The No-Code vs Vibe Code Battle Continues

I'm Parker Rex. I led tech at a startup that sold for 23 million bucks. After that, I tried Airbnb for music. It didn't work, but it lit a fire underneath me to learn programming full-time across the entire stack. Then, ChatPT came out. It was obvious. This was bigger than the iPhone, maybe even the internet. So, I went all in on AI, learning how to use it across coding, marketing, and business. Now, I'm building my own SAS AI media company around it. And this wouldn't be possible without leveraging AI. On this channel, I share daily updates on strategy and AI news. No fluff, no hype. Let's jump in. Cool. So, first we will answer a couple of questions. I know there's a lot of requests that are coming in for the Turbo Kit, which is, if you're not familiar, it's a set of primitives. It's a starter kit for developers, for new developers. NAD is just Zapier but better. I think if Zapier can't keep up, Naden will just absorb market share. Can't really speak to the gurus. I think it's all kind of blank, but the AI agent trend definitely has a future, albeit only a bit bigger than Zapier's market cap. What I was pointing out in this video he's referencing is a lot of the videos that pull the most views are entertainment videos. They are not business value videos, but people mistake entertainment videos about a robot sounding Mr. Superman who can go and read your calendar for you or can check the weather. That modality of then burning tokens and burning time for that request to go through a fancy looking nodebased architecture is not practical. Why would I want to ask this thing for the weather when I already have one that's built natively into the operating system of every single phone on the planet? Or why would I want to wait on the request when I could just literally tap the thing on my screen or swipe my fingers and see it on the right? Doesn't make sense. That's what I'm pointing out is bit about Zapier. Zapard serves a different market entirely. Nad is developer focused. If you go use Zapier, you can do it as a non-technical person. So there are different things. AI agents are going to be a massive massive market. I just don't like when I see trends that are repeating themselves. So you have people that literally got thrown in jail for selling FBA by a by Amazon. It's so fulfilled by Amazon. Hey, I'm going to teach you how to make $100,000 a month. buy my $10,000 course and I'm going to teach you how to make an automation in a store on Shopify or sorry on Amazon that sells mops that nobody wants. I actually used to like Framer because you can make a starter site. There's just going to be this race between the training wheels vibe coding tools what I call them like the replets and the lovables and the bolt news of the world. they were going to be in a race against the drag and drop website builders and then the people who are going higher quality who care a lot about design will go into Figma. I think framer is the least one that's going to get smashed. I think if you look at web flow goodbye that is total totally toast. See, fix my vibe code app sounds perfect. Shout out to Hari. But yeah, fix my vibe code app will be an application that's hosted on ivibeai.com and it will be a react drop zone. So when you drop a codebase onto there, it'll do a list of checks and be a multi-agentic flow that has a bunch of tools that looks for common problems in projects. And this is nothing fancy at all. This is just what a developer would normally do to get something to run. But a lot of the training wheel apps that people are vibe coding, when they try to make the jump from that built-in browser ID to a proper ID and they take the training wheels off, there's just configs that are screwed up. So that just happens. Or maybe it's a package. The common one is they use tailwind four and you should be on three cuz all get confused. or if you're using Superbase, the newest versions of Next typically get tripped up around cookies when it comes to authentication. So, you kind of list those steps out and then work through them. And that's it for today on the questions. Oh, actually this one, this guy was talking about Nomic somewhere. Yeah, I said doing something similar with Nomic. He was talking about embeddings. Nomic is definitely not mine. I have not rolled my own embedding model. I was just mentioning that I've done something similar using Nom. If you're doing embeddings, use Nom. It's free. There's tons of papers backing why it is superior. Cool. Let's jump into some of the news. This is maybe one of the biggest news days in AI history. There's so much stuff going on. First of all, Gemini Gemini is one. Gemini 2.5 is rocking the world. It's better than OpenAI across everything, across coding, across research. It's absurd. and it's free right now. So if you just look at the writing quality, the completeness, the comprehensiveness, the instruction following, and the overall score, this is specific to Deep Research. So Deep Research you pay $200 a month for. You look at Gemini, it's free. They're going to charge at some point. They already have a pricing sheet that's very cheap if you're using it in access or API access. But yeah, that's Phil and I believe he's one of the people at the cloud team. Now, other things that are interesting, this is more on the design generation side, but the cost to make things that were very premium is plummeting. That's great for us builders, that's horrible for designers. Sorry, designers. But be adaptive. When you look at something like this, if I was making Mickey Mouse SVGs full-time for a living, I'd probably be worried about my mortgage every month, first of all, because it's just a tough business to sell because it's hard to justify the ROI on these little images. It's usually like you have to find a founder that has a deep love for creativity, which shrinks your market size. But when I look at this, this is great because these plus up any sort of design. Same thing goes for this image to texture. That's insane. This is Yeah, that's Blender. That's a whole line of skill set. There's a new drop that I'll be adding parts of into MAP. If you're unfamiliar with MAP, that is my SAS product that I had put on ice to get the media brand off the ground and then to also bootstrap the client services so I could dump that capital back into the development of this. But yeah, the master plan around it was basically that personal health management is a miserable issue. It's, you know, over half the people are obese. There's tons of productivity that needs to be applied using AI as well as just all this wearable data that we can access and biometric data that gathers dust and folders. And then it explains kind of how we do it with a bunch of AI. It's a multi-agentic platform that represents your peak self. So when you jump in a map, it gives you the map of what you need to do every day and holds you accountable across all areas of your life so that you can not be guilt tripped by an AI but have what that version of you would do in every given instance and because it has live information because it has all the daily drivers of the notes, the tasks, the health, the chats and the calendar be awesome. But I yeah I was on pause because the multigenic stuff was a little bit more challenging six months ago. But agents SDK helps out with guard rails checks and routes. So in the calendar rolling my own version of Gcal pretty tricky in a lot of ways but like little things like oh I need Gme. So in order to get a Gme link I need to wait that. And then just doing the back and forth with syncing and having a syncing engine and then having a syncing engine for all the biometrics and doing the ETLs on those and then 86,400 seconds in a day or 88,000 one of those. Multiply that by three that's how many biometric points you can get off a whoop per day. So a lot of data and then deciding what to vectorize, what not to. this agent. It's cool to see that this guy's grown an AI hedge fund manager to 20,000 GitHub stars. And each one of these is its own agent that uses a trading strategy. And then it goes from the start to which one do you want to use? So if you're familiar, Ben Graham, the intelligent investor. It's the big bible of every everyone's read it. It's valuebased investing. It's the thing that Buffett was trained on. You have Bill Aman from Persianing Square Capital who made a lot of money on some controversial trades. Specifically, one when he went on CNBC prior to the big old everyone gets sick in 2021 thing and said you you know he went on this freakout tirade media run and then went and played a trade against the thing that he just told everybody. So he like did a he crossed up the whole retail market. You have Kathy Wood. Everyone knows who she is at Ark. You have Munger. RIP Munger. You have Bur who was the main guy in the big short. He's the one that made a ton of money during that. But, you know, clock always strikes once correctly. So, he's made a lot of ridiculous calls since that weren't true. Peter Lynch wrote tons of books on Wall Street and trading. And I always quote him when people are freaking out about markets. He says that a stomach is a more important organ than any other in your body for investing because it's similar to how markets are a scale, not a voting machine. But at the end of the day, a lot of people they vote with their wallets and they get emotional. They get the stomach ache around the price going down even though the underlying assets fantastic. Don't know who Phil Fischer is. Everyone knows who D is. Macro investor baller and then Buffett. So, it's cool just to see. I' I'd be curious if any of you guys have run this. If you've run this, then just drop in the comments what you found out about it. Turbo kit coming later today. Cloudflare workers is encroaching on Versel's monopoly or kind of workflow around the serverless space and making it really easy. There's something called Openext that does an adapter for it, which I think is interesting. And then they're also adding access so that you can run serverless stuff and static sites on one worker. It's not out yet for next, but it's coming soon. It's in beta. This is Superbase. They should sponsor me because I just know a lot of people that work there and if I were to go get a job, I'd work there because it's fully remote and it's just got the insane growth numbers and it's everything is in it. So there's a bunch of roles open if you're curious. dev. A lot of things. Let's see. Harvix. This now has deep research in it. So, if you wanted to go find research papers, gosh, there's just too much in here. Gemini now has AI built in. Pretty cool. Complexity is launching 5K for those eligible for startups. You basically fill out a form. Oh, I'm going to try this out. So, as part of my automation workflow, I want to use MCP's model context protocol. If you're not familiar, it basically gives AI a set of sockets or connections or Lego pieces that fit the square to the square. Meaning, if I say, "Hey, Ableton Live, which is music production, or Adobe Premiere Pro, which is video editing, hey AI, go clip this in this software. Cut the song, cut the video at this timestamp, and it'll go and do it." So, it's going to take away the complexity of clicking around in all these menus, which is really nice. And it's going to give access to developers who are passionate about pieces of software that are not tech forward, that are not going to roll their own agent for years to go and actually implement them themselves. Very, very cool. But I think that the more complex the project is or the software is, the more tricky it will be to do this. But you're seeing it in all these creative tools that don't have the budget or the wherewithal to go and roll their own genic stuff. So now into some strategy. I have a lot of things going on. So this is how I'm bucketing them where I have the things that I'm building. I have the work that I'm doing and then how it plays into this content flywheel. So whenever I do anything, it must play into the next thing. What do I mean by that? Let me explain. So when I do something, let's say I'm doing a client build, here's one thing I do to generate revenue, right? So when I'm doing that, if it's doing a method that isn't interesting or is not pushing what's possible with technology, then the ROI on it isn't very high. Even if I were to get paid $10,000 to do this thing and it will take a day or 2 days or whatever it is for the build. If it's going to take three days, but I'm now getting an ROI of new skill plus cash plus content, then obviously that's better than just cash, right? Like we choose this one all day. And so now I'm bucketing that into, okay, can I make content out of this? What is interesting? What's inherently is interesting about the thing that I'm doing? If I'm not doing something interesting, make it interesting. How can I make it interesting? And a lot of people just lead with the cash part because it's clickbaity and because people are like, "Wow, I could do that." And that's great. But for me, it's beyond that. It's more so how do I get some sort of skill that I can reuse in the future. So one of it might be if in the client build I can use something like taskmaster which I discovered halfway through the build then I'll go redo the build because I want to see can I push the technology so that I'm not touching the keyboard beyond a couple of edits. Can I use voice and whisper flow to write the entire spec prompt while using other prompts that I will make better as I'm going and so that the whole machine is more effective. And so it's similar to how Tesla's value is not in the car itself. In the earliest days, it was in the factory that they're building. Now it's in the software that they're building, in the network that they're building with telecommunications, with other companies that he owns. So all these things kind of play into one another. And as you're progressing, you're trying to make the thing that makes the thing better. So at the end of the day, it's a content factory. It's serviced by agents. So in order to make that content factory work, in order for the pipeline to be clean and the content to be good, because at the end of the day, if the product's not good, it's not worth it. I have to be thinking about the work that I do in these kind of ways. So for strategy, I look at it and I say, okay, well, I have the map turbo repo I need to set up. I know that that'll be part of a module that I do for this for school, but also if I take the long form stuff, which is for the people who have the membership in our community who are trying to see every detail. They don't want like the piffy summaries on YouTube. Then I'll at the end of it after I've made everything and I've recorded it, I'm kind of narrating as I'm going. Then I package that and repackage that into the YouTube content for our main channel. If you want to check out the main channel, it's just Parker X. This is Parker X daily is what you're watching right now. Then I have Xgenerating summaries. So, this is something that would be part of my content factory where there's a trend going on that's very popular, which means that it's market validated and I'm just going to do it better because it's going to be fun for me and I can do it at scale. Where when I watch a video like this, how we build effective agents from one of the guys who builds all the agents at Enthropic, how do I turn it into this, right? How do I take it into the notes, not only for myself to digest, but then have it look like this? So, there's a couple steps in there. it's oh I need to post below the video of whatever video it is with some sort of summary. So if this is on X then I can post the image below it. I can also post the text below it on YouTube. I can also post this itself as an asset across a lot of different places and then I need an a flow that comes back with the findings of that thing of that piece of content. So this would end up being a table with the piece of content, when it went out, how it performed across the engagement metrics for each platform. And then as you build those up over time, you can add more and more to it. But this would be the basic flow. And you do all these manually first. You don't go automate first. Always follow the steps of Elon Musk's five-step algorithm. You automate last. So I do some of this manually, see if it's working. And when I say manually, it's just connecting one to the next to the next. Seeing the outputs, tweaking, see the outputs, tweak, see over and over and over till it looks just like something like this. Maybe even improve it because you can always make something better. And then manually go post those. And then if they're performing well after 10 iterations, then maybe automate one step. So that would be something that would be the school behind the scenes and then the YouTube video where it's shortened and packaged. We have sire. This is a client. So that's a video that I just literally was referencing where I did half the project and realized this isn't that interesting what this taskmaster thing is. That would be YouTube content iOS utility. This would be school content. This is me realizing that when I hold my iPhone next to another iPhone, it uses the U1 chip inside of it to start rumbling and have this haptic effect that's going on to share contact. But turns out that's an API that we can leverage for proximity based features. So then you run a prompt against that and you say what are high frequency tasks that people do every day that people could gain value from from set technology. How would that look? I got a list of 50 of them. The idea is to take this and speedrun a utility iOS app that just leverages all of them. So that would be school content and then eventually it'd be YouTube content like a few like maybe a month after when it's in the store just to talk about and explain it. Chapter markers flow. I do this manually now. I just used notebook LM. I was using something else before with Aonic, but it added too much time to the video workflow, so I reduced it. So, I need to then take whatever the prompt is behind this because Notebook LM with Plus has the ability to paste in a YouTube link and it automatically knows and generates all this stuff, which means they're already training on everything the day that your video goes up, which is they're going to win. Make a simple landing page. This is literally for my barber. My barber was asking me how to make a website and I wanted to as a challenge do Taskmaster plus maybe Origin plus maybe Shad CN plus maybe 21st Dev and see if I can't speedrun him a new website. And so yeah, I look at these things. That's kind of how I plan them out. And then let's see. Lastly, there's just things that I need to wrap up on this, but I don't have time for today. We're running overtime. But how I would do it with Taskmaster, I explained this in the video yesterday is I open this. Yeah, cool. So what I do is if I had this list of tasks that you just saw, then I would start to work on them, right? So when you use Taskmaster, the way that it works is you have a few different files that are created. You have the MDC's, which are the rules for cursor to know how to operate. These are specific and generated as soon as you run Taskmaster and NIT. So you download it globally. You do npm task-master space-g. The g is the global flag. So it'll work in your CLI wherever and you run that and then you run taskmaster init. That's going to spin that up. That's going to spin up the scripts folder that has the dev. This is the JavaScript that operates the thing. You have an example prd. Then you have a prd. So this is the one that will be used to then parse out the tasks that need to get done. So if I had a simple line which is hey I did this with taskmaster only and when it made a change for something then the carousel image quality was reduced because it was prioritizing performance over quality. When you're doing full page high resonance stuff you need to balance those two. So I'd start with something like this. I'd tag in the relevant files with the file names using the hyphens using the relevant paths and then I pass that into this. So I draft it. Right? You're trying to include as much detail as possible. Whatever libraries, whatever components, all the paths, the patterns, and then I do space F1. And then you can see in here, PRD instructions is the prompt variable. And I drop that in there. That's going to run. I'm going to get something that looks like this, right? And then I take that, and then I drop it. I go from being in ask mode, which is read only, into agent mode. Agent mode is allowing you for writes. It's what everyone's obsessed with because it is helpful when guided. I take that and I'd drop it into the chat or sorry I'd say rewrite this draft in the following format and then I'd paste in the format that the team at Taskmaster recommends using and then I'd say and then copy it into PRD. So at the end of the day you'd end up with the PRD that's formatted and optimized for Taskmaster. Then you run hey go and parse the tasks from underneath the hood. It has a CLI command. It's going to run in agent mode. that's going to go call out to Claude. And so you need an Enthropic key. Don't worry, I used it all day last few days. I think I've spent like 60 cents. And it's going to parse those out into the different tasks. And you'll see in the tasks folder all the different ones, both in text file format, but then also JSON. And the JSON gets wiped every time. So don't delete these. That was something that I was doing. But it is crazy because it's doing things that I wouldn't normally think. Like if I gave it really simple thing like, hey, this quote thing needs to look better. and you know yada yada yada. That's a single component request that I'm asking it to do. But when I run it through PRD1, PRD2 and then parse it out. It's like really going deep on things and the quality of the work is stupendous. So really happy with it. Definitely check it out. Trying to think of other things that are helpful in there with that specific workflow. But that's it. That's it for today. There's a ton of value that I'm going to be bringing every single day. So you make sure you follow and subscribe. Make sure you like the video so that I show up in your feed. All right, until next time.