that. Now, JetBrains actually tested all three on real coding tasks. Caveman cut about 8 and 1/2% [music] of output tokens, where it was kind of advertising 65%. And RTK didn’t really save anything, and at low effort, it made tasks more expensive. Ponytail was the only one that saved money in these tests, about 10 [music]%per task, which is half of what it advertises. And on Windows, it has an open bug that can stall every message for several seconds. So, if you want to see where your tokens actually go, a repo called Codeburn shows your spend and flags the waste like an MCP service that you might never use but still loads on every message. And two, I recommended back in May, Claude Mem and Context Mode, I’d maybe hold off on those for now. Claude Mem sends usage data by default, and Context Mode has an open bug that can burn through your [music] usage. So, everything else I tested and left out will be in the description. All right, so if you’re brand new to this, don’t install all of these today. I think that would be extremely overwhelming, cuz if something breaks, you might not know which repo caused it. So, just pick the one that fixes the problem that you’re currently kind of running into the most. And if you’re not sure, then maybe start with the Google CLI or Appify or Squint Craft, and you might as well start with Humanizer, because that one is super super simple as well. But, pull a list of businesses with no website, you know, build one off a sample page, clean up the copy, and then send the link. No one’s going to pay you for a GitHub repo, those are free, everyone can install them. So, what you’re selling is the thing that you build around it that solves their problem. And I’ll put every repo from this video inside the description of the video, so that you can just go ahead and click on all of them right from there. But, anyways, that is going to do it for this one. So, if you guys enjoyed, you learned something new, please give me a like, it helps me out a ton. And as always, I appreciate you guys making it to the end of the video. I’ll see you on the next one. Thanks, everyone.

So, Anthropic engineers just 10x Claude Code by deleting over 80% of its system prompt. But, they also changed everything that we know about prompting. Because a lot of us were told the best way to get Claude to behave the way we want it to was to give it as much information and context [music] as possible. But, with this change, almost all of your information is actually slowing Claude down and slowing down its thinking speed. So, today I’m going to show you how to fix your projects using the engineering approach from Anthropic. So, let’s get into it. All right, so Thoric, one of the engineers at Anthropic, recently published an article called The New Rules of Claude Engineering for Claude 5 generation models. His team removed over 80% of Claude Code’s system prompt for models like Opus 5 and Fable 5 and 5.1 and future models, and they reported no measurable loss on their evaluations of these models. The system prompt is basically the rulebook that comes with Claude Code. So, your own instructions and whatever context gets loaded for the task sit right alongside of the system prompt. So, they’re saying that they could cut most of that rulebook and keep the performance that they had measured. And in my own setup, I used their guidance to review the instructions that I’d accumulated and clean up how my skills work together. And this goes right alongside of Boris Cherny, the creator of Claude Code said himself, which was basically like every 6 months you should delete all of your skills. And we’ve been told that skills are what makes Claude Code much smarter and much more consistent. Now, Anthropic describes that rules they needed for older models became too restrictive as Claude actually just got better and better. One example was comments in code, because they used to put a hard limit on how much Claude could write, even though some projects needed more explanation. Their replacement now tells it to follow the style of the surrounding code. I think it’s pretty easy to end up with the same problem in your own setup. You know, Claude makes a mistake, you add a rule, and then a few months later you’ve got that rule in your Claude.md file or in a skill, maybe a slightly different version of it somewhere else, and you don’t actually need that rule because you keep upgrading the model, but the instructions still stay in there. And the model might not need all of that extra context now. I recently did a video about prompting Fable 5.1. I talked about the idea of defining to the model what done looks like. You basically need to give it the outcome that you’re after, why you’re doing it, and the constraints that it [music] needs to respect. And then you just have to get out of its way. Leave it some room to actually do the work. So, for this video, I asked Claude to look at how I get from a YouTube idea to a script that I could film or an outline that I could film about, and I wanted it to find a small improvement and point me to the file so that I could check the recommendation for myself. So, I’ll show you guys that demo of me actually doing this later in the video. Now, Anthropic’s supply AI team recommends using the smallest amount of context that gets the result you want, and I turned that into sort of my own saying, which is the best context [music] is the smallest amount of context that gets you the highest quality result at the cheapest cost. They also explicitly say that minimal doesn’t necessarily mean short. If the model needs a detail to do the job, then obviously you need to keep that detail in there. And there’s research looking at what happens when you add more input. A paper published in 2025 tested five models on math, question answering, and coding, and performance dropped as the input got longer, even when the models could retrieve the relevant information. The reported drops ranged from 13.9% to 85% depending on the model and the task. Now, those were obviously different models, and that was, you know, over a year and a half ago, and these were controlled experiments, so those percentages don’t tell us what will happen when you trim your own Claude 3 model, for example. But Anthropic also tested loading MCP tool definitions only when needed. They reported an 85% reduction in token usage in their example, and their MCP evaluation scores improved from 49% to 74% for Opus 4, and from 79.5% to 88.1% for Opus 4.5. So, there are measured results behind this exact idea. And in that case, Claude still had access to [music] the tools. It just didn’t have to load all of those instructions up front. Now, in your own setup, you can use a slash doctor, and when I recently did this, the doctor’s report gave me this workflow audit with specific instructions and hand-offs that I need to clean up. So, I’m also going to show you guys that in just a sec. Real quick, guys, a message from today’s sponsor, Hyperagent. If you’re building agents for clients, most of the work lands around the agent, because each client’s data has to stay separate, their staff needs to use it without breaking it, and somebody has to be obviously paying for those runs. So, what Hyperagent does is they give every client their own workspace. So, the agents, the skills, the memories, and the The all live inside one perimeter, so nothing from one client shows up in the, you know, the context of another agent. And your builders work in there with full edit access, and the client’s own staff can get a role that runs the agents, but can’t touch the system prompts or swap the models or accidentally delete a skill or anything like that. So, if that sounds useful for your agency, then check out Hyper Agent through the link in the description. Huge thanks to Hyper Agent for sponsoring this part of the video. Now, let’s get back to it. Now, I have tried some pretty extreme versions of this just experimenting with my own AI operating system. You know, I’ve duplicated it before, I removed the cloud and the removed my skills, and I’ve tried off, you know, a bunch of different prompts to see them do the same job. So, one example that I recently did was I took this interview with Boris Cherney, and I wanted to turn it into a student resource guide. Now, my full AIOS setup that I was used to running every day, it made the prettier document because it had my branding, it had a proper heading, it had links that I normally include, and it had just like more specific context about me, all that kind of stuff. And the fresh version looked a little bit messier. It wasn’t as pretty, but its structure was much better. It had better details, it organized the interview into ideas, and it added timestamps, and it just felt more professional, and it felt like a better guide to actually give to a student. So, what I had was a polished guide from one version and a content structure that I preferred from the other version. Now, obviously, I still want my branding, so what I would do is I would build a new skill out of this, and I would say, “Hey, you know, you should be using this as the header and this as the font and this as the footer and add these logos and things like that. But, I’m going to leave it up to you to actually structure the content in the way that you see fit because I’ve gave you the basically the intent of turn this into a student resource guide.” Because in the pretty version that had my skill already, I had so many guardrails and so many instructions that I was basically just like confining the model, and I wasn’t letting it truly be creative and give me the output that it thought was best because it was reading [music] my skill and it had to follow my instructions. And those instructions were helpful back when we had like Opus 4, but now that we have Fable 5.1, it’s able to just do a better job on its own. It’s kind of like giving an experienced designer the checklist that you would give to a kid making their first presentation. Because the designer still needs to know who it’s for and what you’re trying to accomplish. But, if you dictate the placement of every box and every sentence before they’ve even seen the material, you might prevent them from coming up with something that they might add that you would actually like more. And that’s how I think about reviewing these instructions. I want to keep the details that help Claude understand the job and what makes it specific to me, and then test it whether it still needs the procedure that I’ve written around it. So, when I talk about the context part of my four C’s of an AI OS, this is the kind of information I mean. I mean my voice, I mean my goals, I mean my audience, and the way that my business works. Claude can’t just infer all of that or look it up online. So, that’s what I have to supplement to it. Now, it’s also important to explain why you’re asking for something, [music] the intent behind it. So, if I’m making a resource guide for students, that helps Claude make decisions about using, you know, analogies and simple explanations and using examples. And I’d keep that context when I clean up the skill. But, a long folder listing is something it can probably look up. So, if the same instruction is repeated in multiple different places, I’d check whether I need all of those copies or if I could just clean them up. And if a detailed procedure only applies to when I’m editing a video, I’d rather load it in when I’m actually editing the video. And this is all about getting out of the model’s way, which still means you have to be clear about what it’s allowed to do. So, I’d keep the Claude on MD useful as a starting point. Obviously, telling Claude things like what the project is for, [music] giving it any of the gotchas and the landmines to watch out for, and point it toward the deeper instructions. [music] So, now my whole video editing process doesn’t need to be there when I’m asking for help with a script. So, inside Claude code, you can run {slash} doctor to look through a lot of the stuff happening locally that you probably don’t spend much time thinking about, settings, config files, tedious stuff like that. So, I ran {slash} doctor in a new session, and it checked my skills, my memory, my plugins, my settings, and my installation. And you can see in this table right here, it lists each skill, where it’s installed, and how many recorded uses it has since installation, and an estimate of the tokens that its listing takes up. That basically just means like the YAML front matter. It uses that YAML front matter to find the skill, which is different from loading in the full instruction when it actually decides to run it and read the whole thing. For example, Agent Builder is estimated at 111 tokens, and prove it is 103 tokens. And both show zero recorded uses here. The report flagged 18 older skills with no recorded use, but kept the ones that I’d only created a few days ago. So, it’s taking their age into account as well. It also checks 50 recent sessions from this product and says that it also looked at lifetime counters across 252 startups and skill and plug-in dispatches across 223 transcripts. So, I’d still review that list before I decide to turn anything off or clean anything up or delete anything, especially if I do use a skill through a different tool, for example. But, these are the usage records available to Claude Code in this audit. [music] And then it also checks for things like duplication. So, in mine, the project map was repeating things already covered by the routing map. There were also repeated sections for templates, references, and archiving files. And it identified about 2,250 characters of cuts with an estimated saving of 563 tokens. And you can see it breaks out different places that memory is actually coming from. You know, I’ve got my project-level Claude MD, my user-wide instructions, the auto memory index, and my Claude local MD for instructions specific to this checkout on my machine. The local file was just an empty template, estimated at about 35 tokens. Now, that’s obviously not a ton, but the point I’m trying to make here is that it will look through a lot of the stuff that you probably don’t even think about ever. It also found the Claude MD from the separate video project in my desktop folder that loads into Herc 2, which is it roughly another 2,000 tokens just by its estimate here. So, even if you cleaned up the file inside your project, there can be instructions coming from somewhere else that you don’t even know you’re technically paying for. So, the setup checks caught two broken skills as well. One had the wrong file name. The other had a formatting problem in its front matter, the little description block at the top, like we talked about. So, Claude wasn’t actually properly seeing its description. And those files have been corrected because at the end of the doctor audit, it says, “Hey, do you want me to just fix all this?” It also looks at things like your installation, it looks at things like your automatic updates, and it checked the settings behind auto mode, permissions, and my hooks. And real quick, guys, if you’re learning how to build with AI, I’ve got a free community with a bunch of resources and skills and courses on building AI systems and agents, and you can also ask questions and work through the stuff with other people who are also, you know, behind you or in front of you. So, if you want to join, the link is in the description. Let’s get back to the video. All right. So, after running /doctor, I asked Claude Code to actually look at my YouTube workflow. And this is the prompt that I used. I said, “Audit this repo and show me the smallest change that would make the path from a raw YouTube idea to a filming-ready script faster. Use the systems already here, cite the exact files you would reuse, and do not edit anything yet.” You can see that it mapped the steps from finding an idea through research to an outline to a script and the teleprompter file that I can actually plug in and read while filming. Then it looked at how the work gets passed from one step to the next. The main recommendations about research, my outline skill could gather facts, but it didn’t require saving the evidence in a file that script skill knew how to check. And my script skill had a hard rule telling it to verify the facts again. Now, obviously, I put that there because I wanted verification. I wanted to make sure that it wasn’t making up facts or information. But if we’ve already found the primary source and checked what it says, I’d like the next step to have that work in front of it. You can see the audit counted 25 video folders with a script, but only three with a file specifically called sources.md. Now, this doesn’t mean that all the other videos had no research, but it does show that this particular way of handing over the evidence wasn’t consistent. So, it suggested three edits across the two existing skills, and I was able to once again just say, “Yep, that sounds good. Let’s go ahead and make those changes.” So, now that you understand how you can use Claude to audit your own setup, I would make this something that you’re doing regularly. Maybe once a month, you know, if you’re adding a bunch of skills and you’re changing your workflow all the time, or maybe once a quarter if the setup stays pretty stable. And honestly, what I would recommend is every time a new model drops and you switch into it, you should probably just run this sort of audit anyways, because every model is going to interpret your settings and your skills and your instruction files just a little bit differently. What’s really cool is once you’ve done this a few times and you start to have more trust in the doctor or in like your audit prompts, you can start putting these checks on a schedule. So, for this kind of local setup check, you could use a local scheduled task that can see the same files and settings inside of the Claude desktop app, and you can just leave it open, and then, you know, once a month it’ll run that doctor and it will run these different audits, and then you can wake up to recommendations that you can basically just approve or deny. We’ve talked a lot about AIOS’s or AI second brains, and a big pain point with those is on the humans, right? Human consistency is the issue. Sometimes we forget to update things, or we, you know, put duplicate entries in there, or we forget to delete things, stuff like that. So, if you can actually have your AIOS work with you on that cadence of automating the stuff and cleaning the stuff, then you’re creating this really powerful, sort of like self-improving, self-cleaning AI operating system, which is obviously something that there’s a lot of value in, because once you have your own personal AIOS figured out, then you can start to figure out how do you have your team stay synced up, and how do you have everyone working together, rather than everyone duplicating work and building their own AIOS’s. So, anyways, that is going to do it for this one. I hope that you guys enjoyed or you learned something new, and if you did, please give it a like, it helps me out a ton. And as always, I appreciate you guys making it to the end of the video, and I’ll see you on the next one. Thanks, everyone.

So, Grockbot just dropped two major upgrades, and today I’m going to talk about what those are, why they matter, and what you should do about it. And then I’m also going to tell you guys my number one favorite Grockbot hack that you guys can all get set up in about 5 minutes, and it will truly change the way you work. So, let’s not waste any time and just get straight into this one. Okay, so yesterday Elon Musk came out and said that Grockbot is going to be using the best backend model for any given task, including Claude Opus 5.5 and other things like midjourney, which is a image model, and Suno, which is a like AI music model. And then he said also other leading APIs. So basically Grockbot will understand the task and it will route to whatever model is most likely to give you the best outcome. Now from here someone basically said hey that’s super cool. I have some questions about how this works. Obviously I had a lot of questions too like can we see it? Can we control it? Blah blah blah. How does this affect our weekly usage? And what we now know is that this will be running on Opus 5.5. I assume by default, although we don’t unfortunately really have any visibility into what model it’s using at the moment or what effort level or anything like that. Elon Musk then came back and said, “Okay, so it will use whatever will achieve the best outcome for users. Simple questions will route to small fast models while questions with complex answers will route to large models. So ideally, this is going to be optimizing and getting you to essentially getting your usage or your weekly limit to last longer.” And they said, “However, most bot requests are pretty simple and will be handled by a lightning fast version of Grock 4.8 when that drops. The idea is to give users the best possible combination of speed and intelligence. Now, I think this is a really cool idea. I wish that we had a little bit more ability to control and to manipulate that because then it’s basically like Grockbot is the harness that we could control with any model and ideally any subscription, which would be really cool. And that’s a little bit of foreshadowing to the hack I’m going to show you guys in the end. But of course, some people aren’t super happy about this. People have been asking if they could opt out. They don’t want it to be going to Anthropic. They purposely got on Grockbot because they wanted to use Groc models. So whether you fit on the side where you really like this or you really don’t, that did happen. And unfortunately, we don’t have a lot of visibility into it, but I think that honestly I think it’s a step in the right direction. I think this is pretty cool. And then later that day, we also saw this, which is Grockbot can now search, read, and monitor X, which is awesome because, you know, obviously X, Elon Musk, Grockbot, but normally you had to actually pay for that. So what I would recommend that you do right now about this is you go into your Grockbot. You can see I have one called X that helps me monitor X and find posts and and gives me good like updates and stuff. And I asked it if it saw the update. I wanted to switch that over from our paid API which was the X API, the developer API, which still wasn’t too expensive, but you were still paying every time it pulled in a tweet. And now this is builtin X search and it’s completely free. It is native functionality of Grockbot. So, I then asked it to switch over all my routines and I wanted to make sure that all of my other Grock bots because a lot of these Grock bots touch X and search through X when they’re collecting data, doing research, helping me plan things, whatever it is. And I wanted all of them to be aware of that. So, this bot X messaged all these other bots, Minor, Pod, Eyes, AIS, News, Motion, all these other bots that might need to use X and that do use X and now they all have awareness of this. And then I was just curious about how much it actually did cost us the old way. And like I said, it’s not too expensive. We maybe were at 104 at one point and now we’re about 89 and we were spending about five bucks a week with all the X searching that we were doing. So like I said, not a huge deal, but you might as well take advantage of this and switch over to the native use. So then I went over to minor who I hadn’t talked to yet today and it did get messaged by X earlier and I said go on X and see what people are saying about Cloud Motion and then tell me how much that cost you. So it found these five tweets about Cloud Motion and then it actually said okay this cost me zero. I use the built-in X access. I didn’t touch your paid X API credits, but you do have a free daily allowance of a,000 reads. So, today you have 995 left, and that resets every night around 7:00 p.m. Central, which would make sense because I shot this off right after 7:00 p.m. Central. And earlier today and yesterday, I’ve been doing a bunch of Xarching with this new thing. So, that makes sense. But sometimes you have to take this with a grain of salt because sometimes Grockbot doesn’t always know exactly how it works. Also, this might be running on Opus 5.5 as we speak. So, grain of salt, but this seems like it makes sense. Real quick, guys, I’ve got this completely free kit that I want to give to you all on helping you build your own AIOS, whether you need to build it or whether you need to scale it up. This is the full kit. It’s based on my actual framework I call the Herk Loop. Context, connections, capabilities, and cadence. This thing comes with six skills. It helps you set up a brain. It helps you set up your priorities in your business, and it interviews you. It is exactly how I built my own actual AI operating system, and it’s how I’m able to move so fast. So, if you guys want this, the link for this is down in the description. But anyways, let’s get back to the video. Now, before I talk about this last hack, there is one thing that I think is really interesting to think about, which is the actual model routing. So, let’s say this is using Obus 5.5 and at some point it’ll be maybe by default using Grock 4.8. What I think is really interesting and what I’m going to make sure I’m doing is I’m going to make sure I’m breaking up my skills because think about that. If you have a skill that does like seven things and it’s just one mega skill, that might be fine. Like that might be fun right now because it’s like, oh, run my, you know, YouTube video creation skill and it does everything. It does the ideation, the research, the pulling in the comments, the, you know, the animation, the packaging, it does all that in one skill. And maybe that works best when you have one agent that can just do it. But when you really think about how this is going to be routing on the back end, you want to make sure that you’re truly getting the best experience of balancing intelligence and quality and uh cost. Because if you have a skill and let’s say it’s seven pieces and four of those pieces could be easily Grock 4.8, faster, cheaper, but three of them you might need Opus 5.5 for things that are more reasoning or more design, more taste, then you want to break those up into different skills. That’s at least the first thing I thought about was like, oh well, I don’t think that the model is going to be rerouting or switch routing mid task. Like, it’s probably going to understand load the full context and then it’s going to choose the best model because then you’d have other things like the the cache expiring and you have like this whole handoff. So, it’s probably going to be working where it’s like it’s just one model selection per prompt. I guess. And then we’ve also got the idea of, okay, what if miner, for example, shoots off off my prompt, shoots off prompts to Becky, Dan, Klouse, X, Nat, and Grockbot. And now we’ve got six other types of model selections to be made. I really hope that they start to bring us more visibility. I think it would be great if we could actually see what model they were using, what effort level they were using, and if we could control that, or if we could build skills and say, “Hey, for this skill, please run Opus, or for this skill, please run Grock.” Because that is a big deal. a skill won’t perform the same way for different models and that’s just the truth. So, we want to make sure that we are getting more visibility. I really hope the Grockbot team will bring us that visibility and bring us that customization. But anyways, those are things to be thinking about. Those were two huge upgrades that we got this week, which I am very happy about, at least the direction they’re going in. But now, let me talk about one of my favorite hacks with Grockbot that’s changed the way that I work. This isn’t actually using your Claw subscription. This is still using the Grockbot subscription. Whether you get that through cursor or spacex or what super grock, whatever it is, it’s kind of confusing, but you can have all your Grock bots use your clawed code subscription or your codec subscription, which is awesome because every single one of your bots has their own computer. As you guys know, they they share one big computer. They share a file system, but they own but they all kind of have like their own monitors like their like what’s on their screen is different, but they share the same file system if you open up your file manager. So anyways, because they have this, they have a terminal which means in the terminal they can install cloud code and codecs and other CLIs and they can just run them and all you have to do is authenticate in the same way you authenticate in when you’re on your own machine. So literally just go to one of your agents. As you can see right here, this agent Klouse is opened up inside of Cloud Code in my repo on Opus 5.5 and it’s using my team plan. So it’s using my actual subscription. So now because this cloud computer is always on, you essentially have cloud code always on in your pocket and you can just use um one of your agents to play with it or you could create a new agent that’s literally just called cloud code and you can create a new agent that’s literally just called codeex and now in your pocket you have these subscriptions and it’s really cool because literally all I did here you can see is I said I want you to on the desktop use my cloud code subscription just go to the terminal open that up it installs it. It then prompts me to sign in. So this is where you take over you sign in. You click authorize and then you’re good to go. And then you just want to have it copy in your repo. So clone in your AIOS repo, clone in your projects, and then you’re pretty much all set up because you’ve got your cloud, you’ve got your skills, you’ve got your agents, you’ve got anything that you need in there. And then the final piece to be aware of is you want to transfer over your ENV file because obviously you want all those skills and all of the real value in your AIOS to be transferred over. And so that’s why I don’t like relying on like cloud desktop app plugins or codeex desktop app plugins. I like relying on API keys so that if I switch over to something like this, I can just do one simple migration of one file and now Grockpot can do everything that I can do on my own desktop. So, this has been super great. You know, we’ve we’ve seen Dots, we’ve seen um Muse, and they could pretty much do the same thing. They have computers as well, but I have just been liking Grockbot so much more than those two tools. And I do truly use this thing every day for just really quick and easy and simple automations that I’ve got set up in here. As you can see, quite a few Grockbots working for me. So anyways, get this stuff set up. Hopefully this was helpful. Hopefully you enjoyed the video, you learned something new. And if you did, please give it a like. It helps me out a ton. And as always, I appreciate you guys making it to the end of the video. And I will see you all on the next one. Thanks everyone.