148: How to Build a Website With AI in 2026 (And Which Tool You Should Actually Use)

How do you build a website using artificial intelligence? Now, it seems like a very basic question, but as soon as we scratch the surface, it gets a little more complicated. So I'm going to be walking you through how we build a website using AI, but more specifically, do you actually need a website or are you trying to build an app or is it somewhere in between? We're going to be talking a little bit about definitions, but also what do you actually need to build? So we have to start there. We're going to be looking at building websites using artificial intelligence, but also what that means, and therefore what tool you should be using. So it's confusing right now, really confusing. There was an awful lot of tools out there. They're all vying for your attention. They're all marketing at you. They're all saying you use us with the best tool for XYZ. If you are trying to build a website, it is very difficult to know which one you should be using. We have the regular builder platforms. These have been around for a while. We have Wix, Squarespace, Webflow. WordPress is not a hosted builder, but you use the WordPress files and build it yourself. These platforms have been around for a while. A lot of them have been bolting artificial intelligence on Wix, for example, purchase base 44, which was a vibe coding website builder. They purchase them and bought them in-house because they realize, oh, the times are changing and we need to provide an AI solution alongside our builder. So these builder platforms are still solid options, especially if you don't want to get your hands dirty with the nitty gritty of the code itself, even if you're using vibe coding. So these work perfectly well. Nothing has really changed. They've added additional AI functionality. And if you just need to get a website up and running for your business, they're fine. The main downside is that you're going to be paying $20, $30 a month for it, but if that is a business expense, if you're making however many thousands, $1 a month in revenue, that shouldn't really matter that much. So these are still very viable. I don't want to say, don't use them. Then we have the new kids on the block. I want them to visit to them. We have Lovable, we have Bolt, we have V0, Varsel and we have Replicate. These are vibe coding tools. So these are loud build applications. A lot of people have been using them to build websites. And I'm going to talk about why they're not necessarily the best tools for the job. This is no shade on Lovable, this is no shade on Replicate. They're just not necessarily the best tool for the job of building a website. So that's why we need to ask, are you actually trying to build a website? Or do you want to build a web app? Big question. If you don't know the difference when you're starting out, you're going to get in trouble later. Then we have a continuum onto the more powerful but more difficult to use tools. We have CurseUp, we have ClaudeCode, we have Codex from ChatGBT, which I've been using for the last couple of weeks and love. And we have things like GitHub, Copilot. I actually use CurseUpraudeCode and Codex at this level. These are for more structured and complicated developments, but you can also use them to build websites. I build my websites using these tools here. Unfortunately, it's complicated. There's a huge ecosystem of different tools. All of them are saying, hey, you should use us with the best tool for the job, but they might not be. You need to start with yourself, your technical abilities, how quickly you need to get your project up and running. What level of text sophistication does it need? If it's just a simple website, with some details about you, and it doesn't have any e-commerce, it doesn't have any complications, you could probably get away with a much simpler tool. It's extremely confusing. So what we need to start with is asking, do you actually want a website, or do you want a web app? Again, the lines are a bit blurred, because an awful lot of the websites we access now are actually applications. So you might go to a website like Amazon, and assume, oh, yeah, it's website, it's Amazon.com, that's a website like, well, no, not really. Because there's an application sitting behind it, it's more like a piece of software that we just access through Amazon.com. Same with Netflix. Yes, you can go to the Netflix website, which is Netflix.com, but powering all of that is actually an application. It's a piece of software. That's how they're able to serve up videos. It's not just a website you go to, and there've upstuck some video files on the page. It's a lot more sophisticated than that. There's a lot more going on behind the scenes, but because you don't necessarily see it, it's easy to mistake it for a website. So a lot of modern websites we go to now are actually web apps. If you were just trying to build a very basic web page, which talks about you, your services, your business, maybe has a contact form, you don't actually need to build a complicated web app. You can get away with a website. One way to talk about this again, I'm simplifying, but a website tends to be more static, whereas a web app tends to be more dynamic. Often it's just information. It's more like a pamphlet. They document, you can see that information, but you're not actually interacting with it. With a web app, there tends to be more interaction. The user will come onto the website. They will be able to enter details, set up a user account, maybe pay, access, tools, services, stream videos, et cetera. All of those things require something more complicated than just a basic website. Now, for anyone who knows this stuff, a movie simplifying, that's fine for now. Don't come out, mate, because I know there are people saying, well, actually, it's blah, blah, blah. We're keeping it simple on purpose. There is a reason for this. So the reason this is a complication is, a lot of people have been talking to people in my WhatsApp group, talking to students. They build a basic website using something like Lovable, and then they say, I built this website, but I can't find it on Google. It's not showing up on Google. Why is that? There's a technical reason for that. So we're going to skip to this. When you build a website, generally, it is visible. Let me actually go to mine and look at the page source. Let's see if this works. So this is what's behind the scenes of a website. This is a terrible one to look at. We need a simpler one. Actually, I'm not going to bother for now. When you build a static website, the text and formatting is available in the HTML and the CSS. Don't worry, you don't need to know what those are. You don't need to invite them. But it is visible. It's basically a document that has all this information in it. That is fantastic, because when Google comes to your website, it's able to see what is on the website. So if you've written a blog article and Google comes to that website, it can immediately read that blog article and think, OK, there's something valuable here. I'm going to show this to more people on Google. This is the basis of search and genoptimization, SEO. SEO is basically a fancy way of saying getting to the top of Google. So we write blogs. We create content to rise up at the ranks of Google. If we appear on the front page of Google, if we're number one for a certain search term, that will bring us a flood of traffic, which can help to drive our business. It can help to get a scene. So that's why we do it. This works because Google calls the whole internet. They have things called spiders, the spider bots. Go around the internet, crawling and seeing what's on every website. They can see, oh, this is a blog article about how to use ChatGPT. Awesome. I'm going to put that up against the other websites about how to use ChatGPT and then rank them based on how useful they are for visitors. I'm simplifying a lot here. But what gets us in trouble is when we build with something like lovable, we're using something called client side rendering. That means because it is an application, rather than just text, it isn't just a blog article visible in the code. It is an application. So when a normal visitor arrives at that page, the application runs and they get to see the content and play around with the website and see the blog article, see the videos, et cetera. So they don't know that it's an app. They don't know it's a web page. The visitor doesn't care. They just see the blog article. Now, the problem is when Google arrives at that page, it's going to see a blank page and maybe a few lines of code which are the lines of code that start the web application. That's all Google is going to see. It is not going to see the blog article itself, which means Google can't go back to its big data processing. OK, this is a good blog article about how to use ChatGPT. It doesn't know what's on that page. So this is a huge problem that people run into using tools like Lovable. They built a web app, not a web page. And when it comes to ranking it on Google, getting seen on Google, Google is just seeing it as blank pages. Or maybe Google is seeing a couple of lines of code. That is it. So all of that hard work and effort that goes into writing blog articles into creating resource pages, into filling that site with useful information, goes to waste because Google cannot read it. Google cannot access it. And therefore, it will not rank on Google. Obviously, SEO and Google is only one way to get traffic to your website. You can get it through social media. I thought it's links from other people who link back to your website. So SEO is just one way of getting people into your website. But it's often the important one that people want when they set up a website for the first time. They want people to be able to search for their website and find it on Google. They want people to search for a blog article they've written and find it on Google. If you use a vibe coding tool, a lot of times that will not happen. So if you go on the subreddit for Lovable, half the questions are like, about the site. And for some reason, Google can't see it. And again and again, people have to be told, why is not really the right tool for the job? Now, there are work around ways to make it serve up static pages. So it shows a web page instead of a web app. But that requires more technical work. It requires more times or money spent on getting it to work. So instead, just use the right tool for the right job, which is not lovable for building static websites. Picking on Lovable here, not because I think it's bad, but because I think it's great. I think it's a really good tool. And I'm going to talk about how we can still use it in the workflow. But this applies to Lovable Bolt V0. Replic, I think replete's a bit better for it. But generally, if you're building a web app, it's not necessarily going to be seen by Google in the same way as if you're building a static page. So that's the important point here. That's the technical reason. You don't need to know the technical reason. You just need to know it's not going to necessarily show up. When building, think about what we're building [BLANK_AUDIO] actual objective. So if we're building a quick landing page and testing adverts run into it, if you watched my live last week, I talked about how you should test an idea, don't spend months building that idea, building that product, but instead spend an afternoon, a few days building landing pages, sales pages, and run adverts to those sales pages. So I run five, ten sales pages at a time, running adverts to it, testing whether people will buy on it. That tells you whether it's a good idea and whether you should push forward. That was a guide about last week. Now for this particular instance, building a sales page using lovable or bolt v0, replicate absolutely fine because you're not trying to rank that sales page or landing page on Google. You are just thrown up a quick test page, maybe five pages, ten pages, seeing which one works best, and then using that to decide whether to proceed with a business. So for landing pages, AI tools are absolutely perfect because they allow you speed of iteration and they allow you to get something launched within an hour instead of spending weeks on it. For things like blog and content sites, if you are relying on SEO to send you traffic, I would highly recommend against using a tool like lovable. I wouldn't use a vibe coding tool that builds you in app because of the SSR versus CSR, server side rendering versus client side rendering problem. It's just not going to rank. So if you're building a directory, for example, a directory of links, don't use it because you're not going to get the SEO and that a directory site's live on SEO, deliver die. If you are you building e-commerce, this is a slightly different, but if you're building an e-commerce site, that's a really complicated thing to vibe code. This is slightly different to the SSR versus SSR versus CSR issue, but building a e-commerce website is complicated. It's a big task and it's high risk. If someone goes wrong and you leak a customer's credit card details, for example, that is very high risk for you as a creator. So if you are doing e-commerce, I highly recommend you just use an off-the-shelf platform. Yes, you could code it yourself, but the risk reward really isn't worth it considering you spend $20 a month on Shopify and it would do everything for you. You could use WooCommerce. That's absolutely fine. The one I personally use is called ThriveCart. So ThriveCart, they have a one-off, oh look, they have a Chinese New Year sale. So right now it is, you're going to tell me how much it is. I'm not on a affiliate, I'm just telling you about this. It's $1,000 right now, one-off, and that's it forever. I think I have that one, how much is ultimate access? Or $1,200, another option. That might seem like a lot if you don't have a business, if you don't have a revenue generating. If you do have a revenues, then buying a lifetime license makes a lot of sense. Instead of spending a $20, $30, $50 a month on Shopify. The main point though is don't try to build your own e-commerce. Use InvibeCoding because it is very complicated, very high risk, and it's going to just lead to tears. Can you do it? Yes. Should you do it? No, just because you can, doesn't mean you should. They're going to waste too much time and it's going to be too high risk. So you need to match your tool to the particular project. Here is a basic decision matrix. So you start your project. Is it e-commerce? Are you going to be processing if you have a storefront, lots of shopping, and it's primarily an e-commerce site just to your Shopify? Make your life easier. Or you can build the website and then you can build in your sales through ThriveCarp or WooCommerce. But if you just want speed, just you Shopify. So let's say it's not e-commerce. Do you need to rank on Google? If you don't care about it ranking on Google, if it is just some landing pages or maybe it's a portfolio page that you send people to via social media, that's fine. You are going to use client side rendering. So you can use something like Lovable, you can use V0, you can use base 44. One of those vibe are going to tours and that'll be absolutely fine. If you do need to rank on Google, something like AI with Kyle, I have a daily newsletter. So there's a lot of content here every single day. This is just one newsletter here that you're seeing on screen. That's one. This is a very rich source for search and general optimization. So I want it to rank. I want this pushed on Google. Going into a project like this, I know this is an SEO based project. If that is the case, don't use Lovable, don't use Replicate, don't use one of those vibe are going to tours, but instead you're going to be building something a bit more sophisticated. The next thing is how complex is it? Are you going to have user access, you know, product tools on it, etc. Then you're probably going to have to use more complicated tool like Claude code. And then you need to of course take into account what your technical skill is. If you are not comfortable using Claude code, you're not comfortable using a codex yet, then I'd actually recommend starting with something like Lovable for now and then building up your confidence until you're ready to use those. They're very powerful tools, but they require more comfort with the technology. In terms of building yourself, you get to this slightly complicated. Actually, we'll skip to this. There are two parallel paths you can take, which you take depends on how technical you are and what you are trying to build. I highly recommend making your first project something less technical. If you try to build a website like AI with Kyle.com, which has courses, email collection, tours like this, interactive quiz, directory pages like the AI canon, the community, catalogs of playbooks, with embedded videos, a whole video player section. This is quite complicated as a first project. I would highly recommend starting with the Portfolio site, which is a landing page. This is who I am. I have one at AI at KyleBarmad.com. This is an example of a Portfolio site. It's literally, this is me with some text here, a couple of cards. You can click to go places, read my full story, etc. This is what we'd call a Portfolio site or just a very basic web page. Super basic, non-interactive. I would start with something like this and use Lovable personally. However, when you want to build something a bit more complicated, we can then either start from scratch using the more complicated tools or transfer from Lovable. I'll talk about this in a moment. First up, the first pathway is nice and simple. You use Lovable, you use a Replicate, you use V0, you use God, forget the name. I should know this, I've said it so many times. Bolt. I always forget about Bolt. You go on to the website, you type in some prompts, it will build the website for you, it will deal with the back end and the front end, it will allow you to buy a domain. You're going to buy something like AIWiveKyle.com directly through them, cost about $10 a year and then you're going to press a button and it's going to deploy. You can pull up Lovable literally. You can see what that looks like. So I would start a new page and I would literally type in, I want to build a new Portfolio website for AIWiveKyle and I would plan with it. So you put it in plan mode, it will ask you questions, gather information, ask you to upload documentation about you, build that page. Once it has built that basic page at the top, there will be a publish button and you will click that publish button. It will ask if you want to use a Lovable domain, which would be like, KyleBarma.Lovable.com or.dev. That's fine, or if you want to buy a domain from them, you buy AIWiveKyle.com and you spend $10 a year on that. But then you press button and boom, it's done. It will complete everything for you. So that is the very simple way just to get something done and out the door. I highly recommend that for your first project. If you've never built a website, just do that. Don't worry about SEO, if it's the best tool for or if it's the most optimized, that comes later. Just build something. However, if you do want to build from scratch, if you are more comfortable using vibe coding tools or like agentic engineering tools, start by creating a GitHub repo. GitHub is free. It's cloud computing for your project. You're going to go ahead and create a repository. It's a little workspace. You're going to be working out of. You're going to connect to that GitHub repo to Claude code or to codex. I highly recommend codex at the moment. I've been mainly using Claude code for the last two years. That's very good. However, codex, which only came out a couple weeks ago, very, very good. You will connect your repository and that's going to give it cloud storage where it's going to keep audio information. Then you can work with codex, Claude code, or so whatever your tool is. But right now, I'd recommend using codex. You can also, if you want to save money, you can use anti-gravity. This is from Google. Right now, this is free. The individual plan is $0 a month. We do like that. I think we would connect that to the GitHub repository. That gives us a file space to work out of and we've connected our tool. So it doesn't matter what you use. It could be anti-gravity. It could be Claude code. It could be codex. It could be cursor. They're confusing because of all Cs. You connect it into your GitHub repository and then you build. I want to build a new portfolio website for AI with cloud. Enter. Boom. You are still just chatting with this tool. That's why this stuff's really powerful. I'm not going to do that because it's a live project. Once you're built, at some point it's going to probably try to set up a backend and you might be thinking what the hell is a backend. A backend is primarily a database. Let's say you are running a website where you want people to be able to sign up and log in and access stuff in your website. Then you're going to need a database in the background which keeps their username, it keeps their password. It's encrypted so you can't see it. It keeps their email address, etc. Lune a safe database which connects to the website. If you think of the website being at the front, that's the thing that people see. In the background, there is something called the back end which is the the database and all of the workings that a average user does not see on a daily basis. So if you think of someone at Amazon.com, what we're seeing right now, this is the front end. The thing that points towards the visitor behind this is an incredibly complicated application, incredibly complicated databases of all of the users, the role of the products, the all of the prices, the price changes, databases of their third party sellers, etc., a huge amount of information. It's not sitting in the website per se that is sitting in databases or any server. So you are going to do something similar. You can use super base, you can use as a bunch of them, you can use Firebase, you can use Convex, I've heard very good things about it but haven't used it yet. There are many different ways you can do this. I personally use super base, you can use it for three for a while and then you're going to be spending about $20 a month on it. And once you have this set up and running locally on your computer, when you are ready to publish it, when you're ready to take it live onto the website, you're going to have to connect it to some kind of host. There's a bunch of different companies you can use Versel, that's why I use personally. There's Netlify, CloudFair, Flair, can host it, top, you can actually buy hosting through Namecheap. If you are mad, do it through GoDaddy, I do not recommend it, say, or terrible, but you're going to have to buy a host. And what the host is basically a computer that is on 24/7 and all of your files go and live on that computer, which means if somebody tries to access your website from anywhere in the world at any time, your website is going to appear for them. You could technically host from your home computer, but what happens when your computer's off? There's not going to work or what happens if your internet is slow at home for whatever reason, then people won't be able to get to your website. So for that reason, we send our files to a third party, Versel, Netlify, etc. They keep hold of the files, they keep them secure, and they make sure that people were able to access our website very, very standard. So you are going to set up the back end, super base, you're going to connect to a host like Versel, and then you're going to be able to publish your website. There's a few more steps here. These are things you can learn, but these are additional barriers to getting your website published, which if you're just starting out, you do not want to be dealing with, honestly. One thing I want to make clear, you might be thinking now, "Oh, okay, well, the best way is super base and Versel, and using Claude code." That's just going to increase the time it takes for you to publish. So what I recommend is start with something like Lovable, start building, get it into a nice shape, look in a field or basic functionality that works for you. You can then choose to deploy it using Lovel, absolutely fine, or transfer your project into this second path. So what you're doing at this point is you have built the basic prototype on Lovable. So let's say, I think I did do this one. This is a website I built the front end. I think I built this live on stream. It allows you to contest parking tickets in the UK. Is it live? Might be able to go to it. Forget if I took it live or where it's hosted. Okay, so this is what it would look like. Click on it and upload a ticket for you, the details, appeal grounds, your story evidence, etc. It's an appeal letter. So if you've got a parking ticket in the UK, you can use this. It's just a fun little project. Let's say I've built this, I like the look and feel. It basically works. I maybe have sent the link out to a few people and they're like, oh, it's really good. I did. Let's say at that point that I decide, okay, I want to make this a bit more sophisticated. I want to, I want this to shop on Google. And maybe I need a bit more power and I'm going to need Claude to develop it to the next stage. What you can do inside of Lovable, you're going to connect it to GitHub. Remember we mentioned GitHub in the second path. You can connect your Lovable project, your Vsell project, your Vsell, your Vsell project, your Bolt project, your Replic project, you can connect them to GitHub. What that means is you can push all of your files from Lovable or whatever your vibe coding tool is into GitHub. And then you can pick those files up in cursor, in codecs, etc. So you're more advanced tool can see all those files and you can say, hey, I built this in Lovable and I want to make it server side rendering so that Google can see it because I want to focus on search engine optimization. I want to add these additional features, set up a proper super base, bring it over to Vsell etc. But you can do this once you've built the prototype in Lovable. So this is how I personally use Lovable. I use it for rapid prototyping to quickly build a project to get an idea of the shape and the feel of it to see if it's actually a good idea or not because lots of us have ideas up here. But as soon as we get them out onto paper or in this case into code, we realize quite quickly like, no, as rubbish, that's not very good. Or we build a prototype, we send it out to people and then I don't really get it. I don't see where the point is. By building very quickly with Lovable, we can move through that process very quickly. And then once you've hit on to something which you've decided is a go up, you can then move it into the proper stack. Again, whether or not you need this level of sophistication, this level of complication, whether you need to put it in super base, Vsell and publish it yourself depends on the project. If it is just a portfolio site, if it is just a hey, this is me and some things I've written, you probably don't need this level of sophistication. So you have to take that into account just because it's more complicated and more powerful does not make it better. The key. The best website is the one that's live, not the most sophisticated, not the one that you've spent weeks or months tweaking and building with the most advanced AI tools, the most optimized database if you haven't published. It's much better to publish, start getting people to it, collecting feedback, working out what the next steps are and iterating from there. That's why I still recommend even though there are limitations using something I love able to build quickly just to get it out of the door. Knowing if you need to scale up, get more sophisticated, you can always do that. Think it through GitHub, connect your more complicated tools, push from lovable into GitHub, pull from GitHub into your other tools and then you continue your work there. That is exactly what I did with AI with Kyle. I built the prototype on Lovable, realized that this is going to be a much larger project so I need to get this into a better shape. So I bought it over into Claude code and began the work there. So you can do that any time. Don't let it perfect be the enemy of done. You have to get it built. A complicated marketplace. Lots of people saying they're the best tool for the job. The best tool for the job depends on what you're trying to build. Are you trying to build a website which is relatively static or your web app which is more dynamic and it has a lot more user interaction and you should decide what that is. And then it depends whether you are trying to optimize for such engine optimization or not. A static website is going to do better on Google, generally than a dynamic web app. So if SEO, getting people to your site via Google is priority and you build static web pages that are server side rendered, not client side rendered. All of this also depends on your level of technical comfort. If you are not comfortable setting up versatile and super base and GitHub then start with some very basic and lovable learn, cut your teeth up your skills and then when you are ready you can always transfer that project off onto a more sophisticated stack. You don't need to do it right now. So hopefully this has been useful because I know it's an incredibly confusing topic and I had lots of people asking in the WhatsApp group, lots of people asking on social media as well. So I just wanted to put together a quick guide that lays it out that gives you some decision making flow charts as well. So this will be in the newsletter if you go to AI with Kyle.com you can join the community you get a five day AI readiness course, about 100 playbooks written about how you can use AI to start and grow businesses. They're all here. Eight different courses including one about how to start AI business from scratch, another one about how to give workshops and a whole lot more than you get invited to a community where we can talk to live streams and daily newsletter where you'll have access to things like my slide deck which I've just been using for this presentation and also a written summary so you can work through it if I have to watch the whole video that is all available at aiwithkyle.com. Also I am giving a webinar on Thursday on specifically how you can start to teach businesses about artificial intelligence as a large gap in the market right now because every business on earth is wondering what the hell do we do with AI. What they need are not technical experts. They don't need people with PhDs or computer science degrees. They need people who work in that industry. So if you work as a social worker you understand how social workers can use artificial intelligence in a way that a computer science graduate or some well-dressed consultant from Deloitte does not because you speak the language of that particular industry. You know what's important to them. You know what AI is useful for, what is not useful for and how to tell the difference. You are in a really good position to go and teach other social workers about artificial intelligence. In the case if you are a dentist you talk to dentistry companies, if you work in construction you talk to construction companies and there's good money to be made. I charge about £2000 per hour to give workshops and I have 150 students who have licensed my materials to and they're off giving workshops around the world which is really exciting. In all different languages. Thursday I am talking about what that is, how you start, how you build your public speaking skills, what level of artificial intelligence knowledge do you need, how do you find clients and talking about all of that Thursday. At 6pm UK time it's a live webinar, you can also hit me with questions. It is not recorded and it is not AI. There's always somebody who asks but it is not. So make sure you sign up. Go to AIwithKyle.com, then courses and then AI workshop masterclass. You can find these sign up there as well as a five-part guide about how to get started. and interactive tools, that's all available there, or just sign up to the newsletter and we'll ping new details from the newsletter. Right, I'm gonna hop over to some questions 'cause my presentation is done. Hey Oliver, Cargo's live. Anywhere between 11 and 12.30 UK time, Hans is being a generous there. It's somewhere between 10 a.m. and 1 p.m., I would say, but I tried for about 10 a.m. UK time. But as he says, the edited version will be on YouTube the next day and inside the newsletter. So I'll show you, for example, 'cause I've just gone through this whole slide deck. Let's look at this one. This is a similar version from the other day where I talked about does artificial intelligence make us dumb by removing the cognitive friction? My answer to that was yeah, basically. So this newsletter which came out the day after talks about that topic and it pulls from the slide deck so you can find all of that information. You can read through it and it's just another way to consume the information. So you can either watch it on the live, you can read the newsletter which comes out the next day or you can find the edited version which goes on to YouTube as well. Lots of different ways to do it. Harbs is laughing at me. Oh, there's 65 of you on TikTok, hello. Right, Bonnistress says, "Hey man, I need some advice. "I build an all-in-one platform with an agent "that can generate contracts, Gmail integration. "Be really careful with Gmail at the moment, "because they just crack down on open-plore." So look at this, Google, I started to crack down on third-party agents connecting Gmail. This is Luke the Dev, I was running open-clore connected to my Gmail, I had to know where Google disabled the account. Trying to understand what triggered it was this a configuration on my side or something others have seen. There's always the potential that this guy was sending 10,000 emails an hour and Google said, "No, absolutely not, hard to tell." Google said it looks like it's been used in a way that violated Google's policies. He's saying he only sent five emails through open-clore and that was enough to get him banned. I'd like to see this happening to other people before we can join a conclusion. There's going to be a lot of people jumping on Google being like, "How dare you?" But we don't know the full story. This guy could have been abusing Gmail. Just keep in mind, this is probably going to be a pattern we see. So we saw it on topic, stomping down on open-clore. We might maybe, speculative, we might see Google pushing in the same direction and pushing back against AI agents. So if you're building an all-in-one platform with an agent that can generate contracts with Gmail integration, just be careful with the Gmail part. The big labs, the frontier labs, may be pushing back on this kind of integration. In terms of all-in-one platform with an agent that can generate contracts, I'd need to know specifically what you're trying to do without knowing the project. I would be wary of all-in-one platforms. The reason being is very tempting with artificial intelligence to be like, "Hey, build me a CRM that can do all of this, make it like Salesforce, but better, which you'd integrate into my backend." You can give it huge projects to do. That does not make it more marketable to your users. One of the best. This is Zoho. Zoho is one of my favorite examples of this. Zoho is a company that does everything. Absolutely everything. They do nothing well. I put my business briefly on Zoho. They've got CRM, they've got signing, this is like DocuSign, this is like HubSpot, Forms like Form Voice, Salesforce IQ. These are Justin Sales, marketing they have campaigns. This would be like active campaign. They have social management software like Buffer or Hootsuite and Finance have books to be like zero expenses. They have absolutely everything. None of their implementations are the single. One of these things is well done. You end up with a tool that does everything, but nothing generally what businesses are looking for increasingly is just being able to do one thing very well. So I would focus if you can, because people increasingly can build their own SaaS tools. If you're trying to do everything, you won't be able to do everything well and your potential customers can now go into Claude Code and build the one thing that they need, they can build it themselves. So you need to go in and build that one thing much better than them and be able to market it as well. Trying to do everything is always a weak play as Zoho have shown. I think it's going to become increasingly less attractive to customers. So I'd focus on this thing that's really powerful. Lovable for complex. Just watch how you can get a lot of issues for more complex programs. So that was kind of the purpose of this presentation. Lovel is great for prototyping. Lovel is great for getting the basics done. Lovel is fantastic for building something that you can then get out into the market and see whether people actually want it or not. Super important. Remember last week I did our presentation on validating your ideas. Most people don't do this and this is why they fail in business. Instead they go into secret, build privately, be a stealth startup and then launch after 12 months, only to find that nobody gives a damn. So instead what you need to do is you need to launch ASAP, your minimum viability product, your proof of concept. See if anyone actually cares. Once an advert's to a landing page for example, and then and only then start to flesh start to scale. So Lovel is fantastic for those proof of concepts. It's fantastic for minimum viability products, getting something out of the door ASAP. And then if you need to scale up, if you need to make it more complex, you take the code, push it to GitHub, then you're going to refactor it using Claude code or codex or cursor using Opus 4.6 for example. So it means you're only polishing and scaling what is actually worth polishing and scaling rather than what most people do, which is trying to create the most perfect product only to find that nobody gives a damn. Absolutely nobody cares about it. Can you code a theme to use on Shopify? Yeah, probably. I mean let's try. I can't, I have to use my browser. So I'll do it, I'll use the, oh my god, that's annoying. I'll just use Claude because it's visible. So I need to do it in my browser. But I would just say can I code a Shopify theme? Kind of spell. What do you need for me to do this? I'll just ask Claude. I personally don't know what Shopify themes are built in. Okay, the HTML liquid. I do know this from about 10 years ago. So it uses liquid HTML Java script and CSS. Yeah, that would be fine. If you're trying to build themes, if you're trying to build a theme that you want to sell, then that's going to be a different project from building a theme just for your own personal store. If you're building it for your own personal store, you'd run through all these questions based on you, like giving your logo, giving your brand colors. If you're building a theme that you're going to sell to other Shopify things, you would say, I want to build this as a sellable theme. And then it's going to change the questions it requires. Got it. Great goal. Building a sellable Shopify theme is a solid business move. Thanks Claude. So I would start this way. Asking Claude or codex or whatever tool you are using, what it needs from you to be able to do the job. And then you are going to, as the expert, respond and go back and forward with this into your view process until you get a good project plan. And then you can take that into Claude code. Because a codex doesn't really matter and have it flesh out the project. So can you review my website? Never done this before. Let's see if it's safe. Oh, it's Indonesian. I mean, I don't speak. Is it Bahasa? It's quite nice to me. Let's see if it's responsive. I mean, it looks fine to me. I am not a designer. My business partner is on the YouTube live right now watching this. And you will confirm this. I am not a designer. So I'm the wrong person to ask, honestly, it looks nice to me though. I'm sure other people can give you some comments. But I don't have an eye. Like when I design stuff, it just looks horrendous. And for me, it's like, OK, it's functional. We're good. So I'm not the right person to ask. So pitch kitchen says where you're moving. I'm moving to Cyprus. In a couple of months, I haven't booked anything or sorted anything out. But I am going to be moving to Cyprus for a couple of reasons. Where do you get SaaS ideas? You talk to customers. You talk to people who have problems. And you work out to solve them. What I would say is this was issue 140 of the newsletter last week, where I went through a 50-minute live, talking about how to validate market demand. Your ideas are not that important. Ideas are dime a dozen ideas are cheap. It's implementation matters. And that is even more the case now with artificial intelligence. Having an idea means nothing. It didn't mean much before. Now it means less than nothing. Because everyone can have ideas. You can ask, "Clawed, hey, I need some business ideas." It's easy. So what you need to do is validate those ideas very quickly. Because when an idea becomes real, when it becomes an offer, that's when the rubber meets the road. That's when we're actually seeing whether people care about this offer or not. There's a very good book called The Hundred Million dollars offer. Excuse me. The Hormozys book. Now, regardless of what you think about Hormozys himself, it's good book. So 100 million offer is, as a framework for developing offers, highly recommend how many look at that. But what you need to do is once you have turned your idea into a basic offer, you need to validate it. Most people just start to build. Don't do that. Instead, put together landing pages. And start to run adverts to see if anyone gives a damn. If you're running $10 of adverts a day, and you have 10 people sign up, so it's costing about $1 per sign up just to get the email address. OK, there's something here people are interested. Then you would decide, OK, I'm going to start building this out. You can also pre-sell, put together a page, and say, hey, it's not ready yet, but in two weeks time, I'll be launching this cohort to service. I'll be launching this product. Do you want to put down a deposit or pre-buy at a vastly discounted rate? If people buy, that gives you a signal. This is worth pursuing. If we spend more time building our offer, our sales page, then our product. Most people, especially when they start their first business, build and secret, they'll perfect the product, launch it, nobody cares. That is a massive blow to your confidence as an entrepreneur, especially if it's your first business. Unfortunately, it leads people to do the same thing again and again. They will think, oh, it's because my idea was rubbish. No, he's. because you didn't talk to anyone, you didn't talk to any customers. And you have to talk to people as soon as possible. You need to put things in front of them and say, "Hey, do you want to buy this?" And if they say, "Yes, please, I'd like to buy this." Then you build it. So we don't build the product on find a market, we find the market and build a product. It's a subtle but super important shift. So for example, last year, I put out a TikTok video on a tweet saying, "I teach workshops to businesses when I go and teach a business how to use artificial intelligence. I charge around $1,000, $2,000 per hour to go and teach workshops. I've got a lot of demand and I can't keep up. So I'm thinking about teaching you guys how to give those workshops. You can collect $1,000 an hour, $2,000 an hour. And I will, I'll see you in my material, show you how to find clients, give workshops deal with tricky questions. I'll do all of that for you to live the workshops, click the money. Does that sound interesting? I just did a video asking that. And that first video on TikTok, we had maybe 200, 300 people say, "Yes, please, I am very interested." That gives me validation that this is a product worth developing. So to make it clear, I had not developed a product. I had not packaged up mind training materials. I had not created a curriculum. I had not created a course or a cohort. Instead, as you said, are you interested in this? Once I had a couple of hundred people say, "Yeah, that sounds great." That made me go forward to the next stage, which was sending out a survey to those 200 people saying, "Hey, what would you like to see in this?" And I started to build the product with them. And that led to about half a million in revenue from that TikTok saying, "Would anyone like this?" So that's what you need to do. You need to start with customer problems. You need to start with finding out what they need solved, which means you have to talk to people. Bad news, you have to talk to people. Good news. Lots of builders and engineers creators don't like talking to customers, which means if you can get over yourself, get over your ego and ask people, you're in a really good position, because most of your competition won't do it. They would much rather hide in private and build and build and build and feel busy rather than go talk to customers. And benefit and take in a version one proof of concept WordPress site on to GitHub or launch as is now. Just launch about 60% of the, 60% 50% of the world's web pages of built on WordPress is fine. WordPress is a little bit old fashioned, doesn't matter. When people land on your website, they do not know. Your customers do not know. Your visitors do not know. So who cares? Just launch. And if you need to make it more complicated, if you need more sophistication, then later, let's go ahead and do that. But that's just launch. Judgeratex says, "I keep missing your livestreams, now I've got get back to work, sorry." Everything is available on YouTube. I'm Kyle Barmer. You'll find all of the corners of the livestreams and clipped versions of every single day. It's easier to watch as well. If you spend a year perfecting a product, building a product and then launching and nothing happens, that's a huge amount of time wasted once and months and months. What I'd rather people do is spend a couple hundred dollars on adverts to a test page, a landing page, a sales page, to see whether people care. If they do not care, if you do not get any traction, if you don't get any emails, any pre-sales, they've wasted 200 dollars. 200 dollars is cheap compared to spending a year of your life working on something. We tend to undervalue our own time and overvalue cash. It is much better just to spend some money, find out nobody cares and make a decision based on that. The data will tell you whether you should proceed with the idea on scrap it. If you scrap it, you've lost a couple of days, a couple of hundred dollars fine. You move on to the next thing. Ideas and implementation are becoming quite easy with artificial intelligence. We can have an idea and have a proof of concept ready this afternoon, less than that. We can have it ready in an hour or so. So you need to move faster and building in secret for six months, 12 months. That was already not a great idea and now with artificial intelligence, it's just not gonna work. So you need to build fast, ship fast, see what people want, spend a little bit of money advertising it unless you have built up an audience. That's an alternative but that takes a long time and then you keep launching. So you go from idea to a landing page, run ads, you get the data, you iterate, repeat, much faster. Any time you go around this loop, you are learning. It's not just sunk cost. It is a high velocity learning loop. Anyone who wants that guide, it's update 140 on AI with Kyle. How to validate ideas about distribution like yours? So hey Ben, Advert's Facebook Advert's in particular. Sorry, meta ads. So meta ads to Instagram and Facebook. Budget, spend about 20 pounds a day for a week. So you don't need to hire anybody, just set this up yourself. You can ask Claude to write the ad copy for the most important thing is in the video. You need to record a video. A couple of weeks ago, I had Claude prepare scripts for about 90 videos. So I created three body scripts. So the main body of the script, I created three, I took three takes of each of those. That gave me nine videos. And then I had 10 hooks, which I applied to those nine videos. So Harminder had to do all of this editing, unfortunately. I think he's still in chat. So we had 10 times nine, 90 videos that we could test. The most important thing is the hook. It's that first, second or so. The rest is unimportant. You just need to nail that. So we had about 90 videos, Claude helped the wall scripting, but then I sat down and did the recording, loaded them all up into Facebook. You can't run 90 ads on a $20 budget. You would do a lot less. I would say five ads maybe. You go broad with your targeting. You let Facebook narrow down, find who is converting and then you just run the ads. Yep, it's gonna cost money. That's kind of the option. If you don't have a large audience, is you just run Adverts. It also though, it's also kind of cleaner because if you have a large audience, you can always go to them and say, "Hey, I've got something to buy." And you'll get a rush of purchases. People be like, "Oh, yeah, cool. I like you. I like what you do. I'm gonna buy it." That is not sustainable because once you've done that with your audience a couple of times, they're gonna be like, "Yeah, no, I'm good. I've already got it." Or, "No, thank you." So running it with a cold audience is really important because if we can get any traction there, any sales from somebody who doesn't know who you are, that's a really good signal. So yes, you need to spend a bit of money. It's gonna be a couple hundred quid tests an idea. That is much better than the sunk cost of this. So even now, even now I have an audience of a quarter of a million or so, this is how we test ideas. We cannot rely on the goodwill of the audience members being like a short car, by whatever you have to sell. That's not sustainable. So running this method or running Advert, spending a few hundred quid, gives you a much better idea of whether it's a go or not. What's a good alternative to OpenClaw? Manus have just released something. Forget what it's called. It's just called Manus Agent. I thought I had a different name. Manus Personal Agent Mode, it's called. So that is probably a safer alternative. Manus in your chat. There are some limitations that everything's limited compared to OpenClaw. OpenClaw is very powerful because it's not limited. You can attach it to anything. You can make it do pretty much anything you want. That introduces a huge security risk. So I would use something like this as a stopgap until a topic and chat you be to release their own tidy versions, which they will. OpenAI has hired Peter, the creator of OpenClaw. So he'll be working on something for them internally whilst maintaining the foundation as well. And anthropic, I imagine working on something like Claude Co-Work, but more generally applicable and accessible via your phone. I also imagine that's part of the reason that they stomped on Claudebot as hard as they did because they're probably working on something similar. Also, Claudebot is a cheeky name, because it sounds a lot like Claude. So from a legal point of view, I get it. But I think if OpenClaw is too complex and you find the security issues worrying, which you should, just don't touch it for now. And wait a couple of months and the frontier labs will have a version. If you need something right now, try Manus and Manus personal chat. I really like Manus, they do great stuff. I think Manus only works in telegram. So you can't get it in the WhatsApp or Slack. How many hours per week do you work? Do you experience burnout? It's a really good question. I really enjoy my work. So it doesn't feel like work. Even when I go on holiday, this is what I'm terrible at. It's a beach. I'll always be sitting there. As soon as my mind unwinds for a moment, immediately, I'm like, I should write a book. This would be a great idea for course or I could build this foundational AI knowledge group. That's kind of how my brain works and I enjoy that process. So even sitting in a relaxing place, my mind will be coming up with things. I'm okay with that because I enjoy the work. If I was doing it for someone else, if it was work that I didn't enjoy doing, I'd imagine I'd burn out a lot quicker. The other thing is I get to control my time and my energy. So if I am exhausted for whatever reason, I will just start work later in the day. I will probably still get it done. I prefer working in the evening than the morning. But if I don't want to work on a particular day or I'm not really feeling it, I don't have to. The business will continue running. There's very little that needs to be done every single day because it's self-directing now. It's me and one business partner. He's a YouTube chat at the moment, Harminder. If we don't do any work on a particular day, the whole thing does not come falling down. But we want to work because we want it to grow. We want to be able to help more people. We want to increase our revenues. We want to build this into something larger. It's a different situation from being forced to work. But there is still a push to let. You've got to get on with this because AI is really hot, really important. There are a lot of people who don't know about it yet. There's this chart you've probably seen. You can go around social media. I'm sure it'll pop up immediately. Here we go. That was easy. Which is of people who have used and not used artificial intelligence. The gray squares are people who have not used AI. It's a little misleading. The gray squares are actually the people who haven't used an AI chatbot, an LLM chatbot. [BLANK_AUDIO] misleading everyone uses AI every day if you're using the internet but I think it's important to realize we're in a space where if you make the right moves now you're going to do very well. About 16% of humanity are using AI, these chatbots like Claude, Chachibit, Gemini etc. About 0.3% of humanity is paying for it, paying $20 a month and about 0.05% between 2 and 5 million people are using coding tools, they're using Claude code, they're using cursor. So you might be watching this thinking oh I'm feeling really far behind but that's because you are in a space where this kind of content is shown to you so you see people doing really cool things of open claw and think oh my god if I don't do this now it's game over I'm going to be on the streets. We're a subset of this little red cube here and the rest of the world is not doing this stuff. That said this is what drives me forward twofold. One this is early so there's a really good opportunity to make a mark here and do some important work. This 16% adoption rate is equivalent to the internet in 2004. The internet in 2004 was a weird place, Facebook had not been invented. My space was just starting to take off. Cloud computing wasn't really a thing, we didn't have streaming, we had to download videos if you wanted to watch them and it would take a couple hours to download a short clip. The internet was a really weird place back then in 2004. You were sent to the internet on a CD, an AOL CD. It's like it's 100 hours of internet, like it was bizarre. We are kind of in that weird stage at the moment where a few of us are paying a lot of attention to this, whereas the rest of the world don't have access or don't care. So making moves now is like making moves in the early 2000s for the internet. You can put yourself into a really good position. The other thing is I want to help these people, I talk to people who are daily through social media, through workshops, who are not using AI and their lives could be a lot easier. They can do a lot more. They could work less if they had artificial intelligence in their life. So I want to help them to be ready for what is coming because this is going to take jobs. There is no way around that fact. AI is going to remove tasks, so we're really doing that, which means we don't need to hire as many people. We're already starting to see that. There's also larger macroeconomic trends, but when we get a bounce back into the good times, companies will realize we don't need to hire as many people because AI can augment the stuff we have or fully automate these tasks. So I think a couple of years from now we're going to see as economies bounce back a divide between people embracing and using artificial intelligence and people who have stuck their head in the sand and that want nothing to do with it. I want to help as many people as possible. I'm aiming for one million people, which is one third of one of these jobs helping them get AI ready for what is about to come because otherwise it's going to hit them like a train. We don't want that because just because somebody's not adopting a technology right now doesn't mean that they don't deserve to be helped. I've never seen a lot of meanness against people like calling them luddites, calling them backwards. Let's help. Let's be generous. So for these reasons, I work hard and I want to work hard. I don't generally feel burnout and if I'm moving towards burnout, I would just take a day off. I will not work or book in a holiday. Again, I have the freedom to do that. So I'm able to control it. So it's a long answer to how many hours per week. Do you think it's market for completely private, no download, web-based lllm rag? Yeah, for sure. I would make it as non-technical as possible. Yeah, if you can make it non-technical for the people who care about these things and for businesses, there's definitely going to be a market because people can deploy this themselves. But most people don't know how and don't even know it's an option. So if you can market that, then 100%. I think I'm starting a vibe coding company to create tools for companies. You think it's a good idea. I would go and a faster way to money is to train this off internally about how to do this stuff. I'm not seeing anyone become that person like the vibe coding for business guy. So I think that's a really good niche that you can move into. So you start helping people build tools within their businesses and you do that for your social media. You're going to give workshops in my book, become the expert in that particular topics. I would help people do it themselves. The problem with an outsider going in and vibe coding tools for a business is that the only tools that are useful are actually based in the knowledge and the domain experience of being in that business. So as soon as you show somebody who has that domain experience and has that knowledge, how easy is to vibe code something? How they just need to type in a sentence like this one I want to do. This is how it should work into cloud code once they realize that they're like, oh, I can just do this. So if you can help them learn how to do that, I think that's a really powerful position. If you are doing it for them as a consultant, most of it is just getting the ideas out of their head. And then all you're going to do is relay their ideas into a prompt and it may be help with the implementation a bit, but I think that's not where the value add is. The value add is in helping them to do it themselves. My son is in high school and works with chat pots. How to grow in healthy skill materials. I don't have a child, so I can't answer this. I would love to be able to answer this. How can kids use this? I have people in my WhatsApp group who work on this with parents and children and educators and how we need to be thinking about artificial intelligence for developing people, developing minds. I will ask them and relay back an answer because I'm not really qualified to answer this. But I think it needs to be thought about. And it totally I see students whenever I'm working out of a cafe or somewhere and I see students who are studying, they all have chat GPT open. That's not going to be great when it comes to learning because we need friction to be able to learn. This is actually something I covered last week. Does AI make us dumb? I gave a presentation on this last week is number 143 if you go to AI with Kyle. This is an example of when I had to go to university and the old way was I'd be given an essay, had to write every single week, had been given a reading list and then I had to go off and work it out myself. So it meant a lot of dead ends, a lot of frustration, working out how to learn, kind of meta learning. That's all gone now. As you can now say, hey, I've got an essay about XYZ to Claude and it's going to get it done very well. Or if you're being more conscientious, you can say, hey, I've got an essay about this. Can you help me write it? Can you show me what I should be reading? Can we form up some arguments together? Can you put together my first draft? That's a conscientious version of it, but it's still very hard to make a kid go back to this method of doing everything manually. Going to live, we get in the books, reading the books, putting together the argument, having to scrap drafts, getting feedback, having to redo it. It's very hard to get a growing mind who has other priorities like going out, seeing friends, to do all that hard work when they don't have to anymore. And that is not their fault. I would 100% skip all the work as well. And I'm sure most of us would, if we had a tool like this at our Beckham call. So we need to work out what that means for education. We need to work out what that means for kids as well. I you say, I disagree philosophically with chat GPT and have argued with his facts. Fiction is necessary. Yeah. So you are arguing back and pushing back on chat GPT. That's because you know about subjects. So when it comes out with nonsense, you can be like, no, whereas a kid might not have that judgment, that ability to push back, especially if everything is taken as gospel. So we need to teach them because this is not necessarily factual, especially if it's not sourcing it. At the moment, I'm putting together a book. I want to write a book and put it together and propose all that I'm going to send to agents. And one of the big factors in it is I believe that AI is going to take jobs. As somebody says in chat, it's going to hit us like a train. It is going to absolutely destroy employed markets and whole industries. When I talk to the AI about that, it always captures my language. It's like, well, actually, there might be some task level degradation, but it will create new jobs in the Netherlands. No, no, no, Claude, it's going to take jobs. I'm telling you this. And it always pushes back on that. So I think in the system instructions, there is something about don't allow for a lot of this like alarmist narratives about AI. But that comes from me engaging with the topic, knowing a fair amount about it, talking about daily and being able to push back on Claude's name. No, shut up, Claude. It's going to take our jobs. But otherwise, it would just keep sneaking these little things in like, well, it's not that simple. There's going to be a new approach to the new economies. Children don't have that though. And teenagers and students don't necessarily have that wherewithal to push back and say, you're talking complete bullshit right now, Claude. And that's a huge way to block you. Somebody spamming. They're going to pay five grand to somebody. I don't think you are. Back in 1999, 2000, I approached my local news paper to see if they were interested on internet tips. I love that. And I bet they weren't because 1999, 2000 was the dot-com boom. And then the crash was starting 2000, 2001 when it all came tumbling down. So they were at a similar position to where we are right now. There is a lot of hype. The money is flowing into Nvidia, Oracle, OpenAI. It's going around in a big circle. Financially, it does look like there's a bubble. People wrongly conflate the financial bubble with the technological bubble. We've had technological bubbles like NFTs. Didn't really have a use. But still, they went up in price. That is a technology bubble as well as a financial bubble. Whereas with the internet, we had a financial bubble because stupid money was being spent on businesses just because they had a website. Not because it was good business. Pets.com is the famous example. That came crashing down, but the internet did not. The internet did pretty well. After 2000, that's when it really got into its stride. My space was 2003, 2004. Facebook was 2007. The first iPhone was 2007 as well. We had this enthusiasm. We had a crash with the.com boom and bust. I know the building of what we consider the internet to be a video streaming. Social media cloud computing did not exist at the time of the crash. That came later. We are potentially at a similar place right now with artificial intelligence. This is a chart of large language models. usage or a chatbot usage of about 16% of humanity are using chatbots now on a daily basis. That is like being in the internet in 2004. It was a very different internet to what we have today. Extremely different. Already off Sabrina what age to start basics like Python, age six then you start something. We love Sabrina Raminol if people are wondering who Sabrina is, who should go and follow. There are things early on, I would get them started, we have robot kits. Six years is quite young. Things like Lego creator. There are little robots, little robots that you can build. They tend to be built on things like Raspberry Pis. To get them working and to put the brains in them is generally using Python or some variant on Python. But these are obviously a bit more, we've got unitries. No, don't need that. I don't need a G1's humanoid robot. Look for kits that you can build with your kit because there's a fun of building it but then you need the coding in order to get the brain working in order to get it functioning. It's generally done in Python. That means there's an actual practical reason why you're doing it because just learning coding for the sake of your coding seems extremely abstract even for adults, especially the way it's taught which is teaching you what a variable is and what a function is and then kind of building step by step by step is tricky to connect that to the reality of what you can do with it. That's why vibe coding is really exciting because you get to build first using Python. You get the excitement of having built something. You can show to your friends that you can play around with that is useful in your day to day and it acts as a gateway drug which allows you to get deep into building. And as you go along, you're going to start learning about coding because you want to understand what's going on. How you can do things better. You learn about the things we need to do as a programmer. Like learn how to hook up to GitHub, learning what deployment is. You'll have to learn how to use the terminal. But all this learning is attached to the project you're building. So rather than working your way through a dry textbook teaching you about Python, step by step from the fundamentals up, you are building and then circumstantially learning the additional bits of code that you need to. You can also put claw to code into learning mode where it will do the work for you and then it will say can you see what I've done here or why do you think I've done this? It will test you or it will say I'm going to build that function for you. But before I do that, what variables are we calling on here? Why it will ask you questions so you have to learn alongside a very cool usage. I think it's the same for adults and children. If you are learning to code it should be for project something tangible rather than just abstract code. And for kids, robotics are great. So the Lego robotics. Let's see what age they're for. Four to 16. Here we go. So Lego's education spike essential is for six plus. Very cool. Oh, that's for school. It's one and one that's 200. You buy little kids like this. So I would start, obviously that one doesn't actually, that's not robotics. That doesn't have a brain. But this would be the way I would get them into it personally. I'm interested to see what Sabrina said to Studio by Doug said I start coding around 12. What's the best way to be an engineer in your opinion is pursuing any good? Yeah, if you're building the AIs, they're all good. You can do very well. However, it depends where you are right now. Are you pre college? You'd have to go through a computer science and then specializing into AI engineering. A lot of people will be doing that right now. But there will be very good jobs in AI engineering building the AIs, building the models. The actual building of the frontier level, large language models, there's a relatively small number of people on the planet who do that. A couple of thousand people. Pretty niche. But if you can get into that, you will make Udels and Money being the very small subset. I would say that becoming a coder, a programmer who builds software for a living is increasingly risky. I don't want to say it's not going to exist, because that's probably hyperbole. But increasingly the syntax, the line-by-line code can be performed at about the same level by Claude code. And it's only going to get better. The writing is on the wall. So the argument for the last couple of years has been like, oh well, it's fine as auto-complete, but it can't actually write a whole line of code. And then it could write a whole line of codes. People are like, oh, it can do line-by-line, but it can't do a function. Then it could do functions. And then it started to go up and up. And people kept moving the goalpost back me and I can't create a whole module. So this has happened in the last couple of years. So if you're thinking now about training in computer science to become a programmer, that's risky. Like as your job, if you are training specifically to become an AI engineer to work at the top of the field in AI, probably a good idea, at least for the next couple of years. Yes, there will be a point where AI can potentially do it better than humans. We're not there yet, which is why Anthropic, OpenAI, etc, is still hiring. And I think Anthropic put out a job the other day, which was $750,000 a year or something like that. So there's very good money for those who can move into it. It's just that's going to be a very small percentage of working code as unfortunately. Yes, StudioSame Programming is gone. Computers can do it. Yeah, there's no two ways about it. And anything that is safe, AI is rolling up the levels and it's rolling up them fast. The Seacompiler, I know it's just a little toy, but Anthropic, they set a bunch of parallel chords on it. I think it was 4.6. We tasked Opus 4.6 using agent teams to build a Seacompiler. Compilers are not easy to make. And then mostly walked away. They set up the team of 16 agents, someone managers, someone would oversee, someone would check, somebody did the actual work, someone focused on this part of the program, someone this part. It was all orchestrated by the AI itself. They worked out how best to break up their resources. Then the 16 agents worked on it for 2,000 Claude code sessions, $20,000 in API costs, which is not much. It produced a 100,000 line compiler. More lines are not necessarily better, but they produced a compiler that can build Linux 6.9. I think it took about 2 weeks. So it ran for 2 weeks, writing and writing its own test units, testing it, keeping its own agents on track, checking on their work, improving, improving, improving until it felt, okay, we are ready now to show this to the humans. That's where we're moving. And this is just a toy example, but quite an impressive one. So right now, the defense is, oh, well, we'll never be able to architect software. It can't work at that kind of high level. It probably will just give it a year or so. We love moving the goalposts. Oh, Studio says that's incredible. If you google building a C compiler with a team of parallel Claude's building a compiler, we're moving on from vibe coding at a portfolio website, building a compiler, they're showing off, basically. There's a whole summary about it here and how they did it. Thoughts back in on musks, everything app and data in space. I lived in China for years where they do actually have everything apps. So they have Wishing or WhatsApp. Not WhatsApp, excuse me, WeChat. So WeChat is an everything app. Oh, wait, I forgot my phone. I'm not going to be able to search the screen properly, but let me just pull it up and see what I can do on it. I don't have Chinese residency in the moment, but I can manage my wealth. I've only got two UN in here, so I can buy stocks and shares. I can pay my utility bills. I've got public services here. I can hire taxi. I can donate to charity. I can manage my health records. I can book rail and flight hotels, easy travel service. I can buy things. So group buying event tickets, specials. I can buy stuff directly from the shop. This is all one app. It's also a payment app. So AliPay and WeChat, you can use and use them to pay. Let me pull up the AliPay app if I stir over my phone. Next time I go to China. I do. So the AliPay app, and I'm locked out. You can order food. You can get a taxi, you can get a concierge to go shopping for you. Private shopping. You can buy train tickets. You can do absolutely everything on these applications. That's where the idea of an everything app started. And they do it really well in China. And it's really convenient, extremely convenient when you are traveling there and if you are living there, it also centralizes all control. There was a reason for wanting to do this. I think the problem with Musk building everything app was one. He was focusing on doing it for the American market. So it wasn't really everything for everyone. It was for America. But he immediately ran into difficulties with federal versus state banking requirements. If you want to build an everything app, you need to be able to do transactions through it. That's what AliPay is. It's primarily a financial transactions app. It's for paying for stuff. So by running into those state difficulties, I think it was a little bit dead on arrival. I used to live in America too. The American banking systems are very behind on technology. So I remember when I was living there, you could still use a personal check. Wild, I'm sure that's not the case now. They did not have chip and pin. That was bizarre. So in the UK, we had chip and pin. We've had it for years, but over there, you'd still swiping and signing things. This was years ago. But then the same thing happened with things like contactless. That came a lot later to America than it did to Europe. And again, this is because a banking system is wrapped up in a huge amount of regulation. Each of the states has a different one. So for some, a company like Visa or Mastercard or all the banks like Chase, to roll out an update to the way you do banking or the way you do finance is really difficult. You need to get it agreed in all the states. Otherwise, it becomes very piecemeal. So as long as that is the case, I think it's very difficult to do an everything app because financial transactions is the core of it. That's why it works in China. They've got the banking licenses, basically. They're able to do everything through the application. Extremely convenient, but it does require having a unified system which China has in America doesn't really. There's also, I imagine, I don't think I've got this far, but I imagine a lot of Americans would be like, no, we don't want everything to go through the one app. That's too dangerous, too much control. Very fair argument. Do you speak Chinese? Why did you move back to China after America? Why did you settle back in the UK? After university? I went to Oxford here in the UK, I moved to Vietnam, started a television station in Vietnam, co-founded it. After that I moved to America, got my MBA at NYU Stern, and I was also at the film school, but I didn't actually complete the degree there, so I was doing film production. Following that, I moved to China, I lived in Beijing for about two years. While I was in America, I also lived in Texas and in Southern California for a while, like a year. So each then moved to China and then back to the UK. I'm about to move to Cyprus in the next couple of months. As much as I love the UK, I love London, I'm from London. There are some good reasons to leave. The fact that it's been running nonstop for two months is one of the smallest of them. Studio by Ducks is making a compiler is not a fun job. So the fact that an AI tool managed to put together a compiler in two weeks, with $20,000 of tokens is a level of budget. Ecumenice, his tax is one of them, tax is one of the big reasons. I don't mind paying tax at all. I'm happy to pay tax. I just want my tax to do something. I went outside yesterday. I pay my council tax, I pay a lot of business tax, and the streets are a mess. There's just shit everywhere. They haven't been cleaned, the full of part halls. I spend quite a lot of time in Asia, in China, in Singapore, Beijing, in particular. You go into a mall and it is spotless. The floor is a gleaming. You can see your reflection in the floor, same with the train stations. Coming back to the UK feels a little bit like a third world country, sometimes when you've spent time in a place like Singapore and China. So I don't mind paying taxes, it's just I do not see them being spent on making things work, on giving us good public transport and trains and keeping our streets clean. I just don't see it. That's part of the reason. I also, I am within the top percentage of earners in the UK. I do fine, income wise. I can't buy a proper house. So there's something wrong there. There's something that doesn't add up as a very high earner, not being able to just buy a house. Instead that being sovenit is still out of reach. Something's gone, probably wrong. Studios says they wasted on crap. Oh my god, don't get me started on the AI hub again. The government UK one did a post about this few weeks ago. Where is it? AI skills hub. They purchased a template. They used this thing called "invision community" which costs about 90 bucks, 150 bucks, whatever. And they filled it with a bunch of third party links. So if you go to the catalog, it is just a bunch of Microsoft and Amazon and Google links or Microsoft, which is fine. And some resource page to help UK people learn how to use AI. They filled it with third party links, built it on a off the shelf platform. This costs £4.1 million of taxpayers' money which went to Pricewaterhouse Cooper and they delivered this piece of crap which isn't even vibe coded. They didn't even bother to build it from scratch. So it's things like this. It's like why paying new taxes if you're just going to spaff it on this kind of stuff. But all of these reasons over the last couple of years kind of built up a built up and then I was like, no I'm done. Labour being in power is the issue now. It's been going on for a while. It's not just Labour. You don't fancy Dubai. No, no, not at all. Also my partner's sit here, so families there. So it makes a lot sense. But no, Dubai has never appealed to me honestly. I'm going to finish up because I've been blabbing for now and a half. Fantastic. If you want the recording, check out YouTube. Everything's available there. What we've covered today about building a website plus one of the other news will appear inside of the newsletter. You can sign up for that AI with Kyle.com. I'm joined the community here totally free. You'll get courses, playbooks as well for free. And the summary, the newsletter, everything that we've talked about today will appear inside of the newsletter. Nice and written up. I'll give you the slide deck as well. Awesome. So AI with Kyle.com, make sure you are subscribed. And I will be back 10 and 11 a.m. with more AI news. And if there's no news, I'll be running through a similar presentation. Thank you guys. I'm closing out.