Vibe Coding Office Hours

So today it's all about the vibe, the vibe coding. So we got our React expert here today, Ray, from StateChange, and we want to have a very interesting discussion about vibe coding and React AI app generation. So how do you want to start this, Ray? Because you're the expert.

Well, I don't know about being the expert, but I know that you've had some really recent experiences Finding a newly opening market and world of possibilities as you have just thrust your hands into actual code with the assistance of a little bit of AI help.

Yeah. We had one user who was using Despia and he was using Lovable. The thing is window.Despia is not type-safe in a TypeScript environment. So like you can't really write random stuff to the window as you would do in most no-code tools and vanilla JavaScript because you have to worry about type safety and all that kind of stuff. So I started building a little in Cursor, a little type-safe JavaScript SDK for Despia and published on npm. And because it needs some, you know, specific prompting for it to actually pick it up, I made a video about it., and I posted it on X just for the sake of it because I didn't really first want to post it on YouTube because I was like, it's not really the audience I'm serving. I'm not really serving vibe coders. So let me post it on X. Maybe it does better in there. Maybe I get a couple hundred views on that. And I checked back an hour later, I was like at 10,000 views. And then a day later, I was like at 25,000 views.. And people working for Lovable, I think the product person at Lovable reposted it and people from RevenueCat I think reposted it. Vibe Coders on X reposted it. Everybody reposted it because they were like, wow, I can now make mobile apps in Lovable with just a line of JavaScript. This is amazing. And we had over 100 signups the first day the video was posted for Despia and it was incredibly well received. Feedback from the video. People were very excited about this. And then I still had the thought behind my head, like, yeah, Lovable is good, but you can't build big apps in it because that's just, you know, I think it's a defense mechanism. If you are yourself doing no-code applications and your whole career is around doing no-code, you tend to be more skeptical when something new comes that looks like it is you know, kind of like not serving you well. So I had, instead of accepting and embracing it, I was fighting against it, right? But then yesterday, um, that was when everything changed. Yesterday I had many meetings with people that saw the video. They wanted to meet with me because they wanted me to show the Lovable application and ask me if they can build this into native application with Despia. And I was looking at their application And one guy, he built— this really blew my mind— he built an application for like blockchain and all that kind of stuff. And he did it all with Lovable. He used to have a whole engineering team and he just said, you know, I used to talk to the engineers, now I talk to the AI. You know, he's the one who knows what to say. He did that many years before. And I talked with some other people and they showed me their applications. People who know what to tell other people do really well in Lovable. I looked at their applications and I was blown away that you can do this in Lovable. I always thought it was a piece of shit, but to be honest, I was wrong about that. I think vibe coding, AI-assisted development is really a blessing for many people because you don't have to worry about building a popup and spending 2 hours building a popup. And I remember a year ago, you mentioned that you were talking with, I think it was a lady, and she was trying to get some data into an application and you were debugging stuff in Airtable and how to get it working with, I think it was WeWeb. And you said the value in your business is not in the front end and getting this button to work. The value is in the data that you're selling. People would pay just to get access to the Airtable sheet. And this really struck with me where I said, wow, it's not about making your buttons 100% most of the times. It's not about making your drawers extra smooth. And you don't have to do it manually necessarily. It's about getting an MVP, an interface, a first application out there that solves the problem, and then you can iterate from there. And this is where I really, really, where I misunderstood those AI tools. They allow people who know exactly what they want to get it done fast, faster, lightning fast in a matter of days, not months with no-code and years with writing it by hand, right? So if there is fast, faster, fastest, like, you know, no-code would be in the middle, but AI would be fastest. And this was really a paradigm shift for me where I first started seeing AI-assisted coding as something that's actually viable because I saw all their applications and then I played with it myself and I used my own Despia SDK to build an app. And I built like an application that actually worked in Lovable with backend, with everything, and using my Despia SDK, which was published 2 days ago. And it already was able to pick it up and set it up for me. And I was like, my God, I just had an AI build an application using my framework. This is freaking awesome. So that was the first time I actually accepted that that is a valid way of building applications. And now I got them all as my background. So you see, I'm, I'm, I'm, I'm sold.

Yeah, Bolt backwards. I think that they're inverted.

But I got Bolt back there, I got Cursor back there, and then kind of lovable. But you know, it's the, it's the, it's the last, it's the best thing I could find last minute.

So, and there are, and there are many more at the I'm so glad that that has worked for you. I mean, these tools just keep on getting better.

Yeah.

I think when you and I had a conversation about vibe coding, you know, a couple of months ago, right?

And the, you know, I was against it.

You were, you were more skeptical, but I actually, I might propose that you were less skeptical of AI-assisted coding where you and I were, I think very much simpatico was we were talking about the idea of coding just with vibes. Now, Andrej Karpathy, when he was writing the original vibe coding tweet, was he says, just use vibes. But he's like a serious data scientist. His idea of just use vibes is actually far more academic and more intellectually rigorous than I think most developers on their most serious day. The thing that we landed on, I think still applies here, is you don't get to leave your brain at home. It's not just enough to say, I'm going to write something into a box and then I will have a startup that will pay me millions and millions of dollars. You have to do work. But oh my goodness, you're absolutely right. It's so much faster and it rewards a different skill set. You raised both those points. I think they're worth raising for our audience here. One is that it is much, much faster to do a build, to do a turn, an iteration using these tools. The complaints are often, well, but then when you try to fix it, it doesn't work so well. That's because you need to take a completely different model to how you then iterate on software. It doesn't look like what we used to where you have, hey, I've got this thing and to iterate on it, I'm going to keep on editing it, it actually looks a lot more like what good writing looks like. And like good writing, your second draft starts with a match. You say, okay, I got to here, now I can go build it again and build again. I mean, you will build this again. I know. Yes.

Yeah, you should get that engraved on your background. You've been preaching that for many years.

Um, and it really helps in the world of AI because like even more than world of no-code, the world, your ability to go make another draft happen is actually much, much faster than trying to edit the current one.

I was talking yesterday with a founder of an AI startup. They are kind of like an AI-assisted coding startup because I got some people reaching out to me on X because of the tweet that went viral, kind of like. And I talked with him and he said something very interesting. He said no-code is the future. I said, I would disagree with you, but what is your definition of no-code? And he said, no, no-code is not drag and drop. That is, that is over. The future of no-code is AI-assisted coding. That is the new no-code. You don't need to know coding to build a technical solution to a problem. It's not drag-drop Webflow-like anymore. It is a new dimension, but it is still within the category. The same like Mendex is not like WeWeb, WeWeb is not like Lovable. It's one of the further iterations. You know, you were talking about the cycles. Every few years there is a new way of no-code. Back then it was OutSystems, Mendex, all that. Then it was Bubble. It was kind of like in the middle, but then it was like WeWeb, Nordcraft, Bubble. Now there's Lovable, Bolt, Cursor, Windsurf. All of those, Replit. And that's the next generation of no-code. And I personally feel that that generation is the one where actually no-code provides value to people. Because even in WeWeb, even in Bubble, even in all the other ones, now consider like equaling out WeWeb's AI out of the equation. Let's say WeWeb a year ago, right? You would have to build pop-ups, you would have to build drawers, you would have to build stuff that is in a React ecosystem so— like, like, nobody— no, I know no React developer who thinks about how they should build their drawer. They just import it from a package, you know. Yep. Well, you got to build the whole thing. You got to build the handle. You got to add the event listeners. You spend a day building a drawer. A React developer just does npm install fancy drawer, you know.

That's a really good point though. I think it's worth asking what are the ways you can build? Of course, you can create your own code at various levels of abstraction. There's another part where you can copy and paste, copy and paste whole chunks. We often call that vendoring, taking a piece of code from one of our own brain to another, we call that vendoring. Then that becomes important to use. There's another version that is templating. Using templates that are meant for the job. That's like the npm packages that you were describing. We also have some of those in the no-code environments as well. But yeah, you're totally right.

It's not the same. The React ecosystem is just a million times bigger.

It's huge. It's absolutely huge. There's another level, which is being able to articulate things using an alternative means. This is where the so-called visual development becomes useful., right? And then there's like another level above that, which is trying to give instructions to the AI. This is what software 3.0, as Karpathy calls it. Um, the, um, but, but I think it's worth like saying that there's this whole spectrum of stuff that you can do, and the most powerful tools I've seen are the ones that involve more, that involve more of them, not covering the whole spectrum, but having more options. Like, I think Xano, just taking an example of like one of these visual development type tools, which, which I have some experience with, uh, is, um, I think one of the biggest things that changed the way working with it was copy and paste. Once they introduced copy and paste, actually building with it, like I've been doing some stuff recently with it.

XAML was such a pain before they had copy and paste. I was screaming at my computer for times.

I was like, I got to rebuild this whole thing. It is absolutely transformative in the way that it works. Just take another example that's even a crazy one, Cloud Code. I was actually helping a guy with Cloud Code last night. And, you know, realize that like what we need to be doing was we need to have like reusable paragraphs. So what you want to do is have your— you're going to put together your prompt of instructions and you can sort of have paragraphs you're going to keep bringing into play. They're going to be part of your instruction set for, you know, a given project. It's not about like having it everywhere, like saying, well, actually when you're articulating what you want to create for the app, it's like there's like a paragraph that's going to be different for this app. And then like a bunch of relatively boilerplate things like, You want to use Vite and— or Next.js. You want to use, you know, Tailwind. And this is how you look at Tailwind. You want to make sure you install appropriate MCPs and you only want to use these kinds of approaches. And then when you've given it those things, it works much better. And so the copy and paste actually applies even as a mix-in with something like the AI-generated code. And thinking in those terms of how to mix this up is a difference in mental model, difference in the way that we actually interact with the software.

Yeah, I completely agree. We actually got a question. A question: do these no-code applications that make quick applications work for a real business, or would it just be an MVP? What do you think? I think most of the times it's the same as Bubble. You start out with an MVP. While you're building the MVP, you get a bunch more smarter about how that tool works, you know, all the bells and whistles. And I think there will be an MVP+, and after the MVP+, there will be an MVP++ until you gradually go into actually having a production application. Most of the times when people think they already shipped the real application, they're still at the third iteration of the MVP because it takes time to figure out what your business is really about. And it took a couple of failed attempts to figure out what the market of Despia is. So, you know, most of the times you're trying to pick up the crumbs of a market And you're so focused looking down to picking up the crumbs of a market that you forgot that there is an all-you-can-eat pie buffet right down the lane.

So yeah, the, um, I, I think that, that focusing on the market really is the key thing. I, I, I don't think it's about code or whatever else. It's about like, how do you connect your insight, your understanding that's in your brain with the market and be able to deliver value? But that's the job of the software. Software is a bridge. Between your brain, the supply, and your market, which is the demand. Now, I don't think that the, the no-code versus pro-code versus AI code or whatever, I don't think that really matters in terms of like the, the necessarily the quality of the bridge. The biggest problem is like, are you targeting the right market? Are you actually connecting it with your brain? There's so many people that sort of get lost, right? In like, I'm trying to create a template or I'm trying to create something that looks like a SaaS as opposed to how am I going to create really radical value for the market? And I think someone could do that using any of these three methods. The reason why the difference, the difference in value between the three methods is that now more people can do it, right? So more people were marginally more were able to use the no-code platform. The problem is the no-code platforms didn't really fix the problem of arcane knowledge, right? You had to train up on whatever that platform was in order to take advantage of it. This is actually one of the reasons I like Xano as a backend system, because Xano until more recently has been largely about helping, you know, just creating a bridge to think the way the computer thinks. So you can think that you can do it faster in Xano. That was sort of the big advantage. The no-code tools, I think, are even better because of the point you were making before. The person who knows how to give developers good instructions can give AI good instructions.

Now, when we say no-code, we actually mean vibe coding because now it's official. We are here on No Code Pro Code YouTube channel and we make the rules about what happens in no-code and we officially file a declaration here. With this video that, uh, Vibe Coding will be now called no-code because it's the new version of no coding. That's—

there you go. Oh goodness, you know, I, I think—

yeah, I, I do think they actually go together, you know, like, hey, they all wanted to rebrand, right? Like Nordcraft, Xano, they all said we're visual development. They can get visual development and then we'll give the no-code term because they don't want no-code anymore. We'll give no-code to the AI tools, right?

So Everybody has what they want. No code, no code, live coding, and visual development. Terrible. I don't know why anybody's excited about these because none of them describe value, you know? Like, I'm actually a fan of the, the, the— and it has not taken off because I'm a boring guy, uh, accelerated development, right? The point is speed. If it delivers speed It's great. If it is not delivering speed, what are we talking about? Because the difference between, you know, the, the, the, the, you know, that guy who is giving those developers those instructions, right, and giving AI those instructions, this developers would say, yes, sir, and come back like for their, their check-in next week with like, here's a, here's, here's a status report. The AI grunts away for 20 minutes, and then when he comes back with his cup of coffee, here's Here's the next round. Now we can go again. Jason Lemkin made the same point about working with— I think about working with Replit, uh, where he was saying that like when he had to give, uh, developers the instruction— this is like an ex post, uh— when he had to give developers instructions, he could wait months, maybe even a year before he saw the features he really cared about. When he gives it to AI, minutes.

Yeah, I do think The future of development will be people who have ideas, who are very good at what they're doing and they really know their idea in and out. They just don't have the technical capabilities and they can just address and talk to AI exactly about what their idea is and have it built. I do think most of the times, Lovable, they oversimplify the process Right? They say, you know, everybody can do it. Everybody can do it, but it does take effort and you do have to be willing to learn. But that also applies to no-code. That also applies to Bubble. Like, you can't just go on Bubble today and build your application and have it launch tomorrow. There's going to be a little bit of learning, right? Even though Bubble made it look like that— build an application, no code, no experience, ship quick in a day— they all say the same. So We have to look at this and say, okay, there's always a little bit of marketing garbage on every advertisement statement, and certain parts are being advertisement, certain parts are truth. There's always going to be a disconnect. But at the core of the idea, at the core of the product, it does seem like, you know, AI-assisted coding is probably closer to the process, to the promise that it's making, than its predecessors. We got another comment. I have been building my application 2 months ago and learning everything from zero using Xano and WeWeb. I have no experience with codes and I'm basically all the time studying to apply my ideas. I think that's great because if you know how Xano and WeWeb work, you know the basics of programming, you know the basics of structuring an application. Then if you ever need to go to an AI to get help, you will have a better way to articulate what it is that you actually need. I think that's right. I was talking with someone yesterday and we talked about Nordkraft and that Nordkraft is really passionate about making it for education and all that. He said, they should sell it to schools then and not to developers or to people who want to build. If it's about education, they can sell it to schools and then it can be used in classrooms to learn about HTML, CSS, and JavaScript. And I thought it was fun. Like, I think those no-code tools in the next 5 years are probably actually going to have more applications in education than they will actually have when it comes to production. You know, do you think about that? I'm thinking kind of like Swift Playground, you know, where— I don't know if you know Swift Playground, where you can learn, right? And I think WeWeb, Xano— and maybe not Xano. I think Xano is the one that has the biggest life expectancy of all the other no-code ones, but, uh, the other ones I do see them kind of like right now as Swift Playground where this is where you learn it, but you're not going to build iOS apps in Swift Playground. This is where you get started, right?

Yeah, I think— I hope you're right about Xano. I've been seeing some really impressive alternatives on the backend front coming out.

Also visually, where you can build like with a kind of like a function stack. Which one?

Let me know. I'll— Okay. So one I find particularly— well, there's a little bit of the function stack part, but here's the even more interesting angle on this. I think it's like there's a company called VibeFlow, uh, that's actually— oh yeah, we talked about it. Yeah, they graduated from Y Combinator, and, and the fact they are code generation versus being their own runtime, I think interestingly, you know, separates the concerns. Um, and like, because that's also a big thing for the, for the VibeFlow tools, right, is that they're, they're good build tools. They're focusing on being really good build tools and not trying to also be good infrastructure tools, which I think is critical. Like, they, they reduce the number of concerns. They can move faster with that. The, um, but, but when I think about like the, the vibe code tools, uh, the one of the tricks is how long do you have to spend working on it? And I do think it's a matter of education because the whole, you know, that, you know, we're going to teach people to be better developers, that's fine, you get an hour.

I found a vibe coding tool the moment they're going to launch. I'm going to like, this is the tool I'll be moving everything to if I need to. I think this is really impressive. I want to show that to you. I want to get your opinion on that. I'm not getting paid to say this. I really think this is the best one I've seen.

Do you see my screen? I see it in tiny form.

Yeah. Okay, it's called Radian, and it lets you design, develop, and ship faster with AI. But look at this canvas. This is a combination— the best of Webflow the best of Lovable. So it's like you take Webflow and you, you know, you, you, you marry it with Lovable and gets a child. That's essentially what that is. And I really like that you can visually style stuff and then you can vibe code the thing, and then you can— but you can switch between chat and editor. It kind of reminds you of WeWeb, but it's full stack. So you have canvas, backend, and settings. I think it's really impressive. Look at that, you got the complete product development platform. We optimize Radian with the visual tools needed for more balanced approach to app building, helping you take over. You can visually create tasks and all that stuff. Then you can have either accelerated development, but you still have the visual control of it. You see you have your containers, your layers, your styling. You can really make it look the way you want because The biggest thing I hear when I tell people, hey, you should use Loveable, they're like, yeah, but I'm a designer. I want it to look very specific. With this, you can outsource the concern of logic backend technology while still having it visual, but you can really style it in interface and it's very close to Figma. I think this is very cool. They are going to be soon in beta, so I hope that they're going to give me access to it. But it's really cool. You can bind data. So like you can visually bind data and you can make it a list. So it combines the best ideas of WeWeb, Nordcraft, Bubble with the power of Vibe Coding. And then the features are amazing. You got a built-in fast database with edge functions and all that kind of stuff. You got integrated authentication, you got server functions and APIs. So you can, it even takes on Xano possibly. And then you get fast page load times. So you have server-side rendering. SEO. This is really like, this is really competitive to Nordcraft. I think if this is, if this launches, it's probably over with them. You got custom domain, you got branching. Again, another take on Nordcraft, um, because this is like the main differentiating factor which makes Nordcraft very attractive, is that you can branch and SEO benefits. This is kind of like what makes people move from WeWeb to Nordcraft. They have that. And then You got starter templates, you got cron jobs, and you got code mode. So I do assume they don't have their own proprietary runtime. We're probably also talking about a code generation tool such as VibeFlow. You know, I was looking at VibeFlow and as a designer or a design-focused, design-obsessed person, I did miss the visual part for the design. But being able to visually build the database, visually bind data with the database that you're building, visually style it, and then still see the code because you can obviously switch between code and no code, I mean, this is like my prayers have been heard. Right.

Assuming it works.

It has to work.

There's a Dilbert comic strip from back in the day about office culture. There's this one panel I remember that to remember all my professional career, um, where, uh, where I think it was a pointy-haired boss who says, it was such a good idea in my head.

I mean, to be completely fair, I really wish this thing to work out because it sounds too good to be true. Like, imagine, think about it, you got the flexibility of Webflow with the power of Vibe Coding. Oh, and that's coming, you know, and you got full stack, you got full stack And you got branches. Like, this is a dream. Like, if they do this, I think Nordkraft needs to really pick up the pace if they want to. And WeWeb. I think both of them, Nordkraft, WeWeb, and Bubble, all the, the, the second generation of no-code, they really got to pick up to compete with those.

I think you just put your finger on a really important problem, which is pace. Um, I've been seeing this in the, um, you know, I, I talked to these, uh, you know, low-code teams, um, I, I'm sort of hearing the same thing of, you know, that they have multiple priorities and they are a small team, so we can't really do this that fast. Uh, and so, and that becomes a reason like, hey, we're moving really fast, you'd be amazed what we're doing with a small team. Fair enough, but I don't care, right? Uh, as much as I might love these companies and people who work at them, the, um, the, the, the pace of the market, that the market moves at its pace, it doesn't care what other burdens are carrying around.

But I also do think If you think about it, what stops Lovable on adding a Webflow, Nordcraft, WeWeb-like interface on top of it? They definitely have the funding. They definitely have the engineering resources more than any other no-code tool out there. What's stopping them? The fact is that they don't want it right now is probably because they see it as a limitation to their goal that they're trying to accomplish because they know the moment we add too much of a visual interface to it, it's going to limit what the AI can output and it's going to put the user in boundaries, right? The only way you can get visual control in a development interface is by putting the users in boundaries because you can only do this but you can't do that, right? That gives you visual control. So I do think Lovable intentionally is cutting back on the whole visual control because while WeWeb AI in comparison follows the rules of WeWeb better and produces a good result for WeWeb, we can't overlook that it produces a good result for WeWeb. Not compared to general code ability that exists in the market, if that makes sense. Where WeWeb AI generates the best version of a WeWeb app for you, Lovable generates the best version of an app for you, which is a difference. It's not within the constraints of the company itself.

Yeah, that's a really good point about like the rails that these things have. That said though, I think there's like another level of this. Which is the AI-based tool— what the AI tools provide that the, uh, the, the no-code tools don't, at least in their current iteration, uh, is to your point, flexibility and dynamism. And that's actually something I'm talking about a lot with these companies, like how to be more dynamic, right? Um, because the future of software is in dynamic— is being dynamic. Like the, the idea that like AI could generate software so fast actually kind of implies software doesn't need to last nearly as long. Right? Because crunch all you want, we'll make more software. We'll give software that's good for the next 5 minutes, then for the next 5 minutes. It doesn't need to be the same piece of software serving all the way through, right? That it can be dynamic based on your circumstance. But you can apply that even to the idea of the build. Like, uh, you know, actually, this is a project I'll be working on a little later tonight. Um, I'm working on actually, um, educational BringRout video thing. And the, um, the— what we realize is we're— we are bytecoding tools to help build the tool, right?

I do do that too. I, I have to do that too. I built myself a little, uh, bytecode tool to validate, uh, keystore files when, when people upload a keystore, so we can just go in there, validate it, and then, you know, make sure that everything is okay that people upload. Because if you're like letting people upload stuff as a file and then run it on a development server, there's quite something they can do with it.

So, so just roll that tape forward just a little bit, and what you get is vibe coding the tool that provides you with a visual interface for being able to manage the thing that was just vibe coded for you. You then use it, apply whatever intent you meant you, you need to do to fix it, and then you can go back to the AI, it goes rebuilds. The tool doesn't need to last.

Yeah, to be completely honest, I do think I could build probably a copy of Webflow inside of Lovable in a day. Like, I never thought about it, but I just thought about it. I was like, if I don't— like, what I don't like is you have to render lists manually. You have to. But I know how the DOM works. I could think of building my own runtime. I mean, I did it a few times already. So I could probably go into Lovable, come up with my own JSON-based programming language, build a visual no-code tool with a preview for it in a day, and launch it as, I don't know, a mini LinkedIn profile builder and then hook it up with Xano or Suabase.

Yeah. And then how about like if you can, if you can make the builder or either the components of a builder, yeah, that then is being assembled on demand based on helping solve the problem?

Because like, hey, what I think would be really cool if Lovable adds the functionality to have web components like Notcraft did. I think that is going to be very interesting. So you can build your web component in Lovable and then ship that hosted web component into your Webflow website or anywhere else where it makes sense. So you can compose technology together. I remember when we had the video with the WYSIWYG stuff and with kind of like the fallout, I remember Robert in Office Hours asked, what's the best way now to build an application? Because he said, obviously, WYSIWYG isn't the best way to build a full-stack application at this time anymore. And you said, It's about composing multiple frontends together into one frontend. You don't have to think linear where you have to make a precision choice, this or this or this. It can be this or this or this and this. And that really struck with me. And right now, if I evaluate what technology I'm using to build an application, there are certain factors that make me move towards Nordcraft. There are certain factors that make me move towards WeWeb. And there are certain factors that make me move towards Cursor. And there are certain factors that will probably make me move towards Lovable. For example, with Lovable, if I'm validating an MVP, I'll use Lovable, right? If I'm building some JavaScript framework that will be used by my Lovable application through an npm package, I'll be using Cursor. And if I'm using a building kind of like an internal tool quickly, and I just want some prebuilt templates, and I need to control the way the calculations work, and I want to make sure I have control, probably want to do WeWeb with that to just quickly spin that up if it doesn't have authentication, just to get a dashboard going. And Nordkraft I would use if I'm building a very complex application, like a complex CRM, a complex no-code builder, a complex visual interface. Um, Nordkraft I would use for complex custom— if you need custom data visualizations, that could be a no-code editor, that could be some complex rendering of charts, graphs, those kinds of of things. Whenever I need to be in the driver's seat and I can— and can only be me in the driver's seat, but I want to do it visually without having to scream at a computer because it doesn't get me. So there are certain things that are easier just to do than to tell AI because you don't even know how to express it. I think Nordcraft can also be very great for the discovery phase and actually figuring out what you're trying to do because you can do it visually. Sometimes it's just easier to doodle on a screen.

That's an excellent point. You know, actually, I— that's a great idea for a vibe coding tool or a vibe coding kickoff. Doodle on a screen, you know.

And we should, we should typecode one in Lovable, and we'll call it— we can't call it Radiant anymore because it's already taken.

But, oh, and then you know what you do? You don't have to generate the code. Oh, what you do is generate the prompt.

Oh, exactly right.

Sketch a prompt. From sketch to prompt. Yeah, well, there's a thing called prompt folding. Have you heard this term before? No. So prompt folding is basically prompt improvement. So there's actually something called prompt fold. I think it's even a service. But the idea is you can give it a prompt and give it a bit of your context, and then it basically makes a better prompt. Like, I think you kind of meant this, right? In order to basically be like a whisperer to other AIs. And then does it sort of in that sort of, you know, lots of excess word, you know, AI language that AIs like to talk in. They also like to be spoken in, spoken to in, and it does produce better results. So the idea we're talking about is basically an AI engine for creating better prompts. That, that seemed very, you know, meta for Software 3.0. And I think there, there's a lot of money to be made in there, I think.

I think there— that is probably a very good point. I am really stunned by what you can do in Lovable because the ability for me to just say, hey, do this and follow those rules, and it actually does it really well. Pretty surprised by that. Now, would I build Despia in Lovable? Probably not, because with Despia It's my business. It's my baby. I want control because I know exactly how it has to be better than any AI. You know, I do use AI to come up with ideas for stuff that I don't know yet to learn about it. I mean, again, it really depends on how you see the business that you're doing. If the business that you're doing is something you heavily rely on, you need to know everything because, you know, you need to be able to manage a team. You need to at some point then tell people what they got to do and you want to be in every business process, visual development can be great because it will allow you to move faster than code but still have the own ability and still be able to move stuff. If that is a very important business to you, that also requires what is your investment, your time investment towards your business. If that is your passion, what you love to do, and you have the ability to do it the best by expressing yourself visually because you're a visual person Visual programming is just fine. That applies to me and Despia, right, for the front-end interface. But if I'm trying to build a calorie tracker application just for the sake of getting something on the App Store because they have a silly idea and I think I can get 500 people to sign up for that, why go through the loops and hoops to build this visually? By the time I'm done with it, there are already 5,000 competitors on the App Store. I just might as well just do this in Despia and Lovable. That was hidden product placement.

Let's make the subtext text. Could you just put up the QR code for Despia?

I appreciate it. Where did I put that?

Look at that. There we go.

Right here.

It is good and impressive technology and I'm so glad for your success finding the 5 coders. I think you're going to find there are lots of people who are comfortable making a web app. It's easy to develop, sort of fast.

And then I was actually surprised. There's one web coding company that is selling like their React-based programming tool to build mobile applications. And I was very curious. I was like, why are you not using Expo for this? Because you can just use React Native. And they were like, oh, React Native is a pain. Like you're not really getting it on the store. You got to work with the Expo Go app, and it's not really a great experience when they need to test stuff. And, and they had some troubles in setting that up, and it wasn't kind of like a real native-like nice experience, like where you just push a button and it's on the App Store. It was— you had to go through quite some hoops to get it there, right, to set up external services. They couldn't fully white label one-click publishing. So we actually created something that is Despia White Label, which allows people to use our Despia technology for their existing vibe coding tool to just plug it in there as a REST API. So your AI tool or your no-code code can generate a React app, a Vue app, an HTML app, anything web-based app, and then you send that URL with the configuration JSON structure to our API with the API, like with the administration access, like the app manager key for the App Store account, and we'll build it and deploy it directly to the Apple Developer account or Google Developer account with the push of a button. And they can just plug that in. So they're actually now in the process of getting that into there. We already have another person who's actually just using Despia technology for their startup, for their business, into their application builder. But I find it's very interesting that people, uh, that people that look at Despia see the value in it. And I didn't have that before, before, you know, we were trying to reach the no-code people. The no-code people are like, okay, this is great, but I like when people ask the hard questions, and I'm getting the hard questions and the hard problems with all the people that want vibe coding. And it's, it's, it's, it's, uh, it's, it's nice to me, um, seeing that people actually want to integrate Despia into their product to ship applications to the App Store. We also had the CEO of Lovable create an account on Despia.

That's nice. You know, it's nice to have the little—

the celebrities. I mean, he, to be fair, he sent us a support request saying I can't press the button because while he was signing up, I was maintaining something and fixing something. So there were like 5 minutes where the button didn't work. So I do have on my intercom now complaint about the lovable CEO not being able to click on the button, but I fixed it immediately once I saw that. And, you know, I was just temporarily working on that. But it's nice to see that they actually see the value in it.

Sure. What you have is Product-market fit, man. And like, you, you had, you've had a good product, you've worked hard to make the product as, as, and, and that is, that's exactly right. You have a lot of people look at product-market fit, think I gotta keep on tuning the product. In fact, what you kept doing was tuning the market. Like, you went first from being a little bit more Nordcraft-centric to being—

it wasn't a little bit, it was like hardcore Nordcraft.

Uh, yeah, uh, and then, you know, you got whatever you got out of that, and then you got— went to WeWeb, like, right? And then you go to, like, the next platform up that's the— which is not only, you know, uh, you know, uh, newer, right? But it's also have a whole lot more users. I mean, like, the, the number of users in, uh, Lovable, right? Number of developers. Number of developers in WeWeb looks like this, and the number of developers in Nordcraft looks like this, right? And that's not a statement about the quality of any of their products. It's just a statement of like the size of their book of business. And if you're offering something that's going to be a multiplier of some fraction of their book of business, well then going after the bigger market is going to make you more money than one going after smaller market. And that, that just kind of stands to reason. Now you, now you've got that, it's like, okay, great, what's next? Um, I, I, I think you're going to find even those platforms that also support their own path to mobile You will be attractive. Uh, you— they will be perfectly happy for people to be using Despia because in order for them to use Despia, they need to be building, they need to be hosting.

So, uh, I talked with people who were using Bolt.new and they have like this React Native Expo integration and, and they were like, we don't like it, it's horrible because with Bolt.new the integrations aren't that nice and, it's not that nice of a process to get it in there. Where with Despia, we spend so much time being able to have one API that just does everything for you, from building, signing, publishing. And once it is on TestFlight, the only way you test is through TestFlight. It's on there, and then it's one click to get it on the store, and you meet all the requirements, and it forces you to follow the requirements, and it, you know, forces you to implement best practices, implement some of the native best practices when you're building hybrid applications automatically. Correctly for you. And we implement those things because it was made for no coders, because I didn't fully trust every no coder to set it up correctly. So we do have some magic happening on the backend that is making your application feel extra nice. But now that is actually a big benefit for all the other actual development tools who are building actual code, because they see the benefits of being able to, you know, if it is already a little bit complicated to build with AI, at least it should be simple to get it into the App Store.

I concur. And then, and now look at all these other build tools that you can be creating a video on. Hey, let's talk about Despia plus Bolt, Despia plus B0, Despia plus, uh, Radian, the one you were just showing us, right?

I mean, I'm really happy about Radian. I wanna, I wanna get access to that.

That sounds like, that sounds like a dream come true. Uh, and then, you know, whatever the next dream come true is, you know, there, there are lots of different ways to make these apps. And I, I gotta say there may be some interesting ways and like time to think about as a service.

Less Windsurf, less Cursor, less Cloud Code. I mean, I'm already having the domain, like you can go to web.despair.com to lovable.despair.com. The next one will be, I think, replit.despair.com, and then we'll have a way of doing Replit videos. So it'll be hyper— like, I want people to know exactly whatever they're using, they'll be able to integrate it with Despair. That's the biggest competitive that we have towards React Native where you don't need React if you don't want to. You can use Vue, you can use anything, Svelte.

Yep.

And Next.js, do all the nice Svelte stuff in it.

And just keep on, keep on making that. Hey, I, I, that all what you're describing to me sounds wonderful. Yeah. And I would wish that you just keep on.

And I remember you saying it shouldn't be for Xano, like it should be, uh, it shouldn't be Xano for X, right? Um, and I thought this is very important because right now it was Despia for xTool, right? And now it's everything works with Despia. It's now Despia for everything, meaning wherever you are, whatever you're using, you can use it and you can build applications. Uh, and you know, it doesn't limit you in building it. And then I think also this is a very important important thing that is probably I need to put my effort into advertising is let's say you start out by vibe coding an application in Lovable, but now it is already nicely working and you want to bring it to the next level, right? You can then move it to Cursor and your app will still work and it will still be on the App Store. It will not disappear. And you can move the stack because the Despia framework itself stays the same. Just the origin where the files are being dynamically loaded through updates over the air may change. So with Despia, it, it is your companion on MVP, it is your companion on MVP+, MVP++, MV, and then the production version, the final version. You can use it for the whole lifetime of your product, your project, as long as you prefer building stuff in web tools. It doesn't matter what tool you're using. So you can start in Lovable, you can continue in Replit, and you can finish in Cursor, and you can whatever you want to do, and then go to Norcraft when you're up for that.

It all has— there's a lot of power, you know, in those, in, in, in that promise, you know.

Yeah.

And, uh, and, and this promise that you are, you know, capable of of, of fulfilling. And that's what makes it attractive. And that's why you have Despia, like you say, Despia for X, where there are lots of X's. And, um, and, and for people who are looking at this like, hey, what kind of software company should I build? I think something looks a little bit more like what you're doing. I think that kind of glue aspect where you are providing a key service that provides a lot of value— we're talking about like under what circumstances do you use no-code, right? Yeah. And the, uh, and, and Despia is an example, something where you did use a no-code tool, Norcraft, or at least doing part of, and it, it allowed you to accelerate being able to make a real thing in the market that's making you real money. The, the point wasn't using no-code versus not using no-code.

The point is your connection with market. Yeah, I do 100% agree with that. And I do also think that building software that doesn't limit people into where they currently are is probably the way moving forward because technology is just going to change even faster now with AI. So if you build a product that is like Chrome extension for QuickBooks, it's probably not that good of an idea. It should be Chrome extension for any accounting software. You want to build your product more general because the bigger your market is, the more customers you're going to reach. Because I think I see it most of the times, um, people limit themselves. And I actually was talking today with John, because we had to— because we got some requests from some big companies to talk about Despia with them. So I was like, we gotta— we should probably sync up before I go in there. So I don't say anything stupid. So I was talking with him, and he's my CTO and my business partner. And, and he was saying, why don't we build a vibe coding AI, no-code kind of interface for Despia. And I said, no, it's not our business. Our business is two things: connecting with every AI, vibe coding, no-code tool that ever exists out there, and selling our service white-labeled so nobody knows it's actually running Despia to other people who do what you just told us that we should do. Because I'm rather making money from all my competitors that would be my competitors and all my competitors' customers that would be my competitors than trying to go on my own like WeWeb. Well, you know what the slogan was of the most profitable company of the 1990s? I did not know. You got to tell me that.

I'm too young for that.

Intel Inside. Yes. Yes.

Now it's Despia Inside. That's right. The, the, the, the secret success is being great middleman. That's from Layer Cake. And it's, uh, and you can— that, that, that is a great position for you to be in with state change. Actually, I'm looking at the same thing. I've got like our, our system that we just made like last— the end of July introduced state change system, which was, uh, which works amazing.

I've been testing it. Everybody who is considering building AI-assisted vibe coding applications, you gotta join Ray's community. It is what got me to the place where I am right now, and I am the living testimony.

I super appreciate that. And, and yes, I hope people do join. Um, the, the— but that assistant, its, its obvious future this fall is not to be an interface, but to be like a component people are using as part of their other interface. This is like where MCP becomes interesting and to be experienced as an MCP.

So I think the moment you imagine Lovable would let you connect— I think they're working on it— if they let you connect to an MCP and you can get Ray in Lovable, like you can have like your little Ray living inside of your Lovable, uh, application, that would be amazing. Because then Lovable isn't fixing my mistakes, it's Ray.

And he's way smarter. Well, that's the— it's on their roadmap to have MCP.

And yes, that's the idea. Like, that's good.

We can make a Despi MCP as well. Yep. Uh, and like, the— I think the more you can integrate those things, I mean, there's just such power to it that instead of saying, oh well, I— you need to use my app, but then you're going to use that other app, the other app, like those apps just come to you. Uh, that's that idea, like having the component that's sitting on the inside, it may not be as glamorous, but delivers a lot more value. And that, that's like the, that, that idea of componentization, along with the idea of dynamic software is the big, the big trends that I'm watching for the end of the year. Right.

I also think it's very important to, you have to prioritize when you're building software, you have to prioritize on where the market is. It's like we were, they made a complete 360 on what they, everything they stood for. But you have to do that if you don't want to fall behind, right? For example, with Despia, we're not too concerned right now about, you know, writing new documentation, making new videos for our JavaScript SDK. We're more concerned about adding new stuff to the WeWeb plugin, making more videos about the Lovable, the, the AI-assisted development integrations, than, you know, pursuing on our plan to integrate with Bubble because it's a dying market and Norcraft is a very small market. You can integrate every— all the new features that you have available in Despia into Norcraft with a line of code, but it's very unlikely that there are going to be any videos about Norcraft in the future and how to do specific things in Despia just because they're going to be videos be made about how to do it in vanilla Java. JavaScript, and that will give you sufficient information how to set it in Despia. And we're going to add in Despia, in, in Nordcraft, a command function that will be like command and callback, and then you can define yourself what you're getting back, like the JavaScript SDK. So I don't have to make a video for every native feature anymore, and I can focus my time on the market that actually creates revenue and not on the market that just costs resources and doesn't create revenue while still not having people that were in that market and purchases stand in the rain. We're going to keep maintaining what we're doing, but you know, every video, every tutorial you make ultimately is not essentially a tutorial. Every tutorial, every video you make is more— is an advertisement essentially, right? And yeah, our focus is on making as much advertisements for those AI tools because that's where the market is. And they will support us making at some point more for the no-code ones. I mean, WeWeb, I'll count them in as an AI tool, right? Because they do have AI, and I think that is very appealing to most people. But if you look at the signups, we have like 10 users using Nordcraft, and we have hundreds and hundreds of them using Lovable and WeWeb.

So Yeah, I think we, we get to, to end where we began. Like, what is no-code? You know, we talk about live code and no-code, and like the visual development kind of is no-code, but the, um, but, but what you're getting to at the end here is that no-code is no-code, which is to say doing your marketing, right? All this stuff is about getting the technical part out of the way so you can focus on things that drive the business. And in the business you're in, there's a technical insight that you There's good technical insight, but then a lot of the rest of the thing that drives value is content, telling people how to use it, and then distribution, which is the marketing. And fortunately, those two things are really, really well aligned. And now because of the glory of whatever you wanna call no-code, right? The, the, the, the low visual development stuff, the AI assisted development, the whatever else, um, you're able to focus on the part of the business that, that really matters and creates value.

Yeah. And I also do think generally, when someone is choosing a tool to build their business on, you probably want to choose a tool that is going with the future and not going with the past. And I think that is very important because you don't kind of like want to build your business on something like Builder.io, where, you know, the project is kind of like sinks down. I mean, you can rebuild and you can, you can iterate over that. But I think it's very important to consider that you're not married to a tool. And when you build something on, you're going to probably rebuild it in something else in a year, in 2 years, in 3 years, right? And, but also, if you will have to rebuild in a year or 2 years, what's the point of going with the tool that you probably already know won't be there in 1 to 2 years for anything that comes your way? Right. And I think people should really look into AI-assisted development. I mean, there is a reason Ray is using it. Oh yes. Right. And there's a reason why I'm now completely amazed by it, not just because it's a bigger market, but also because it actually adds value to it. And, you know, the fact that it is a bigger market shows that it's adding value because success has traces, right? So I think it's a very interesting topic, and I'll of course I'll keep Despia in Nordkraft for now. I don't know what will be the best tool in a year, right? And because maybe Ray Deck, it looks very promising. I like also that it includes the word and the name Ray, makes it very promising.

Well, as, as the other Ray would say, you know, and we said at the beginning of this You know, I think all that is true. Like, the tools can come and go, but you will build this again.

Yes. Wonderful.

It was a pleasure with you today. Same here, man. Look forward to it again. Take care.