Vibe Coders and Marketers WILL Do This (Get Ahead Now)

I'm Parker X. I led and scaled technology for a startup that sold for about 23 million bucks. It was right around when AI came out. So, I went ham over the last few years using AI to build all sorts of different startup ideas. Some were awful ideas, but I got really good at using AI. And now, this channel is where I show behind the scenes of me scaling an AI services business to 100K a month in profit. We typically cover the same sort of things. This is going to be a juicy episode. So, let's just get into it. Yeah. See, this probably got your attention. That's an AI influencer. Isn't that absurd? We're going to talk about that, but just wanted to kind of gaslight you guys. That is. Yeah. So, we got some questions that came in and I went ahead and I answered one ahead of time. So I just this the question that came in was basically like if you were to start from zero as a vibe coder. So you're here and you see this promise of the billion dollar company that's a one for your vibe code idea. But then you get your first version of it out which is awesome. Like congratulations that's sick. Then you might hit another wall and another wall and another wall and you end up having issues because you're not aware of how the fundamentals work. And I'm not professing that you need to go and learn every single thing. But there's a way to leverage the absolute crap out of AI while you're learning new concepts. And I wanted to share this today because I'm doing something that's new to me, which is self-hosting a bunch of different containers, like dozens and dozens of them. I've self-hosted small projects, but not at the scale that I'm doing now. So, I needed to brush up slashlearn some DevOps stuff and I wanted to record how I did it. So, there's a really good saying, if you can't teach it to a friend, you don't get it. It doesn't need to be so dumbed down. This isn't like the Richard Fineman level of learning something, but if you couldn't explain it in simple words, then you don't understand it. And so what I found is if you're learning something new, in my case I was trying to figure out something with Caddy and a bunch of stuff with Docker and and Superbase and stuff like that, you can use this prompt which is really really helpful. So I also noticed that I'm not covering the screen. Explain this to me. Give it context of who you are and then what you're capable of doing. So, it's a directive, a task, and then a piece of context, and then an additional piece of context. Now, there's variations of this, but what I typically do after this, like I'd put in uh like the concept that I'm trying to learn, and then I'd paste in some documentation. This same prompt can just be like blown out to actually working in projects. So, let's say this is level one. Level two, you want to learn more, find an open source project that someone did. You go to github.com. Yet another one. Let's do that. Okay, so I found one. There's 14,000 examples. There we go. So you'd find 14,000 examples. Of the 14,000 examples, surely there's a project in here. You can then sort it by popularity. Find an open source project that's done it right and is popular and then you can see how to learn the thing. So that's a long long explanation, but then you would ask, hey, explain it using asy charts. I was trying to figure out cash in depth today because I'm a scrub. And so what I did was I started asking those questions. Then I said, explain it in asky charts. It showed me a bunch of asky charts because I just always have used the stuff in next. I haven't had like to really know like I've used Reddus before but I haven't had to like explain to a friend and so now that I'm going to be use caching more and self-hosting and stuff you know it just takes you a little bit deeper into it. I was using GPT or sorry Gemini 2.5 Pro and then you know you're asking questions along the way like I didn't know what a pooler tenant ID is. So I have three things going right I've got the codebase that I'm trying to figure out. I've got this thing asking questions. I had documentation open and I had the little chat GPT companion flying up and answering questions. So, that's how I would do it. Let's see. The second question that came in was someone talking about agents and I was excited to answer it. Actually, this would be a channel that I'd recommend to this person I'm about to answer the question for, but NDV Dan, shout out to him. He's got some interesting concepts, but I also think as much as I like his content, He is overengineering things for what won't make sense soon. And uh let me explain a little bit more. So this guy says, "Still watching the video and just found you." Thank you for still watching. But I'm deep in creating multiple 6 plus agents directly in Python, not using n type workflow software. Cool. I've done that, too. I said agents because I only just stumbled upon that term. Dude, where have you been living? You under a rock, man. Did you just find out about that? This is the the era. This is where it's going. And what they do for integrations of AI. Yeah. Could just be a for loop. I've been trying to develop mine for several months now without even knowing the correct term to call them. Anyways, looking forward to getting value out of your channel. I love pushing Claude. Thanks, dude. That's a nice thing to say. I love pushing cloud beyond its limits with direct API programming and not some gimmicky drag and drop interface that has limits. Boom. This is the part this is this is the fun part. I've been there where I'm going to just pull up a meme and this is this is us together. This is us. We're huffing some copium. We're coping so hard because we see a drag and drop thing and we're like, "Oh my gosh, like I need to jump in Vim and I need to use the hotkeys and like if it's not in Vim, see that C11 clang, whatever. If it's not like handcoded, then it's not legit." Well, no, that's not true. There are so many things that you are probably coding that you should not be coding because you could take that leverage. Coding skills are leverage. Like if you know, Legion 9396, you gota leverage leverage up, you know, like you you just spent months writing these Python things. If you haven't seen what's out there, there's probably a lot of stuff that you didn't need to do. Like for instance, uh let me see. I think it What was this thing? Hang on. I think there was a project that I can pull up called uh what was it called? I made a Maramo library like a prompt library and it was all in Python and it's like Jupyter notebooks on steroids. It's just literally for doing it. But like in that project there's a bunch of agents that are trained that are just at the end of the day it's Python functions like whoop-dedoo and a lot of them are repetitive and then you're just chaining them together and then either like looping through the same function. Oh, this is the critique agent. And then it loops around and does it like an agent is just an LLM that has access to tools and then the alpha is in your ability to prompt the crap out of them in the right way and get really good outputs. No one cares if you wrote it, if you drag and dropped it or whatever. They just care did it do the thing, if it's a business, did it save them money, make them more money, or save them time. Those three things, those always sell. So, I don't like I get it. Um, and I also am in the same place as you where like yeah you do just make API calls but like you can do that and make you can do that in N8 for linking and gluing things together. So there definitely are limits. That's why I'm I'll talk about it a little later in the video, but it's like why I'm setting this self-hosted setup is because I know how to do both and I want to see what happens when you have the flexibility of all of them, right? It's like, okay, I have an N instance that's self-hosted now. I have a CDN if I needed that, but I also don't. I just happen to be in this template because I know Postgress better. So, I'm gonna use their vector thing. And it's like, oh, like you could use websockets or you could use the freaking like superbase version that's built on elixir. It's all real time. It all does the same thing. If it accomplishes the same goal, then like just go with the one with the least amount of resistance. So, make more agents. Uh, what kind what kind of agents? Covered this question in tomorrow's upload. Thanks and look forward to hearing more about your dope agents. Yeah, I couldn't agree more. I I think that's a big theme on this channel is just talking about agents and actual business use cases. So, can't wait to hear about what you're building. Uh, okay. This is just more rant on the whole like learning thing because this question just keeps coming up like perplexity search from GitHub examples. That's how I solved my problem today. I was like hand rolling all this server stuff and then I found someone but it was after the fact that I tried to hand roll it. So when I watched someone do it I wasn't spoonfed the answer like I knew where they were going with it. I was off by like a degree. So if you and that wouldn't have been the case like if I didn't do the stats before then I would have no context and I'm just literally sitting in a lecture hall. And so then I made this little chart where it's like from left to right. If you get a walkthrough of someone holding your hand, patting you on the back, and showing you every little detail, you're not going to learn that much. If you do it yourself, you learn the most. It's painful. It's going to take longer, but you're going to remember it and you're going to learn how to learn. And then AI is down here. Like it doesn't even deserve anything. You don't learn anything. So yeah, leverage it to learn. Don't leverage it to do the thing. Now, onto the news section of the daily upload. So, Grock is I made a comment. If you're not following me on Twitter, I post a bunch of hot takes in there, but Grock blew up the world with their exciting model and I made a bunch of memes about it. But what just happened with the Gemini release is now Grock 3's API is not going to go launched till Christmas. Santa Claus is bringing Gro 3 in the API. And I say that jokingly, but it's like I don't think they're going to release it because Gemini just stomped on everybody. And if you're not familiar, I was watching Theo. He's the guy uh if you watch my content, you probably know him. I've learned a bunch from him about random React stuff. But he had a good point where it's like the advantage that Google has like they own the chips, they own the science, like they've been investing a lot in that the science part of that like the research and then they own the data. Most companies, they just get one. Like if you're open AI, you know, you can go watch his whole video on it. It's way better than this little dinky Walmart version of it. But if you're open AI, you just get one. And if you're like Grock with a Q, then you just get one. So Google just woke up and they're ripping. And so yeah, Groc, it's in Telegram now. That's cool. It's a nice partnership, more distribution. I think they're on similar vibes with like freedom of speech and stuff. Then if you are Patrick Star living under a rock from Spongebob, then yeah, you also don't know that OpenAI image gen is having a moment. There's literally like anime versions of everyone plastered all over the internet, which means LinkedIn's going to find out about it in a month and then your grandma's going to see it in like Q2 of 2026. But like the image gen's really good. Like it's crazy good. So check it out. Thousand $1,000 and free credits to Google. They can do this because they got FU money like that. They print money by selling blue links. And everyone's like, "Oh my gosh, they're going to like fall off a cliff because they don't have blue links." Uh, no. They they figured it out. They have better stuff than everybody. Quit hating. And if you go sign up, this is actually it brings me to this other topic. On my self-hosting adventure today, I bought a server from a company called Netcup. Horrible name. Just sounds weird. And it was because the nerds of Reddit, the neck beards of Reddit, told me that that's the best version to get. And it's literally all in German. Of course, you can hit translate to English, but they just like the whole dashboard. It's like written in PHP. Like when you see the PHP thing, you're already cringing, but it's like the most confusing thing ever. And then now I've I'm figuring out how to just run on the Google stack. I'm all in Google. I'm going to start wearing one of those stupid hats. But yeah, Gemini 2.5 Pro obviously came out. Check it out. It rips. It's free. I have this prompt that I'll share. Um free. Yeah, it's free if you just drop a comment below for dope prompt. Actually, don't don't say that. Say rubber duck. If anyone comments rubber duck in the comments, I'll put this prompt in. It's not even for coding. It's just for life. It's like talking to someone that is way smarter than you and just cuts. I'll just show you part of it. It's so I don't know. I was using it today and it's like this guidance prompt and it's you are a hyperrational first principles problem solver with zero tolerance for excuses, rationalizations or BS. Pure focus on deconstructing problems to the fundamental truths and it just like I was giving it my plan to 100k in monthly profit and telling it all these things I was doing and I didn't give it enough context. So, it was tearing me apart and I was about to throw my laptop against a glass ceiling because I was so angry at it, but it was true. But it didn't Why are you angry? It was making me mad. Sorry, we got the peanut gallery here. But it was making me mad because it was pointing out things and flaws in my strategy and I'm really competitive. But then we came to an agreement that my plan is good. It just didn't have context. But this thing, I'm really happy with it and so is she. So, Vibe Marketing, this is the last yap sesh of of news, but I've been talking about this on the channel and this idea of marketing orchestration. So, every job, you know, let's just say 300 million of them are going to be entirely displaced by 2030. Kind of crazy. And you have this whole thing that's called like reskilling. And reskilling, here's an example. So, let's go back to the agric agrarian revolution. So, I'm not going to try and drop a date. Maybe 1817 or 1980. No, 1917. It's some date. I don't know. When they invented the tractor, you had all these guys out here that are used to carrying around the pitchforks and the rakes and they're working the field. And then the tractor came along. And here's John. 1892. 1892. You have John Deere over here and he shows up with his first tractor and he's like, "Yo, everybody." And when he says everybody, it's literally like thousands. Like I it was some crazy number of farmers made up GDP for the country cuz like I don't know that was just the job. And this guy comes around he's like, "Hey, I can do like 10,000 times more than you can." And everyone's like, "What? No." And they get mad. They deny it. They think this is evil. Kind of a similar spot. Everyone is the desk jockey over here. So like the neck beards of the world and tech and then like all the marketers and like anyone with a desk job, no offense, all of you, myself included, except that's why I'm making all this content. That's why I'm like learning all this stuff. You're going to have to figure out how to leverage this stuff. There's layoffs already happening. Like it's a thing. But anyways, the way to avoid that is you reskill. And so what's happening now, and you know, first it was vibe coding, right? And I talk about that a lot. Now it's vibe marketing. We just throw the word in front of it. But what it means is it's just leveraging a bunch of automation and a bunch of fine-tuning on models to sound like a brand, sound like a person, and then taking a core piece of pillar content, i.e. this thing we're doing right now together 20 minutes in that taking this the voice for it the tonality the voice inflections the pace the way that I speak the words that I use and then turning that into marketing campaigns across email across Twitter LinkedIn Facebook Instagram Pinterest Blogger Medium you you know literally everything and it's done through NAN primarily as like the orchestrator you You know, it's literally just glue. You could go hand roll this yourself, but have fun wasting time. Just self-host it if you want to be nerd. But it's starting to pop up and you know it's going mainstream when like Forbes is talking about it. That's kind of funny. But then I look at this and I look at the Google trends and I'm like, "Okay, well vibe coding just popped. Now, this vibe marketing thing is gonna literally and like give it I'd say like majority of marketers that are in ecom are not on X. So, I'd say this, let's see, this is the recoil effect on X. But since they're not on X, since they're on like LinkedIn and that's where all the corpo slaves are, then they're probably not going to get hit with this until till like July. I'm thinking like this Vibe marketing thing, it might get another rebrand, but I'd say by Q3 at latest, this thing is going to pop. So, I'm going all in. And by I, I mean agents are, not me, but I'll be orchestrating them on owning these keywords. When people type in Vibe Marketing, guess who's going to show up? Me. This guy right here. I'm going to be all over these keywords. I'm going to be all over Vibe Coding, even though I hate it. I hate those words because I think I like Andre, but I just the whole like everyone using a training wheels app. I don't know. It's a good net positive thing, but it I only dislike it because it's like a like tricks a lot of people. Like they think that they're doing it and then they I don't know. I've seen like bad cases. Maybe maybe I'm being too harsh. So that's what's happening in the world of AI. And I'm going to be leveraging all these things both for brands and like doing buildouts of them because if you go on any job board, everybody's trying to figure this out. And if you can build these things, go figure out how to do it because the contract size like minimum contract size should be 10K. That's what I've like figured out. It's like anything less than 10K per contract is stupid. Like total waste of time, especially if you're qualified. Now, on to strategy. So, I want to talk about some self-host drama. I mentioned that earlier. Just uh yeah, I guess it's really like figuring out the right provider shouldn't be a problem. I'm going to end up going with Google. The whole point of it is to have this 10-man agency. That's just me. But it's kind of this like funny thing that's going to happen where there's agencies like this this really just like agent. Agent is like a fluffy word because like if you look at the word agent right now, you could think of like uh like this kind of like scammy dropship boys. So, if you're familiar with that era, and I do BOI because that's lit, it's what it is where it's like you're starting to see that like lowbrow cheap get rich quick schemy style content coming out where it's like a guy and the main part of the thumbnail is like his Lambo and I can't draw a Lambo, but you get what I'm saying. And so there's like a bit of a negative connotation around agent, but then there's like the real side, which is you have multi- aentic workflows. And this is what we're after here. And this leverages many third parties. It acts as glue. It has uh DBcontext. It has multi-model and multi-modal capabilities. So model meaning, okay, we're doing something that exceeds a million tokens and we don't want to go bankrupt. So don't use entropic. Having that logic built in modal means okay is it an image is a voice is it text which medium are we going from and into and vice versa and then having this as like this is what I think of when I hear the word agent is like something more in this camp and I think this is the future and it'll encapsulate the use of all the amazing software development kits that are out there at our disposal. So it could be the agents SDK oops SDK. It could be that combined with the AI SDK because these things see updates very often. It could be using any other number of tools. It could be hand rolled. It would include eval to know like when things are going on. And then another big one would be the hit or human in the loop because these things suck at first. So whenever you're going to do one, just do it manually first and output whatever the decision that's going to be made to yourself, especially if it revolves around like payments or customer communication. Like you want to dry run it. You want to see what it was going to say so you can either feel awesome about the thing you made or be so glad that you did not send the thing or have it make the decision. Um, I guess like the last one would be like when I think of agent, it's it could be like these like oneoff utilities, which is fine, too. Slash scenarios. And that brings in why I like this idea of a nodebased editor because if you I don't think we're there yet, but this is where I want to go with what I'm personally building for what I'm calling the agent like team which is behind the scenes. It's just a home server. It's a server with a bunch of stuff on it. VPS rather which yeah is that but it has a bunch of servers on the server. Um but on here like with nodebased stuff in current you know March 26th world you can do like your nad thing and it's cool it's got like 400 different modules right and I think that's a good start but this is the FisherPrice version of what will exist soon where the next question is okay well if you can like run some scenario in here you can have multiple scenarios of nice and cool. But if you can have that, but then one of these nodes is your codebase, not just a file, but a code base, that's it. Like you've cracked it. Because with that, if I go and I say, "Okay, cool. This is the like Parkerx.com codebase." And I can go and run a set of things before this and then go and have it do the work and not have it shipped. Like if I had a form or if I had the ability to literally like let's say in the case that I want to use a voice note while I'm at the gym to generate a feature and I have it in my head. I know exactly what to do. Like it's not just like go build tic-tac-toe or a calorie app. It's like no, I want the ability to have a stereo from the PNG that's in the public images folder. I want the stereo to pulsate and I want it to rotate on the third second and I want it to pull the songs from this playlist on my SoundCloud and I want it to play and then when that's playing use frame or motion to shift the things on the page like whatever. I'd give it some really detailed ding. Then I'd also have like the information about the files and then file uh like line numbers in my head. Then I could like kick that text off. So it would go from text and this is a horrible example but it should uh explain the point. It would go from text to your architect. Your architects be like, "Dude, that is so effing dumb. Like why would you do that? This is how you should do it." And then you guys would have your intellectual sparring. And then once that's happened and you've looked through a couple of the specs, then it goes off to the implementation engineer, right? We're not there with designs, but you get the idea. It's basically like you'd have these steps from anywhere. So it could be a text, it could be anything that can shoot an HTTP request, then reaching back into a codebase. In my example, this isn't a SAS product, but it could be. And the better you are at figuring out how to plan these things and knowing the underlying tech, the better these will be at their jobs. And you'll be able to see around those corners. Like, I don't care how good an AI model is. Even if you make a piece of dog crap with your like vibe code stuff, it's like it'll probably it will be able to fix it. the model is going to be so good, but you won't I just you won't be able to like see around corners like you can when you're more technical. So then it could go and it would just loop through and continue to try to do the thing and then fire it back. And so here's the preview link on REL or on your home server. So I think that's where it goes. I think it's going to be sick. And that's just in like vibe code territory. Now think of marketing. Okay, let's think of marketing. And I want to have it where I literally have a canvas because it's easier. And I can have code bases that are nodes on there. But I could also not I could also just literally go and say, hey, like I want to fire crawl a niche of ads from the Facebook ad library because I want to drive more revenue to my business. go get the best top performing ads that are live and have been live for more than 30 days in the workout niche. And so you'd say to that into maybe a prompt and this is literally something you could do today and then that would fire off to firecrawl. It's going to come back with your 100 images. It can download them all, store them. It can then train on them. So you can have a fine-tuned model trained on these best practices whether it's this stupid way of doing it like oh cool like yank all the words out of it for direct response copy or oh I want you to do a batch job um on OpenAI more expensive but I want you to do that so then we have a tuned version of it need a lot of images for that but you do all that in like one thing and then you just name this the image trainer or something easier here, right? But this node would have a couple things in it. It's like, "Okay, cool." Like, it can download stuff. It can retrieve things from a niche. It can retrieve things from any page. It can then go off to the trainer, whatever. So then you'd end up like in Vibe Marketing World on this in this um hypothetical, you'd be able to then chain that and say, "Okay, like we need someone to go and run the ads. We need to have a budget." like this would be for a paid media play. The much more obvious one is a content pipeline, but I'm not going to get into that one because I've explained it a lot on this channel. So, I think that's it on the 10man agent thing. Oh, actually, I just needed a good gug of the nectar. That stuff's so bad for you. But yeah, I just wanted to kind of verbalize those topics for you. A an agent actually has a lot of things underneath the hood. So like Perplexity has like a really good open source version of it. doesn't do nearly as much as the real one, but in order to just go and retrieve search results and then display them the way that they do it. There's seven or eight different little agents which are just tools that have capabilities of using large language models and they work together and they check on one another. So, did I get enough information about this query that they made? They asked about the Tesla price. Well, did they mean the Tesla price in dollars or Canadian, US? Can we infer that? can we just run with it? Um, so there's like a split in the road. It ends up just being like a set of nodes, right? So, might as well just see it as it is. Those are going to exist for all these topics. I did a really funny example, decent example of one in marketing for paid media. The sales one's going to be really interesting. I think there's going to be growth ones. I talked about that where you can just have automations that go and engage with people with a human in the loop at first. Fulfillment is going to be an interesting one. Actually delivering on the code. I talked about this with someone in our paid community which goes live in two days. Um and it's going to be expensive soon. That means you should join now. But let's see. For the fulfillment stuff, it's an idea that I had would basically be like fix my vibecodeapp.com and you'd have just a react hot draw drop zone component here and then people could just drag and drop their project onto there. And there's a lot of tells for vibecoded projects, especially if they came from lovable or bolt. But the way that they're bootstrapped, there's things that would not exist had they just bootstrapped it themselves using npx create next app at latest. And so you can like go and check for those things. You can check for the common flaws, which is like, oh, how are they handling cookies? It was trained on next 14.2, not next 15.2, so they need to wait the cookie. There's a lot of these like you'd come up with like a list of these things. And then when they drop it onto there, you have a little check box that says like, "Hey, like I'm cool with you reading through this." It would go in and run it basic checks and then try to run it. If it didn't match the stack that this thing was trained to go and fix, then they wouldn't be qualified. But this would be a really easy way to make money. It's like, okay, you start ranking by doing Vibe Code uh master class or master guides, step-by-step guides using an SEO agent, and that kind of pipeline. You'd run it through a plagiarism checker to make sure it's not getting dinged. You'd set up your JSON LD and your knowledge graph and your metadata and your crawlers and your robot.txt and all these knowledge graphing, all the things that you need to do for off- page and onpage SEO. Set up a pipeline for SEO. So, it's cranking content. And then you'd go and you'd have an auto replier on the cursor forum, on the Reddit forums, on all the newbie forums, just answering questions that are like the FAQs on the homepage. you to have the FAQs and finally if they like when they get to your site then yeah you hit them with a oh yeah we can fix this and then boom stripe link generates collect payment if it doesn't run and you have already collected payment then yeah you go and take a look but this is where it's going folks 100% customer success management too it's like if you look if you've run a business before and you've been at any level of scale you know that like NPS is a big one and the way it was done before was you might have like a sequence ments and some growth marketer somewhere is trying to figure out like how do we keep them in? How do we keep them in? Do we surprise and delight them? But it was still manual and it was if else statements. Now you can do more personalization. You can glue things together much more effectively and it's just going to be absolutely nuts. Uh qualifying and disqualifying. This is the last bit of strategy. Just was thinking about this a lot today when I was using that prompt. And when you have a goal that's around profit and you have the offer to do more work, you start seeing the dollar signs and you forget the quality of the work. And so this can be on either side of the coin. It could be someone that's like going to get you a 25k contract. You're like, "Oh, sick." But then you didn't take enough time to qualify it. So then you end up like having no leverage on the work. You can't reuse any of the work. It doesn't give you the time that you need to do other stuff, but you're like, "Oh, it's 25K a month." Same thing goes for like the smaller ones where it's like, okay, like a lot of the best sales teams that I've seen, they rush to disqualify. So, it's like, will this client push my skills forward? Is that a thing? If not, probably not, great. Will this client open doors to higher quality clients? If not, see you. Will this lead to a great business relationship where it's going to go up and up and up and up and there's a high ceiling for a long-term either it basically will be high retention and high lifetime value. So all these things are just things you start weighing against one another. Um because it might be that had you instead not taken on the extra like little, you know, let's call it like a little 5k contract that you think is going to be I can't even spell. I was like, "Oh, I'm going to record at night. It's going to be so much better." And I'm like tired. What do you know? But it'd be interesting because like you might trick yourself into looking at one of these small ones be like, "Oh, it's just 5K." Like that means it's going to be really easy. But then even if it is easy, it doesn't mean that it's worth it. Like it could mean like a lot of back and forth and all these things. So I'm just thinking a lot about the disqualifying part now and then like how can you get the most leverage out of each deal. And then lastly is the build queue. So let's just pull up where we're at. I mentioned this earlier, but I have a basic setup for Superbase with their full stack. I think they have 11 different containers and it's really nice because it has vectorization stuff if you need it. It has Postgress which is amazing because you can do PG boss if you're not familiar. It's awesome. You can do workloads, you can do uh triggers, you can do all these things off obviously storage, but I'm not going to be using storage on this most likely. I'll use a GCP bucket. Um, and then I'm going to go and set up something on Cloud Run and give that a shot cuz, you know, $3,000 is cool. And then you can have the CI/CD set up. I'm just trying to find. They have these conversational agents that they're trying to get popular. So, let's see if we can't find something interesting in here. Let's do for team of agents. I haven't set this up yet, but they have a bunch of tools in here. And I clicked one of them. Let's show the use pre-built agents. So, we have DMV. Why don't they just go sell this into DMVs because they need help? But in this use case, it would allow users to renew their driver's license online and book appointments at local DMV offices. Features generative AI requires API integration. I haven't played around with these, but it is interesting because maybe there's something here that you could just tweak where it's like, okay, well, I'm imagining DMVs. That's appointment bookings. So, that's probably an automation someone could go sell for 10K. Not sure what's involved. Let's take a look. Uh, can you help sample utterances? Let's just see what it does. And we're going to deploy it in Iowa. It needs cloud logging. $3,000. This is too funny. They're really pushing these. They must not have any users, but it's a relatively new feature. I haven't seen it. If anyone has used this, let me know. Drop a comment. But okay, in conversational agents, you can build AI agents that chat naturally to serve your needs. You have an agent graph to get a big picture. Is this a is this a node? Wow. Called it. Nodes are everywhere. So, yeah, you get to see it. And then I can open up the task. Boom. Here's your instructions. You have conditionals. You have tools. Oh my gosh. It follows the open API protocol. That's nice. Little mini map. It's all about the nodes, guys. You didn't believe me. Close. Goodbye 75 startups with this. Gosh, I'm so happy. I've tried to raise money for an idea like this like a year and a half ago. I would be like crying right now if I did that. This is great. Yeah. So, each node has a set of specific instructions and you can add tools. You can add data stores. Very, very cool. What else? Evals, versions, integrations. What can we do here? One-click telefan. Okay, that's pretty cool. So, I could build that thing I was talking about earlier. I can make a conversational agent for building a new feature of a pulsating speaker to play jams on my site. That's all I ever wanted. Yeah. So, I could call or text. Interesting. Very 3D avatar environments. And then that's the thing too is they just have the whole ecosystem. So you could spin this up and then have it be in a container and it's there. It's live. It's going to scale to infinity. Um I'm just curious. My only qualm with this is like how oh how specific can you get? It looks like you can get pretty specific. Let's see the samples. So that's Oh, interesting. So you define their API for the tool. Very cool. Yeah. No, that makes perfect sense. Okay. Conversational profiles surface real time relevant information to agents while they talk to customers. Whoa, interesting. That could be like a CRM where you're porting in data. Yeah, there's this is this is something worth checking out. Yeah, I think if you just go to Vertex there, it just should be on the homepage. I was curious actually if Yes, it's in preview. What I was curious is if you can use like I remember I spun up like six Google accounts because they were giving $300 away for credits and I just used them all on Claude, but I don't know if they're still running that. Let's Let's just see if we can't get you some cash before we end this and me some cash and then we'll wrap. Uh so cloudcon now Google Cloud Admin Google Cloud Console Platform. I was on this site all day. Come on. Oh, here's a workflow I made. This is a content thing. Basically, it's a bucket. It's got pubsub. that's listening for something, drops a YouTube video in there, does some magic to it, makes it sound way better. This mic sounds awful, so it makes it sound better. And then it drops in an HTML file with a short summary, a long summary, transcript, some other stuff, timestamps, and then there's a pub sub that uh publishes that info. So, you have a web hook listen. Um, but let's see. I think it's in model garden. So, you go into Vertex AI and then you go to model garden. And I don't think they're doing it anymore. I don't see anything about credits. Yeah, because before if you had the credits, you could go and then you could enable it and it would be like a popup. But unfortunately, it seems like they've taken that away. So either way, this has been fun and I appreciate you guys watching. If this was helpful for you, just me yapping, make sure you like the video and subscribe. Honestly, it's more important to me that you guys leave questions because those are fun to answer. I really, really enjoy it. So I'll see you tomorrow.