#26 Amazon Legal Threats, Vibe Coding Three.js, Design/Dev Dream Tool

All right, we are live. Kabad doesn't Kapaza doesn't know that we're live. We are live. I'm still searching for my headphones. We're live. Yeah, I heard I don't have headphones. Wait, where where are they? We are live and two people joining. Welcome. Yeah. Join everybody join and I will search for my hip on the come in. Come in. Come hither. If you're not on YouTube, scan that QR code. Come over on YouTube. Join in the chat. Shall I get up on this side now? Let us know where you're in the world you are. Whether it's morning, evening, afternoon, night. I found them. I found the headphones. I like the depth that this creates. Like the depth this stream creates.

Can you can you hear me just fine? Okay, cool. So, let's I'm gonna I'm going to play a video which we're probably going to hear between you and I. We have a brand new deposition that just dropped featuring Ilia Sudskver, former founder of OpenAI, one of the YouTube to see if we can hear it over on YouTube or let us know wherever you are. Interesting bits. 62page document which I read in full. I highlighted the most interesting bits and I'm going to break it all down for you. The cast of characters include of course Sam Alman, Ilia Sudskver, Mera Moratti, former CTO. We have Gregman, the former boarding against Maybe it's in the recording. It doesn't found it like I'm gonna give it a like.

If you're the one person that's on YouTube going on, I'm going to read through it. Let's get into it. And this video is brought to you by Recap. Great. Yeah. One person on YouTube. Give us a like. Whoever you are just liked our own video. Um, and if I remove from the stage and then find the right tab. God damn it. Got loads. Okay. Okay. So, I'm also trying to get our live link on our Instagram so people can join us there. Good. I'm good to go. G to G. YouTube join. Okay. So, cheers for the like. I see you. I see you. I will find you. Okay, cool. We're good. So, last Yeah, last week we had Halloween.

Uh, what were you dressed uh as last week? Didn't do Halloween. I didn't do Halloween. No, I didn't dress up as anything, unfortunately. Yeah. Um, did you? Yeah. Yeah, it was a lastm minute thing. I I didn't really plan for it, but my friends and, you know, my siblings joined and we had a gathering and I didn't have any shirt, so I just quickly made one. Um, you probably could guess what it what it was. So, they brought some blood that you could like, is that that scary to some Yeah. Or like showing your API key. No, I did a I did a 404. 404. I mean, I would argue a 404 isn't as scary as a 500, but you know.

Yeah. Um well, 404 for not found, you know. Yeah. And somebody asked if it's the the zip code or something from, you know, the city that I'm coming from. And I was like, I am officially too geeky about this. I went to a I went to a Halloween party once as a backend developer and I had like I think I had like a can of Coke and a hood and I didn't shower for six days, but that was that was a bit extra. No one knew what I was. I think I had another accessory as well. I It was like a can of Coke or like a water bottle. I don't know. Something. Anyway, um a what?

Can you say it again? A water bottle. Yeah. I went to So, it was my birthday this week. Actually, it was my birthday on Monday. Oh. And um Cool. Congrats. Uh, happy birthday. Like, guys, can can you can we have happy birthday in the chat, please? Orchestrate happy birthday in the chat. I went to a place called Cahoots, and it's a 1940s themed um bar, cocktail bar, and one set in the underground and one set in a a post office. And the the menu, which is just over here, actually, I've got it. I got it with me. The menu is is an old newspaper and then down the bottom there's a cross word for cockney rhyming slang.

Do you know cockney rhyming slang at all? What is that? Okay, I have to I have to Soy rhyming slang is a way I I think it's a way to speak in code. Okay. So you could you say words that are different from but they rhyme. They always rhyme. So the classic is apples and pears. which is stairs. Um, okay. A ruby, a ruby murray is a curry. Oh, let's let's pop down the old ruby and get a curry, you know. I'll have a ruby or whatever. Um, yeah, my old china. No. Yeah, my old China doll is is like a mate. Is it a China doll? Anyway, it's your friend. Um, so yeah, coty rhyming slang.

So like, and it's funny because like I talk in cotney rhyming slang, but I never really thought it was Cotney Rhyman slang. You know, Jack Jones, I'm on with Jack. That's when you're on your own because it rhymes with Jack Jones, you know. And the most modern one that I know is um AM Amber Herd. Do you know who Amber Herd is? I think so. So she she took Johnny Depp to court. Yeah. for def defamation. And one of the things that she did was take a poo on Johnny Depp's bed. Do you remember this? Yes. Unfortunately. And so the cotly rhyming slang has emerged to when you want to go for a turd, you go for an amber herd.

Oh my. So that's more modern. That's more modern continu. But, you know, there's lots of there's always a classic classic ones out there. But, yeah, there you go. That's a short lesson on copy rhyme slang. I've been always saying that I'm learning English from you and I don't know if it's a good thing or a bad thing. Yeah. No, I don't know. Yeah. I don't say much cocky rhyme, but you know, bottle of water. Anyway, cool. So, you will be um we'll you'll be celebrating. Well, I've already celebrated. I went out two nights ago with some old friends as well because my friends opened up a a theater here, so that was good. Um yeah, and I'm potentially organizing a another birthday thing in a couple of weeks, but I need to pull my finger out and get that organized.

But other than that, yeah. Cool. Do you do anything for bonfire night? Is that a thing that you do in Germany? Do what? Not sure. Bonfire night. Bonfire night is like I suppose. No. Yeah, it wouldn't be. It wouldn't be something you do in Germany. So, Guy Forks was was a guy who was planning to um infiltrate the Houses of Parliament and he gained access and he was going to bomb it's called the gunpowder plot, right? And he was going to bomb the Houses of Parliament. It was a political statement and he got caught and they just put him on top of a fire and killed him because he, you know, he got uh he got caught for for trying to bomb the Houses of Parliament.

This is a big deal. This is like Yeah. you know. So then every year on the 5th of November, we we stick a a man on top of a I mean, you don't tend to do that anymore, but like Yeah, you would stick like a like a a doll on top of this big pile of wood and just set fire to him and have fireworks and stuff. Okay. No, we we don't have that. Or at least I haven't seen it. Yeah. Yeah. Wow. We have some weird tradition. It's like 5 500 year old tradition that we still have for some reason. But anyway, yeah, we're learning about the history of the Britain tonight, aren't we?

We are. I will check that out for sure. Cool. In the meantime, we have a show, too. Yeah. Unless you have any news you want to share of the week. Well, it's kind of like a jump scare. Uh, it should have been better for last week, but I didn't know of this last week. So if you share your screen and click the link that I shared in the docs uh with you the title is like to show to show okay like you click the link and you open it on on your screen and I would like to see okay right share this tab instead. So I'm sharing my tab add to stage. So you have that turned off.

How? What do you mean? How? Like you knew this already? Well, clearly. Clearly. Yeah, cuz I turned it off. Yeah, that that scared me last week. So, this is so Okay, this is crazy. Like for anybody watching, please go to your LinkedIn and go to data privacy and turn off AI generative uh improvement whatever because Microsoft who owns LinkedIn by the way if you didn't know by default they take whatever you put on LinkedIn to to use it to train their models. So you are basically a guinea pig by default to to train their models. And this is insane. Like this is actually illegal in Europe. And I don't know how they are getting away with it.

I don't know. I Well, yeah. I don't know. You can't like steal my data to to train their model. Well, clearly they can. Well, they're getting away with it. And I don't get what like the internet has turned back into the wild west, hasn't it really? That people just say and do whatever the hell they want. Oh, this is a this is a 10x return on your, you know, um, oh yeah, revenue and like, oh, we can we can quadruple trillion your I don't know, you know, and this is another instance of that. It's like, "Yeah, we're just gonna we're just gonna turn this on and we're just not going to tell you about it."

And you know, have you have you researched this? Like I I think I saw a video a few weeks ago. I saw something on this a few weeks ago. I was like, "Huh, that's I'm going to turn that off." God knows how long it had been on for. Yeah. Yeah. Um, and I'm not sure if even if turning it off if they really uh, you know, if they will actually two months ago, this was up this was updated two months ago. So, at the very least, it's been going on for two months. Yeah. Yeah. Um, Primagen in one of his videos, he said he trusts some of it seems like some of the AI stuff are even less trustable than uh the the schemes you have with crypto, which is like a crazy take, but at the same time, it's not crazy at all.

Like some founders make it sound like exactly like crypto, even worse potentially. Yeah. Yeah, it's just such a how would you even call it? Like a just a a desperate panicked struggle to just, you know, I agree. Yeah. But a lot of you Yeah, we have good stuff to share. We have like cool stuff to share and we should probably start with those. Cool. Uh well, my brother's just being hacked. So, shout out to my brother. I'm just helping him on the side here. So, a few words. How he Well, how do you think? I don't know. Like password. password. He's just changed his password. And it's like the password is what he does. So, it's like why have you changed the password to the thing that you do?

I'll get him to change it right now. But anyway, um okay. So, I will show something cool. I'm going to prepare my screen guys and I will we'll start with the website of the week. Great. Not as much of a website, but let's have some fun. By the way, we have Well, we need to do the intro, but just to Yeah, we'll get Oh, yeah. True, true. Lot of cool chilled out stuff this week. There's a few like other than the jump scare you just gave us. There's another couple of jump scares, but most of it's just like cool stuff that's been going on. Get ready, boys. Cool. So, I will prepare this and then we'll go with the intro and then we'll talk we'll talk about this stuff.

All right. Ready? Ready everybody? Kia, we have also a few people on on Instagram. So this week we have PewDiePie training his own AI. Typescript is now number one programming language uh in the world. Um, the creator of 3JS, the most popular 3D library uh on the web, is vibe coding his updates. And Freepic announced spaces, which might be the funny, the most fun and the most creative way to generate anything with AI, video, pictures, etc. uh all of this and a bunch of other cool uh web and AI stuff in this week's uh uh episode. This is Command AI. I'm Kavarza. I'm Samuel. And this is a week in AI design and dev. Cool.

So, let's start with the I'm going to add mine to stage. All right. So, this is probably one of the coolest things I've seen this week as for it's not as much of a website. This is more like a component, but something on the web 3D and super duper cool technically and visually. And we'll talk about the technicality and how it was built. So, the entire thing is based on this thing that you see here. This is not uh on the web. This is not an interactive component. What you see here, this was a video. Somebody made a video and it got 12 million views and it it's really aesthetically pleasing. It's really beautiful. And then somebody uh named Conrad, he made it like for real on the web and it's interactive and we can actually come here and play around with it.

like run on the web super duper lightweight. You see it loads almost instantly and it's super like 120 frame frames per second. Super interactive. It's so nice. And you see all the details and I can change the color. I can change the light direction. Um and I can even add blur and yeah it's super nice. So, we can even see the code for it here. And here's where it gets really interesting. And I need some of your help to explain the code to me because this entire thing is built on type GPU. And I was like, what is type GPU? So, you know, WebGL and you know web GPU. So for those who don't know all of these like graphical things 3D stuff that are going on on the web they are based on webgl well mostly and now we have webpu which is an updated or like an better version of webgl webgu is more performant but these are very very very difficult to write code for because like in these languages these are very low level and it's very difficult to do anything with them.

Uh therefore there are libraries on top of them and type GPU is apparently a type safe a TypeScript library for web GPU TypeScript library that enhances the web GPU API allowing resource management in a typeafe declarative way. How does that sound to you? Um, so yeah, I mean it sounds like it's uh Well, we're talking about 3JS later on, but it's just a it's a library on top of WebGPU, right? Exactly. So just like 3JS being a library on top of WebGL, this is on top of WebGPU. So this is in a way an alternative to 3JS but this is from what I understand is not as of a high level like it's more difficult than 3JS and it's not as complete because 3JS has been out for I don't know many many many years and this is quite new uh but I read somewhere that they are working on supporting even 3JS being able to I I don't know how that would work.

No, but to use 3JS here as well. What might come sooner is 3GS supporting web GPU. Interesting. So could you like explain to me what what is Typescript because we'll talk about this in a second as well, but what is a type safe language? Um so with JavaScript you can declare a variable with var right var um fu equals bar a string with bar right yes JavaScript is so loosely typed that just in the middle of your code you can just change that to the number three which could then break your code because if you're expecting it to be a string of bar down here at the bottom of your file but in the middle you change it to a number or accidentally gets change to changed to a number you're going to break your code.

So in its simplest form, typescript is you say constant well actually constants are native now in in types whatever but you you type that variable fu as a string or it gets implicitly set as a string once you assign it to a string. But then if that changes later on in the code or where it becomes really powerful is where you say this function let's take web flow um component web uh co- components as an example you say this co this component accepts three values and one of them's a string one of them's a number and one of them is this. If you try and pass it three strings or three numbers or anything else, TypeScript at buil at runtime will just say, "Hey, uh-uh, four or four not found."

Yeah. So, TypeScript is is is a way and it all it does is compile down to normal JavaScript, but it's but that happens at build time, but at runtime, you're writing in in a type safe language and you're going to get these warnings. If you create an object with certain values inside and use that object later on, you'll actually get code hinting and various things like that. It's a really enjoyable and it works so well with AI because the AI can reason about the code a lot better. It can understand the code a lot better because it's like okay well this function only takes a string here and blah blah blah all this kind of stuff.

So yeah. So it's more guard rails around it. Yeah. Yeah. Yeah. Yeah. Cool. Cool. Yeah. No one really writes JavaScript anymore really. Um, TypeScript is kind of like the de facto way to it used to be a real headache cuz it's there's it's a lot more laborious to write TypeScript, right? You got to type everything. Well, you don't have to actually because you can actually write JavaScript in Typescript. It's backwards compatible, but it is laborious and to get things up and running, you have to it's you have to learn it. But now with AI development driven development, it becomes a lot easier, you know. Interesting. Cool. So, Typescript doesn't really add anything to this. Like, you could do this in JavaScript.

You could absolutely do this in JavaScript. You don't need TypeScript to leverage WebGPU. Mhm. So, but I could have guessed that this was essentially 3JS. So, this isn't this is not accessible. You know, you couldn't you couldn't have this as part of a form. You couldn't. This is not a It's not practical, but it's a it's a flex on um physics and and web. What do you mean it's not practical? I like don't you mean like every website has to have this functional? Maybe then functional. It doesn't it can't work inside of a form, which is where you would expect this. it can if you built your whole experience with WebGL, which again isn't um accessible, then you could probably have it do something.

But yeah. Uh so couldn't we like have a duplicate of the form or like a couldn't we have this as a kind of like a layer on the top? Uh, but if I change the values here, being able to change it in an like valid HTML like form input and vice versa if somebody using like like the keyboard to change something um show it show the values here. Can't we make it essentially I'm saying can't we make this um accessible? Is there a way to do that? I don't know actually. I don't know whether there's a because I know you can pass values into shaders from outside but I don't know whether you can pass values to and from the canvas because this all this is is a canvas.

I don't I don't know. I don't think so. Okay. Well, maybe sometime in the future, but for now this is just a cool experiment and it's technically very impressive. And uh here's the the the tweet. Um oh here we see the type GPU it's the it's not even the the V1 release and for people who want to uh like do to learn more about it I found like a few uh YouTube videos like you can check it live coding liquid glass with type GPU so yeah that's pretty cool and this post got a million views like this one which is pretty Good. Right. A guy to follow is a guy called Ji. Do you know Jhei?

No. Uh, give him a search. J HY. Maybe it's even uh in Twitter. In Twitter. What is his name? J H Ey. He does some crazy stuff with CSS that's all accessible. Oh, I think I know he's Yeah. Yeah, of course. Of course. Yeah, I I'm already following him. And yeah, so this guy, so I saw this actually. So this is also really interesting. So this is this there is this kind of trick if you wish uh to visually uh make elements you know space correctly because we don't want to mathematically align things. We want to visually optically uh have them aligned. So when you look at them they look align. You know, the typical example is when you have the play button inside of a circle because it's a triangle inside of a circle.

It has to be uh optically aligned. And this is a really interesting way of optically aligning things. So yeah, and yeah, a lot of the things that I do uh are in some way or another are like coming from him as an inspiration, you know. He's his work is very very inspiring. Yeah. Very cool. Half the stuff I I can't even read half of his CSS. It's it's crazy. It's insane. Like he he Yeah. He does so much with CSS and most of what he does is no JavaScript. Just CSS. Yeah. And when he does JavaScript, it's just like, oh, here's a line of JavaScript. I mean, It's very cool. Yeah. Yeah. I give him a follow.

Jay is insane. Well, thank you for like mentioning his name so specifically. So now I know him by name. I just I just knew him as the CSS guy who was at Versal. Yeah. Yeah. He's staff design engineer at Shopify. And design engineers, by the way, this is what I'm like following the most these days. Uh I'm like trying to find more design engineers. Maybe Conrad is a design engineer. Oh, he's a software engineer. Yeah. Very cool. Uh that's an old trick from graphic design. Oh, Matthew Munger is on the uh on the chat. Uh we got some Web Flow news today. Yeah. um stare at the layout, cross your eyes, and blur. So he's when you had that um blurry thing thing.

Oh yeah, true. Yeah, you kind of just Yeah, you kind of just let your eyes just get a bit lazy and then you just you can kind of see it a bit better or Yeah. So this bit Yeah. Yeah, it does work. It does work. Cool. So, next should we talk about spaces? Let's talk about spaces. All right. So, let me bring something on my screen. Maybe something cool. Maybe this one is the coolest to talk about. Yeah, we do. We do this one. So, this is potentially the coolest way to generate using AI images, videos, anything. This is a nodebased system on a canvas like Figma which we'll see in a second from Freepick.

Um so here as you see I have a I have an empty canvas and I can start generating or here I have a template. uh if we load it we will see how it is essentially showing us how we can do stuff and I haven't given it a try but I was thinking maybe we can actually right now do that so we have a box which is a node and here we can add a prompt for example this prompt I assume if I add it and run the prompt maybe later can I do it for free oh no it doesn't let me to do it for free. Ah, see. Well, maybe we should explain a little bit about what it is first.

Um, yeah. Yeah, I'll explain. So, this is all a nodebased canvas that allows allows you to prompt an image for example, create an image and then add a style to that image, add an edit or make a video out of it, and then kind of like stitch these together. all with these nodes. So these nodes you can imagine them as input output. So the input of this or the output of this node you connect it to the input of the other. So it's kind of as a reference. So you take this as a reference and then here you say what do you want to do with it which is this is a bit unfortunate that I can't do it even though it's using nano banana which is it's free isn't it?

No. Yeah, we can't do it. Uh, maybe I should have uh bought some credits. Anyway, so you can do this to not just make images as I mentioned. You can even make videos and you press run and you make versions and all of these are connected. So, it's in a visual canvas that looks really, really nice. It's really well designed. Um, very Figma-like. And I truly think this is the way to do uh like AI image gener image and video generation uh generation just looks really cool really easy to explore. It looks like a mood boarding in a way. And funnily enough, they have even in their templates, if I find it like here, they have templates.

And in their templates, you can do where are the templates? Um, if I find the templates, I don't know where the templates were. You could do uh exactly that that um like create a canvas just for ideas. Well, I I would I that wouldn't be something I would do with it at all to be honest, but like in my head I would describe this as a multimedia um transform nodebased multimedia transformer tool. So taking you know you might take an image and say hey make this im make this flower blue and then you're done with it all done download the image. What this allows you to do is you can take an image and you can say right make this blue and then in the next stage of the transition make this 3D and then it's you kind of got this multi-stage um manipulation of an image or a video or I don't know what the inputs were on there but you it could be a 3D model or something like that but it it just allows you to build these pipelines to manipulate images.

And what I like is you could let's say you apply a specific kind of treatment like maybe you take a concept image from an artist and you always make different 3D models that you then put on your website. This is an amazing way to just have a pipeline where you can dump your image which you always get and it will run it through that pipeline and generate like six or seven different formats. And you know, maybe it's something for a poster, maybe it's something that you know, your website can use, maybe it's something, do you know what I mean? So, it's like it's a really nice way to just get different types of media output from one base of media.

So, so it's continuous as well. Yeah. Um, I think it's a really cool tool. I think it's really, really cool. I can't see any for me. I can't see any use for it, but I think I just need to simmer on it to think, well, how can I use it? Because I only just found out about this like before show. So, um, yeah, but it's it's definitely an interesting way to manipulate, change or create stuff. Yeah, this is the new way to generate probably like marketing assets if we generalize it. But you'll have tools that do one to the other. Do you know what I mean? You'll have a tool that probably takes an image to 3D or take a video and make um you know, whatever from it.

But it's that multi-stage you can pipe through via prompts. Every stage I'm I'm guessing this is all 100% AI. So it's different stages are then prompt. Uh you prompt the stage. You can pipe the same prompt into multiple things or you can provide different prompts to the multiple nodes. It's cool. Are these the templates there? Yeah. Yeah. These are Let's have a look at one of these then. See what they got going on. This is generating a 3D model which is also it sounds really cool. Yeah. Yeah. So they start out with an image. They got a prompt to make it. But this is like this is probably the simplest thing you can do. You could then take this 3D image then well probably the the final output would be a 3D image but you could have you could have changed the color.

You could have different Yeah, that would have been a perfect example actually like the um Ray-B band glasses. Maybe there's color variations on the Ray-B band glasses, but you take a you take one source image and say generate a red, a green, and a blue and then take that and then make it more into a 3D. So, you got three 3D models. Um yeah, actually to be fair, that's that's kind of two because they've made it 3D and then they've spun it around into making it into a video. I don't know. I don't know. I mean I don't know what nodes they give they make available to you but yeah so here we see the input it's just kind of like separate boxes and then you you use these as input for the AI model and then you have like different versions of the essentially the main image that you had here.

Yeah. Yeah. That's where I'm gravitating to like product shots and product imagery, you know, where you just need to throw it. You need stuff for posters, you need stuff for web, you need whatever. Yeah, these used to be so freaking expensive to do. I actually did a course, a photography course on like how to do product photography. And yeah, you know, you'd use like really expensive cameras like 100 megapixels and then have like I don't know like 50 shots of the same product. You'd do like 50 layers with different lightings and different levels of focus. It's still done. Like when you see the the picture of a Rolex, that image is not one image. It's like probably like 50 images layered on top of one another.

In one image, one piece of the watch is in focus and has the best lighting. And then in another one, the other layer and the glass, for example, the reflection is a different layer. And you put these layers on top of each other. And this is like very very expensive and it has to be very detailed. I'm not sure if this like I'm not just talking about free pick if in general AI can replace that for product photos on that level. But obviously if you have a normal like a less expensive product maybe uh you could take a normal picture with your phone and then run it through something like this to generate environments around it and different lightings.

And you see how the lighting here uh reflects the the sun and you know creating these interesting shadows. Um but like looking at this this is still a bit too soft to me uh for me. Uh it does look a bit AI generated the the glasses but that that's a different topic. Yeah. Yeah. Cool. Pretty cool. How much was it again? Did I see like5? I'm not sure like which one. Uh let's me actually see where did we have it. Um okay let's the pricing here we have the pricing. So this this one has the AI kick off your AI journey. Okay so we can do AI here. Don't think I'll get your lot right, but probably play around with it for five euro.

7,000 AI credits a month. I don't know. Whatever that is. Yeah, he's 50 videos a month. That's okay. That's not too bad for five. Oh, sorry. Eight year eight. Yeah, if you do monthly. Oh, I see. Got it. Got it. Got it. Yeah. Cool. Yeah. Yeah. But it looks really cool. Here we see also this is this is where AI is going with image and video generation. We have these models like the Google Vio and Cling and Sora and what else. And these companies are building tools around them. And I'm really excited about the the tool being that place where you have all of these different AIs, you know, the best of all and then do something with them.

Anyway, super cool tool and I'll probably maybe I'll um I'll try this instead of paying for what's it called? Midjourney. I'm still a paying user to make midjourney. You use it a lot then? Not myself but like my uh my team uh my designers they use it for and actually we use it for our product like flowing as well because the images we use there we want like a static images and that's like an easy way to get really nice images uh for let's say mockup shots in a way. Yeah. Maybe we'll use this. Yeah. Go for it. That's how you get on. Oh, actually I will because I asked I asked my team to create a video of a frame like a thumbnail that we have.

So we have the thumbnail designed but we don't have it animated and I don't want it in a specific format. I just want it to a few things to move. Um actually I will sign up and use it for that. So yeah. Nice. Now I have a use case now. Now I can now I can actually buy this and have a use case for it. I mean for eight euros I you know you're not exactly stretching the stretching I know. Yeah it like because I have just so many like why why why am I paying for mjourney like we're not using it that much? Not really. Most too much money in those tents. That's what you got.

Yeah, we have month that we don't touch it. Yeah, but I don't pay for adob anymore. So, that's another reason to sign up for this. Yeah, good riddens. Good riddens. I had been playing with um I still haven't figured it out, but I had been playing I bought a subscription to Canva and I've been using the AI thing, but all I could like I was trying to because I was trying to generate a thumbnail of you know the Neo robot. Did you see my post on LinkedIn? No, I don't use LinkedIn. I go to LinkedIn from time to time to post something and then I leave. So the the thumbnail I used for one of my posts about that was this old man stood next to Neo with his arm around him or whatever.

I just wanted my face put on that old man, do you know? So I was trying to select the layers and prompt it, but it just kept on generating a robot. So I I need to figure out what it is. But um yeah, I'm hoping it is literally as we hope and well what's actually practical which is actually modifying the image with AI rather than just shitty image generation but we shall see. We shall see. Yeah. Um let's move on. Let's move on. We can talk about this for ages. Uh few add to stage. So let me share the right tab. What kind of world will we live in when AI is not only responsible for building applications riddled with security flaws while the very systems designed to protect those applications from breaches they themselves are powered by AI.

Well, open's latest tool Arvark embodies this paradox using both as a builder and a security researcher. So I came across OpenAI launching Arvark which is a autonomous security research agent powered by GPT5 and um I don't know I mean I use AI systems I use SERP actually to be fair I don't know whether that's AI powered but I use AI to MCP you know trigger the SE grip review and whatever but it's interesting that you've not only probably got codeex writing the code, but you've also got the the Arvark telling you everything's fine. And it just feels like this is a way for AI to conspire against us and actually just start building software that's uh I don't know either full of security holes or um very well uh hidden security holes that AI can use to breach into the software.

I don't know. But dystopian views aside, I quite like these security tools that specifically you know it's their one job to do this. Um, I've recently used Test Sprite, which uses AI to write, you know, just specifically for tests. When you've got these tools that are really, really specific for one thing, they tend to be very, very good. So, I would be tempted to use Arvark. I don't know if it's free actually. um works by monitoring commits and changes on code bases, identifying vulnerabilities, how they might be exploited and proposed fixes. So yeah, would highly suggest adopting kind of specific tools that you know specific use cases because when you're coding, when you're vibing out, you're asking AI to take care of so many things.

Accessibility, security, um you know, design, uh development, all of this stuff, you're asking for it. when you really dig into the minutia of all these different disciplines, you start to realize again it's just like average at best. So, welcome to the world Arvar. I want to find out the price to be fair um to see if there is a price or it's probably just included in the in the uh subscription to chat GPT probably. I'm not too sure, but yeah. art Pretty cool. Cool. Pretty great. Yeah. Um, let's go into the comments section. We have stuff. I've got stuff. Have you got stuff? Uh, not really, but I I see like So, we've we've got this video.

And actually, this isn't specifically comments to be fair. Um, this just re there's a lot of there's a lot of negativity on this video. A lot of coping on this video in my in my opinion. Like a lot of a lot of like nah nah this is never going to happen and um fake news. Uh you don't need an engineering degree to vibe code. However, um helps to have some knowledge. Like a lot of people see don't see this ever being a thing. And and if you don't know this video, this is me talking about the uh how cheap it is and how easy it is to now build software. And oh, are we even talking about it today?

Actually, I'm just going to flick through our notes. Um it's it's the sponsor. This is the sponsor that I told you about. And uh it's making it this I'm dealing with a sponsor at the moment who's making it even more easy for people normal people to build apps just by prompting. Okay. Yeah. The first thing I want to address with the comments here I think they've I think a lot of them have misunderstood what I was saying. A few actually a few of them seem to think I was talking about that people can build their own people can build SAS applications and then sell them and and there's a lot of commentary around uh the the the maintenance that's involved in building SAS you know the the uh yeah security databases or and all the rest of it.

I'm not talk I'm not talking about rebuilding SAS. I'm saying SAS is dead because people will be empowered to build their own software. They they're like instead of going online and being like oh you know uh zero as an example uh instead of paying for zero why don't I just just build my own little tool and I don't need to pay for anything ex you know external. This is what I'm talking about. people will be empowered to build their own software. And I've got a sponsor sponsor that I'm chatting to right now, which makes it even easier to use your iPhone to build native apps. So again, that why would I why would I pay £3 a month for this meditation app?

I'll just build my own meditation app. It's like it's it's easy now. So this is this is what I'm saying here. But the reason why I bring this up is it's it's now very surprising to me and you know I want you to challenge me on this, right? It's now really surprising to me how people lack and I think I did a tweet about this which you might have seen. People lack this ability or or people just don't have this ability to think progressively. They're kind of like there's two aspects that they're one they're locked into the current because AI has faults, right? They're locked into the current way that AI messes up and it just it hallucinates and it it gets things wrong, but for whatever reason lack the ability to really just project into the future and say, "Well, actually, a year and a half ago or maybe two years ago, Will Smith eating spaghetti was an absolute mess."

But now, you probably couldn't tell the difference between the two. Well, you probably could, but it's exaggeration. But the exponential progression, all you've got to do is extrapolate that into the future and say, well, why can't why won't we be able to build apps? Why won't normies be able to build apps in the future just by prompting? And it's just so surprising where people are just lack I don't know seem to lack that ability to actually just project their their thoughts into the future. I don't necessarily um enjoy or agree or want this future that I'm I'm speaking about. Heck, I'm frantically trying to become a better AI developer beyond. Well, I mean, I kind of enjoy it, but like if it was my choice, I would just I would just keep with my current skill set and just have make money off my current skill set, but no, I've got to retrain now.

I'd prefer not to have to retrain. I'd prefer just to be able to but I'm read I'm reading the tea leaves and I'm thinking right I need to prepare for a future that I see whether I want it or not. Massive coping I don't know massive massive coping mechanisms that that seem to take place when you're challenging people on their you know their norm. The other aspect any points on that by the way I think a lot of it is because we had chatp 3.5 which was the best um like the the the the thing that got the most attention at the start right because we had other versions before um I believe that was like the good model that we started you know noticing and since since then have we gotten like a like a true big update.

I think it not really not something fundamentally different because anything from the you know we went from the old AI to the transformers that Google introduced and then the new transformers since GPT 3.5 I believe if I'm not wrong they haven't gotten a massive massive upgrade. Maybe that's why people don't see um these AI tools being able to do anything like truly meaningful and big. Right now we can do a lot of prototyping and it seems like the rate we have a we definitely have a progress in the complexity of the prototypes that we can make and the even MVPs in some on some level but to create something like making notion you know that's not something that AI can do anytime soon right so maybe this is where the disconnect is On that point though about like anytime soon like no one's saying it's going to be soon like no one's no one said that it will happen soon.

I'm saying it will happen right and that that sort of level of thinking is like because like it's it it's not that it's not progressive enough like in my eyes anyway the fact that you no one's saying soon and you know we're not I'm not saying with the AI that we have now I'm saying just extrapolate what we have how far we've come in two years and paint that into the future on a undetermined Okay. So, can I can I play devil's advocate and be like advocate for the comment section? Well, if you do the same thing with the moon landing, we should have been now somewhere else in the galaxy. But we haven't gotten back even to the freaking moon.

we we are not uh we can't say for sure like tech will evolve the way that we imagine it will. Uh the same thing that happened with uh cars. We thought we will have like flying cars but we still don't. Um but it's physically impossible to have flying cars. And I don't mean like yes we do have flying cars right now but it's physically impossible for all of us to like have it as a norm to have flying cars because to move that amount of air is extremely loud. I have a drone here. Like my neighbors get annoyed by this let alone with you know a freaking car. Yeah. So we um get over excited about so many things and we we think that okay take the the current thing and just scale it in whatever direction just scale it up we'll get to the moon but we don't always and with software I'm honestly not convinced that the current um models can get us there not sure because we are we have literally run out of data.

Now we are using the models to kind of generate data be agentic and we are building around the current model models because it seems like we are we have reached the level that the current models can um give us in a one go. So we are building around it. For example the the free pick example that we showed it's not like why are we doing this? Because the the image and video generation models are not that true like they can't read our mind and we tell them to do something and they just get us that thing that we want or a set of things that we want. We still have to build tools around these tools to get there.

And the same with back going back to the comment section and building software. I I think it's not going to happen anytime soon. To be able to do anything like that big with AI. When I say that big, I mean creating notion. When I say that big, I mean creating like Google Chrome because you know Open AI, they didn't build their own browser. They cloned Chrome. they cloned Chromium and they just built a little bit on top of it and not much and it's you know buggy and has issues. So I don't think AI is there yet and I'm not convinced that it will get there. uh when I say it there's an asterisk like anything that can happen obviously will happen if we given enough time you know but I'm I'm not sure if let's say in the next 10 years I'm not sure and Anu Karpathy I don't know if you've seen that video of him that I mentioned last week um he's the AI guy he's an AI scientist he's done research he knows how to code uh these AI models.

He has done the work in the field and he's saying that these agentic models are a decade away, not a few years. Cool. Well, I'll leave it there. I'll I'll go on to my next bit because yeah all all actually all I'll say is like I didn't say a timeline so know and the only other thing I'll say is like I actually I think we we can absolutely go further than the moon but it was it's not technology that's stopping us it's literally we sorry we can but like with AI I don't know we have other issues with AI we have reports on on there are chips from Nvidia laying around because there is not enough electricity to turn them on which is crazy to think about.

Yeah. So, which will go into what my second point is, but like Yeah. So, with with the moon landing or you know go flying to the moon, it wasn't it's not technology that's holding us back. It's actually learning that well there's no reason to go to another planet really. We need to be looking at um Sirius a whatever it's called, you know, the closest star. Um so saying that the secondary yeah um so the reason why I bring that up as well is like and this this goes both ways is is mo more so a so socioeconomic factor which is once again this rest in my opinion restrictive thinking in that we apply our current understanding of the world around us and are unable to like we we we put the world around us into the future and it's like well you can't do that and I'm always reminded of what um Rhyar was saying he what was the phrase he said like we're a we're ancient hardware running in medieval software living in the future or something to that effect like we're always kind of playing catchup and we're always like we're always kind of you know we're never um fully in the I mean for that matter we can't imagine it but like I think what we care about will be different you mentioned a browser this is a very stupid example and I don't want to labor on the browser thing but there there will be a few there's reason to think of or care about building a browser anymore right in the future it might be that people don't feel the need we might be living in a future where SAS dies because you know what AI is off doing everything and there's universal income and we're not working anymore.

SAS might die because of that, not because of um people building their own applications. We we we need to understand that the the world we operate in, the world model that we operate in is not going to be the same as what we what we operate in now. The things we we care about, the things we do, the thing the thing our dayto-days just will look very very different. And the the quickest easiest example I can give on that is COVID. How that came a year of our lives maximum two years completely has changed everything and now we just know no different and we just don't even think about it. Yeah. This is a now a new world model that you could not have one you couldn't have anticipated it but you also could not apply the things we did and you know the things we like now people care like I know mothers who care so much more about the being there you know raising their kids they care that so much more because they have access to it does that make sense like this whole the whole world model does it it totally makes sense and this is what I'm saying about predictions.

We did in the past overpredict and also underpredict things. We we sometimes over overestimate and underestimate. A good example and it's quite funny is uh wireless power. There was a time where people thought every device will run on wireless power that we won't have any cable to plug in anything. it will be just charged wirelessly because we can do it. we can actually do it and it's like to some degree but it's not it's not efficient like we we could not figure out a way have not at least for now and I'm not sure even if it's possible physically to find a way to efficiently move electricity sort of from a source to another even like the short distance we have with our phones we lose like 20% minimum to the to the heat and things like this technical stuff But there we did think that we'll have wireless power everywhere and we don't have it.

It's not even close and it seems like we won't uh have it at all with AI. I'm not saying it's literally the same. No, it's going to obviously change the world. But I'm not super convinced as somebody who I I would by the way call myself like very optimistic but I'm still not very convinced by AI being to that level that some altman that is trying to sell it to us. Um let's move on because of timing. But it's it's it's an interesting debate. I'm I'm just we'll see I'm just so I'm so like amazed at how certain people are about their uncertainty if that makes sense. It's like nobody knows. Nobody knows. But there's so much like all all I know I know enough to know that we don't know but all we can do is predict and we're always looking for the to the future but some people just can't do that.

Very quickly because again we'll um and it's not really much of a comments thing by the way. Let us know if you if you have comments like let us know because we will read them and we'll create they create such amazing discussions um for some of the stuff we go through. But this is less of a comment more of like an evolution of um a video we went through or sorry a topic that we went through last week. So, uh, Elon Musk is taking Open AI to court. And what this has meant, I think it's Elon Musk. I I might be misremembering the beginning of this video, but here we go. Musk versus Alman. It is Elon Musk versus Alman.

And Ilia Sutska, which is one of the OG founders of of Open AI. He he's Yeah. pretty big pretty big cheese. he has had to come forward and be very um honest with a lot of the inner workings and how stuff went down with um with with uh Sam Alman because if you remember rightly Sam Alman got kicked out of Open AI. Yeah. And then a week later came back and it's just like it's just kind of like drama. Yeah. huge drama that's just been kind of like just brushed under the carpet. But now what's come to light through Ilia Scog's kind of reveal is that actually he was a big proponent behind Sam Alman and by the way I should should have prefaced this.

This is again off the back of some AGI videos which have just gone out in the last 20 minutes or so. Um Sam has been h has seemingly been uh demonstrated as a very tyrannical leader. You know, we see a very soft spoken person on stage who seems to want to do good and this and that. Uh who was the female the the CTO? Uh the Italian Italian girl, you know her. Forget her name. Ilia. We're all discussing and Sam Sam and again lied to the board. And by the way, I think the the entire board has kind of changed now because they all lost faith in Sam and all the rest of it. This does not paint Sam to be a very very nice person at all basically is where I'm getting at.

And off the back of our conversations last week about AGI, you know, and I'm I'm I'm there saying in the video, I'd like to think that we are uh you know, we're all good people and whatever. This kind of me up a little bit thinking, you know, is is Sam really the right person to be behind the helm of potentially one of the the winners of this race? Is he someone who could be trusted with AGI if he's been this lying and deceptive? And I know we're only seeing one side of the argument here, but with this video goes through the different accounts of those people and discussions that took place that inevitably led to Sam Alman being fired.

So, it seems like he's a different person to different people. I'll kind of leave it there, but it's Yeah, it's pretty um it's pretty telling. This only came out a couple of days ago. Yeah, it doesn't seem like he's a really specifically nice person and it seems like he lies easily which makes it really difficult to to have faith in having you know AGI in his hand but I'm pretty convinced that we are not going to see any uh to we're not going to see AGI anytime soon at least to the definition that we all had a few years ago. And one of the things that makes me not really like what Sam is doing is that he's changing the definition of AGI rapidly.

Like every time he's being asked, he's lowering the standard for what AGI is. It seems like nowadays to him AGI is an agent that can do a bunch of things um more than what we have now. So he can claim that they have reached AGI in I don't know in a year or two. It seems like that because I honestly don't see any like intelligence uh emerging right now like the the magic of these LLMs has gone away. It I don't see it anymore. Yes, that when you start using them very like amazing but then they become like still amazing tools but not intelligent creative conscious tools uh or beings not tools which is the definition for me for AGI.

AGI should be creative. It it has to be creative. It has to be intelligent and it probably should be conscious. I'm I'm not sure about that last one because, you know, do we need suffering? I'm I'm not sure. It's philosophical, but the thing that Sam is portraying as potential AGI is not. To me though, this is less about AGI and more so having someone potentially being quite tyrannical and and and disingenuous at the helm of something that whilst it might not reach AGI, it it that's enough of a driver to get somewhere where it's very they've built something very powerful and very Yeah. powerful, I think, is the only way to put it. Yeah. So, yeah.

I don't know. It's interesting. It's funny. It's not funny, but it's uh Yeah. You see these because Ilia left. A lot of these people left. Board member board members left. They clear uh uh Dario from Anthropic is in on this, you know, he's he's been wrapped into this web and I just again I just worry that Chachi PT might be the winner or OpenAI might be the winner and they're not going to they're not going to stop until they have something very very powerful, you know. But anyway, that's that's all I'll that's all I'll say on that. Should we talk about PewDiePie? We should talk about PewDiePie. Yeah. Yeah. I had no idea who I don't I still don't know what why is was he a gamer?

Is that how he got so many followers? Uh we'll we'll talk. Yeah. Yeah, we can talk about that. All right. Okay. So, you know PewDiePie? Yeah. It's kind of funny to talk about uh this. Wait, let me Okay. Uh we also should bring it on this. So PewDiePie is now training his own AI. Like no joke, this is this is kind of crazy. You know the guy, he was a gamer. He started as a gamer. He started streaming and also sharing his videos on YouTube and he become big before T-Series. I don't know who knows this. If you know this, comment it like tell me that you know that part of internet history. This is this is a very important part of the internet history.

Uh PewDiePie versus T-Series and then T-Series took over. That was an Indian channel that got bigger than PewDiePie. PewDiePie was at that time the most subscribed uh YouTuber on in the world. Anyway, so he got really big and he made a bunch of different videos. This is what you just asked. Yeah, he he didn't just make gaming videos, just a bunch of different funny videos. But recently, he's been making videos on how he's building his own PC and started like with uh with Linux and things like this. And now he made a video called Stop Using AI Right Now. And it's quite funny. in the video he he shows how he is running his models locally.

So this is how it starts. He's just yeah he's just running a bunch of models locally and he's in this video specifically saying he's not touching any image or video generation uh tool. He's just talking about like LLMs for Yeah. infrance. Exactly. And it's quite funny the video because that the most fun part of it is when he he shows that he creates he he creates a console of his. If you don't want this to be spoiled maybe you don't watch this part but but um he creates like eight agents that talk to one another. when he asks the AI something, these eight different AIs think and then they talk to one another and through a voting system, the the winner vote gets, you know, outputed and the ones the agents that uh perform bad consistently, they get ang he essentially removes them.

He kills the AIS and the AI figure that out. They see uh that and he reads the the logs and he sees the AIS talk to each other and like what who invented this sick game that kills us for you know having a different opinion. So the AIS come up with a way to compensate for that. So they vote in a way that nobody gets killed. So they get not really useful because uh they change their opinion so they don't get uh collectively killed anyway. He he shows that and he shows how he made his own web uh like interface all vibe coded to do to talk to his AI to do web search. Um yeah to do a bunch of things like this and it's it's really cool.

The video is really cool and he continues by the end of the video he shows that he's starting to work on training his own AI not just inferencing but like training like a little he's just fully getting in and just enjoying becoming an engineer in a way. Yeah, he is becoming an engineer like do a bunch of like his last videos are around these things. Yeah, it's funny because like honestly don't don't get PewDiePie at all. Like he's huge. I just do not I enjoy his He's not someone that appeals to me. I don't know his history. You just he said he was a gamer or something like this. It's not the point. I watched his um he anti- Google video deleging himself.

I watched that video and he inspired me to start looking into this stuff and you know in in in some small way even though he wasn't specifically talking about AI he pro he probably had some impact on inspiring me to start to look into local LLM stuff because I remember looking at that thinking I'd love a remote server that's stored in my thing that that I could access all my drives and I started looking into it and you know trying to figure all that out and then I was like oh it would be cool to have my own own cloud LLM running all this that and the other. So it's so funny how a quote unquote normie is doing things and doing things well clearly like I haven't seen that video by the way but he's yeah you know if he can figure it out I can figure it out and fair play to him you know didn't you sayo covered this video as well?

Yeah, he uh I I think so. I I didn't watch the stream, but his stream name was like the stream title was reacting to PewDiePie's video or something like this. And a bunch of people started reacting to him. And I saw posts on Twitter, on Instagram, people praising him essentially saying that this guy he went through he did you know video video funny videos on the internet gaming and now he's just learning uh programming. And he actually says in his video that he truly enjoys going in the zone and learning coding. and he says it's so hard, but he's enjoying it so much because he has to concentrate and just learn so much. Uh, and I I I do feel the same like I'm pushing still like code and still learning and it's really fun when I can get the chance to get in the zone and just focus on learning it.

So, uh, a lot of people saying AI is coming and don't don't learn this, don't learn that. I think that's like the worst advice. Uh just learn whatever you are passionate about and if it's coding even better. I would highly recommend learning it. Yeah, I say again gatekeeping is just for some reason people just want to gatekeep you know any sort of discovery and enablement of people learning stuff. And PewDiePie is an interesting case because like I say he built his own server. He built a Linux server that he could run his email, store his files, photos, all that sort. So, he's obviously a very talented, clever clever person. However, AI has enabled him to essentially become a developer, a interested in it, you know, only for his own personal projects.

And you know, I won't dwell on it too much, but it just goes into what I was just saying about this idea that people are building their own apps and they are being introduced to it and it's not as hard as they once thought and it's more fun than they thought. You're going to capture people just by lowering that bar. So, fair play to it. Good on him. Yeah. Cool. Next, uh should we go back on something? Which? You heard me typing. I'm I'm sorry. You heard me typing. So, I was I um I bought Freepick. I couldn't help myself. I bought a plan from Freepick. Let's We were We were quite long on that subject, I'll be honest.

I'm going to say let's save that to the end. Okay. We'll save it for the end. Yeah. Okay. Sure. And I will show you the results. Yeah. Yeah. Yeah. Stay to the end to see the results. Engagement. Yeah. Yeah. No, definitely. Yeah, because we we're on it for quite a while and we've got a few to go to and then then we'll pick it up depending on how we feel. But um yeah, so let's talk about we'll do we'll go quick with this one. Um just because it's not too interesting. It's bit it's a bit political. So, Amazon have gone to war with AI agents, sending an aggressive legal threat to Perplexity, telling them to stop agents running on their website, Amazon.

Let's see what Plexi has to say. So, basically, um, you we you've seen GBT Atlas, I use Comet. These agents, they go off and browse the web, do things for you. It's your own personal assistant. It's uh yeah, it's your own personal system going out onto the web. Amazon have sent a legal threat to Perplexity telling them to stop receive a legal threat from Amazon demanding we prohibit Cobbit users from using their AI assistance on Amazon. I wonder if Chat GBT Atlas and Opera Neon and DIA have all received the same threat, but it's uh to stop using on Amazon. Why do you think this is Cavaza? Why do you reckon when when you say Amazon, do you mean Amazon.com as the shop?

Amazon.com shopping. Why do you think that they want them to stop? I don't want to reveal it. I just I just think because they are evil they probably have their own version of comet something similar and they want users to use that and they don't want any competition that that would be my first guess what would be your second guess second guess probably I don't know server cost or something like this would it cause uh way more surge in search and using their essentially services. That would be my second. What's the biggest money earner on the internet? AWS, isn't it? Like Oh, money. No, no, no. What? On on the web, what's the what's the biggest ads?

Ads. Ads. Oh, ads. Oh. Oh, ads. If you got an agent running off and doing these tasks, they're not there to advertise to you. They're not there to upsell you. They're not there to manipulate your buying behavioral habits. And this is why they don't want them working on specifically Amazon.com, which they're still evil. Well, yeah, exactly right. There's Yeah, there's no two bones about it. They're more interested in serving you out sponsored results and influence your purchasing decisions with upselles and confusing offers. If you got an agent doing it, you haven't got to worry about that. um cancelling your Prime subscription is is difficult or confusing, you know, because an agent will go and do it and it might take them two hours to do because that's how long these agents take nowadays, but like it will still get done and you won't you can just focus on what matters.

They will earn less money through agents. Now, I'm not sure how I mean, I would like to see the uh the actual letter, but I I I wonder they obviously won't have spun that angle. That's this is something that Comet or Perplexity have inferred, but it's just interesting. And I just I hate this. I I I hate this just capitalist greed with such passion that we can't just give an authent someone someone an authentic experience by giving them something that they actually need or want. It's like I mean it's your technology is rubbish Amazon to be fair because I'll just do a search on something that you clearly just need once like a kettle or whatever.

Here's loads of kettles. You must love kettles. Even though you've bought a kettle, they could sort your algorithm out. But do you know? Yeah. Yeah. Go. Go on. No, no, no. I wasn't going to say anything. I wasn't going to say. Okay. Do Do you know the It's kind of like an old joke about Apple and Amazon. Like uh Apple has proved that design matters to be successful and Amazon has uh proved that design doesn't matter to be successful. Why? What do you mean? Because a lick like have you have you seen the design of Amazon.com or the whole reason Vers exists as AWS sucks and Amazon.com is trash. Like it's it's funny like how comically bad it is when it comes to design and how good Apple is.

Yeah. Anyway, well they make up with it. They're a digital service. Apple aren't a digital service. They're like a hardware company, right? People don't use Apple. Well, I suppose you use iOS, I guess. But like the the feature set of Amazon is just so good. As corrupt and awful Amazon is to society in general, we still are like, ah, but it's just so easy. You know, the fact that, do you know, and I did I tell you this? So the the lens that you're seeing right now is is a is a new lens. I was I was I use this one for my YouTube videos and um I use this one for the podcast, right?

So I bought this lens 50 nifty50. I also bought an 85 mil um 85 mil. The 85 mil was something like 400400. This was something like 150. A very cheap lens. Um great lens for In fact, do we have the same one? Do you think? Do you think we have the same lens? Probably not. a nifty50 as well. Do does yours when you hear it focusing, can you hear? Yeah, the same. It's a nifty50. It's It's not It's not amazing for video recording if you need the mic to be on the camera because you will hear the focusing, but like for what we are doing, yeah, it's amazing. It's crazy how good it is. So, I bought these two lenses and low key I used them both for a wedding shoot with my friend and I was like I tried them both out and I used this one.

I was like I quite I do quite like this. It matches your one all the rest of it. Anyway, so I send the the expensive one back, the 85 mil. God, I wish I had chosen the 85 mil because it's much more expensive than this and I would have sold it and then just bought this one for cheap. Anyway, I've ruined the surprise there. So, I sent it back and there there was some technical issue inside of Amazon. They said they've received the this lens, the one that you're seeing right now. They've received it and I'll I'll expect to get my refund in three to five days, £150. I like, whoa, whoa, whoa, whoa. No, no.

I sent you back the expensive one. Yeah. And whatever. And so, I chatted on Amazon. Five minutes on the chat interface. Don't worry. Keep the lens. We'll refund you everything in the next three to five days. You know, they are they are good, you know. They will they will I've had a lens once. They refunded me £900 for a lens once that went missing. They just gave me 900 quid, you know. So, that's Yeah, that's crazy. But, um, a lot of this, I mean, for a lens, it's really crazy. But for smaller things, they are not doing you, you know, a service. They are not doing it because they are good. Because if they have to take it back like if they have to take it back and get rid of it or fix it, it's really expensive.

Yeah. So worth So it's cheaper for Yeah. Like even if you are a merchant and if you sell like thousand and thousands of your products, some of it will be damaged. uh and it's just the cost of the business and the the collectively the customers are paying for those things as well. It's a it's calculated in the price. So yeah. Yeah. It has to me as well. Yeah. Yeah. But point is is that they they their service is very good and we we don't care for the design because it's like I'm just it's function over form and Apple is form over function or whatever, you know. Um anyway, that's that's that that was uh be interesting to see what happens with this whole case.

I don't think they'll win because you can't stop anyone doing anything. Um, but yeah, let's move on to Mr. Doo. So, you know Mr. Doo, right? Yeah. 3GS is one of the most well-known JavaScript libraries on the web, arguably. Um, other than React, 3JS is is huge. It's what a lot of these 3D applications are built in. Some of the stuff we looked at at the beginning of the show, all this 3D stuff, a lot of it is built in 3JS. It's responsible for over 70,000 websites. And it was a startup. It was started by one man, Mr. Doob, and he's seemingly been completely converted to Vibe coding updates to this very important library. Now, this is just something I picked up on.

Let me share my screen. Um, he keeps chatting about claw code and like adding features, talking about adding features and things like that. All done coding. And I'm just so surpris him. I'm just so Oh, he does a lot of reposting, doesn't he? Um, yeah. Can we cancel out this? Can we cancel out reposts? Uh, no. You're just, you know, there's this one here on Chloral Co for the web, which by the way, you get $1,000 of free credit this month using Claw Co for the web. So, he's obviously Yeah. So, you can see his credit there. What's he got left? 981 credits. I see you. I see you there. Um, yeah. So, $1,000 of free credit if you're on the max plan.

Um, but yeah, he's he's making updates to 3JS done through VI. Can you imagine? He's got a lot of trust in in claw code here for sure. Yeah, he knows how to prompt. Oh, he knows how to prompt. He's got one in 3JS here. He's got one in 3JS there. And I'm just seeing that, you know, he's probably making very very he's probably making very specific changes to like one file. Do you know what I mean? He's the creator of the library. He knows the ins and outs of it. And but it's insane how such a amazing technology has been developed by one guy. Yeah. Yeah. I mean, he's got there's over 2,000 contributors to to 3JS.

Uh, in fact, there's exactly 2,000 contributors to 3JS. Um, but he's he's the one vibing out on it, which is pretty pretty impressive. Um, that's all I had to say about that, unless you want to say anything else. No, it it but yeah, it's really cool and was surprising. Surprising. Um, I wonder if math is still in the audience because we've got a uh we've got a we've got a uh web we've got two web flow updates to be fair. do let's do storage after this one. Yeah. Um so Finssweet our old friends have announced dev mode for web flow. Yeah. Um I really up that intro, aren't I? I'm going to redo the intro for the for the uh I wanted to actually show it uh but for some reason Let me Let me Oh, go on.

Can you sort that out? And if you can do it, just give me a shout. Web flow is f Webflow finally has a powerful code editor that allows you to edit your code just like you're in VS Code. It's Devmo plugin by our old friendsuite and it's pretty great. Um, did you get it? Yeah. Cool. Yeah, I I have I have it on my screen, but maybe we want to watch the video. Yeah. Yeah, let's watch. tools that make web flow even more powerful. Finsswuite extension is used by more than 30,000 developers in web flow extension right dev mode. Quickly update your HTML and CSS using a code editor inside web flow. Some use cases.

Instantly find classes, attributes, tags, and styles. Hit control or command plus F and type what you want to find. Quickly add attributes without the need to select manually the elements. For example, add attributes list load to your CMS list. Type FS list element equals list and FS list load equals more. Done. Bulk text change. Imagine you have a lot of buttons in your page and you would like to change all of them at the same time. Select the parent element. Find the text content of the button. Select it. Controll or command plus D. Change the text. Save changes. Done. That's really cool. Bulk classes change. Now imagine you have all these titles of the page with heading style H2 and you would like to change them to the class heading style H3.

Controller command plus F. Add the class you want to find. Now add the class you want to replace and click to replace all. Save changes. You get the picture. Yeah, that's really cool. So, not quite, you know, the fullyfledged dev experience, but this is already better than the in-house Web Flow editor. It's awful, the in-house web editor and code editor, sorry, in Web Flow. What I quite liked there is that you saw, you know, they they chose the parent wrapper. So now this gives you the entire this is the code that's that's been created under the hood as your no coding. So this is this is obviously something we've never seen before. You you you get this sort of editor when you have an embed element or well yeah just an embed element, right?

You don't get it for the page, but you can see the whole thing now. And I think this is great and I I foresee um AID, you know, updates to this. You use AI to vi vibe it out. Um it would be interesting because of course Web Flow creates HTML in a certain way. And I'm not sure how malformed HTML will, you know, how that will render out on the canvas and whatever. But regardless, I can see AI adding here. I can see just more VS Code like um EMTT predictions, things like that just being added to here. Really good first step to this this dev mode experience. And my heart kind of goes out to Miyagi who gave up on the original vision of Miyagi which is a vibe code experience with them um web flow cannibalized there.

They were going to focus on the coding experience. Now dev mode has come out. Now this is not to say that workflow would come out with their own version of this but you know this is this is already they've launched you know at the end of the day Miyagi has not launch their coding experience. So well I guess they have but they just haven't doubled down on that just yet. So yeah really cool. So is that on your screen right now? Yeah this is and is it free? Are you paying for this? No it's not free. That's the point. So, I thought this would be a part of the unlimited uh finuite pro that I got.

I'm not sure if I'm signed in with the same account or not. Uh but it seems like I I did try a few accounts and it seems like in all accounts it's asking me to pay if I want to use it more than 10 times a month for these uh features like merging combo classes. I really do like this. So, this is when you have, you know, you put too many styles in too many classes on top of another and you just want to merge them down to one class that holds all the styles. That one is paid. Um, and then the dev mode is and you you've got that, right? You've got that merge.

Yes. But only you can use it 10 times uh a month. Yes. Yeah. Yeah. Yeah. Yeah. Yeah. So here we also see their marketing website. So you have the dev mode, you have the fluid design generator. So this is a way to calculate responsive like fluid designs. Um then you have the merge combo migrating from pixel to RAM. Uh organizing classes and folders. This is also really cool. This is based on the way you name stuff. it automatically creates like some like like folder so you can easily see a component and its classes within that component. Uh you can navigate between pages based on your history and this is probably one of the most like useful ones that is kind of underrated.

uh the web flow breakpoints like the bigger break points they are super annoying and they are not useful at all and don't use them ever never and if you do like tell me why why would you do that to yourself and with finite you can remove them if you by accident add them web flow doesn't let you remove them but finite just access them uh and obviously you can duplicate uh remove the duplicate styles so that's that and we can also go look at the pricing. So the pricing is six bucks a month for everything and if you pay your is this across all sites. So this is you can add this to any site.

Good. Yeah. Seems like they've addressed the pricing here because this is this is doesn't look so painful. Yeah. Do I need one for each side? No. Finsu extension plan is per workspace. Yeah. Okay. I don't know if it's like per Web Flow workspace or per their own workspace. Uh Finsswuite I I don't know how to say this nicely but they are not really good with naming because this thing it's called Finsswuite extension. This is a web flow app. So we have a web flow app called Finsswuite extension. And previously we had the finite extension which was a finite um chrome extension. This what you see on my screen this one on the this is so they move the chrome extension into an app seemingly.

Yes. Yes. and they still call it extension and it's not maybe maybe to not confuse people but this is I argue is more confusing because they should be looking onward and like to the people that will join who are going to see it as a web flow app not as a chrome extension. Uh anyway, with the old thing, you can still access the head and footer code to your project from anywhere you are in the designer, which is really sweet. And the bulk redirects. So, the rest of the stuff has been moved here um if we like it or not. I wish they would not do this to us, by the way. I wish that they would keep the extension as it was and add this as a as something that extension a new extension because right now I need to also install this on every uh site.

It it takes a click or two but still the extension was just there once you would install it. But yeah, it's pretty cool and I'm sure I will be using it. I'm not sure if I will pay for it honestly. It's not a lot of money, but it adds up. These things add add up and I don't know how it works with my team. Uh, do I need to pay per person? Yeah. Is this is this a thing of vibe caring your own extension? You know, what's the process like of getting an extension into web all this? Honestly, yeah. You've got the you've got the um web flow documentation, API documentation. All this is doing is leveraging the documentation and saying, "Yeah, yeah, site scripts scanner is coming soon."

What What is that? Don't know. Bob Bobjack Jam says it sounds like they need 3D assets. A 3D assets thing. Oh yeah. This is This is So Bobak is from my team. We are doing the It's the I told you about the 3D and you've seen. Oh, okay. So he's name dropping is he? Yeah, man. Sounds like you guys need 3D assets, huh? Well, we can recreate these in 3D and make them like look absolutely amazing. Um, we have some work done for a client which we will share in a few days time uh once they once they go live with it. So I don't know what is sight script scanner. I guess it just looks through your site and identifies or collect scripts because some people just add scripts here there and everywhere.

It might pull them all together so that you can you know put them all in a a logical place like the header or the footer or move them to wherever you feel comfortable that you want them. Do you know what I mean? Maybe something like Yeah, but you could go to inspect and see all the scripts here and then find them in but you'd have to go look how you got to go through everything because with the embed element you could in theory put scripts can be just be anywhere you know and yeah but like you can't see them here. Yeah, you can see them there but you have to go and find them. Yeah.

Well, you search and then but yeah, it is definitely if you do a lot of these things, it is definitely worth like five bucks a month paying like 60 a year. Uh but I don't think I would need that unbite CMS. This is crazy that this is still not something that Web Flow could do um natively and import SVG. I believe there are a few apps that already do this, so I would not necessarily need to kind of buy buy this extension for this. Uh the the dev mode by far is the coolest part of it. Yeah, very cool. Should we talk about webo flow storage? Yeah, definitely. Take us through it or at least you know I don't know if I uh actually I have the link here so I'll copy the link.

All good. I'll just command T and paste it. So we have bind the data storage to webflow cloud and the way it works. This is something that you should explain, but it seems like this was not a thing before. Did Did you have access to this? Yeah. So, I've had I've access for this for a little while and you know, shameless plug on my course, my Web Flow Cloud course, I've got tutorials recorded on how to set these up. They might one might even be live because I remember getting um uh SQL light set up and running. So, what these are, they're they're an aspect of Web Flow Cloud where there's three there's three parts to it.

You've got SQL light storage, object storage, and KV storage. Now, storage is a funny word in these in these uh you know to name these from Webflow. They're not really there's the only one I guess they call it data storage there, but object storage is things like you know it's an S under the hood. It's an R2 which is an S3 um storage. It's a place for your assets, images, videos, documents, whatever. It's a place to put files and you can through web flow cloud you can pull and push that data and store and sorry not those those um assets you can pull and pull pull those assets so that's object storage KV storage and SQLite storage are very very similar and that they are databases essentially they're databases SQLite is a relational database which means you can have several tables and they can reference each other they're called relational data databases.

So you could have a users database, you could have a um let's use blogs as an example, you can have a blogs database, you can have an author's database or sorry tables within the database and a and a and a blog can be linked to a user and it can be or and it could be because of tags and all the rest of it. Okay, KV storage is very very simple. It's called key it stands for key value. So it's simple this equals that this equals that and there's no relationship between them or at least you would have to do something clever to establish those relationships but it's just a it's a it's obviously a lot quicker you know reading a whole database table can be very uh costly on on performance but KV it's very very quick sort of you know simple data basically so it's these three aspects are available on um web flow cloud I'm assuming this is now out of beta.

So, it will prompt me now to look into the documentation and see if the KV storage, which is the one I was having trouble with, is seamless and I can just create some KV storage. Um, you know, I don't what you call them just yeah, KV storage. So, that's what it is. And I'm guessing this is just it coming out of beta. Seems to be chugging along. I mean, I don't know what is the what is the hunger like on Web Flow Cloud at the moment, you know? I What do people do people like Web Flow Cloud? Do people see understand Web Flow Cloud? Do people I'm not sure actually. I see a need. Yeah, I don't I I don't use it.

I haven't had the need for it yet. I would like to, but I I'm like busy doing my own coding stuff mostly like front front end uh let's say design engineering direction. So no, I haven't had the need for that. Um, so for those for those that don't know, Web Flow Cloud is a way to essentially merge your Web Flow project with a with a Nex.js JS or an Astro project and and so there's a there's seamless experience between the web flow project and the Astro project and um with that enables you to build very sophisticated apps that are on one part managed by designers and marketers who typically don't have code. Mhm. You merge that with an experience that's handcoded by developers, very advanced.

You doing things that Web Flow could never do and the user will have that seamless experience. You know, they they could just log in and then they're part of the Web Flow cloud experience, but it just looks and feels completely seamless. So, that's the simplest way I've um yeah, I've uh just worked out how to describe it. But it's cool. It's cool. I have a course on it. Um I think I think there's definitely a use case for it and uh it just depends on the actual um yeah how widespread is that use case. Yeah. Cool. Cool. Uh GitHub report. Yeah, let's talk about that. So, so there is this new GitHub report that shows Typescript is now the number one language uh in the world, which is pretty interesting to see because we talked about Typescript uh earlier in the show being your favorite.

You you you said you like Typescript or something along those lines. I love Typescript. Yeah. So I saw this and I found that uh to be interesting. Somewhere here they they showed the amount of users using TypeScript versus it's a really long report by the way and I have not huge. Is this just on TypeScript or is this just GitHub? No, no, it's just like GitHub. So here this one I was looking at this one. So we see JavaScript was for a long time the number one programming language on GitHub and then Python was for some time. You see 2024 start of 2025 Python was number one and now Typescript out of nowhere is number one.

Python 2. It's not out of nowhere though is it? Yeah, it's not over. It's it's there's a very specific reason why and that's because of AI. And I was I was looking at the report and I was looking at some of the numbers GitHub are reporting on just their overall usage. And in the past year, they've had 36 million new users of GitHub. And in total, they've had 180. They have 180 million users in total. So that's to 20% have joined GitHub in the last year. And no doubt that's got something to do with Vibe coding. Mhm. Um and coding language of choice from vibe coders is TypeScript. Typescript. So yeah, definitely AI has something to blame for this, which is a good thing.

You know, Typescript is is was I used to you we used to kind of sigh when it was like, oh, got to write us in Typescript or people suggested or insisted we use TypeScript. There was a little bit of like h because there's so much upfront cost and and features just take double the amount of time because you got to type safe them and this and that. Whereas now, well, but the the and the gains that you got from that were only seen much later on in the in the project. Whereas now, because of AI, because it's all type safe Typescript first, one, the AI can work a ton better. And two, you you generally are just able to um work on your codebase a lot more efficiently because things are set up in a way that everything's type safe.

you know what types and and things the functions get passed when you create a React component. You know what props it needs all through the power of TypeScript. So yeah, very welcome. Yeah. And it makes it far more accessible. So yeah, pretty cool. Yeah, that's really cool. Year overyear growth you see over here. Cool. Uh builder.io or you want to go or you want to do your result. We got 10 minutes. No, you you can do builder.io. You understand that better maybe or I don't know. I can I can dodge it together. I can definitely take you through it. So, this was found by you obviously. Um do you want me to I'll I'll show on my screen review and I'll put it online.

Yeah. So, um and watch it. But it's essentially your classic shoot maybe stunning websites and applications. It's your classic vibe coding tool with anyone on your building websites building agent but the addition here any slack conversation or any and in fact it's more like and code agent which is this is interesting your existing code bases. It's the only agent that actually understands your design system and designs UIs and prototypes exactly like you do with deep design system intelligence using your components throughout. It's the only agent that lets you prototype on real products in real context with real data with your actual UIs you can get internal and customer feedback on easily. It's the only agent that lets anyone in your team send pull requests and automatically addresses build failures, PR comments, and any type of automated feedback.

It's the integrates with everything agent that you can connect every single product you use, data source, tech stack you have, and every team member in one place. It's the no handoff agent where instead of having three different tools going on developments, everyone works on one source of truth. It's the agent that learns how your team works, storing memories from feedback from product design and dev to make it better hone the way you build from end to end. It's the only agent that can visually build your native applications sooning iOS, Swift, Android, Flutter, etc. We've already So, I will just say this is where it lost me a little bit because there is a lot of promises going on here a lot.

This tool seems insane. Um, you know, learns how you code. By the way, it's not the first to do this. Like I say, it feels very much like Devon. It's like an integrated environment. Um, the nice nice aspect is this obviously here. This doesn't look real. This looks like After Effects, you know that you probably won't be able to see it. Lot of problems. But he says here builds apps storing memories from feedback from product design and dev to make it better hone the way you build from end to end. It's the only agent that can visually build your native applications. Soon supporting iOS, Swift, Android. The only app that can the only AI that can build your native apps soon to be Swift, Android, iOS.

like how is it how is it an app that can build? But then you're saying the only languages that is soon to be using those that that just lost me and that that then made me go back over it and then question kind of like well hang on this this there's so much going on here. Um if they can deliver on it amazing. I'm also a little bit disappointed that one of the um biggest features in my opinion is this design development integr this integration seemingly that you can have designers doing all this that and the other with a with a Figma-l like flow here you got developers vibing on the left so this is what we hope and dream of right this is everything we hope and dream so yeah seems amazing just a few little I don't know little little red flags on it for me, but it does look very very very cool.

Yeah. I I also looked at their website and I was a little bit confused. By the way, this is this is really cool this stuff. Yeah. So, this builder io and then you go to their platform, then you go to fusion like what what the he what is the difference? Is it is it is it the design aspect different or is this a strict vibe code experience but actually this is all that integration stuff with MCPs and you know linking with um pull requests and do is this more like Devon is this more like a is fusion more like Devon whereas builder.io is more like lovable I'm not too sure. So if you know, let us know.

But that's my overall understanding and assessment of this tool. Um, yeah, but apparently Yeah. Go on. Sorry. You go. When I saw the video, I was also really excited because it did look like that, you know, a mix of all of these tools. But then I was like, is he making a new tool? Is he like abandoning builder? But then I came to the website and I got more confused because it seems like it's not a new tool. It's like an addition to the already existing tool which which by the way before all of this AI if I'm not wrong it was a tool that could you could do like visually develop things with it similar to the way web flow is but more powerful on the development side of things but then it got all AI and now it has more AI on top of the already changed to you know have a lot of AI so it it kind of ex um yeah it doesn't make sense on that end.

And when talking about like having design versus dev not versus but like alongside dev mode to me it shouldn't be like because there was a time it felt like design mode is like web flow and dev mode is code but I feel like that is changing. Uh I think the design mode should be more like Figma like a free canvas that you can have versions that's the whole point of it to have uh variations to have exploration and then the dev mode could be similar to the way that web flow is you know the web flow designer with a view on the code so you could see the code and obviously AI to write the code and then maybe have something like the web flow cloud for you other developers uh you know having their needs taken care of with the cloud there.

So yeah, but this is I don't know this is cool. I would pro probably give it a try but the video showed it to be I don't I'm looking at this now and I don't think this is aimed at you know one person freelancers. This is not aimed at those people. This is this is enabling teams. You know what gave it away for me was this note at the top here that PMs can use AI to write tickets in Jira. Designers use AI to create mockups in Figma and developers use AI to build IDE. Three teams, three tools. So I think it's bringing teams together. I don't think you'll see Well, I mean I would have you signed up to this?

Have you you know have you No, not yet. But okay, fine. Also found it. But I don't think this will be this is this will be a watered down design experience. It won't be like an experience where you can iterate and have loads of different versions and this and that and the other. It will be a very locked down rigid design experience like you like you'd see in web flow just because that's the way that you know um if you were to take away a lot of aspects. How do I explain this? This is going to be really difficult to explain. Um, if you take a lot if you take away everything that doesn't work on the web inside of Figma, you'll end up with something very much like WordPress uh WordPress web flow.

Yeah. Um, you know, auto layout is what you need and and this and that. I suppose you've got the free flow canvas a little bit and this kind of thing, but like you'll never to be end up with a lot of basically what you're seeing here. for it to be useful, you'll you'll end up with what you're seeing on the right hand side here. So, I don't think it's necessarily going to help designers design. I think it's going to help designers who know how to build websites in something like Web Flow contribute to the development of the website, if that makes sense. I don't think designers will feel they're not going to be, like you just said, they're not going to be able to iterate.

They're not going to be able to free form canvas. So, not be able to explore and experiment with designs just because that's practically impossible in the context of a web tool. Um, it might as well just be a separate page. It might as well just be a separate tool. Um, so yeah, inevitably you're always going to end up with this very locked down and rigid design tool. Like I say, um, not it might be very useful for someone like yourself, but I think it's going to be it's aimed at kind of teams and and enterprise solutions, but regardless, it'll be cool to see where it goes. They're promising a lot. They're promising a lot. Uh, should we see what you've got then?

Should we Everyone's been hanging with baited breath to see. Are you even ready to see that? I don't know. Oh, are you too fast? Too fast. So, where did you get this? I'm not showing it. I'm not showing. Yeah, I will explain. All right. So, we we talked about Freepic having a nodebased video generation. So, here is a need and I signed up and I have a specific need. I need to animate this piece of whatever like thumbnail. This is a thumbnail. Uh I wanted to animate this manually but then I thought let's use AI to animate this. What what what do you see here? So this is made in Figma by the way. This is made uh manually in Figma.

How would you animate this in a animation software or even with code? This is after effects probably. Yeah. After Effects, I'd probably use GAP. Probably it would be easier for me. But you you'd have the 20 24 and then the four goes out and it gets, you know, slow smaller and the five comes up with a nice easing and something like this, right? And then maybe the five goes and the six comes. And this is what exactly I prompted the AI. And first I tried to prompt it directly using uh just just using my own imagination and my own words. Are you ready to see the result? I want to see the prompts. I want to see what I want to see the word before the result.

I Well, I can't do this. Well, I'm sorry, David. I can't do this. I can't zoom in more than this um without revealing. So, I mean, I've seen the design. So, like, you know. Yeah. It's like create a smooth number animation. Okay, the 20 24 going to 20 25 and 6 and 7. Uh only change the last digit and the rest stays the same. Okay. Okay. Yeah. Okay. Yeah. So the space time continuum is um broken. the flux capacitor is is not quite ticky. Yeah. So, I was like, "Okay, uh let's try it again." Okay. So, this is another try. And it's just like, "What? What the hell is this?" I was like, "Okay, let's ask Chat GPT to write the prompt like make a really long prompt."

And not much better. Like, okay, like the the two I've already got I've already got a hunch. I've already got Is this a use case or is this you just saying how rubbish the tool is? No, this is Well, the thing is I know like with so many other things these tools can do it but this is the use case that I write right now have and I just paid eight bucks for this. Yeah, here's another version. And different prompts. I figured it out. I figured it out. This is the use case for this tool. Okay. Create your first your first image will be the number five, right? Just the number five. And your first prompt will be have this have this number shoot off and and a gray number four or something like that.

Have and get that locked down. Get that first animation locked down just in isolation. Your second image because this is the beauty of nodebased piping, right? Your second will be just a flat image of the 2025 or whatever. Then you would say merge those two. Merge the video and the image together. into a video. That's I think there's there's where you got to think of when you talk about node, you've got to break down the the path in in logical chunks. What you reckon? But if I do this as well as I would just do it in Figma prototype that I never touch or I would do it in code, you know, you know, the the whole beauty of this would be to oneshot something like this because it does seem simple and I'm pretty sure there are enough examples of this exact ticker animation.

Yeah, this this is nothing like novel, new or groundbreaking, but maybe because of how minimal it is, it is even more difficult because of the the minimalist tech. But you you're approaching this in the same way that you approach the web flow maybe box central, right? You're you're you're asking something that can be done very simply or whatever. Well, you got to think about is the scalability of something like this. Is this something you do time and time again? Is this a long workflow that yes, it might take you long to set up in the first instance, but does it pay dividends in the in the future because you're just doing the same thing over and over again?

Are you creating 60 videos out of one image? In that case, something like this would be really helpful. So, whilst I I I know you're only showing like an example, and yes, you're right. It's not a valid um representation of why this tool sucks or whether it's brilliant because the you you're not wrong in what you're saying. It's just this that was never meant you're comparing apples to oranges if that makes sense. Yeah. Um, I see this as as mass um mass output variances uh constantly doing the same thing or constantly doing multi-step processes and just chaining it all together and that's where I see this coming in clutch. Yeah. saying that I'm not clever enough to even think of a use case, but I can just I I just No, I can't imagine like ads uh or have like let's say what yeah product shots like we we looked at the finite stuff here.

What if we could like create layers and 3D and make them look like they were rendered in 3D or I mean with UI it might be more difficult but like with actual physical products That could be something. Um, nonetheless, this was my use case right now. And I have far more complex UI to animate uh for thumbnails. And I thought this one because this is a custom code, let's not animate it in in the UI, just let's do it in After Effects. It's faster uh for my team to do it. But apparently, yeah, we'll go that route because AI can't do it properly. And by the way, I noticed for all of them I was using the auto model and I don't know which model that was.

Maybe this was my mistake. Maybe I'm I suck at prompting. But any of these all of these prompts were different. Yeah. Um do they're VO because I think VO is is a good place to start with again. Yeah. I can't animating something that you don't want to affect the rest of the image with. So that might have been a good to have. Yeah, I'm out of I'm okay. So, this one already Yeah. In the direction. Yeah. But here's the issue. It's not like when if I write code, by the way, shout out. We We are over 100 um viewers. If it's with code, it's so much easier because of the deterministic uh nature of it.

like you see how the code runs and if you have a bug you can change the bug and you know obviously make it work but when when we are talking about generative generative nature of things you can't control it it's generally you could tell it out code that does this though or you could tell it to output like I don't know whether it can but will output a uh after effects file that does this though there there probably ways to do it but you're totally right? Like it's a black box that you can't Have you seen this videos on Instagram where a director is directing um AI actors and they'll say a script and they'll you know there'll be a husband, a wife and two kids and then they got to read a script and he'll be like um so lower energy.

Um and then one of the one of the kids like morphed into a woman or the mom and it's like um hold back on the morphing. Don't do the morphing. Less energy and um just just read it as if like there's no you know just read it as if you're just talking to each other. You don't need to shout and then he'll play it again and it will just be like totally different again. It's like um so yeah just don't do that. I'll I'll send I'll send it to you and please send it. That that is good. And this is how I feel like with this image and like visual generators because with code it's still kind of like more stable.

It's still not very stable, you know, but it is a bit more stable. I'll vibe code this, I don't know, in a few minutes or if I manually build this in web flow. Even with the old interaction, I could do this probably in 10 minutes. Uh, but I don't know. I spent €8 and 20 minutes just to do this, which is I don't know everything for content. So, this is the cost of our business. $8. Yeah. Yeah. This is for content. Very good. All right. Well, that's all. Yeah. Go on. I I wanted to say I spent eight on this so you don't have to. There you go. There you go. Yeah. Well, that will be title.

That will be the title of the clip. That'll be the title of the clip. Yeah. Cool. Well, that's all we got time for. Make sure to follow us on Instagram and YouTube, Commandai Show. Like, subscribe, and all of that good stuff. And join us next week for all the news in AI design and dev. I've been Samuel and I've been Kavaro. Keep on vibing, baby. Peace out.