Vibe Coders Should Do This (Learn Faster, No Training Wheels)

Hey, I'm Parker X. I led tech at a startup that did $73 million at one point, $47 million at the sale, sold for 23 million. We'll cover that one day. Very dramatic story, but in this channel, I go through my strategies for growing AI services company to $100,000 in profit a month. And we cover a couple things. to cover news. So things that are relevant to actually making money because there's so much AI news. You can go get some spam from somebody else, not me. We go through strategy of things that I'm doing, things that I'm testing in the field, whether good or bad. And then we go through the build pipeline. So things that I'm building. So without further ado, let's jump in. Had some piano going. I don't know if you can hear that, but now it stopped. So, let's start out with Q&A. We're getting some questions on this channel, which is cool. I think I've been posting for a little bit over two weeks now. And so, the first question that came in is from Aiden, and he says, "Thanks for the awesome content. Our working session was incredibly helpful." Question for folks like me and Itchy Chicken. Shout out, shout out Chicken. Kind of hard to say. Should we stick to learning coding with tools like cursor client or windsurf? I've noticed I'm getting pretty good at prompting. Cursor builds exactly what I want, but I'm not really digging into what's happening under the hood. Do you think that's okay? Or should we ditch the LM crutch and write code manually, even if it slows us down initially? You touched on this during our session, but I feel like it's a question a lot of the Vibe coding community might be wrestling with too. Great question. And yeah, that was fun. and we had the working session where we just kind of chilled and we were both working on stuff. But the way that I would do it is the same way that I learned which when I was starting to take coding really seriously because I've started and stopped learning a ton of times. So let's just go here. So I'm 32 and I started coding started scripting when I was like 14 and this was just really simple stuff hosting a CS1.6 server. So it's like changing configuration and learning how to run a server with FTP. But then you know there's an 18ear span there. Again at 18 I said I'm going to learn HTML and CSS. So, I did that with Treehouse, which was like a popular learning site. It's probably around still, but it sucked and it was grueling. And then again, at 22, I tried to force myself how to take classes on like linear algebra and all these classes that I had access to through a colleague who went to Brown. And so, I was getting like fundamentals of this, but it never really worked. Like, I did all these other things. Then at 27 I did basic Nex.js. So that involved React. But again all these were failed attempts and not you know that failure is a bad thing. It's just like hey it never really stuck with it. It never was fun and that was the issue. So I think one beautiful thing that's happening now is it's really becoming fun because you can go and you can learn fast. But then when I started getting good at it, so that's when I was 28, I was starting to actually make things that I wanted and that was the key for me. And then I think 29 was when uh 35 came out. Basically, if you can frame things for solving your own problem, then you will push through the inevitable struggles. So, if everything is through this lens of solving your own problem, then you will push yourself to do it. And on the LLM subject, I think that I would do how I learn now and how I learned and how it stuck, which was it didn't involve using any sort of composer. So I guess yeah, for noobs, this is what I'd recommend recommend. For noobs, I would say you have your docs of your stack. This is one resource. You have a stack that you define. You pick one and you don't jump. And in that stack, pick one that's popular. Pick one that is tried and true and doesn't change every single day. Don't jump on something just cuz someone on Twitter is is saying to do so. But the one that I did and how I felt the pain was when I really started taking it seriously was right when Next had launched uh something new. It was called the app router but I started adopting next and I was learning one way to do it and then they came out with a new way to do it. So I really like halflearned one part which was called the pages router and then I had to go and basically start again. So I would pick a stack that is well known for me. I kept with it and I kept with next. So that's a stack, but you'd pick one of each of the following. So you'd have call it a meta framework. You'd have a language or like a framework. In this case, it's React. In my case, it's nextJS. And that's I think it's like a 15. know you'd pick a styling library slash uh setup. So in my case it's tailwind CSS three not four and what else? Shad CN UI and lucid for icons. So that covers your bases everywhere because tailwind is the stuff that's inline. You don't know how to write CSS, that's okay. Like you shouldn't have to learn how to do that. It's kind of a moot skill. Like just so you know, if you wanted to style a box, and this thing was just called box, you would have an HTML that's drawing the box, like an HTML file, and then you'd have a CSS file, and you'd link the file back into HTML, and then you'd do a selector, which is like box, and then you'd have to write like background dash color space, you know, fff or whatever. Whereas in Tailwind, it's just all in one file. So you're like box and then you pass it this thing called a class name and you just do it there. So you'd be like bg black and that was tailwind. So yeah I would pick one of those. I'd pick a database as a service just to start and I would always pick superbase because it includes a whole lot of stuff and you can outgrow if you chose to the self uh the cloud hosted stuff and then self-host it. So the room the ceiling's awesome. So I'd pick those and then I'd pick a way to deploy it. So you could do Bracell or Netfly, Netlefi or Cloudflare or GitHub pages, whatever one you want. So now you have all those. You have your stack. So this would be a nice web stack. If you're doing iOS, it's going to look totally different. You probably wouldn't use any of these. You'd probably, if you want to do native, you do Swift or Swift UI or UI Kit. Pick one of those. In the case for web, I would pick one of these. I would then figure out what it is that the problem is solving for. So, this brings in the docs part where it's like, okay, uh if it's Stripe, I need to learn about Stripe. If it's something else, I need to go look at those docs. And you don't have to read them because that's really boring, but you need them as reference. No developer has them learned perfectly unless they've used the thing a lot. Like for instance, me, a problem I was trying to solve had to do with calendar. So, I know the Google counter API docs by heart. I know all the methods and all that, but that only comes from practice. So, now you're like, "Okay, cool, dude. I picked a stack. Like, I just want the answer." Well, then you have your problem, you have your stack, you have the docs that are related. So, then when you're actually in practice, what I would do is I would treat the composer as a bad habit. And it sounds crazy, but it's just one of those things where like if you can just use chat mode half of the day and then the other half is composer. Like the the less that you use the composer or the the agent, the more you're going to learn how the fundamentals work. You're not writing it right, but you're at least like giving yourself a chance to say something that's right. It's about the topic. Because if you just say like make this thing and then it just runs off into the distance and does all of it and you don't even know how the thing works that it's spitting out. Even if it does work, then you can't make additional specific changes or if an issue comes up, then you won't know how to fix it. And a lot of people will say, "Well, hey, these models are going to get better." Yeah, they are. But guess what? you're not going to be as good as a person who does know how it works. It's just how it works, right? Like if you know how functions work, if you know how databases work, if you know all these just basic fundamentals that get boosted by having chat mode, then you can see around corners and you're more technical. The more technical you are, you will be able to leverage more agents in the future. So right now, the way I think of it is because I'm technical, I can have all these little agents, which are essentially employees for pennies, working at a high level of leverage. So in this case, I'd get a 6x multiple on my productivity if these things were like uber efficient. It's not going to be like that, but as I make them better, they will. Whereas if you don't know how to leverage them, if you don't know how to train them, if you don't know the task that is the thing you want them to do in and out, then you might just have one and maybe like one and a one and a half. So maybe you'd have like a one and a halfx multiple. So yeah, I think it's just a matter of of actually asking it. Treat it like a 247 tutor. If you're using a split screen, I always do it like this where it's like I have a third, a third, and a third on a wide screen. So, actually, it would look more like this. Boom. Boom. And so, I might have the website I'm building and then the ID and then the docs and then the IDE. You have your little chat and you're asking a ton of questions. It's just so much better that way. Um, I think that there's a best practice as well for any LLM project that like let's say you're up and running, you're you're using chat and you're learning, then as you get more comfortable, sure, go use the agent, but then I would also use what I call like the aid link method, which I'm going to make more videos about because it's working really well for me, where you have eight hours to work on something, you should spend 75% of whatever time period planning and 25% executing. So let's bring that down to 10 minutes. If you have 10 minutes to make a feature, 7 and a half minutes should be going towards the planning of that. And that's where the value is, right? Like if you can be a good planner which involves both PM skills, project manage or product management skills and architectural skills then you're off to the races because then you get one shot. That's how it happens. You don't get one shot by sitting there and saying help help and then like threatening it. So it's just how it works. As the models get better, as you're better at this, you get more leverage. this 6x turns into 100x or whatever. I mean, this is the future is you're building all these things. So, if you want to make a fleet of agents that are dope, learn how to code and do it with leverage. Okay. Um, it also it doesn't even mention the fact that like you have cut offs. So, these LLMs won't know about the latest version of Superbase. If you're coming from a training wheel app like Lovable or Bolt, then there's already configuration that's messed up that is going to make you have a bad time. Cool. Thanks for the question. Next up is news. We've got a lot of things that I came across, so let's just start drilling through them. Uh, this is something that's interesting. What's the best CLI for crawling URL and turning it into LOM ready text? Well, I know it's fire crawl. What I didn't know about was this cursor tools. I that's wrong. I had seen it before, but I had not run it. So, I'm going to do this in a future video, but it is interesting where it essentially gives cursor. It's a CLI tool that you install globally, and it adds in a few things like the ability to call other models that already have prompts in them. So, if I wanted perplexity to go plus something up with some web research, I could go and do that. If I wanted Gemini to be the thing that's only used for getting full context, then I can go and do that. I think it's a uh catch 22. I like it because it does provide more functionality out of a box, but I also hate that it's in a box because I can't know exactly how it's working. I don't have the opportunity to improve it as much because it's going to be relying upon what they think is the best thing. Um, so anytime like I had this question from a friend like, "Oh, the new version of cursor is buggy when when my friend used it and it's like, okay, I asked him like when did you get a new or when did he use it?" He told me and it's the same day that 4.5 came out for uh GBT. And every time a new model comes out, like cursor gets a a little grace period, I'd imagine because they're one of the first checks that OpenAI invested in where they get the model first and they can go and try to figure it out. But it's not like they just like have it all figured out the day that it arrives because the day that it arrives, this thing has new context. It has a new uh system prompt. It has a new way of giving the info. So when it arrives to anywhere, let's say cursor cursor now has to figure out this new thing like it just figured out the the old one. So like let's say it had uh sonnet 35 and everyone's like, "Oh my gosh, it's on fire." And they just figured this out and everyone's happy, but then this one comes along and it's like, "Oh my gosh, it's buggy. Cursor sucks." It's like, "No, it's just they haven't figured out how to work with the prompt." So, if then you added another one that was like, oh, now you have cursor tools, well, it's just more room for error because you have more that's hideous. You have more spots that it can be changed and you have you have uh colliding system prompts. So, it's an interesting thing, but I wouldn't love it. This is crazy. The This looks like a real person. Like that's kind of kind of wild. It's called Rev Reev. But these realism prompts that are coming out. Like this looks like real. Like this looks like some indie photo. Um and I just think it's it's kind of wild. Like that one doesn't look very real, but it led me down a rabbit hole actually. Let's see if I can pull it up. Uh history Yeah, I was curious because this guy was talking about how he uh was using the new image gen from Gemini to take a photo of somebody and then take the clothes off of the figure and then create a layout, a flat lay flat lay of the outfit. And how you could do that basically to have an e-commerce store, which is wild. like it took this photo and then it extracts all these pieces of clothing and gloves and all these things and then you can wire it up to an agent that finds the items and puts them in your shopping basket. So you I just think it's getting crazy and it brought up this topic I hadn't heard of before which was what a control net is. And the best way that I could explain it if you're not familiar what control nets are because I wasn't until like 11 pm last night was when you have an image generator and you say, "Hey, make a Whoops. Go make a bear with like sunglasses on." Well, it's starting from scratch and then it's using a bunch of noise. Like if you're familiar with photo editing, it's using noise. Then it's using all the stuff in its head to try to reduce the noise into that bear. It knows what a bear is. It has all these things, but it's still starting from scratch at the very beginning. Whereas a control net is a specialized neural network architecture used in AI generation, particularly with stable diffusion. It gives users more precise control over the output. So in the easiest way possible, basically instead of just having a base model where you're like, hey, go make the bear with the sunglasses, you can include a control map. And that control map could be a sketch or a depth map or an edge map. And then you use that basically as the starting point. So you're not just starting from scratch. And u why is it a big deal? Well, as it points out without it, let's see. Points out this process used to be quite a bit more complicated without the new one, but okay. Required this. So, I guess the control nets are built into Gemini is what this guy's talking about, which is just kind of nuts that you can get it. And um the next thing I wanted to talk about is actually this guy here, Fabian. I was looking all over his page because he's doing some wild stuff with his business called Glyph. And I wanted to just take a quick look at it where I can basically have things tuned to a certain look. And I think under the hood, this might be a Laura. I'm not totally sure. Um, but it's like, okay, I want to have an N64 version of Parker. Like, what would that look like? Or, yeah, let's see if we can't set one up. So, if I sign in to run this, let's sign in. It's probably going to charge me a credit. We'll see. Uh, tinkerer continue with free skip. I hate when that happens. You go to do the thing that you want, you authorize, and then it kicks you back. Come on. Let's see. Uh, I can drop an image. Let's grab a head shot. And let's see. Run. I have 10 credits. I should be able to do this. So, it's calling Comfy UI under the scenes. Yeah. So, Comfy UI is a nodebased image generation tool that lets you drag and drop the nodes. Kind of looks like N, but like a thousand times more complicated. And you can add all sorts of Loras, which are kind of like image nets, but they're like they're like filters. So, Be careful if you go on any of the Laurel websites. Remember like one of the first times I went on it and they didn't have a new filter that's like the not safe for work filter and I opened it and it was just like whoa, where am I? Um, okay. So there's me. That's N64 Parker. Cool. Let's download it now. It's burned into my memory. Yeah, this glyph thing does seem pretty cool because it takes all of the like complexity away. It basically productizes the amazing technology that is, you know, image gen and Laura and all that. So, I think I had one more. Yeah, this glyph chrome extension. So, you can remix the web with this Chrome extension. So, let's try it out. We'll go to Reddit and let's see. I probably don't have a credit to use, but let's try. We can go to I think maybe map map porn is like one of the most popular subreddits for showing you like different data visualizations. Funny name, but okay. Let's take this thing. The concentration of castles in Europe and let's open up this. And where is it? Glyph, where are you? Glyph. What can I do with this? I can sign in. Okay, sure. Um, I am signed in, bro. My brother fail. Okay, well, whatever. I think that was all the things I wanted to cover. Oh, if you're trying to learn about LLMs, this site's really good for it. which is called Baby Coft, BB Y Coft. But if you hit the space bar, it'll actually explain how tokens work and it like walks you through all this stuff and it's really if you know nothing, it might be worth checking this out. But lastly, I just noticed a lot of the stuff that I'm talking about with automating your processes around sales and marketing and you know, taking a video like this and having to go chop chop chop and distribute it across everywhere is being talked a lot about a lot by this guy, the boring marketer. So he says, "Hey, the central stack for it is N8, Geets, AI, Google Cloud Console, data sources, and open router." And I agree. I think I'd change some of these. That's in the video I posted yesterday where you can basically have superbase kind of collapse the option two, three, and six. Maybe you can still call out through open router. But yeah, I totally agree with this where it's basically you're training this orchestration that's happening on the way that you communicate. And then he was talking a little bit about how the future of search marketing isn't really optimizing for LLM visibility. It's optimizing for agent decisions. And agent decisions are going to be based on the same stuff. So it doesn't totally matter like if you use one vernacular versus the next. But at the end of the day, agents are going to be the new way that people make decisions and source information. So it'll still come back to domain authority, page authority, all these metrics that Google uh basically defines your ranking on the internet for SCPs or search engine result pages. But as we know, Google is only a small percentage of the total pi, but everyone does a similar thing, right? Like either way, the algorithm is trying to just find the most reputable person that's putting out answers to people's questions. So, that's a big motivator for why I make so much content. Okay, now let's talk about client stuff. We're going to talk about strategy. So, these are really exciting. Actually, this is my favorite part of the video today. So, I wanted to just hammer this thing home about client strategy and how the SOP is non-conditional. It's not like Chick-fil-A if you go to one in Kansas that they're going to say something different to you because you know them or like you basically like they act the same way everywhere. Like they have a baseline level of service that they provide. Like they'll always say my pleasure and if they know the person, they might say my pleasure with the person's name, but there's still a baseline. And I just had to remind myself of this because some of the recent client work I'm doing is with people that I know. And I had thought, okay, I'm going to follow these normal SOPs that I've put in place, but you know, they're not getting back that fast. So like, let me just lay off on it's a friend. No. Like business is not emotional, and you are doing them a disservice by not following through, not following back, and not pushing them to get the thing that they need. So accountability goes both ways. And so I just wanted to like hammer that home. And the way that you do these SOPs is you define them. There's standard operating procedures around every area of a business from sales to marketing to the delivery or fulfillment of said thing to CSM or the customer success management aka like after the fact. How are you dealing with them? How do you deal with payments? Comms is a big one. I talked about this in a different video. I think the one yesterday. So, if you're watching this now and you want to see it, go back a day. But defining all of those, each one of these is going to have like 20 to 50 line items. And that's only because not it's it's not because it's annoying for the customer. It's so that you thought of all the different like branches that go out like how could I surprise and delight this person? How could I uh nurture them such that we can hit them with more value later and keep the business uh relationship going and growing? So, that's just something I wanted to point out. I also wanted to talk about I got pitched on an enterprise opportunity last night. Um, it wasn't necessarily that I was going to be like the guy to go and do it, but it was just a friend of mine who had just done a project for he was just explaining how he has a friend that works in a big enterprise company and how they just don't know that any of this stuff exists. And you know, if you have a couple thousand person org and they're doing stuff and like if we have a a spectrum of like know AI knowledge, if you're even watching this video, you're at least like a three out of 10, but they're like a negative three. Like there's people out there, they haven't even gotten to the starting line. So take advantage of this. Like think about it. What are the lowest hanging fruits for an organization that you know? I think if it's enterprise specifically, leveraging a relationship would be the thing that gets you in there. Selling that in would be a lot harder if you didn't already have the red tape cut. So, next is up on the plan. So, I think a lot about lifestyle design. You're like, "Wow, that's that's jargony." But basically, like if you have if I sleep like six hours a day, then I'm left with 18. That's Yeah, that's right. So if I'm left with 18 hours and I need to make money, I need to then focus part of that on direct revenue generating things, right? So maybe you call like I do like 4 hours in the morning like direct money generating things. Maybe it's the client that I need to finish work up for. Maybe it's uh going and finding new clients. You can break it apart in a lot of different ways. If you wanted to follow like the Ryan Seirant thing, his is the finder, the keeper, and the doer. And he breaks up his day like that. So the finder is going and finding new business. The keeper is doing the uh like the finances of the the business, making sure it's good to go, working on team. Then doer is doing the things that you said you do during finder. Like if you're working on new clients, then you're closing them. Then you're doing the initial project. So maybe you you split it that way. But it's like, okay, if I have x amount of hours, I've allocated this many to that. like I'm always working backwards to what are the things that I could be doing that can like first of all make money then also align with something that I really like doing. So if you can find that that's magical um because you'll get better at it too. That's another thing. It's called icky guy. Icky guy. And it's a Japanese philosophy where it's like you basically want to find the things that you're good at that are fulfilling and that make money. And if you can meet there, it's good. So how it applies here now that we've covered a philosophy lesson is well I have these three channels. What are things that I already need to do that I like doing that I can post and repackage and make content out of? So, with the three channels of daily, which is what you're watching now, the behind the scenes kind of me brain dumping, the builds, which is satisfying the thing of I like to build stuff, and then the main channel, which is to drive attention towards AI for Main Street. I just thought of it like this, where it's like, okay, well, if I just at the end of each one of these videos, I have a build pipeline, which is in orange and it's down here somewhere. I had a build build pipeline. I would talk about it on here at the very end so I'd get like a little bit of an idea and then I could also get feedback from you guys on like questions about it. I do like doing it. I already do it. I still record it. If I go and I build the thing and I record it, it naturally puts a timer on me accomplishing the thing. It's like a pressure to get it done. So that's gonna be better because I know someone's watching me, so I have to be better. And then with the outcome, it's like, well, could I package up the thing that I just built in a way that's attractive to a wider audience? Well, yeah, I could. So, I was thinking about that. I'm like, all right, well, that'd be sick because then I just l literally do two videos in a row. I've done my planning for the day. I've done my build for the day. And then I just need to do some post to package it. And so one of the things I need to do is I actually need to build this NAN so I can automate my business. You can automate a bunch of the administrative parts for those SOPs I was talking about. So for my daily thing maybe the title would be like uh how I'm I don't know like how how I'm automating my automation or my AI business. It'd be something like this. And that would be a big portion of the video be talking about that. If there's something more interesting that I'd cover in there um and you know like road to 100k if there's something more interesting in there then yeah sure I'd cover it. But that would be really awesome because then when I go to build it, I just say building N8 super uh NA uh NAN and Superbase automation platform, something like that. Like I posted one yesterday and ideally this would be one part, but that one's two because again I'm doing it for the first time. A lot of these things I won't be doing for the first time. Like I won't be from a technical perspective I won't be doing for the first time with this. I've never self-hosted. So getting it to work with other things. Uh I've done that conceptually of like integrating databases and services but not this one. Anyways, yeah, I do need to do this. So, it ticks those boxes and it, you know, I also think like, are these things on trend? Well, yeah, actually that's a bonus. It's not necessary, but it is. It's on trend. So, then I'd say, okay, well, what is then the packaging for that? Well, if I end with this awesome self-hosted thing that runs on a server that I did from technically scratch that can then connect into and run these growth agents that go and leave comments with the chapter markers on every popular YouTube channel that doesn't have them and drives traffic back and then listens to the videos that I create to come up with plans for future content that then comes up with monetization. uh methods based on the things that I'm producing. Like I have all these different agents that are gonna be then plugging back in. They're gonna go and there might be like a node server that goes and gets info and does a thing and then comes back and then it works. Like if all that's happening then there's some cool title that you know it could be like I automated my business or something. But that would be you know I'm not going to do it here but that would be like the idea where it's like what is the juiciest part of that? It's like I automated my business from I fully automated I gave AI my 90% of my business. It'd be something you know that's like what does that mean? So this one is actually where the work would come in like how do you package it? And I'd probably be thinking about that before I did it. So maybe that'd be another qualifier or like YouTube package. So at least I'm like thinking about it. And then that same thing would be for stuff that I need. Like I need this is so simple, but it's like I want to have better landing page for all my links. So this probably wouldn't be a title for it, but you guys get the idea. It'd be something like this. And then it kind of like funnels its way through. And uh I think that is going to be great. So on the actual build pipeline side of things, I kind of just covered it, but in my head, just to be sure as to what I'm doing next is N8N plus Superbase part two. I need to get that. And then I'm probably going to put up this link tree thing because I need links. That could also be a package. Like the fact that link tree is like this billion dollar company is just wild to me. So I'd be like, "Oh, I made a billion dollar tech company thing in 30 minutes." Because with an a link in one shot, that's you could you could literally do at least the non like editor drag and dropy stuff, but you could build the link tree hardcoded with your stuff. Maybe even with database with a one shot uh link tree. I have the YT uh chapter marker agent. I have the cursor forum agent. So, those are what is up next. I'm going to be going after this right now. You can see the status of it. It ain't working. These two are dead. They're dead as a dog. But yeah, that is it for today. You guys keep asking questions. I really uh really appreciate it. I'm going to say Aiden, this is so clutch. Thanks for the question. I answered this in depth. See you tomorrow with the answer. That's March 26th. See you soon. Okay, so reply to him. That's a weird reply. Okay. Yeah. So, ask questions. The channel is actually it's growing. So, that's cool. We've got 116 people that are subscribed. We jumped how many in the last day? We jumped 24 subscribers yesterday and 14 the day before. So, it seems like you guys really like this video I made about basically calling out vibe coders. I said, "You're not that guy, pal." I don't think people got it, but it's uh it's a funny meme. You're not that guy, pal. This guy, let's just watch him and then I'm going to take You're not that guy, pal. Trust me, you're not that guy. Okay. Okay. Oh, man. Okay. Thank you for watching the video and this is going to be called the ultimate YouTube content pipeline. Ah, there it is now. Okay, see you. Like and subscribe.