Software 3.0?

Hello, Ray. How's it going? We haven't been doing this since for a while, uh, because I've been hunting iOS bugs, but yeah, it's been a couple weeks.

I'm so glad that we get to do this again.

Fantastic. So today we have a very exciting topic, Software 3.0. Yeah, how do you want to start with that? Because you came up with the whole idea.

Well, yeah, I mean, last month, um, Y Combinator, uh, over the accelerator of San Francisco, um, held what they called AI Startup School, and their keynote speaker was Andrej Karpathy. And he gave a just marvelous talk. I think everyone who's serious about building for AI should go check out, you know, wait till the end of this vidcast, but, you know, then go check it out on what he calls Software 1.0, 2.0, and 3.0 and like what that means for how we create software and metaphors for thinking about it. Sort of calls it the new operating system. And the, you know, the crux of it is he said like for most of his career, right, and indeed most of the history of computer science, We've been creating code, right? We've been writing these instructions, the computer executes on those instructions, and the languages for making those instructions have gotten better and better and to a certain extent easier to work with as well, from like punch cards to assembly language to C to JavaScript to some of the visual programming control planes we've seen in the course of the last 5 or 6 years. But, he says, A second thing has happened, which is that starting actually about 8 years ago, so even probably more thinking about low-code, software programming changed dramatically from writing code, which is a set of instructions, to setting up structures which would allow machines to learn themselves. It's about the numbers. Then this led to things like computer vision and basically trainable models so that instead of writing code, We wrote structures that allowed computers to learn essentially what the proper instructions were in terms of computations. And but even that, you know, was, you know, it's less explicit but still relatively hard to do. And then just in the course of the last— it's been, what, 33 months, I think, since ChatGPT came out and the— and probably, you know, good maybe 5 years since we've seen like GPT-2. And in that time, really just in the last maybe 18 months, you've seen an explosion of what he calls Software 3.0, which is using natural language to describe to the machine what we want it to do. And then it does those things. Not like AI-generated code, but the AI is the code. And that idea of going to, you know, that different way of describing to the machine our intent so it can execute for us. Is what he describes as the new operating system that we're all working within.

I got like— can you see the meme I put up here?

Oh yeah, yeah, yeah. And I think this raises a really important point that like the point of Software 3.0 isn't the same thing as vibe coding. Now he's actually the guy who invented the term vibe coding, or he coined it as, you know, in a tweet he did like whatever it was a year ago. But this is about using, if you will, natural language and intent to go create old-school code, as opposed to the new-school stuff is the code. I think that that idea of we're not trying to just generate the same old code that we're making 10 years ago just using new tools, but rather this is a new way of working with machines is what makes it really exciting. That's why he calls it Software 3.0.

I mean, to be fair, the lower one, that's more like me if I ever— every time I tried Lovable, but I mean, there is an interesting point to be made in also creating code that is primarily used by AI, like Xano doing right now. Like Xano, I think, I don't know, XanoScript is probably one of the first and few programming languages which have only been created for the intent of being used by an AI. You're not going to write XAMLScript yourself.

That's what I was saying about a year ago when, or maybe 9 months ago, I think is when they announced it, but a year ago is when I first heard about it. I think that's its rep and we'll see how that works out and what code the AI really wants to work with and how optimized it is for it.

I saw that Xano is really like pushing Lovable to integrate their XanoScript with it. Like Daniel has been on X and he's been saying, Lovable, you gotta add Xano as an integration. Like Lovable has been like, why should we do that? What's your use case?

Which is fair. And also a question of scale. I think there's a thing in the world of low-code, no-code that like has an enthusiastic fan base, but a really small one. And the, um, and, you know, Lovable says, hey, how can we point this thing towards big markets? And, and that creates an interesting opportunity because I think Xano certainly could be a wonderful complement to, you know, vibe coding type platforms like, uh, like Lovable. Indeed, you can even think about, sort of think differently entirely about how to code with it. Uh, as a market opportunity, it's still, it's still a minnow in the ocean. And the, um, and that, that leads to a little bit of a you know, a circle, right? They need to get around, right? They grow, get more attention from those folks. But then once they get that flywheel going, I think there's a lot of opportunity for Xano to really take off because of what you're pointing out, like the fundamentals of the product really lend itself to be successful in this domain.

Yeah, I mean, I think nothing is— I mean, actually, I just wanted to say nothing is stopping Soupbase from doing a Xano competitor. If the market wants it, I think there is a very narrow market who wants Xano, but I think that market matters a lot because we're talking to people who have the ability to think technically and turn those things into procedures that a computer wants, but don't have the ability or the time because their time is too valuable to write the code by hand. It's a very niche product if you think about it. It's like, it's almost a luxury item. It's like you're an expert at the specific thing you want to build an application for. You don't want to write the code yourself and you don't trust another developer to do it as good as you would do it yourself with XAML. It's a very niche market and it targets only a specific person, specific kind of person, but I would say that that person is probably willing to spend a lot of money on it. And making the barrier to entry easier and removing a medium plan like Xano had for $9.99 and pushing everyone who pays $9.99 now to $2.49 makes sense because it's a niche problem that they're solving. It's a niche market. And I think that is ultimately how we build software because most of the use cases, everybody I know that's building remarkable stuff in Xano is building their engines in Xano for a problem that they're an expert at. So it's essentially software for one specific group that they're generating. So everybody now has the ability to create software no matter if you are an engineer or not. You can be an expert in real estate, finance, and you can build this like Kamyar. And like, he's my prime example. I mean, he's not using Xano anymore, if I'm not mistaken, but he started out with that.

Well, and I think that, that creates sort of a category, uh, competition question, right? Of like what is the value add of these low-code environments if you can be getting assist, right, from AI? And certainly our friend Cam here certainly uses a lot of that and gets a lot of profit from that. He's a guy who did not come in with like any sort of formal technical training ahead of time, but he climbed the curve. He worked his way through a few different low-code platforms, learned more about how a machine worked, got some help from StateChange, little plug for StateChange there. And then he went, and then he started using, you know, things that would help him create code as he was looking to do more advanced stuff. And then Cameo's story is, I really think, an interesting one because, you know, the history of SaaS, as people might have described it over the course of the last 10 years, is really about 3 things. There are 3 kinds of resources that cloud providers really give you that SaaS then sits on top of, and those are storage, networking, and compute. And for a lot of the history of what we think about for SaaS, most of the game has been in either storage or compute. You can think of Netflix as storing a bunch of movies— as a storage and networking, because you can think of networks as storing a bunch of movies and then delivering them over pipes to people. And this caused a lot of controversy back when of, you know, who's gonna actually pay for all these Netflix subscriptions are using up all these, you know, cable internet pipes and whatever. And the But the AI stuff really changes that. And where Xano, I think, really specializes— and Supabase actually would be an example of a company that's really good at storage, right? Because they are in the business of providing databases. The base in Supabase is supposed to be for database. And Supabase was kind of a riff on the name Firebase, which in turn was a hosted database provider. Supabase was like an open source alternative to it. And now it's sort of a relatively popular cloud-based alternative that runs on PostgreSQL databases. Databases. But that's what they do.

They provide databases.

Now, they have some ancillary stuff. They do like serverless functions and stuff on the side. But their core is they are a database company. You have a database, you want tables stored up in the cloud, you know, Postgres, Supabase becomes a great way to do that. Xano, though, I think is really interesting because Xano has a database, which I would posit isn't very good. A lot of people who are growing up their systems, I'm sort of helping them migrate off of Xano because the Xano database is not world-class, but their compute approach is world-class. And this gets to what you're talking about, like what, what Cameo and some other folks are using it for— the engine part, right? The engine isn't the gas tank. It's not where you store your energy. It's where it converts energy into value. And that transformation and that execution is really where the value is. I was talking— I was actually talking with the VP of marketing over at Xano a little bit ago describing to him a particular market opportunity I sort of saw for it, and he said, "Oh, then that'd be great because they buy more compute." Because even Zano sees that they're really in the business of selling compute far more than they're in business selling storage. And in the AI context that we're living in, the difference between AI and older versions of software is older versions of software is basically what we called CRUD, right? Which is a a mnemonic for like how we interact with databases. And so you have a database and so you have a front end for it and data must be flowing between the two.

Yeah, right.

Right. And that's where Supabase really shines because you have storage and then you can have somebody else providing the front end and then everybody's buying a Supabase database for it and Supabase is happy all the way to the bank. But where Xana sits is in the middle of that, right? Is the providing the value-add over the storage of that and the compute and the analysis. And I think we're all of greater appreciation for what, you know, we can do with that because of AI. And I think if they can grab that opportunity, which right now is smaller than the rest of the market, but it's where the growth is, that I think there can be just tremendous potential.

I tested their agents and it's very wacky. It works today and it's broken tomorrow and it works today after tomorrow. I'm not getting constant results from their agents and it feels like something that was shipped for the sake of being able to show it. And before they showed it on camera, they had to prepare 3 hours to get it right for the 30-minute presentation. That's how their new agent features feel.

I give them a lot of credit.

Great job. They did a great job. It's just, it feels like they shipped it without doing like very, very rigorous QA. Yeah.

You know, I, I, you can, you can criticize or praise the, the quality control and the speed at which they shipped, but the, um, but I do think they took a risk and I think it's the right kind of risk. Xano certainly should realize that, like, if they did not make a change, people were just going to move, move on from it. Um, and that this allows them to be leaning into where the future of software is, and that frankly plays to their strengths. I do not think that Xano's agents are a good competitor to Flowwise or to LangGraph or to ADK or any of these other, you know, really world-class systems that have been working on this problem for a while. It looks very much like, I'd say today they're maybe about 12 months behind where the state of the art is from an agentic point of view. They do not yet have a UX theory of how to make that context-driven development, i.e., the text, like a first-class thing.

It should look more like Gumloop but with multiple function stacks put together.

Maybe, but, but Gumloop is an automation system much more than an agentic system. And Xano, I think to their credit, is saying, well, maybe we need to leap over. We can't compete with the automation systems because those are automation systems, even AI-enhanced automation systems, because there's N8N, which is getting a lot of ink right now, but Zapier, which is doing it really well, and they're a big company. And then there are a bunch of players on the enterprise side too that are really quite impressive.

It's interesting that it ended like a complete revival because I remember 3 years ago, N8N, it was kind of like this, like the little brother of Zapier who nobody really was using, but only people used because it was cheap.

Or because they could self-host it. That was kind of the rap on it.

It was like Zapier used by developers who like self-hosting stuff on their own computer or cloud infrastructure. Structure.

Yep. And, you know, and now, and I think that that shows the advantage of not, not be fighting for income because Make and Zapier were fighting to build out their businesses. And then N8N, which sort of was adrift in that market, I think had the room to say, well, let's try something wacky. And they did. And I think they've done a good job. Now, I don't know about Make, but certainly Zapier has come around. Uh, and so like, it's a, it's a, it's a, it's a straight out knife fight, I think, you know, for, for doing that stuff well. And they're all getting better at it, right? And I think that the winners are the customers.

It's pretty much still the same how it was 5 years ago.

Which, Zapier?

No, Make.

Oh yeah, Make is Make. And, and the difference between Make now and Make 5 years ago, I think one might fairly say, is the name. Right, it's called Tegrobat as opposed to being called Make.com.

It's one of those products that never has changed, like Xano in the first 3 years.

Well, there's a little bit of, and you know, my apologies to our viewers from this part of the world, but it is a little bit of a European thing that, you know, American companies tend to change their identities and pivot much more sharply and much faster. Make is a European company. And then there's Norcraft. You know, I just don't worry about the front-end companies like Lovable. I mean, I, I think very well of people who are doing this.

There may be a backend coming to Norcraft.

Oh, very cool. I think very well of Andreas, but I, I think the category of these front-end builders is it's not where the future is. Because I look at the future of front-end as being starting to be smaller. I'm starting to work— I don't know if you've played at all with Comet or Dia or Neon, you know, these AI-first browsers, which are increasingly—

I'm still waiting on my Dia invite from you.

Sorry, what?

I'm still waiting on my Dia invite from you.

Oh, goodness gracious, I'll get you that. Okay, but the, um, you teased me.

You said this is amazing, you gotta try this. Ray sent me the invite.

It is great. Two weeks later, no invite. You know, I actually just found out where the invite's for just the other day because another friend was asking about it. So now, now I know where to go to go get that for you, so I'll get that for you. The, um, but, but the, the way that these tools blur, right, the experience of say a ChatGPT app or Cloud app for that matter versus that of a browser is really eye-opening. When you go into Warp, right, and the way Warp blurs the distinction between a console program, right, like a terminal program versus an agent, an agenting system is also really remarkable. Like the The way that AI becomes a first-class consideration just changes the way you use these things. It's no longer about having, I've got my site and then navigating to another site. The sites become smaller in front of you, just like these phones significantly changed the way that we use apps.

I love you got a QR code for StageAgent on your lock screen. No, but if you think about it, Norcraft has a very unique opportunity with the web component.

Oh, sorry, what? Oh, yeah. Oh, you know, that's totally right.

Like, I think they're not doing that at the fullest extent, because you can bring a web component into Webflow, you can bring it into WordPress, you can bring it into a native mobile app, you can run— render the web component in a WKWebView on a mobile app and extend the functionalities of your mobile app with a web component that has native access, all those kinds of things. Like, you can do lots of things with web components and that those are those smaller environments, those are this AI chats that you could integrate on your site. That is the Chrome extension that you can build because I have been shipping Chrome extensions that are built in Nordcraft using a web component. So it's just a web component that gets injected in the site and because it attaches on the shadow DOM on the side that it's in, has access to everything on the side and can do everything. I've been doing Chrome extensions like crazy in NodeGraph. They work fantastic. You can self-host them. It goes down from $30 per million requests to $0.30 when you host it yourself on Cloudflare. What I like about it, they have the same position like N8N because it's open source. They do have the ability to shift as crazy as they want to, which I think, you know, maybe something for them to do. But I think those web components are something that a lot of people are sleeping on because they allow you to build great things. It's pretty cool. I'm right now building a web component that is running on a custom Android kiosk, which is essentially a sales tracker for Despia. When it will be shipped from China to my home, I will have a custom Android kiosk that's like a clock, running a web component in it that is a counter that will give me the daily sales.

Cool. First of all, the hardware idea is super cool. Yes.

With hardware, because you can turn a web component into an Android app with Despia. Then once you have it, you can install it as a launcher. We're now experimenting with a new setting in Despia that lets you build your Despia Android app as a launcher app, which is really cool if you're building kiosk, if you're building custom hardware.

That is really interesting. I love that because the more you get off of the, I'm building a traditional web app, the more interesting the business becomes.

No, I mean, this is one of the investments that— I mean, most of the investments that we make in Despia are things that no user ever wanted. It's just things that I think are cool. So if I have that, if I have the, the drive and the passion to build a little sales counter clock for my table, we will go through hurdles to make sure that Despia supports that. I mean, we also added app widgets for the people who wanted it, but the number one person who started a mission That is—

I love the creativity you're putting into that. I like your point about Nordcraft and web components. Now, one of the things I recall about Nordcraft is I remember getting excited about Nordcraft web components a year ago. It just never went anywhere because my impression was that the folks at Toddle, as it then was, We're really discouraging this as a build mode. Is that sort of not the case anymore? Or was I misreading it to begin with?

I think here's the thing, what I truly think about their web component approach. I think when you take a look at their advertisement for web components, it was you could bring this into your Framer site. You could bring this into your Webflow site. I think they've been trying to mooch off their communities to get people started into the editor. That's why they added that component support.

Okay, well, the web component support though, that's been around for quite a long time though, hasn't it?

That's been like the core functionality of, of Notecraft. But I think because they implemented it the moment it was very new, I think they added it because they're like, okay, we need users, how do we get them? How about we'll let users who use WordPress, Framer, Webflow, build their components in here so we get them into the editor.

What's your impression of adoption on that front? Is that how people are using it?

I've never seen people use web components in NerdCraft at all. Never seen that. And it's very sad because I can think of a million SaaS ideas how you can make money with web components. Like you can make an AI chat. That connects to the Webflow CMS and ship it to your users as a web component that every time it gets loaded, it uses like a custom rack that you can build in Pinecone or Xano even. And every time it gets loaded on the page, it's calling one, it's charging this one request. And then you bill your Webflow users who have this AI CMS chat on the page every month per request. So like you could turn a SaaS business for people and solve niche problems in their existing websites using web components.

Yeah, I think that the, you know, the drop-in widgets are a wonderful business that they have had.

Like Elfsight is doing this, you know, Elfsight for Webflow widgets.

I don't know what you're talking about.

It's a subscription. It's like $20 a month and you get access to like 500,000 loads or so of widgets on your site, and you could pick the widgets for calendar, event calendar that can be linked with Google. I think it's a big business opportunity. I actually talked with Philip, who's building a Gantt chart. I think he's also in StateChange. And I said, if you have a big business opportunity, ship this Gantt chart as a widget like Brightum and let people kind of like work with it and sell that to WeWeb users, to WYSK users. Primarily to WYSIWYG users. That's— WYSIWYG users, I found, is the market out there that is worth— that they are willing to spend the most amount of money on JavaScript libraries that make their life easier.

Well, I was gonna say, maybe you could talk a little bit more about the WYSIWYG market, because it's been a little spell since we've had talked about that over on the, on the vidcast.

I think that the WYSIWYG market is very unique. I have the ability to compare markets And I think the Nordkraft market is probably the least amount of market where you want to sell templates and so on and so on, because I've been doing it and I can tell you it's very hard selling in that market. You gotta, you gotta put like 5 times as much effort into selling something to the Nordkraft community than doing it to the Wiz community, because it's a different mindset. People who use Nordkraft are developers. Developers are rather running against the wall for the next 15 days and figuring it out themselves than paying $50 for a template. That's what I found. In Wist, it's like, "Hey, does anybody have any templates to sell out there?" It's like they're waiting for it. It's like, "Oh, there's a new template. Let me buy that," because they're already used to paying $300 to FinSuite for a video component. Component.

That's a really good point. You know, I remember when I was first looking at the no-code market, um, back in, I don't know, '21, the, um, my— the line I sort of landed on about it is that no-code is developer tools properly marketed, because you're not selling to professional developers. Professional developers believe you should— they should be— you should be using them, right?

But you talk to developers That's why it's hard to sell. So, yeah, he is a very tough sell. The fact that we have 50, 60 people in there, it's like nobody would have ever thought that this is possible because we, you know, like the only people that are in my community are not Norcraft's core user base. Those are the people who used Wist who couldn't go beyond what they wanted to do in Wist, and then they saw me making videos about Norcraft and they converted, and then through through that conversion. They've joined the community to get through the conversion, and now they're, they're helping, and I'm helping them build the applications. But I have not one person in the community who was first part of Nordcraft and then found me through making Nordcraft videos and joined the community. It's always the other way around. It's always they are looking for a solution, they want to build an application, they find me, through me they find Nordcraft, They use the Nordcraft, they get stuck, they join the community. It has never been where any of my business came from Nordcraft. It was I had to bring business to Nordcraft, and some of that business that I brought came back to me to be my business. So that's how it works with Nordcraft. With Wiz, it's different. You just gotta be on YouTube and scream, hey, I got a video about Wiz. You got 1,000 views and you get 3 people emailing you saying, hey, how much is your you out of the race.

And you have people who are already primed and sort of, you know, richer for that. It's interesting because, you know, I think of Wiz as being a little bit longer in the tooth, right, as one of these platforms goes. Certainly younger than like DwarvenCraft is. But to your point, that same factor means that there's, you know, more people who are more comfortable with it. They have grown with it. Certainly there are plenty of people who come into StateChange and they launch in with their Wiz questions. And there's a lot of power to it too, because you can just sort of inject and drop in JavaScript wherever you need it. And that has, that's about a lot.

I made a video today how you can add native contact access into your Wist application using Despia. And I'm now starting a new documentation page only for Wist. It's wist.despia.com. If you go on that website, you will see all we have only the appropriate documentation tailored for Wist for DeskJet SDKs. Because if you go to any other SDK provider, they just tell you how to do it in JavaScript. And then you're like, okay, but how do I do that in Wist? And you got to ask Ray how to do that. Now we have specific documentation only for Wist on wist.despia.com. And we also have specific documentation for DeskJet for WeWeb. So you have— we have separate documentation for DeskJet for WeWeb. WISP, separate documentation for Despia for WeWeb, and we're going to have separate documentation for Despia for Norcraft.

And you know, that kind of brings us back to the whole Software 3.0 thing, because Software 3.0 is about describing. It's about using natural language to describe how things work.

Simple idea, because our documentation is made for AI. It's not made for any users.

Well, I think there's a lot— there's a high overlap Venn diagram. Right there, right? Because a lot of the same principles that make it easier for a human to understand make it easier for a machine to understand. Yeah. Completeness, repetition, right? These, you know, I'm not saying it needs to be Shakespearean writing, but it just needs to be out there. And if it's explained and it's, you know, people can go search and find it, which is how people use documentation as well. It's how AI uses documentation and they can read it because it's just written, you know, in English. It will work very well. And I think that what you're pointing at is like the— I think that people who get into the software business, particularly in software tooling business, spend a lot of their energy creating code, right? When in fact, the way that's going to be able to connect to other systems is going to be through that documentation. And I don't think you need the same sort of documentation that you needed before necessarily, But you need a lot of it because that's going to be what the AIs are hoovering up in order to be able to put things together and create, you know, the compositions, right, that actually connect to business value.

And it's also how do we connect the applications that we already have built to AI? How do we allow that? Because most of our users— AI is getting into more and more people's daily workflows. So how can we integrate AI into what our users already use so that we keep them as users, like Xano is doing it right now. Um, and, and, and that could be if you're building— and I would not suggest anybody, if they, uh, if they like a stress-free life, to start a company that will be a no-code tool, um, or anything that needs documentation. But the question is always how is your product being perceived by the person that is asking AI? Because most people are using AI to make business decisions. So it's not just how do we add AI to our product, but how do we get the product discovered by AI? For example, if you now go to ChatGPT and ask it, is Despia performant? Because it doesn't have the context we want it to have yet, it's going to say Despia has a shitty performance. It's a terrible thing. It's just a WebView wrapper. Use PWA Builder. It's the same performance, which is completely not true. But that's how AI sees the product because AI doesn't look at the YouTube and sees the benchmarks of us doing 60 frames per second and it doesn't see my Twitter posts and it doesn't see the video. So here's the thing, we need to shift how we're doing this. When we're creating video content about a product, when we're creating advertisement that convinces humans. The next thing is we have to convince machines and they're not going to watch videos and they're not going to get the advertisement message and they're not going to say, "Oh, he's a nice guy. He's always happy smiling at camera. I'll trust him," right? AI looks at the facts and if you don't have your facts straight and if your copy is not descriptive enough, AI will think differently about your product and hallucinate about it.

I, I think that, that there's a lot of— right. And, and I'm—

every day I'm asking like, like ChatGPT some trick questions about Despi. I was like— I, I asked it today— I was like, is Despi a good way to build mobile application? Uh, because it is based on a web view. Um, how many rejections are known out there? And then it said something like, "There are thousands of rejections for Deskia apps. It is criticized by Apple that it's a very bad way of developing an app." Completely made up, it's not true. We have a 99% approval rate unless you're building an app with explicit content, which happened before a few times, and of course Apple didn't allow that on the App Store. But it just makes up the facts. So what I found you want to do is you want to put an element LLMS.txt file on your website, which serves a markdown version of all the facts that you want AI systems to have about your website, because they scrape for the specific file that you can host on WeWeb and Nordkraft if you want. You can't, I think, on Webflow. And that will give AI more information about your landing page that you want it to have and not just rely on the copy and fancy animations. So the future of advertisement will not be advertisement to humans. I mean, you still have your appealing visuals, all that, but the major decision that people will consider and that will get more and more important is optimizing your website so AI understands your product and it understands all the ins and outs. Because if we were to put some keywords to this video, on the Deskia landing page, AI will perceive it completely different than it currently perceives it, right? If I were to put GPU-powered rendering up to 60 frames per second, GPU-driven UI rendering, or runtime caching for better performance, or native execution engine using window.Deskia calls, or UX optimization layer to prevent default WebView behaviors, and the ability to create native targets, WebView, and WebKit widgets, applets, 3D shortcuts, and CSE distribution tools and code signing to the website, which is right now not on the website, the responses that ChatGPT will give the user if Despia is a valid solution will be completely different than what it tells people right now. Right now it says, oh, you should use FlutterFlow, don't use Despia. Now if you add those add keywords to the website and ask ChatGPT again, which are not lies, they're completely true. It's just it doesn't have all the information to make a smart, informed decision. It will say, you know what, actually it is personal preferences. You can go far with both platforms. With Despia, make sure that your web application is not shitty, but also with FlutterFlow, make sure you're not doing, you know, too much custom code and too shitty stuff in there. So both should be considered carefully, and ultimately it's up to you which one you can build better in. So it influences the way AI sees your product based on the context that you provided, which brings us to the new thing we have to do is when we are out there selling our stuff, we have to provide as much context that's in text form crawlable by those bots so we're not sinking like a ship down like the Titanic on SEO. Because software 3.0 is more than how users interact, it's how users find it even, because the way people search is changing.

Yep, uh, that, uh, I've been seeing some people reporting like, what does, what does AI recommend? Um, and that, uh, which is kind of cool because it means like a lot of the SEO tricks, they just don't work. It's going to be what creates value.

Documentation for WeWeb, because right now ChatGPT is going at the core Desperate documentation and it's getting all the things wrong and saying, do this, do www, like it invents a method for WeWeb. That's why we have weweb.desperate.com. AI at some point, hopefully, once we have it correctly indexed, will be able to find the specific code snippets that we wanted to use when you use WeWeb AI.

Yeah. Can I ask, are you looking down the road of making an lms.txt or an MCP server for supporting that integration?

We're doing that, yes. We already have an lms.txt file that will be published on every page and documentation pages in the next update we're doing on the landing page, which we're also testing the messaging because the messaging right now is— the landing page isn't as good as a product.

Okay, fair point, fair point. Sometimes that, that actually— I feel like that often happens. Um, and I, I've come around to the idea that like, you know, I am where I, I feel like the wine can be better than the bottle, you know? And the, um, I feel like that's an okay position to be in. There are people who are really good glaziers and they know how to make really good bottles, then they're looking looking for some wine to put into it, and they can sort of shine it up. And that's one kind of skill set. And my skill set is a little bit more as a vintner, right? I'm making the wine, and then I'm looking for an adequate bottle that, like, doesn't completely hide the value of the wine, doesn't detract from the value of the wine, um, but is, is just a conveyance to get from, from here to there. So, like, the, the question I asked about the website— my website, I think, right now is detracting from state change. I think it makes it look worse than it is.

It does. I'm sorry to say that you should consider rebuilding it in Nordcraft.

Uh, you know, whatever. Sales decisions might be part of that, but I think a big part of that is the copy. It's just not telling the story correctly. And the, uh, but there, there's an opportunity to tell the story. Um, and, but, but I think if I can get it from bad to merely okay while making the underlying product great, uh, there is a lot of— there's probably playing to my strengths.

I think it goes around the message because that's a desperate landing page. Ship your web app truly mobile native, a native runtime to compile web to native. I think what I found is our questions that we're getting is, what's the performance compared to Flutter Flow? Is it performant? People don't care that it's one-click publishing. They care about two things. I can turn my web app into a mobile app and performance. I found actually performance is a bigger question than all the 30+ native features that we're giving you, which I think are completely amazing, but users don't care about it. I found that the actual users care about two things, performance and one-click App Store publishing, which is sad.

Oh, oh, now I'm thinking there's like a, there's a marketing opportunity in here. Call it triple speed. Fast to develop, fast to ship, fast to run. People are asking about that performance question, you get to handle all three. That becomes your promise, triple speed.

Triple speed. That's really good. I got to trademark that. Triple speed.

Speed with a cube. This would be a little more mathy. There are a number of different ways to to portray it. But like, I think that's awesome, right? The thing is, and I think this is generally true of marketing, you put up something— there's a law, I forget, but it's one of those joke laws on the internet, um, about how like the best way to ask a question on the internet is not to ask the question but to throw out what you think is probably wrong and have everybody crap on you while telling you what the right answer is.

YouTube and Twitter.

Yeah, you sort of put out an opinion and then boom, people give you a different perspective.

And I had the WeWeb community help me figure out how to run Despia. I made a video, said this is how you run Despia. I was unsure how I would run Despia in WeWeb. I just said you're going to call wwWorkflow because I saw a video from Ray that this is possible. And then I got like many comments like where Beatrice said, like this is completely bad, you should use promise.resolve. I was like, I don't know, it's supported in WeWeb. Now I know more. Then someone else says, why don't you just make it a component that can just dump it in there? I was like, I don't know, we can do this. So I got all the information and now with that information I gathered by educating the WeWeb community in the wrong way how they would use Despia, I got all the right information. So now we do weweb.despia.com for the documentation with all the right documentation and I made sure to put their credits on there.

Yep. Then I think the way to get to right is you start by being more wrong. I call myself a data scientist. I learned more quantitative approaches like social science problems when I was in school. One of the big ideas I got from my stats professor my freshman year, a guy named Ed Tufty, was that we're always wrong and that like stats is always about being wrong. What you're trying to do is reduce the error. Error never goes down to zero in any reasonably complex system. So it's like you start with the idea that there's some wrong, you're just trying to become less wrong, then that's a journey that we can all get on.

Yeah. Yeah. And I mean, it's been very interesting. Like always, what you always think your people will, your users will like about your product usually not what it is. So I found you got to offer a variety of things that can be appealing and then you see what has the biggest pull. Because the number one question I'm getting is this or FlutterFlow. And to be fair, it doesn't matter. Despi has 60 frames per second. Most phones don't have more than 60 frames per second.

Yep.

Unless you're having some crazy Android phone that has like 120. So it does matter. Despi has 60 frames per second. FlutterFlow does 120, but it doesn't help you having 120 frames per second. Both have GPU accelerated rendering, so both use the GPU, the hardware, to render it. So no issues with that. But on paper, it looks better having 120 frames per second than 60. But in reality is they probably don't even have a phone that can do 60 frames per second.

That's entirely possible. This is where you have multiple kinds of speed that are all driving the bus really helps. I think FlutterFlow is a wonderful product. I say I just don't use it anymore. Particularly in the world of AI-assisted development, it's like, oh wait, there's stuff you can solve through languages that have broader uptake, I think that's one of the advantages that Despia has is the fact you can be working in these JavaScript-centric environments. AI really knows its way around JavaScript. You can make a lot of progress, then make use of that infrastructure deployment. Whereas Flutter, like I say, I think FlutterFlow is a wonderful product, but it's the only way I'd ever recommend someone use Flutter.

It's good for them. They're making more money probably than Google is doing from Flutter.

Well, yeah, I mean, Flutter is one of Google's many, you know, technical initiatives that it hopes is going to make it some money someday because, unlike say Swift, it really is Android first. But Flutter Flow, I don't think is particularly Android first. I think a lot of people who are shipping on it want to ship iOS apps because there's a lot more money in shipping an iOS app than there is shipping an Android app.

I don't think you should ship an iOS app on Flutter, but I also don't— I don't think you should ship like an Android-first app using NestJS. Soon we'll be able to have the best Android experience. Right now, most of our effort has been making the platform for iOS because I think it's where the biggest market is. You want to have iOS and Android. That's where NestJS is great. If you want to have Android and iOS, you may want to use something else.

At least today, right? Because you're not quite there yet on the Android front.

I mean, we have it running very smoothly on Android, and everything that we can extend from the Android system, we're extending to the fullest extent. But I am personally just not an Android fan. So right now, if I look at a Despia Android application, we're doing everything we can, but I personally, as an iOS user, feel that's not enough. And maybe if they've been Android users their whole life, they're like, oh, I can live with that. I can live if I click on a button and it has this blue thing, like this blue highlight coming, which I could never live with. So we're now working making Despia running in Android more like iOS, just to meet my standards and what I like.

You know, we have a comment from Laura Curtis talking about how— and maybe you can put up on the board here, which I think is an interesting point about the cost of working with iOS versus the cost of working with Android. And as long as we're on the subject of mobile development, maybe you had a couple thoughts on that because you sort of been— you swam in both oceans. Yeah.

I mean, I think it's like besides having an iPhone that you can run to test, I think it's cheaper building and less time effort building an application for iOS compared to Android. Because here's one thing that we have to deal with on a daily basis. Android is changing things all the time, especially all those mods, all those custom launchers. For example, we just shipped an update for Android that targets null Android, like Android without all those crazy skins people can throw at it. And it works on the Pixel phone, it works great. But then on Samsung, on One UI, or what they call it, there is this little issue that is only on Samsung One UI and on every phone that's not a Galaxy. It works on a Galaxy, but if it's not a Galaxy, it's like a low-budget Samsung phone, we have this one bug in there. And now we have to go in there and fix it for all the users. Because that's the whole thing of using DeskSpace. We do it for you. So we just fix that bug. But this happens all the time with Android. You have those compatibility issues because everybody is using their own version of Android. Android is open source, and then Samsung is taking that, and you can get Android running on your own phones tomorrow if you just purchase 2,000 OEM phones from China, and then you go to Android and license it on those 2,000 phones. You have a minimum of 2,000 phones and you can be tomorrow an Android phone factory. The reason that it's so cheap, because now if you can make a phone for $50 in China, it's, it's low barrier to entry to create a phone company tomorrow. And then you put all your mods on there, all your skins and custom launcher, all that. Now you have someone with little to no capital, and then it's like half a million or so, but you have someone with comparably in this market, almost no capital. It's like you can get a tent for that in LA or something. So you have very little capital and someone is a new phone manufacturer who is now adding more complexity to the Android ecosystem because now there is another one who may not be following all the rules and being all up to date on their systems. And surprisingly, this happens a a lot of times on the low-budget phone versions from big companies like Nokia low-budget phones, Samsung low-budget phones. So this is a problem that we've been having to deal with is it works great on Galaxy and then on a Samsung A15 there is this issue because they are doing something different on there and then we have to test it on a specific phone and a simulator to see, oh yeah, this is this issue and then we gotta patch it you gotta update your application to stay compatible with this one Android phone out there. So this is why I like iOS because it just works great on everything because it all goes through one manufacturer, through one OS system. That's why I'm saying Android does not meet my quality standards yet, but I think no Android app will ever meet my quality standards because Android does not meet my quality standards. But if you are are within the Android quality standards from Google, Despia gets 100%. It's just that 100% of Android standards is like 50% of iOS standards. But I may be opinionated on that.

So what do you think about a new third-party operating system then?

I think it sounds like a wonderful idea, but it's probably not possible to do that because it would, if there would need to be, it would need, I think, for that unified operating system to work right, we would need to have one manufacturer who's carrying the whole load. So it would probably be like Apple opening up iOS to Android, but then we have like complete different technologies, like, or Apple opening up iOS to other manufacturers, which is probably never going to happen because if we have this unified operating system with Android And the issue of having this unified operating system like Android, which every manufacturer can use, you just need to buy 2,000 phones to license it correctly. I think there are ways how you can license it without having to buy 2,000 phones. It would be involving the manufacturer's license and kind of them white labeling their license for your phones, which people do. And so that's why sometimes if you have an old cheap Android phone, it may just stop working. Are getting updates. But then everybody is having their own mods and their own skins, and they're the problem. It's like if you run certain SDKs, I give you an example. Right now, this is the latest thing we're working on. Success haptic is something that should be standardized on every phone. Like you have error haptic, warning haptic, success haptic. Haptic, right? And then you have heavy haptic and light haptic. Every phone supports that. That's default Android.

I'm sorry, could you explain for the viewers what haptic is? That's a term of ours.

It's like your phone vibrating. It's like if you do light haptic, it vibrates lightly. If you do like heavy haptic, it vibrates heavily. And success is like a sequence and error and warning. Now, some Android phones I think we have the issue with lower budget Samsung phones, like Samsung, the A series or whatever that was called. And some Chinese manufacturers, I think it was like Umidigi or something. They had the problem that they don't support Success Haptic. And people brought that up. So the users that use Despia, their clients or their users had that problem that the sequence wasn't vibrating. Now, why is that problem? It's not because it's a Despia problem, it's because those phones don't support, don't have the sequence set up for Success Haptic even though it's an Android default. So this is what I'm saying, I think we'll never be the best Android, but I think it will never be the best in general. So our new strategy is we're going to, we already purchased the top 10 brand Android phones, so like the Samsung flagship, the Google phone flagship, the Huawei flagship. It was hard getting a Huawei in the US actually. So we got all the flagships and we're testing it on the top number one manufacturers that share like 80%, 90% Android market share, and we're going to optimize Despia for those to work 100%. That's our new approach in making sure that we have the Android compatibility. Because if we're going to focus on every small little person modding Android, they shouldn't mod it. If they don't have the capability to maintain their system, they should just use the open source version vanilla without having to put all their crazy skins on there. But that's hard. That's a tough thing to do with Android. And Flutter will have similar issues there. React Native, like depending on the devices. Like with Android, there's so many devices and so many different quality standards, especially for certain Chinese manufacturers overseas who don't have the QA team or something. So you, like, if you're building an Android application, you're not building an Android application, you're building a Samsung, Google phone, Huawei Huawei, Xiaomi, and like some Nokia and some 3 other companies' phone application. That's essentially what you're doing. And I would suggest only testing for those in the simulators. Because if you focus on the whole Android market, that's like a full-time job in itself, making sure that it works with every Android phone. Got it. Cool.

Cool.

So that's why I think it's easier and probably cheaper in terms of time going with iOS and then having— we have an iOS and Android app. Your Android app will be the companion app to its iOS app and you will focus on the top 10 Android phone companies.

Okay. Cool, man. Let's do this again.

On Android, I always suggest never do sequences. Build your own sequences. Call like heavy haptic 3 times in a row with a setTimeout in a loop.

Oh, in the JavaScript, you mean inside the web?

Yeah. On iOS, you can do success haptic because they all run on the same principle. Like it would be brrr. And on Android, I would just do a loop calling this one line of JavaScript with like a 50-millisecond delay like 10 times or something, or like 3 times to create the sequence.

Very cool.

So that's what lets you do that. But you've got to know that because then people use Success Haptic and say, oh, it doesn't work on my Samsung A15. And I was like, can you try it on a Google phone? And then it works on the Google phone. And then we're like, why? I was like, yeah, it's Android. So the lower, like, because there are differences in the OS even within Samsung, even though it's one OS. Like, the flagship models tend to get a better OS version. It's more maintained.

And I think that also becomes a really key, key issue. Like, when you have lots and lots of tiny stuff, like, how well do all those little tiny things get maintained versus the main brand by their company. They threw a bunch of stuff out there. Some of it sold well, some of it didn't. Stuff that sells well, that gets more attention, but then people who bought—

Is not Despia being maintained? Despia is maintained perfectly for Android. It's the Android companies that sell the phones that are not on the latest up-to-date setup that Google sends them as a framework to use, if that makes sense.

It does. That's a really interesting point about being fragmented because entering a fragmented environment is also very much what we're doing with AI. That brings us back full circle to the Software 3.0 thing and this idea of being compositional and the web components and being inside the AI, etc. Because having control over your platform, I think maybe something that's becoming a part of the past and the experience you're having with Android today, could be very relevant to the kinds of chaos challenges that we're all going to be facing in the coming months and years.

I can't wait to go through all the chaos. I think just one idea that I think is very cool, there should be an AI for end users to use to build web components that will have access to the backend. Imagine you can go to an AI and say, I need a web component to convert files and it could build it faster and you can— it puts it in your library of mini tools that you can always open again. Like, if you think about, like, you know, like those websites where you can, like, that are halfway legal where you can download YouTube videos if you ever need to. Imagine you can have AI build it yourself and make sure what you're doing doesn't have a virus.

Cool.

You know, like sometimes I do it with my own videos because there's a website, it's like NoGPT. I put my video and I get a transcript because I don't always get a transcript from YouTube that I can download as an SRT file. So I use those websites. I'm sure I probably don't have 50 viruses from those. I was like, if AI could build it for me, which it can, and that's how most users are using Lovable. And that's why they have a big market share.

Oh, that's an interesting insight.

Yeah. They're building software for one. And then back to circle to software 3.0. It's software that you built yourself through software. To solve a specific problem that you have, which also branches out if you want to do that commercially to help other people like you. That's when we go to Xano and, you know, whatever frontend tool you want to use.

Yeah. And you know that, that theory of, you know, software for one, software for some, I think that could be a great kickoff topic for our next vidcast.

Amazing. I have like 3 mobile apps on my phone that I use only for myself.

Very cool. Android Kiosk app.

An Android Kiosk app.

Well, and the kiosk, we have to see the kiosk, right?

I'll let you know once it ships. I'm also doing right now another project for a client where we are building an Android kiosk that's in the wall and it runs the Deskia app in it. And it will be used for certain things like for checking in and all those kinds of things, which is pretty cool. So like we're going beyond your phone and we're going now into hardware and making Deskia into a framework. I think like Kiosk is like the best thing that ever happened to Android.

Yep.

It's way better than making Android Kiosk Launcher app than making phone that supports every Chinese manufacturer. Because those devices, they just run pure Android. Those, like, if you go at the companies that build Android kiosks, nobody at the company has any, any will or any— I would ever have the idea to say let's make a custom launcher for this whole thing and customize the way the file picker looks and then stop maintaining it in a year. They're just saying Android open source is fine for us, we're just using vanilla Android. So that's why those kiosks is actually how I like to test Despia because if it works, if Success Haptic works in those kiosk environments, every SDK works in the kiosk environments, which it does, it'll show you that it's the phone manufacturers screwing up because they're not maintaining it properly. Which happens all the time with Android. With iOS, you never have that problem.

Yeah, but with iOS, you don't have that very cool kiosk business opportunity too.

Use the kiosk. But I think the kiosk thing is a really cool thing. I remember there was one person in Office Hours that was in your Office Hours and your community that said he wanted to use AirPlay to make a display that can be updated from an iPhone. And he wanted to do that through an Apple TV. And I remember, I was like, why doesn't he make the TV as an Android kiosk using an Android TV box, build it as an Android launcher application for the Android box that would just be like a kiosk connecting with the TV as an HDMI cable, and then he just builds a mobile application for his iPhone that will use that in real time to update the kiosk.

Right, that becomes sort of more the Android way of doing it versus using, you know, Apple TV. It also has app support.

It would be simpler to build because Android has like this kiosk functionality with the custom launcher.

Yep. No, and I think you put your finger on it, that the, the, the, uh, the opportunity to build different kinds of things with Android, um, that beyond just building an app for a phone because That is what interests me particularly in Android.

It's not necessarily building Android mobile application, even though that's pretty cool, but I think if you want to make the most money building mobile applications, you're going with iOS, and if you want to make the most money building Android applications, start building launcher applications for in-person businesses, like ordering systems and all that. There's a lot of money to be made in that. Consumer market, iOS, enterprise market, Android, but do it as a kiosk.

All right, yeah, we, we definitely should be getting into this more, man. I think there's, there's a really neat business conversation to be had. A few weeks, we're going to have a kiosk in the back, and then we will have— and then we'll have the demo of the kiosk. I think everybody who's watching this is waiting for that to happen. Wonderful.

I think that was a, was a great way talking about Software 3.0, maybe it's Android Kiosk as well.

I think we went from talking about Software 3.0 to potentially Hardware 3.0.

Hardware 3.0, I love it. I'm— you know what, I thought actually about, because we do have this launcher functionality coming, I wanted to make like a knockoff of the Rabbit R1, just like as a video, how you would build the Rabbit R1. It would like— the video on YouTube would be rebuilding the Rabbit R1 60 Minutes with Despia.

I think that there's merit to that. Like, hey, what does a very small launcher, something smaller than a Raspberry Pi, look like? And then put the software on it and go.

There is a big OEM market in China where you can just buy those things in like small quantities with Android running on it. So it's pretty cool.

Superman.

Let's take care.