I'm Building a Vibe AI Agent Swarm Server in Public (Everyone will have one in 2025)

Hey there, I'm Parker Rex. I led tech for a startup that sold for 23 million bucks. After that, I raised a little bit of angel money and tried my hand at building Airbnb for music. It was pretty silly idea, but the technology was great. Then after that, I've just been paying myself through doing AI services to go all in fulltime on this AI space. So, I think it'd be hard to find someone that's spending more time thinking about the future, thinking about how to actually use AI every day in both your personal life and in your business life to advance. So, I'm all in on this stuff. And on this channel, we go through a couple different things every single day. We go through some questions and answers that come in. So, any questions that people have about how to use AI, any the topics that I cover, you can drop a question. I encourage you to drop a question on any one of these videos and I'll be going through them. We go through some AI news through the lens of someone who's using AI every single day to try to get leverage. So, I only talk about the things that I think are high impact. We go through some some of the strategy that I'm doing to try to get the most leverage out of AI, building both a SAS product, a media company, and doing AI services. And then we go through a road map, which is just some of the stuff that I plan to build soon and how I think about it. So, cool. Let's start with the questions and answers. We got a bunch of them that came in, which is great. Let's see. Am I on the wrong channel? Yeah, I'm on the wrong channel. Whoops. That's the main channel. I have another channel that I'm going to be posting a bunch of content on that will be specific to like different use cases. So, inside of the community, inside of the brand that I'm building called Vibe with AI, I'll be going through all of these different use cases of how I use AI, and these are actual ones that I use pretty much every single day. And then I'll be picking the best ones out of there. So, if there's like 30 in here, I'll probably take like the best five and post them on that channel, just so you know. So, worth checking that out. And then this is the daily upload channel. So, questions, we have a bunch of them. Hey, Vive TV by Basil. Thank you. So, we're gonna start out with Aush Patil AI. Hey, Parker. I'm 17 and I've been hearing about agents everywhere on my X feed and YouTube recommendations. Let's make this bigger. That was a question or a comment I got. Make the stuff bigger. So, there we go. About agents everywhere on my X feed and my YouTube recommendations. I know that the best way for me to learn would be to build something, but I don't know what's possible and what are the right tools to be messing around with. I'd be experimenting with all the new models and Gum Loop. Any suggestions? Yeah. So, we DM'd already on Twitter and I just wanted to say like first off, if you're 17 and you're thinking about this stuff, that's awesome. You're ahead of most of the world, which is crazy. And it is true. The best way for you to learn is to build something because that'll force you to learn something. So always start through the lens of a problem that you personally want to solve. So I find this most helpful and I got this from Peter Levelvels who if you're not familiar he's an indie hacker and he wrote a book called the make handbook. But if you have a problem that you personally want to solve and it could be really small it will push you through to figuring out how to do all the things involved. So, I'm going to use a silly problem. Let's say I want to my problem is I want to put together an ongoing curriculum of AI topics to learn. So, I want to have I need to learn AI. Let's let's use that. I need to learn popular AI technologies. That's my problem. And that's actually your problem. So, this is great. So, then you'd ask yourself, okay, well, I could learn it the oldfashioned way. That'd be silly. What are better ways to learn? Okay. Well, maybe I can use chat.com or chatgpt to set up some basic tasks where I could go in and I'd figure out, oh, I know that this thing can proactively message me updates because I can feed it the things that I want. I can show you an example. So I'd go in and I'd switch this over to scheduled tasks and I could just say I'm trying to learn AI. These are the following technologies I think that are replicable and these are the things that I'm trying to build. Then you can set this on a schedule where we'll message you and you'll get a notification with a certain topic. That's one way to do it. But you've then by having started with that problem, you've come up with a solution and then you've gone and maybe I always refer to the docs. So I would go to open AAI docs scheduled tasks. I did this when I was working on something with a friend for crypto stuff. And you could then see the help center. How do these work? What do I need to do? Then you can become the best person or one of the best people at actually learning how tasks work and coming up with interesting use cases, things that help yourself. And then you can go and talk about it on the internet and document that. And then people come to you to learn about that. They'll follow you. You'll gain friendships through it. You'll gain a network out of it. And that's kind of the combination which I really preach, which is code plus media. Those are the highest levered things especially in the AI era, the era of abundance. The media one becomes more and more important. So I would start with again like coming up with some sort of problem that you have and then go find ways to solve it. There's going to be 50 ways that you could try to solve this. There's another one you could use. It's called the fisherman prompt and that's in my air table. So, if you go to parkerx.com and then you go to the AI tools database under here, there will be a prompt. And I'm going to change this because I hate that Air Table just takes absolutely forever to load in, even if it does. Here we go. Come on, Air Table. But there's a prompt in here that's called the fisherman prompt. If I could find it. And what it does is it basically breaks apart topics that you're trying to learn into bite-siz lessons. So, let me see. Maybe it's an obsidian. Second, this prompt should be helpful. If you want the fisherman prompt, then drop the comment. Or actually, just say I want the fisherman prompt below. And let's open up what was it? Fisherman fisherman prompt. Here we go. So I'm a insert name of your level profession. So these are tags. These are XML tags which you should use for prompting and you can learn more about that in the program or just watch one of the videos when it comes out on my main channel. And I want to learn topic so I can objective. So it' be I am a beginner developer and I want to learn how to use AI so I can get leverage for businesses for myself. So follow the rules below to generate a comprehensive yet mini uh concise miniourse and then you basically list out these subtopics. So yours might be that you want to do some no code stuff that has AI in it. So you might put nadn or make or you might say I want to learn how python applies to it or I want to learn how the versel AI SDK works. And this would be something that you just tell to your chat GPT instance and then go through and make it a practice of 30 minutes a day as a habit. I'm going to go and do this. You'd be surprised. I did this for TypeScript because I didn't know everything about Typescript. And you don't need to know all of it, but there's certain things that you should know. And um this does a really good job of getting that 8020 or the PTO principle. So that's what I do. get really good at figuring out how to identify problems in your day. And Peter Levelvels from his book is basically like no problem is too small. Just make a list of them. Keep it in an Apple note. Keep it wherever. And then when you want to go learn, use that as the starting point for uh your learning journey. So, thanks for the question. When it comes to different models, like you can try them all, but there's just a couple of main players in there that I really think only matter. And I talk about this in the video I'm going to make for Vibe, but it's basically like there's Google Open AAI and Thropic and X. If you just stuck with one of them, you'd be in a great place. So, I would use Google because it's free and they right now have the best stuff. So, I' I'd hone in on that. And then Gum Loop is just another no code tool. Maybe pick a no code tool as well, but try to kind of uh like narrow it to don't jump tools every single day, especially when you're learning. So, thanks for the question. Uh Lionus says, "You're my AI news now, by the way." Awesome. Love to hear that. The iPhone cat guy, he has like maybe the most epic uh avatar in here. I need to make a badge for most epic avatar. So, people that comment that have the most epic avatar get the badge. That'd be kind of cool. Uh, Aush, thanks for the second question. What do you use to record? I use OBS Studio. So, it allows me to change the setup and the scenes for it. So, in order to make a video every day, the production needs to be really quick. So, OBS lets you just switch between scenes. So, if I go full screen like I do at the beginning, then it's just a click of a button. I don't have to edit that. And uh other things you can do with it when I get a better machine. I'm on an M1 Mac Pro. Like I maxed one out when I got the M1, but you can add more stuff on top of it, but I already have a lot of the things running. But I can add like a browser as an object on top of there. And then I can put code over the top. Meaning like I can have actions that occur based on keystrokes or I can have timers. I can have a progress bar. So similar like right now you're watching this video and there's going to be chapter markers that are generated with AI and I want to in the future use OBS and a browser and a prompt injection, not prompt injection, code injection that takes those chapter markers and then shows the progress bar across time programmatically rather than editing it. So you kind of brush over Google Cloud. I think it's a great tool. Yes. What other things does it have or what other alpha tools do you know about that Twitter is not focusing on? Yeah, so the thing with this is I can't go super in-depth on these every single topic or I could but it would be like the majority of the video. So what I can do is I'll make a video specific to using this. This is the AI studio. I think it's good where you like you go from when you're prompting something if it's really basic you can get away with just using Gemini on a web app. I wouldn't because I like the control of it. But then the second level to that is the AI studio where you can do a little bit more. And then finally you have Vertex which is just the best one. So I'll make a note to make a video on Vertex AI as I use it more. But yeah, it's it's to me the best one because it ties together all the things. So it's like if I have uh you know just basic Gemini web app and this applies to all the different places you can use all the different model providers but typically it's like you have this is like level one and then level two is the playground and you get a lot more customization you play with the temperature and whatnot and then level three is actually the cloud provider level and this would be through something like Azure or Bedrock through Amazon or the Vertex AI through Google and these ones are the just the best ones. Um, I've built stuff in Azure. I've built stuff or I'm starting to build stuff in Vertex and I'm going all in on Vertex because Azure is such a pain in the butt. The only reason I used it was because I got $10,000 in credits and I ripped through those by basically running a bunch of uh betting stuff against a bunch of data that I had. So, I'll make a video about that. And okay, so I think that brings us to the news. Yeah. Oh, no, there are more. Wow. Um, okay. Enjoyed your vid, man. Thank you. What automation should I build on Edna to understand what's possible? It starts again with a problem. It's kind of like the thing that I said you can go and another tip for this is to indoctrinate yourself is really what you're doing. You're like marinating your brain in this sort of content is if your Tik Tok is full of garbage then you're going to fill your brain full of garbage. But if you can go and change your algorithms on all the social media apps that you use by in Tik Tok you can actually reset your algo. What I recommend is then just typing in like n typing in make typing in Python typing in Nex.js JS typing in verscell typing in chat JPT all these keywords and then go and like them and don't like and then leave actually let it play out watch the thing because that's the engagement metric that they use to tune your algorithm and when people think of the algorithm in a negative way they're like oh like especially on the creator side they're like the algorithm it's this black box no it's not these are the most state-of-the-art AI models that are trained to reflect what you like. So if you're seeing stuff and you're wondering why they're showing you it, it's because you watch it. They know you better than you. So go reset it and then start doing that and then you'll eventually find a real use case. You can watch a bunch of them like these long courses, but it's, you know, the best way to learn is actually by doing it. And so make a basic automation where everyone has email. Go and try to figure out how to set up an email that figures out all this information about a person that emailed you. That'd be a good one. That's just a lead enrichment thing where it's like you could do that every time you get an email. You have the option to run this. You could run it once where okay, everything that's in my inbox that's not archived, go run a scrape against their personal website. Okay. Well, how do you get that? You lay out the things that you want to do with it. And the end goal being in this case, I want to know a bunch of stuff about this person. I want to be able to craft a personalized email. I want to be able to figure out what is an asset that I can create for them or what is a piece of value I can make for them in exchange to work with them. That'd be a really good start. And you just pick one of those and you go and build it. And yeah, it's the same thing as before. The best way to learn it is by doing it. So my question is what is the thing I should build to understand highle automation? Anything. Some of the best automations actually just look like two nodes. Like you'll see these videos when when you go and you and uh reset your algos on these platforms, you'll notice that the ones that have like 90k likes are probably like more simple. But the ones that get like 9 million likes are the ones where there's like a gajillion nodes and it's like I've built a version of Superman that I can talk to and it has like a funny voice and it's really cool and I can ask it to update my tasks and play a song for me. It's like okay well actually that's going to cost more be more annoying to use than just doing the thing yourself most of the time. So, I will just say that that like a lot of the sensationalized ones work better for generating clicks and then they funnel them into a program that teaches you stuff that like is really just a bar trick. So, be careful of that. Go build something that actually solves your own problem. Turns out your problems are probably shared with other people. Uh, what are some tools that I should be building in terms of having high reusability and having great build business value for users? I actually talked about this with someone on our channel the other day. Um, so what I would do is I think everyone needs to get more technical. I don't think that becoming like the make wizard is a long-term solution. I think you need to code. So I would go and I would look at these versell templates and try to reverse engineer some of the ones in AI. I would go and figure out how does this Nex.js AI chatbot work. I built one of the first ones of these in Swift literally the week it came out. And because I did that, because I open sourced this tech and I made like an AI SDK for Swift, I got friendly with some of the people at Apple. I got a bunch of people to follow me and then become friends. And um I would just go and get more technical. Like again, it's going to start with what's the problem you're trying to solve? like, okay, if you're trying to get more technical, go and take one of these, pull it down, and then rebuild it from scratch. And that's how you'll learn. And it's not like you don't have the end solution, right? It's like if you have a test and you have the quiz solutions at the end, it's like, okay, well, how did they get there? And you work backwards from it. Some of them would be like this next JS live transcription. You can just see like what's in here. So like this uses deepgram. Deepgram is like the best at basically transcribing stuff. Uh maybe it's not the best anymore. We used it at a client project like three months ago. So could be better. But I would do that. I would uh use their templates. I would go on GitHub and I would search different projects that I want to figure out. Like there's a really popular one called Kaj K hoj. And this one will teach you basically what would like a model aggregator look like with saved uh prompt templates. It's your sec AI second brain self-hostable. So like this would be like level five out of 10. Like you can go all the way to like super nerd, but this would be okay. I want to learn how to make different versions of chat bots. I want to learn about self-hosting. So yeah, that's that's what I would do. Thank you again for the question. And then do you have other YouTube channels? Yeah, I do. I have just it's Parker Rex. So just take the daily off of the username and you'll land there. that was primarily uh product management content. So if I look at channel and I flip to this, I go to oldest, you'll see that this is all product management. So this is me building in public and like lessons learned from 0 to 50 million. Um this one was really interesting. I like live vlogged us building a new set of features for a new business line that we launched to 500,000 people in three days. So I've been in this game for a while now. I'm just again like full-time on leveraging AI. So I I really would find it hard to find someone else that's spending as much time on it as me. Um so yeah, thanks again for the questions. Love your videos. This is my fiance. She's awesome. She's sitting here listening to me yap. So she's putting up with me. And then that's it for the questions. Let's go into news. So AI news, this guy Greg, let's let's start here. This guy is awesome. He writes a lot about communities and building out uh stuff in the AI era. You've probably seen him on YouTube. He's much larger than me. But I wanted to go through some of these. So, this is what keeping me up at night these days. And I bet you this this thread is going to have like a pazillion bookmarks in in a little bit. So, ChatGpt 40 images as big as the chat GPT launch probably birth a thousand one to million dollar vertical software businesses. I totally agree because you can look at something like this where people are taking this to make like pretty awesome landing pages like just off the rip and uh very cool. So, it's only going to get better. And you can imagine that you take one of these and you build a product where it's like, okay, you upload the photo. The photo gives you this, but it's a flat image and you can get the JSON, but then you need to turn that into Figma node code so that you can take those Figma JSON outputs and then put them into Figma because you want to touch them up. You want to make those changes. And so from there, then you'd have the Figma. You'd be able to make it look unique, which is still like there's gonna be a premium on not looking like it's generated by AI because people are everyone's going to have this soon. So then this the alpha will be making it better. So you're still going to need to know how to design. It'll probably be tools in Figma. They're starting to launch some AI stuff now. But having each one of these steps where it goes sketch to prototype flat and then flat prototype to Figma, Figma to plus up ideally using AI and then ultimately taking that and then doing code gen on it and then plus up. So there's human in the loop on every step. Eventually two three years from now you could just upload it and it'll give you the code. But right now I've used every single one of those Figma plugins and they're pretty pretty mid. So, we're in the MP3 Napster era of content. I agree. This is why it needs to stand out. So, in order to basically stand out, you have to be unique. You have to have a proof that people should listen to you. You have to have interesting stuff that stands out and you like you are the the moat. It's you. Like, no one can AI you. They're going to rip your content most likely and remix it because everyone's making these parasite systems. That's a huge reason why I'm building a community. That's why I'm like leaning into both code and media. So, every calendar will be rebuilt from scratch. Totally agree. I actually rebuilt an inbox, rebuilt calendar and a CRM. So, I know that. Um, I thought I was building digital employees, but it's more like digital employers. The first AI systems can manage human workers. will cause a restructuring of the labor markets. Yeah, that is interesting. But this isn't totally new. There's a thing that was called uh Mechanical Turk and it's an Amazon service where they basically used labor arbitrage in different countries at scale. meaning like you could just flip on a switch and then have a task done by what was called mechanical turpp, but it was really just people in different like low-wage countries. Um, what I do think is interesting here is, you know, you have task dependencies in any process. And so the difference now is that they're just going to be much cleaner. So if we think about it like when I ran a bunch of engineering teams when I would run our entrepreneur operating system for our company like I'd set up systems in a sauna and in a sauna you might have a project and in that project you'd have the goal of the project and then you'd have all the tasks of the project right and so what would happen is you'd have all these and then some of them would be red because they're dependent on someone before them. And that way each one of these would be assigned to someone. So it's like, okay, Parker, Scott, Jay, uh, Kale, like they all have like different things that they need to do in here that were are waterfalling to the next one. It doesn't mean like, oh, waterfall versus agile, like we're not talking about product management stuff. We're talking about just like general project management, which is different. But that system still required somebody to go in and tick these things off and do the check-ins. And then the same thing goes for weekly business operations. So if you had a scorecard, this is another thing that I implement and I do. It's like you'd have your scorecard and you'd have your stats and if something was down, then the owner of said thing would need to go and do some analysis and try to figure out a reason why that thing's down. We obviously want it to go up in most cases. Most KPIs you want to go up. If it's conversion rates, obviously um if it's churn, you want that to to go down, but conversion will go up, right? So you have it's dependent. But that would be another one where it's like, okay, well, you still want maybe part of it automated where you can pull that in. That's not new. That's just reporting. But maybe the analysis piece would be something that an agent can go and do. And then that would be in a way the thing that's managing the staff underneath it. So this could be an interesting use case where it's like at the maybe at the daily level even you could have a KPI agent and that role would be to check stats post stats analyze the changes in the this and then finally you would actually need to go so it's going to see oh like this thing's off like we had an expectation here then it's to come up with what the root cause analysis is and a lot of people are really bad at this because some of it can be what I call squishy or subjective where you ran an offer and you also released this new technology and you also release this new uh policy within your staff of uh sales folks and if you do like a multivariant test at the same time it's really hard to identify it but you can still like speak around it come up with a thesis um so the squishy ones I think would be more challenging but you can still prompt around that whereas the objective ones where it's like hey like conversion is down guess what you guys launched a change that went from having two screens and you'll notice this like when you check out on Amazon, they have a cart review and then they have checkout. And these pages are always the ones that the PM and the team are always really focused on. So it's like you guys made the button on the checkout page, it actually is no longer stuck to the bottom. So if you had a lot of items or if you had uh all the details of a new payment method because you launched that, this is now below. So you could have that be tied in. So, I do definitely believe that this is where it's going and it's going to stomp on a lot of the product analytics tools because product analytics tools are they're you know they're not like all these are laggered metrics meaning like they just report on the thing that came before it but they are built around going and doing what's called an event taxonomy which is like every step in a funnel has an event and then those events are then collected and run into funnels whether it's something that looks like this or line charts or histograms or whatever. But I think a lot of this stuff is going to go away because you don't need to go and find the insights. And I know the product analytics tools in this in this day and age can find the insights for you, but it's still like you have to go to this thing. They're really expensive. So I do it all leads back to my thesis which is like every company is going to have their agent server and the agent server is going to host all of their stuff. So they're going to have a database which is their business thing. They're going to have a CRM. They're going to have their business apps. So like the actual like core software. So like maybe they're using Next. They're going to have their next apps. They're going to have marketing uh sites that will be separate from those. So those might be using Astro. They're going to have their own instance of Rag. So you might have Rag plus uh chat and all this stuff will be used internally. And then on the like the biggest one will be basically like their agent team. So you'll have a box that's set up that's just agents and you might have another one that is agents within it then. This is the future fellas. So this is what I'm building inside of Vibe with AI. And when you join you get access to this. We'll talk about that in a little bit. This isn't like total promo but it's like because this isn't even built yet. I have like the view with it. But this is where I think it's going. And I'm basically doing the research inside of this media/ education company so that I can have people give us feedback, too. It's like this might not be right, but with feedback, you'll make it better. So, yeah, that's where I think it's going. Um, there's obviously other stuff in here, but I I'm not going to go through all of them. Yeah. What happened with the spaghetti video? Backlash won't come from just replace workers. It'll be everyone who realizes their digital identity is being converted into training data. Yeah, I mean this just uh is something that's happening like it goes to this concept which is if you don't make your own identity, someone's going to make it for you. So you need to put yourself out there and have a stronger digital footprint. Cool. That was that one. Let's go into We talked about the web designers one. Oh, Superbase launched this. This is nuts. They're going after uh clerk if you're familiar apparently makes it incredibly simple to build powerful applications by giving you things like authentication, file storage and real-time subscriptions out of the box. But sometimes building your own custom UIs to fully utilize these powerful features can be timeconuming and also errorprone. Today we're excited to announce the superbase UI library, a set of convenient battle tested components that can be easily dropped into any application. We now provide custom Superbase clients for React, Nex.js, Remix, and Tanstack Start, so you can get up and running with Superbase immediately. For authentication, we have custom blocks for signing up or signing in, email confirmation, and password reset. For real-time subscriptions, we now have a component that shares cursor positions to build collaborative apps. To easily integrate file storage, we now have this drop zone component that allows you to upload multiple files to Superbase storage. For AI code editors, we now have a set of rules to help steer that agent. Yeah. So, this is a really big deal because it goes to my point about get more technical. This is a game changer. So, if you're unfamiliar, Superbase basically has taken backend as a service to the next level. So, bass this was originally, it's not the fish, it was originally like Heroku was a good one. Actually, no, that was past that was platforms. But back end as a service takes all the things. So you write your app over here. So we'll just say app. It could be web app. It could be iOS app. But then you obviously need other things. So in my case I use next. And this does the back end and then the front end. And if I need something that's a little bit chunkier, I might go use Express or Go for back end. But for basic stuff, you can get away with next. And then you need all these other things, right? You need the back end uh or sorry, you need the database. And what the database has is it has an API gateway in front of it. So when I'm writing code and I want to fetch something or I want to do like a crowd operation, then I can just have like a folder that's called queries and I can have a folder that's called mutations. And then in each one of these folders, you'd have, you know, like uh delete log.ts. And then you'd have a singleton, which means like there's one file that represents the instantiation of the connection between the two. Meaning like instead of each time like having its like a a bazillion lines of the same thing, you just say this is our connection to superbase. And then you'd in each one of these mutations which are like changes or queries which are lookups you go and you call that thing. So that's just one like example what superbase does but it's just been like this guey that you go to and it's awesome because you can see all the tables. So right, it has the database has edge functions which let you do interesting things like oh I want to do a a trigger where when we add a new blog post we want to do a trigger that listens for the change and then goes and or like remixes the blog into LinkedIn post and then stores it for whatever reason probably wouldn't do that but it lets you make calls to external uh APIs and in this case it would be like chat GPT But them adding stuff that's not just the kind of administrative backend tooling. It's really helpful. And then it kind of gets into the area of like if you look at clerk.com, their whole business is built around just doing that where it's off and user management. And it's cool because you can drop in these components. A lot of people hate on them. I don't use them. I just use Superbase. Now it's nice cuz like I can just grab like instead of using uh Clark which I don't use but I just give Subase a shot and make it really easy for starters because most apps I mean you have to have O right. So you'd have that and then you'd have like the React drop zone. So instead of going and grabbing react drop zone as an npm package, you just use the superb basease one which most likely under the hood is using the same thing but it's tied in and they'll probably handle the right like best practices for cookies and all that stuff. Um what else? Here's another example of the wildness that is 40. That's like a really good thumbnail. And then last superbase one is around their types or sorry the language server for Postgress. This is really helpful. So if you're unfamiliar with language servers they are built into most IDs. If you use Vim then you know you've had to install it with Mason. But a language server basically gives you the syntax errors. So instead of like just raw dogging code and being like I don't know you're doing like a P tag and you do capital P it should be lowercase P. It'll give you like the little squiggle. So that little squiggle is a language server and by Postgress having one that's from Superbase is going to be really helpful because you write migration code which is the code that is like a an update to the structure of your database. And I've always run into issues with the ones that weren't super basease related. They just don't work as good. So, it's nice to see them going with that. And that brings into the topic of like type safety and language servers and tests and all these guide rails around your codebase. They matter more and more and more in the LLM era. So, um is this Murea AI tool? We'll just try it really quickly cuz I don't I only have like 10 more minutes I can record. So, let's try this out. Music production. I've tried the music production ones and they're horrible. So, let's see. See, library. We'll do this another time, but more music production stuff. Let's wrap the video by going into this. So, I mentioned that I'm like full-time doing this code and media stuff. So highly recommend you follow along. Made this tweet about Naval talking about it. So it's like code and media are permissionless leverage. They're the leverage behind the newly rich. You can create software and media that works for you while you sleep. Totally a thing. Like I think the combo of the AI marketing orchestration aka like vibe marketing which is going kind of viral plus personalization equals AI mode. So no one can be you. You need to have you at scale do stuff that's interesting and worth talking about and then leverage your technical skills to make the marketing orchestration layer really awesome and then make the coding bit even more awesome. And that brings in this agent server that I'm developing. Oops. Apparently it crashed. Okay. So there's a community called the cyber range and I thought it was really interesting where this guy he's in cyber security and he built a system where when you join the program you join it automatically provisions backend access for the members. So if you're trying to get into cyber security, he basically opened up a container as I understand it to the open world. And so you can basically I mean just read it like you can like it allows people that join to see real time like cyber security attacks. I'm not going to even pretend like I know enough about this because I don't. But like he made it so that bad actors can hit this engine that he's opened up and then the members can go and see how that works. So, I'm not doing this obviously because I'm not a cyber security guy, but I'm an AI agent guy. And so, what I wanted to do was build something similar like I mentioned earlier where I believe everyone both personally and and uh in business will have one of these agent server that has everything on it. And so I'm building one and right now I just rolled it from an open source project but it has superbase edge functions Kong storage reddus qrint which is vector storage nadn superbase oer open web UI which is uh basically chatgpt skeleton kind of like that thing caddy which allows you to do the deployment so we're going to have cicd through github where whenever I make changes and I push domain it'll just kick off those runners and then it'll deal with reverse proxying if you're not familiar with that. It's like how do you get all the traffic that's coming into here? Like if we have traffic that's coming in, how does it know which one to go to? Right? Because one of these next apps might be mapthemap.com, the other one might be Parkerx.com. So we'll have reverse proxying in there. So you can do that. And then the real time stuff, this is through Superbase, but it uses Elixir under the hood. So if you need real-time chat, if you need anything real time, Elixir is like really awesome at that. And yeah, that's what they use at WhatsApp. If you go read early WhatsApp days, like that guy's obsessed with it. Uh Superbase Studio, which is the guy you see, uh this thing Crexen, I don't even know how to say it, but this is open source perplexity. Flowies is another uh no code automation tool and then you have image proxy all this other stuff. So what I'll be doing is I'm just adding more things to this. So the next one that I'll be adding is the CI CD pipeline. So it feels like Verscell if you've automated or if you've pushed code to Verscell you know like why everyone's obsessed with it is because you just push and then you get a preview branch and all that. So that'll include preview URLs across all the different apps within it and then obviously production. So that's what I'm going to be adding there. And then I need to add the actual Nex.js apps. I need to add Astro support. I'm really interested in the idea of having Astro as dedicated marketing sites because they're just blazing fast and I want optionality. Um, other stuff that I need to add into there would likely be just some sort of uh like age top or PM2, but something that you can know it's like the maintenance support. So that's a different topic, but like monthly maintenance. So you're making sure that your box is always up to date and that all the containers, like if Superbase comes out with something, then you want to make sure that you're on the latest version that there's no issues. And then monitoring. And for anyone that's made this far and knows more about this than me, what are you guys recommendations on this? because I've just done self-hosted stuff for like single apps and now I'm doing this one and I'm planning on having um the ability for people to take this and then duplicate it for every member that joins school. They'll get access to the repo. But I want to know what the best way to do this would be for both provisioning them their own ones automatically like having it like spin up the box so that like as soon as payments received for this offer this thing's done and they get access. So I need to figure that out if you guys have ideas and then obviously on the monitoring side I'd be curious. So please please comment below on that. And then what else? Yeah. So that's the plan there. And then the road map. This is for the road map for the media content media stuff that I'm doing. So I think that I'm I'm basically treating this like it's a Twitch channel where I'm going to be live streaming slashrecording most of the stuff that I do. So you can kind of see where it goes and I'll play around with the idea of like showing the numbers. What's been interesting is this channel's been up for 23 days and it's acrewed 100 or 290 people which is awesome. But then including revenue metrics and all that is something I'm playing around with. Um and then I'll I'll be building basically different offers within there. So behind the scenes stuff for the media brand that I'm building which I'll be really excited about doing and then behind the scenes stuff for building out map which is this product which is multi-agentic. It has five tools in one. So you have your notes, your tasks, your health, your chat, and your calendars. And then each one of these has sub agents within it. And then there's a master agent that runs the whole thing. So that's that. And the way I'm doing it is basically by having a set schedule. So it's like Netflix style content for nerds. And for me, it's awesome because it when you put yourself on a schedule, when I make the thing, I have to do it. Like I have to stick to my word. So, I'm going to be trying to push myself to do the prep ahead of time, which is the research, and then the development will be the episodes, and I'll be doing them live so people can join and ask questions. And um it should just push everyone forward, which is great. And then the home server, like the agent server gets better. Um people that have questions about the agent server can then get those answered. We can work together on it. And then there'll also be this like road map thing that I'm making which is just like what content am I consuming off these videos, you know, besides these little news things that I think would benefit people that are interested in getting more technical in the AI era. So even things like hacker diaries is really awesome to like learn about security. Um, and I think I need to wrap this video up, but in the next one I will, yeah, I'll likely wrap this thumbnail to prompt the Vertex AI into the video that we got the comment on because that would be an example of a container that's tied to an app where it's like if I go to Vertex Vertex and I go to home server, you can see whatever. But basically what it allowed me to do is take a prompt have prompt variables line those prompt variables up to a form if it's a reverse engineer product where you want to get a a PRD based on reverse engineering an existing product you answer like six or seven questions it'll generate that and give you that as an output and invertex you can make those prompt variables and you can just oneclick deploy it uh as a container so I think it's going to be really really interesting and I think it's going to be really helpful if you guys want to join and be a part of this journey of building the agent server. Make sure you check out the link below for the school. You can join there or request to join. We are going to be going live with it today. And make sure you like the video. That's so more people can see it. And it's like your way of basically saying, "Hey, this video is good." So like the video and uh I'll see you guys in the next one.