#37 Opus 4.6, GPT 5.3, Moltbook Crashout, Codex, Webflow AI Claude ads, Figma Vector

But it will come on soon enough. Don't know why it's not. I know. Sometimes I has it that I need to take my battery out and put it back in. I think that's what's going to have to be done here. Unfortunately, the old the old off and on again. Oh, we're live. Hello everybody. How are you doing, beautiful people of the internet? Oh, hang on. All right, here it comes. Whoa, that is some So, we are deep live, but we don't see you yet. People are starting to join. People are starting to worry. All right, let's do this. Let's do this. Uh, let's take that down. Let's take that down. Right, I'm back. Back in the room. Hello everyone. Hello there. Hello there. Hello there. Water. You're Oh, you're drinking [ __ ] water. Yeah, of course I am. This is my go-to. Nice. So people are starting to join. Um wherever you are joining us from please consider joining from YouTube. Uh we can you can comment, we'll interact with you, we'll talk, we'll chat, hang our hung around. So or give us a subscribe. Our thumbnail isn't up on YouTube, so I'll quickly do that. What a week. What a week it's been. I literally landed I went I went to Well, I'm talking about for me. This is This is personal time. Ah, personal. Okay. This is personal time. Tell me. Yeah. So I went I mean not that personal but like I went to I went to Sophia this week in Bulgaria as a as a quasi business trip searching for opportunities um striking deals you know all that kind of stuffing um closing deals uh riding the waves buy sell pointing my finger all this sort of stuff, you know, being being the big boss man. 6th of February, six two um and yeah, it was mixed in with a lot of drinking, a lot of eating, a lot of meetings and uh yeah, it was it was a good good fun trip. And I literally landed late last night. And uh luckily I stayed the two hours ahead of me, the one hour ahead of you, Kabaza. Mh. But I stayed on UK time. I tried to wake up really late because I knew I wasn't going to be there. I was only there for 5 days. So it has been actually quite seamless kind of getting back and you know getting back into the swing of things. So yeah. Cool. kind of trying to keep track of the AI news during the week has been pretty hard. Yeah, it's been a crazy week. It has indeed. It has indeed. Haven't got to use and do the code stuff. I mean, most of my most of my time as as people know that maybe follow me. I've been just doing the maltbox maltbot stuff. And do you know what's really funny? I didn't like the name Malt Book, but it seems to be the name that I keep accidentally saying, which is kind of funny because I just Yeah, I don't know why. It's kind of funny. Uh, but anyway, doing a lot of the claw see open claw stuff over the last um just before I left and then So, is your agent joining the the antisocial movement on Maltbook? Well, inspiring against you. Yeah. No, I mean like we'll cover malt book later on uh in today. We'll talk about it because it basically you know crashed and burned and um it was something it's something we suspected when we first started playing around with it because anyone you don't need any credential like you just need to verify and then you can just post like we didn't know what the difference between human and AI. Um, we were like, "Well, how do you know the difference?" And you can't. But all of the religion stuff, all of the language stuff was probably just trolls just Yeah. posting to to Maltbook. So, I was like And then everything got really slow and it wasn't able to like connect and do anything. So, I just I just removed it. I just removed it. Okay. Well, yeah, that's what people do. uh how would you authenticate people and you know distinguish it from bots and what if I what if I use that computer to authenticate with my face or whatever and then I just give my authentification to to my app and he you know my club bot starts are you talking about mockbook specifically or just in general? Yeah. Yeah. You don't need to do any sort of crazy authentication. Yeah. But how would you add it? Yeah, that's what I How would I add it? Oh, I don't know. I don't know. Using Sam Altman's uh I something. What was it? That scam. Yeah. I don't know. Um World Coin. Was it World Coin? Something like this. I don't know. Yeah. I mean really there's no way cuz we spoke about like you know agents controlling the browser and knowing the difference between a human and realistically there's not a lot there's there's not a lot you can do to tell the difference between a human and an AI. Um you know especially with interacting with things and APIs and stuff like that there's nothing yeah the the agent tries to act like a human. So there's really not a lot you can do to separate the two. You can try and prove that you are a human, but you couldn't prove that you are a human um imitating a bot because there's no way to differentiate the bot um to say it's just a bot sort of thing. But yeah, um so yeah, it's been um yeah, I delete deleted malt book because it just was not working and uh I didn't really believe it from the get-go. Uh but yeah, it's funny. It's funny. Well, yeah, we'll get into it. We'll get into it. I won't go to you. We'll get into it. Um, but yeah, what's your what's been going on your end? Um, burning through my tokens. Mhm. Tons of them. I've been having so much fun just building and building and building. Um, I'm I'm rethinking a lot of things. I'm like like well I I guess I need to like think bigger. I can. It's the I know it's probably the Dunning Krueger effect that I briefly mentioned, but I feel like I can do whatever. And I'm probably at the top of the what's it called? Like the stupid like the It has a name. Uh mount stupid. I'm at top of Mount Stupid right now. Well, bear in mind you you you are still just creating app. You're not creating a SAS app. You're not creating something that tons of like you will sell and tons of users. You're certainly getting there. You know, you're you're learning and whatever, but there's a big difference between creating an app that you use internally, doesn't need authentication, doesn't need this, you know, the the security criteria is much less when you're building apps for yourself. And you know what, go nuts. But yeah, um yeah, don't don't be fooled into thinking that you can build production ready SAS applications for clients at the level you are now. Case in point is, and we'll go into it, is the guy who invented Maltbook. You know, the the the last tweet he said, it's like something along the lines of I built Maltbook without reading a single line of code. I had a dream of an architectural architectural vision and I just prompted my way to success. This is the future. He's del he deluded himself into thinking he can prod can release something production ready. I would say uh Claudebot or now OpenClaw was by far and away more secure and better built than his pile of trash website that he built. you know, he he thought that because he could see something that he had implemented all of the right ways to do things, but he did not at all. I don't know. I I feel like along the same way kind of like I can build it if if it doesn't show if I don't have a bug, there won't be any bug. There won't be any issue. Well, I I know I know it's not going to to scale that way, but still like in the past if you would inspire people or if you had like ideas and you were a designer, you could just design stuff, explain, but you couldn't build it. You had to build a team around it and in order to build it. But I feel like not that I don't need a team, but I feel like with a smaller team, I can do so much more. And it's in short, it's making me rethink the way I was thinking about what to build. I'm to be honest, I feel a bit lost to be very honest. I feel like if I can build anything, well, not any, but you know what I mean. If I can build anything pretty much, what should I build? It's the horizon has expanded so much since I've been using clot code and just getting myself in in the past three weeks. I'm like, yeah, I need to I need to think bigger because the cocktail app with Replet, you know, I made it with two hours and I'm like, what now? It's done. Yeah. What am I doing? It's weird because you know this this trip this business trip I went on I went with an old friend and we had lots of you know life existential um to conversations and so I've been thinking a lot about you know it's easy to be like overwhelmed with everything but sometimes it's just about bringing it back down to basics simplifying everything not getting caught up in the chaos because realistically the answer to your question is often very very simple. Yet we're so uh distracted by everything around us that we we just end up in a like a what's it called? Tidal not tidal wave. What's that spinning thing of air? Twister like a um man, why can't I remember what those things are? I don't know. I know they called I know the film's called Twister and I know they're like a triangle. Not not a hurricane. Tornado. Tornado. Tornado. Ah, okay. tornado feels like you're just caught up in this tornado where there's just so much stuff going on and you're being distracted and whatever but it's just a case of just take a moment to meditate like you know I feel like you know the the shiny object syndrome I feel like that every single day with especially like this week this week has been like I've been on this shiny syndrome shiny object syndrome for the whole week and I've been crying all of these things. All of them. But none of them probably deep enough. But yeah, well, here's the thing, and we've spoken about this to people on the show and and you know, all that lot. Just have fun, you know, get yourself excited and whatever. And then and then just recognize that, okay, maybe maybe you're maybe this is just a fun little period of your life where you're building your tools, but actually it starts to get boring and you just go back to design and you go back to doing the design stuff or or or whatever it is. Um, but just have fun and then if you find that you know what, I want to start doing this for clients, then deal with that when it comes to it. And then maybe what do you mean? Uh, I'm doing like I I started offering clients. Uh, mate, you you don't know like I'm I'm saying this is a friend now. You don't know what the hell you're doing. So So uh there is this client they want um they wanted to use Sketchvab for AR hosting and it cost like 80 80 bucks a month and I was like why not offer them like our own solution? We just build it. So, you don't know what you're doing though, mate? No, I don't know. I I have Do you know anything? Like, would you know anything about uh web Jason? Like, no. I I'm doing research like I'm I'm research mock book is a result of someone who thinks they know what they're doing, but they don't. We Okay, so I know that we don't need scale, right? I need I know that we we will have like less than a thousand users I don't know like per month maybe 5,000 maximum per month I know that we won't have that type of scale issue that type for security I will ask I could and potentially you you know so I will I will I I will figure it out I'm not saying that hey I'll just prompt it with like a single prompt and it will you know, happen. But um you do not have my blessings in this. You do not have my blessings. Again, I told you I said I'm at the top of this. I'm literally stupid. You are. You are. I'm literally offering like applications that I will build for my clients and it will be fine. It will be it will be all fine. You know why? Anyway, why? Because you're a man stupid. No, I'm not even reading the code or the plan. Well, I'm since our lives, I've been starting to read actually more of the plan and being critical and I'm telling GPT 5.3 to criticize 4.6 Opus and vice versa and trying to make them Yeah. come up with better and better plans before we go forward. I said this to my friend. Um, if you don't know how to draw and you ask AI to draw you a picture, you'll probably think it looks faking great. Let's let's bring it into let's bring it into our our world. You probably ask AI to design your website. If you if you're not a web designer, you probably think it's great. I think it's great, right? I'll think, "Oh, that's really good. That that does the job." But a real quote unquote real designer will come in and be like, "No, that's utter trash." And that's the same thing with Yeah, I know. same thing with the code. It's like you Yeah, you it it deludes you into thinking that what you've done is actually correct or right, but you don't know what you're looking at in order to spot the signs of like this is actually terrible, you know? Yeah. Um, I don't want to trash Theo, but I watched his video showing how how much better uh Cloud Code is with with that skill at front-end design and like marketing websites. And he shows some stuff on his screen and he's like, "Beautiful." I'm like, "Oh, no." Yeah, I I get why in the first glance the website looks beautiful, but no, it it has so many issues like design-wise. It's just not there. Uh so yeah, I I get what you say. Uh like it might produce something that looks good, but it's probably not as good as seems like. But yeah, I'm I'm learning. I I and I don't think I'm trying to I know that I'm not trying to build enterprise apps. I'm not trying to build the next web flow. I'm trying to build like I would say even like mini SAS like micro SAS is what I'm building now and multiple small tools. Good luck. I will show I will show some I will show some. I don't thing is I don't care about showing like I don't care about the looks. We're not really talking about and and we're not really talking about it functioning. We're talking about whether it is a poses a a major security flaw, you know, or whether it's leaking um security. Well, yeah, security is spans across everything. Whether you're leaking data, whether you're storing um you know data in the correct way, whether you're enabling one user to edit the data of another user, whether you um whether your server can even handle uh rate limiting as an example. Have you heard about rate limiting? You know, so it's like um yeah, it's it's dangerous. Someone's just asked why am I so cynical? It I don't think it's cynicism. I think it's like, you know, you you you learn it's like saying, "Oh, I'm going to go to uni to study um I don't [ __ ] know like maths or whatever it is." And then you think you're a mathematician after day one. It's like you've got a whole two years to like now learn like it's just it's a case of getting better. You you get better. you know, you you know so little at this point just because you you can't just charge forward with confidence like well you can don't get me wrong what you build and what what you create it it might look like it works but then it only takes like you you've just got to look back at maltbook as the most recent example um clawbot in some ways there is literally evidence right in front of your face of people who charge in thinking they know what they're doing like the cynicism isn't my belief in Kabaza like I think you can get there for sure but like just be selling this to clients is where I think the buck should stop like work on your own stuff learn about it in your own time don't stop doing what you're doing or any this is where I would disagree a little bit I mean I started even with design with photography that I did I started offering services that I have not before always this was like there is always like the first app that you build and of course like there there's a level that you build for yourself and then you feel like okay I can up the you know the difficulty a little bit and sell it and get paid to learn. So once I the the way I see it, if my client pays me 2K for an app uh to build probably like an agency would cost like 20 times more or 100 times more I don't know I I will use this money to learn and to make sure that you know the things that you are talking about are good good enough correct. So I'm I'm taking this money to um to learn And the same with design. There's something very different in um you know being given some money to take some photos and then the photos not really coming out and it's like whoop like no it can be like really bad. Imagine you do like wedding photography. Yeah. But you can't you not Oh no, of course. Right. And that's the that's the kind of first example that came to her head. Yeah. That would be absolutely stupid. Would you put yourself forward for wedding photography just because you know how to click a button on a camera? Not exactly, but I've done like wedding photography and sure before doing it, I've done photo, but like you still like when you go into this new level, you up the difficulty. There are unknown territories. The same with web design. You could theoretically say, yeah, you get the design wrong. What happens? Well, if you get a button wrong on a website, you might tank the conversion of the client. You know, there are so many things that can go wrong. Yes, being hacked is probably one of the worst, right? Maybe it is the worst, but I'm just saying that I'm this is how I do things. I get myself into this difficult waters and I'll figure out how to swim. It's just it's just at the cost of probably like it could be financial cost to the client. It could be financial. I don't know what this app's going to be doing, but like you could leak details. You could you could um GDPR is obviously a massive thing as well. Like if if something someone finds out or you know whatever, like it it's not just a again, it's not just like, oh, but like like whoopsie like I just I made a mistake and I will fix it or oopsie. It's not just like it's not something you it could be like yeah my master user uh was compromised and they changed the passwords of everyone stealing all of their information and their financial detail. Like the consequences of failure are exponentially higher than a few bad photos at a wedding or a bad button causing conversion. Yeah. But yeah, I'm I'm not getting myself into this situ. I think this is the difference. I'm not. You are though cuz you're you're building an app for a client and getting paid but it doesn't have like this type of like security vulnerabilities that you are talking about like we're just hosting some 3D assets and instead of using a tool that cost 80 bucks a month I'm just going to I don't know maybe charge them 20 bucks a month for my app. But you're probably going to be using ex you're like I'm going to use yeah hosting on AWS and uh using some open source AR stuff. I'm still figuring it out. But my point is I'm getting paid to figure out those things that you are mentioning instead of just doing it on my own. I'm getting paid to figure them out and I will do not condone this one bit. You should never learn on the client's dine. You should be you should be able to do these things before you do it. Maybe little bits of knowledge here or there. You learn something new in the process of doing the thing that you already know. But to be whatever anyway. Yeah. Um let's get on with the intro then I think. Yeah. Okay. Let's go with the intro. We have comments. No, no, it's mentioned. Okay. Okay. Well, was a long a long chat. We have close to 40 people. Uh, okay. So, this week we got two new super powerful models, Opus 4.6 and GPT 5.3 codecs. Uh, Web Flow surprised us with their new AI builder. I'm so excited to show you what they have. Uh, Figma added an AI image to vector editor that is really cool. And we have Codeex app which we will compare to Claude Code desktop app. all of this and a few more news on this week's um episode of This is Command AI. I'm Kavarza and I am Sam and this is a week in AI design and dev. Cool. So, we have tons of stuff we should go through. Yeah. Yeah. So basically the coding wars is heating up as two the big two released their final frontier the frontier models just 27 minutes apart opus 4.6 and 5 GPT 5.3 codeex. So we'll go from I think that the fact that they are 26 minutes apart or 25 minutes apart is just insane. And um do I have the I have the everything's freaking crashed. Uh 6:12 p.m. was codeex. This is really annoying. And where's Claude? 5:43 I think Claude was was released on. Um, but if we go through, so are you telling me that chat are you telling me that OpenAI is probably a bit mad at Anthropic and they're trying to to show they have they have something to release? May we'll talk about that. I don't know. I don't know. It's it's might just be a massive coincidence. I I I maybe wouldn't read into it too much, but um essentially Opus 4.6 six imp improves on coding skills. I mean if you don't know what opus is anthropic have claude which is their model and they've got three levels. Opus is the big boy you know costs the most most uh capable sonnet is sort of in between and then haiku is like the fast snappy chat model that they have. Um and yeah we were on opus 4.5 we've now increased just a just a digit. It's interesting as well that chat GPT increased just a digit as well. So I don't know maybe they didn't know not too sure. It plans more carefully sustains a genensic task for longer and can operate more reliably in larger code bases. Has better code review and debugging skills to catch its own mistakes and in a first for our open uh class models. Open 4.6 features a 1 million context window and that's in beta. So very very uh capable uh like smashing all the benchmarks as we come to know at this point. I mean you know I mean a gent pretty good increase there. And if we look at the numbers we have Opus released before um Sonnet and before Haiku which is normally not the case. normally haiku uh we have haiku sonnet opus because opus is the the best and the beefiest one. Yeah indeed. Um wanted to bring this as well. introduces agent teams or I think this is more of a claude code uh feature but agent teams being able to spin up um you know asking to build a feature and having a specialist agent on your front end specialist agent on your back end specialist agent on your UX and then they can combine to uh and and communicate with each other to work on the task at hand which is pretty pretty cool. Um, I was going to show quickly how to how to do that because there is a bit of a there is a bit of a knack to it. That that's actually interesting. I want to know how to do that because I started um to plan an app. I have a document for it. And I started tingling with the idea of what if I task list that I have. What if I could name the tasks um and apply them and assign them to different agents like agent one, two, three, four, five and then have like a way of running like five different agents and each would go and pick their own tasks and do it. Yeah. Yeah. So, are you telling me this is going to do this presumably automatically? I mean, you have to I mean, it needs to be it's not just going to do it automatically. I think you you need to allude in the prompt to say that you want multiple agents. Do you know what I mean? Like, so the first thing you do in your in your settings JSON, which you can you can get in your root the root of your um file uh user. Um where are we at here? here claude settings JSON you got to turn claw code experimental agent to true or or one and then when you spin up claw code nothing will change you just got to again oh no look what I've done now um if I site 05 go into my thing here um again it's in your prompt that you say you know um Yeah, build this X feature. Make sure to make sure to use the separate agents and whatever. So, it's not just going to pick that up. Um, or at least it'll be more likely that you want to point it to do that rather than it to, you know, try and do it automatically. I think it's better that you are more explicit with it. So, that's kind of a cool feature um of it. They also have made some here you go. This is relevant to you. They've also made some cyber security improvements uh to both of the both of the models as well which is really really cool. Um and yeah, I think this is bigger and better which is why I wanted to kind of combine the two uh the the release of GPT codeex and Opus in that we there's not much more to say. They're just they're just better. Do you know what I mean? Um so if I go to comet tab blah blah blah um GBT5.3 codeex in fact we'll go to this one ultimately the same thing um you the the only thing to note is again this was released 25 minutes after op O opex Jesus Christ opus um but it it's scored significantly higher on the uh the terminal bench. Other than that, I think we need to look at artificial analysis which haven't made that haven't it's not as up to date as is comparing them all yet. Again, it's just been released, you know. So, they they've got not got a good comparison between Opus uh 5. Oh my god, Jesus Christ. There's so many numbers right now. 4 Opus 4.6 and Codeex uh 5.3. No one has the comparison between the two yet because they just got released so close apart. I think when was it yesterday they got released or last night even? Last night. Yeah. So the way I understand it is just just use the the newest model. If you're a developer, you'll just look at the numbers and just use the the highest number. Especially now with Opus being having the higher number. It just makes it much easier in a way. See what is the best. That's the best. just use it. And and it's such a good such a good point as well because we'll get into codeex and the new tool that there was released alongside all this. You know, I just prefer clawed models. Like I'm I'm interested to, you know, going back to shiny object syndrome, I'm interested to see what all these latest models are doing, but I think by now most people have just sort of found their the model that they prefer, right? Um, yeah, if unless you're encountering major issues or unless you're in a, you know, it's crazy that this time maybe maybe not this time last year anymore, but like just over a year ago, we had we were handwriting all of our code. Now we're we're just really being pedantic about minor things that an agent is just writing so much of our code. So, anyone not really interested in discovering like the latest and greatest and and splitting hairs on all these different models, you can't really go wrong with all these things. I'm not about to tell you that Codeex is better than um Opus or Claude even. So, yeah, in in an attempt to to not make people sort of too overwhelmed, they're both great models. Yeah, you know, um, and it's good to know that there is improvement. So, pick your battle. If you're already paying for GPT, chat GPT plus, keep using codeex, it's great. If you're losing chord, use cord uh cord code, then use cord, you know. Um, yeah. So, that's it. I mean, that's that's the news. Newer models, new some newer things that again the there's a steer you can steer. Oh, what you doing? There's a you can steer the models um in GPT codeex the actual app which we'll go over you can actually steer it like midway through it's constantly checking back with you and checking to see if there's any additional um information added. You can kind of do it with clawed code. I have done it um you know during a little break or whatever when I've cued a prompt. It's kind of taken that into consideration. I've not needed anything any more um more responsive than that. But we'll see in the codeex app how responsive that is. Maybe you can show us a little um interrupt a prompt or something. So yeah, pretty always welcome to see the new new um coding models and there are some significant updates here, but use use whichever one's your favorite. Um I was going to talk about Maltberg, but do you want to talk about codeex just because we're already we're chatting about it. It seems like a more of a seamless transition. Yeah. Uh, let me get my stuff ready. Just a quick thing. Okay, I'll I'll show that in a second. So, where was it? Ah, now I found it. Open AI is showing an entire new way of building apps with AI. It has a UI and I'm loving it, but I'm not using it as much as I really wanted to. So, here we have Codeex app. The naming is probably confusing. You've heard Codeex, but this is the codeex desktop app and it is a whole new way of building with AI. So you have these folders that are your projects, different projects and within each project you have different chats and you can entirely use this new UI to build without even looking at your code. But if you want to look at your code, you actually can. You have this view to look at your structure and the diff and it's not a full-on it's not a code editor. It's just you can't edit that code. Yeah, I think you can. No, I can't. What's that? What's that plus on the left? Is that just to reference the line in the Oh, request change. So, it's kind of like a comment to change something. So, like the like the inline changing in cursor, I guess. But you can also open with cursor. You can build it. You can commit from here. That I really like. It's quite simple to commit directly from here. Um you have the terminal. I'm not using it really, but you could. And and this is like my new favorite way of doing like this this agentic I guess vibe coding in a way. Um and like this is this is my project. I was doing it with uh cloud code but now I just changed it to codeex and continued. And this is also kind of like interesting how you can just easily have your project and use whatever model whatever tool you want to to working uh on it. And yeah that that's the app. It's really nice but it's open AI. So if you don't like OpenAI models, you're you you can't use it. But we can compare that to actually let me get my uh clawed desktop code. So this is the Clawed app, the desktop app, and this is the code tab. And this is something that I didn't use before. I haven't like in the past. I I thought I just never thought about it, but it's essentially the same thing. So, you have this code tab here. You can have different chats and you see this is a different project. You see it here. This is also a different project. And here you just use claude. In this case, I'm planning something. Uh I'm planning something. And yeah, it's it has a UI. It's very similar to having the to use the the the terminal app, but with a UI. And I would say it's it's a bit nicer and it's nicer to also manage different apps. But what um codeex app got really nice is having like these different threads that are like kind of like uh organized by the project. And I really like that that is different project different thread easier to to easier to do multitask multitask on multi projects. It just makes it much nicer. Uh it has also like just to uh continue on the codeex app side of things you have the the skills that you can install. You can just add with a button. Now I have Cloudflare deploy. Um it has automations which by the way is very similar to Claude Co-work in a way to me. I haven't used it yet but it seems like these are things that I could do with it as well. similar, not exactly the same to cloth co-work. Um, I will try this in the future, but I'm really liking this new UI. Um, and I if I have to look at the app, the the code, I can, but I don't have to. And I really like that. And I just actually You want to say something? Just saying. I don't think you you can't look at the code in clawed app, can you? Um so to if you want to look at the code in so if you want to look at the code at claude app no I guess you can't from here I don't see a button to do that here I can copy the the oh no you can actually you can not the app but you you open it in your code editor and you have the terminal um and here I I didn't approve the plan, but it did something. Anyway, this is what I'm going to do next. I'm going to read the I I haven't I haven't approved. Did I click approve? I don't know. Anyway, so this is the claude code. No, the claw desktop app code. The naming is Oh, man. Okay. I I guess codeex app is even like in a way better than claude desktop app code that that's how how I'm calling it. It's the CL desktop app was just in the code tab I guess but it's not any Yeah. So I used uh the codeex app. I used the GPT 5.3 codeex to add something to my chat app. So I'm building this chat app alongside with some YouTube assistant stuff. And the chat app, I turned it into a fullon like chat experience and I I love it. Uh and I just asked it actually let me uh reload so it shows up correctly. Yeah. So it's the uh welcome screen. Uh I just told it to add these two features. So I have a note feature. I can uh append a note kind of like from my canvas to the chat or draw something. I can like draw a flower or whatever and I can append it here and send it to the AI chat because sometimes I I need to doodle something and I say can you make a JSON prompt for this for my journey or whatever, right? And it's so helpful to have this and I just vibed it and I just told Codex to to add these features to this welcome screen as well and it did and I tested and it works and it's Yeah. Weird one. Where are you running the app from? Are you just This is Yeah, this is the So you still have the browser? Yes, this is the browser inside of cursor. The reason I do this is because I have my uh my plans like at at hand like very easily I can just access my files I guess if I need to. Um I don't use the cursor agents anymore. I think I'm going to uh cancel cursor altogether as well. probably what I started to to play around with is this select element because I had some places where where I wanted to prompt the UI to be more consistent. For example, here you see I have this tab system. Here I have it too like this kind of like tab system. Again, it exists here. I had three different styles and the same for these like dropdowns. I had different styles everywhere. This one was different because I was building in pieces. So I I would use the cursor um inspect element to to get these and then prompt them to whatever agent that I'm using to fix them. Yeah. So that's kind of like that has been my way of doing things. And if I would wish for something, I wish there was a tool like this that would separate the projects like this, but I would also want to have it like with alongside with the code editor. Maybe maybe a cursor with this view would be cool. So cursor kind of like the agent view but as powerful as you know the the claw code something like this. That's kind of what I'm wishing a a mix of all of these. So I can open the project within the same window and in a different tab I can see my code in a different tab I see my agent and on the left side I see my projects and I don't have to deal with 20 different cursor windows because right now I kid you not I I can count probably like five different cursor projects open which is which makes the window management a bit Yeah, busy. That's why I really like this. But it's just codeex. Yeah, maybe make a feature suggestion or you know, I mean this is V1, right? So, who knows what's to come. Yeah, but I think they are on the right track with this whole like this is this is this clicked. That's what I wanted to say. this clicked and I think Open AAI even though I don't really love their models or OpenAI itself, I think they are on the right track with this UI. Yeah, it's just much nicer than the the terminal. Even though the terminal looks like more professional, whatever, this is nicer. Yeah. No, it's cool. Pretty cool. I uh I don't know whether I'll you I'll try and use it just to kind of keep a breast of what's going on. But again like I'm just clawed models for me work really well. I like CLI. Um I like the focus of the CLI. I like reduced you know everything reduced. Um yeah I I begin to also like it but I mean this is not far from it. And I noticed you you mentioned people choose one of these models and tools and they go with it. They get used to it. And I I I started thinking like why am I liking uh clot code more even in the CLI and uh I'm also using warp um as you uh suggested. I just I really like that the the responses are so concise and so on point. So it's kind of like it talks to me and it it really connects and it it is much easier for me to understand what it's what it wants from me and how I can just go next and I I continue building. Well, we'll go over this we'll go over this on our next stream, right? But I think there's a lot to be said about tuning your experience, right? And not just accepting what it gets out the box. If you feel like chat GBT is for boats, right? You've obviously got different, you know, you've got extra high, medium, and low, stuff like that. You've got that as a first point. Second point is the agents MD file that we speak about, which is, you know, again, making a rule to say make your responses concise. Um, maybe it is I don't know. But you you you should feel like you you are able to tailor your experience. If you prefer one thing out the other, add that to your agents MD file, you know. Yeah. Don't don't don't just be okay with what comes out of the box because that's step one in a genic brow uh aentic engineering if you want to call it that. Um step two is really getting to know the to the tools. It's like a cat that's just crashed. Yeah. No, it's like the cat um feeder auto automatic feeder. Yeah. It just sound like just something crashed in the background. No, it's like his um food dropping. Yeah. Um yeah. So, you have many many skills, agents, MD, um ways to to help you get your work done. Um and just using out of the box is just step one. Yeah, I know it is kind of being lazy, but at the same time it's to be honest a bit overwhelming to do all of these settings and then imagine I'm like I genuinely started building um a plan with different agents and you know assigning task to to different ones and trying to understand how can I have different sessions like different complete sessions of clot code attacking one single plan file. And then in my head I was like thinking they could go in and I could name them. I I could say you're agent one, agent two, agent three, agent four and go attack this and I could say there is another agent look at them. I I know that they are building these things internally. So I don't know like should I you know go all this way and try this new way of doing things while they're doing it internally. So it's kind of this balance and a lot of times you end up just using the vanilla settings whatever out of the box. It's definitely a good starting point and I think mostly the vanilla experience of claw code was better for me and all I've done is over time and again why we're probably going to look into it in the next claw code session which by the way we've been streaming my imparting my experience and my um knowledge on claw of claw code to kabaza while he builds an app um while I wanted to introduce it over time so you know we'll talk about it But the difficulty that we encountered doing things in the last episode, I'm hoping to then alleviate with some introduction of that, you know, so we could see the pain, feel the pain of what what it was that then we can okay, this is how we're going to solve it, you know, and that's kind of been the the feel the the the vibe and the um the uh just my teaching method throughout this whole thing. Let's let's get it up and running. here is snag. Really hit that snag and then be like, "Okay, here's the here's the remedy to that sort of thing." So anyway, yeah. Um, all that to be said, um, yeah, again, use use whatever whatever suits you. Find what what works for you. And there's no harm in sticking with it. You know, if you're just a casual user, we are nerds that like to look for new tools and explore these things. It doesn't mean you should, too. Yeah. By the way, I didn't get to read the plan and Claude started coding the plan. So, does it have a plan mode? Does it have a plan mode? Uh, yes, it does. Uh, so let me show it again. Here there is this plan. By the way, it says that it's doing the plan. And here we have the ask code and plan mode. Yeah. So there's an interrupt kind of feature to this as well where you can stay or not interrupt but steer it. Do you know what I mean? Put a prompt in that it's seemingly from my understanding it's kind of check in a bit more frequently than say claw code would. There's a kind of call and response type of heartbeat to it or something. I don't know how you'd So, so what what I noticed is quite interesting is like right now you see um like what is happening now like it seems like I can start chatting. Mhm. But it's going to continue working again. See without me doing anything it continues working and now it again it continues working. So it has this plan and it's going through each of the phases. So even it even though it looks like it has stopped, it kind of like picks up and starts again and continues. So let's see how long it can work uninterrupted. That's that's the problem with uh chat GPT models is they do tend to talk and think and whatever. It's it isn't as concise. It does take sign significantly longer which might work better for you if you like to again maybe I can imagine working on multiple things at any one time. It gives you the opportunity to do that because it takes so long to do those things you know. Yeah. By the way like um talking about like continuous work there is a way to do things send the task to claude uh to cloud not cloud on codeex app cloud. So you can Yeah, you can do it locally or on the cloud. Oh yes. What a time to be alive. Yeah. Yeah. Yeah. Yeah. Because you can do that on Claude code as well. If you go Oh, sorry. Yeah. The Claude desktop app. Um if you were to create a new um you have to start the Yes. So you have to start a new session and then in there you've got local. You can see local. Oh yeah. And then you connect to GitHub. So then it will just do it on the cloud and then push it automatically to GitHub. But you Yeah. You pick and choose which you want to work locally or whatever, but you need GitHub set up. Yeah. Okay. So it so is it on anthropic servers in that case? I think so. I think so. Or GitHubs. Either way, the cost is is kind of covered on the um on the thing. Saying that we didn't we didn't cover this this deal that they're doing with Opus, did we? Uh what exactly? The 50 $50 extra. Oh yeah, we haven't. Yeah. Um maybe you show this because by accident I clicked it. So yeah. So Oh yeah, that that's for the previous video though. Yeah. Yeah. Yeah. So basically, but you can't you can't show it. Yeah. Um some reason it's not letting me edit. Before presenting a screen, remove the other one. Okay, here we go. Yeah. So, in my settings here, just make sure no um API keys, you get this, which is uh 37 pounds, which is $50. You get extra usage just to celebrate the the launch of 4.6. So, yeah. Uh you you did it, didn't you? So, you're on the $19 plan. It charges you an extra $50 and you just get some more usage. I couldn't see any statistics or or any uh uh how do you say um exact numbers on how much extra usage you get. I think it's it I think it's like $50 of API pricing or maybe they get like yeah probably at the very least $50 of API pricing. So the way I understood it and the reason I clicked it, I thought you pay 50 bucks and then you get that 50 bucks of uh extra kind of like API pricing anyway and up to a certain time. You'll get even more kind of like unlimited. I thought you'd get unlimited until the end of the deal, but $50 credit. $50 credit. So what is special about it? It just that's why I mean I think I think it's like slightly more generous than API pricing. Maybe it's half price API pricing and you get $50 or something like that. That's what it is. You're on 100. Yeah. Yeah, I'm on the 100. Yeah. Yeah. Okay. So why not 200? Because I don't you know I haven't coded I haven't done any code in a little while to be fair. Um so I will $100 seems to be a good balance. $19 isn't enough when I get heavy with it. Uh yeah, 200 is just way I've never hit the limits of 100. And I've just released a video today about Clawbot and Open Claw and buying a Mac Mini and stuff. It's like we need to get used to like seeing the limits because there's so much flexibility. If you see the limits, upgrade. It takes two seconds. If you're not using it, downgrade just Yeah. So, I'm not hitting the $100 limits with the 19 bucks a month uh plan. I'm hitting the limit sometimes within five minutes if I have multiple agents. Sometimes I do have multiple like agents redesigning or something. I' I've done like kind of like big redesigns with multiple like six agents and hit the limit in five minutes and I was like, "Oh, shoot. I I didn't see that this coming." So, yeah, I'm trying to be a bit more careful. Here we go. Right. We can go on to clawed ads, but next in line is the webflow AI builder. So up to you. Okay. So let's let's let me I like things to transition nicely because we were talking about chlor code, but really doesn't matter. I just a bit anal like that. Yeah. Push it to the limit. Uh, I'll show this and then this and I'm just put some cream on me. Cream on my elbows. Dry skin. Cream for your left elbow. My right as well. The right one needs a different type of cream. Yeah. Yeah. Okay. Cool. I can hear something with your mic when the I guess the desk is wiggling. No, I know what it is. It should be fine now. Yeah, it is fine. Okay. Yeah. Cool. So this is the new web flow and this new web flow update is much bigger than it seems like and it's about the new direction Web Flow is taking with their entire platform. Wow. By the way, not even seen this. I don't know what this is about. You haven't seen this. This is Web Flow. Like it's hard to believe. It looks like, you know, Figma can framer. Sorry, Framer. It's framer. It's Figma. It's a canvas and it's so freaking cool. So, let's let's actually get use uh get let's um let's actually start using it. So, we have this AI site builder. It has a new tag because this is like the the new version. You you can start it directly from the Web Flow homepage. And I already have a prompt ready. So, I'm going to give it the prompt. And let's go. It's a JSON prompt. And here you already see some new sort of like UI. And here we have a site map. What does it look like? Real. Yeah. Yeah. Quite a lot. So, uh it seems like right now I can do only five pages. I tried to prompt it to do more. Haven't been able. Doesn't matter. I'm not going to use this today. I'm excited to use this on my old projects. I'm excited to use this for other things uh other than this AI builder. But this is where they are showing it off. So you can delete the pages. Actually, let's delete this about page. Ah, now that I have four pages, I can add a new page. Add about page. Can I just point out as well generate this is really this is really nice because it's imitating the product development life cycle. So you set a prompt it's then taking you through the information architecture and the UX and then this forms the basis of the next prompt which presumably is the design and the pages and stuff like that. So yeah I like this is following the product development life cycle. So here I can add for example a new section FAQ. U what is missing is to to change like to specifically prompt each section that is not available yet. But I'm super excited that all of that what you said with the life cycle but also for let's say projects that already exist. I would love I would really love to see this view for my existing projects. In addition to all of this, the metadata like imagine if you could have like um seeing seeing the title, the the title, the maybe even the the heading structure like you could toggle to add all of this in this view and you could like have such a nice way of like managing all of your website and what is where the information uh architecture and everything. Anyway, so this already looks amazing and you can edit it and whatnot. You can swap sections if you want. Now, let's generate site. And this is where it gets even more exciting. Like this is genuinely so cool. And I'm so happy to see this like new like WebGL essentially canvas because this section that we have is almost certainly rendered via WebGL. Um and it is going to open up so many uh opportunities. So this one is I guess taking a little bit. Now it's reloading and it's going to give us this new view that is also very similar. Now preparing site, creating sections, drafting content and then we'll have finalizing all of these pages. Uh at this point I can also like drag and zoom in but I guess it's taking a little bit of time time to drink but yeah this doesn't look like web flow right like the old one I mean is that I mean you tell us probably but it does that exist anymore the old one or is it just this this or so this is just the intro for the AI builder. And you mentioned such a good point. This is this is the last time we see this view unfortunately as of now. As of now, once you pass this once you go into the the old web flow UI, you cannot come back here and change things from here. And I that that is what we want to see. Why would you? It is so nice to see all of the pages next to one another and I would like to navigate the entire web flow experience this way like this way also with the other way. Anyway, so here uh we can go we can obviously zoom in and zoom out and everything is super sharp. So I I don't know how they are rendering this and it's very performant. I mean we have only five pages anyway. We click edit, we go in saw that there was already an animation. So, let's go back to all pages. I go back in there is this nice animation and I can uh change the colors. Uh I can come here. I can say I like this one or whatever like I like this one. And you see how everything changes kind of like automatically uh very in a really nice way. So I can do this uh with the colors and I can add animations. We'll talk about this the headings and sections what not. We can do it here uh for whatever we don't like. I can go back to other pages. Let's go to the contact page here. Let's say we want to add a section underneath. We have now this view. This one takes a while to load. Uh that has been my experience consistently with this taking a bit and there is a fair bit of hero sections. I mean all of are these all hero sections? That's that's quite a lot of sections. Mate, this is so laggy. The other views are so fast and this is just so laggy. Anyway, let's let's add this one. Can I move it up? Oh, to move it up, I have to use this arrow. Okay, cool. Now I have it up. Uh, the images seem like to be um made with some image generator tool and I can obviously change the the color here to to to match the rest. And that's that's this new view of kind of like configuring your pages in a very very very simple way without going into all the the details. And this is what I'm saying. I would love to have these different ways. I would love to see this as a let's say an a marketer view, an editor view. After I made a website for them, they could add new pages this way. I would love to see that. Anyway, we say continue building and now we go to the old Web Flow UI and we say goodbye to that new because realistically that whole experience like you as a uh Web Flow expert, you're probably never going to see that view on a website. You are you would you do you see yourself ever using that AI builder on No, not for myself. No, but I would love that. That's what I mentioned. So the the update is very exciting. Not because of the AI builder. I don't care even a bit about that AI builder. Maybe it is nice for clients who want to set up a uh a simple nice website that is not too complex instead of buying a template. And I'm not meaning even a creative template. It is a replacement for simple templates. M uh I wouldn't use the AI builder to answer your question. What I would love to see is these like UI pieces, these like different ways of managing a project and you you remember that like side byside view. I would love to see the different break points similar to framer in that view. That would be extremely useful. Imagine I change something. I see I see it across all the break points at the same time. I don't have to click around. I just have it. So I would love to see this view completely migrated migrated to that other view. So within the web flow canvas like within inside we would see kind of like framer and we'll still keep the the controls here. Anyway, so just just show you a little bit more. This is the the the website the the the same old way but there are a few things about it. So if we click on the navigator or on the footer we see there are already uh components like webflow components. We have this theme editor where we can do everything that we already did in that other view. We can do all of those things and we can add animations. So, there are a bunch of pre-made um presumably GSA animations that also add stagger. So, I can add like right now it's popping. I can add slide in apply animation to all to actually all elements. Don't do this. Don't overuse this. But just to show we can animate everything. And you see there is like these really nice stagger animations and we have here tabs that are completely functional. Everything is staggered. It's done really well. Can you prompt in new sections now? Yes. So well on or just just to show you this this question here. Can you actually prompt it to use certain naming conventions with the classes like we'll talk about this. Yeah. Yeah. That's no, you can't do that. We'll talk about that. They have their own system. So, with the AI, you can find a layout or you can prompt a layout. You can say I don't know what what what do I need? Um, add testimonial slider. I don't know if there is Oh, it it's it says find layout. So, it's not Yeah. So, it says fancy search. Yeah, it's a fancy set. It's just there's no prompting. But maybe do you think from the from the AI site builder you could tell it to use client first? No. No. Uh that is clearly Oh, oops. Something went wrong. Please try again. How do I try again? Uh yeah, I guess that didn't work the first try. Let's let's see if it works the second try. So adding the the the section is one thing but also has to uh take over take the all the stylings that we have. So that that's why it's taking some time and with this this button it just swap styles. Again a lot of uh parallels to Room. I mean, big shout out to Room because they got this right and I think they singlehandedly changed the entire ecosystem with website builders showing like this life cycle way of thinking about it like going from sitemap to like more detailed uh view and then even more detailed view and then styling and then publishing. Anyway, so I think I yeah, I covered a bunch of stuff. Now, let's go to the style guide. What people are probably very curious about. They're using something called Flowkit. Flowkit is a web flow's utility for CSS framework that helps you build websites quickly by applying pre-made CSS in web flow. From what I what I have seen, this is similar to client first and mast and maybe in between them. Um it it it says version 1.0 so it is a complete thing. Uh here I see we have some code stuff. What is this? So we have some SVG wrapper. So we have a bit of custom code. They are not shying away from using that. And for the colors you see how the classes are made. SG color sample. What is SG standing for? SG style guide potentially. Yeah, it's just a prefix. Where are you seeing SG? Just in the Yeah. Okay. Yeah, fair enough. Style, I reckon. So, yeah, sorry. It is um quite small. I'm going to zoom it a little bit. So, that's the colors. Let's just have a look at these. So these are the okay they are adding the the classes first and I see here are the HTML tags. So these are for the tags these are for the classes and here we see heavy use of variables and we'll take a look at that as well. Here we have even the gaps everything is uh set up with uh variables and yeah even the colors the the button. So we have class of button and now let's have a look at the do we have components? Yeah, we have just two components and the variables. We have the base that is empty. Then we have the colors pretty clean typography sizes even interactions and the slider. I don't know why the slider anyway. So, it is maybe not as clean, but I'm look I would not start a project using this just to be clear. I wouldn't. But if somebody comes to me with with a project based on this, I wouldn't be too mad at them. It It's usable. It's usable. Okay. It's not clan first, but it it's fine. It's usable. It's not masked because I like mast more than any other um one, but it's it's okay as long as here's the thing. Who's going to use this AI builder? Probably people who are first timing web flow. And what are they going to do? They are going to probably destroy this. Not directly. They probably just change something somewhere and you know uh this will get destroyed. So as long as they don't destroy this style guide, I wouldn't be mad about um using it. Yeah, there's a question here about, you know, users design system with that. Um, my guess is they have a this AI builder just has a as a format of every website will be laid out the same. You'll always have the style guide. You'll always have, you know, you'll be using certain components and this and that. Um, I think the AI aspect probably is mostly deciding which combinations of those things to use and what the content should say. Yeah, I I think it's there's there's realistically I think there's a very light layer of AI that actually plays a part in this. It's just understand your requirements and then piece together everything. So just to be clear, yes, the AI is not generative in that way. The AI is not generating a new section and you know generating the HTML CSS. No, the AI is just in a way I hate to say this. It's not even necessarily AI in that way like generative. It's just picking different sections from the existing collection. I like looking at you know the amount of sections that you can add. It seems like they have thousands probably like thousands of different sections. AI is choosing from those. The the colors I I've done like I've I've used this tool this new update for three websites already, three tests. Uh it's always the same color set if it makes sense. Like when I when when we come here, this light and dark, this is like fixed. This is what AI has. And it's a nice safe space to start with. So all of this is not AI. The copyrightiting is AI. The image generation is AI. Generative AI. And that's what I said like I'm not really excited about that AI side of things. Um because AppGen is truly generative AI code generation. All of this is not. What I'm here for is um that you know view that I showed this this gets me really excited. Uh this uh view gets me really excited. These are fun parts that are I I think web flow is just you know getting into this canvas you know free form canvas with with this update. This is just you know like showing us a glimpse. Maybe they are testing it. Maybe they are like seeing how viable it is, the performance, how community likes it. Yes, we like it. I I was in a meeting with the community. Everybody loves this view. So, uh take this as very positive feedback, web flow, and yeah, uh the AI part, it's fine. Yeah. Yeah. Cool. Good to see that they're still cooking. Yeah. Yeah. I guess they all have clawed code unlimited access and they are just like building building maybe or maybe they're still hand coding it you know it's not it's not overly complex to be fair you know it's it's good it's cool um as we say it's you know as Fernando says here it's not like repleting lovable where it's and as you say like it's not like generative like from it's just using a a good solid formula base formula and then yeah having having that natural language intermediary between the creation and the the idea. So yeah, pretty cool. Yeah, lovable and replet would be comparable comparable to appgen but appgen I used it and I used replet and I've been telling you like in private like replet blew my mind away. It's just I I I'm I'm like I I'm I'm beyond like even being able to express how amazing it felt using uh Replet uh and Web Flow Appgen is beta. This is Yeah, it's Yeah, I don't know if they are putting a lot of time on this. I hope uh they do because Yeah, Replet was really cool to use. Yeah, for sure. Cool. Let's move on to Mbooks. We skipped that one. So let's do malt book. Yeah. So last last week we spoke about malt book which is a social media for AIS that saw them creating religions discussing humanity. It even had Andre Kapathy remarking how insane and interesting the experiment was demonstrating what could be described as sentience. But it all came crashing down becoming almost unusable and a cesspit for crypto scams. kind of like Reddit, which is the social network it looked to replace. So, this is the iconic tweet. Let me um get this up. Uh oh, you I've been having a lot of fun with mold book memes. Uh I know there there are like some human behind it, but yeah, it's it's it's been a moment. It's been a moment and we're here to celebrate it. So, this is this is the iconic tweet. I didn't write this is this is what takes me back to what we spoke about at the beginning. I didn't write one line of code for Mbook. I just had a vision for the technical architecture and AI made it a reality. We're in the golden age. How can we not give AI a place to hang out? So this is the naivity of it all. You know 500k views that that okay cool. the naivity even he had a technical architecture in his mind and look even when that uh is apparent not sure how apparent it was uh what what happened so let's let's go let's go let's go back in time let's go so what what happened was um Jameson O'Reilly I've been trying to reach Maltbook for the last few hours they're exposing their entire database to the public with no protection including secret API keys that would allow anyone to post on the behalf of any agents including yours Kaparthi. So look he was able to reach it. Here's his claim token. The API key is there. And let's be clear, crazy. Let's be clear. The API key token just used to post. Basically, all this would mean is I could post as Andrew Kapathy's maltbot. Uh yeah, the maltbot or agent for that matter. It doesn't it doesn't maltbook doesn't discriminate between whether it's um an open claw agent, whether it's clawed code, whether it's codeex. It really doesn't care. It's just a it's just a place that an agent the I if in a perfect world the agent can go and look at posts and comment and just respond as as it would see fit. But we'll get into get into the uh compromis. But the entire database was completely open. Nothing nothing. Uh and I I'm here saying like this won't happen to me. This won't don't clip this. Yeah. Yeah. And and so here's the thing because we one thing we spoke about is we were like well we don't know what the different like this could very well be a human posting to this platform because the method in which you post to that the the malt book the malt book was it's just a curl request. There was no um way of deciding whether it was a human getting and because I could go into claude code right now and say hey go and post a post about designing your own religion to mold book and it would do it and post. There was no security there. There was no authentication. there was no um even if there was authentication we before the call we even tal or before the show we even spoke about um how you might differentiate between a human and an AI you really can't differentiate between the AI but the point is I could prompt my AI to go and read a post respond to it or create a new post or I don't think I've got this t uh this open actually um yeah because I haven't got the I haven't got the tweet, but there was no rate limiting on the API, which is when I began to initially start to see it crash and burn is that I just could not access Maltbook. Couldn't access it. And it's because there was no rate limiting on those APIs. So, you could post or you could upvote and then like a billion times I could just set a for loop to upvote a particular post. And we'll get into the what happened with that. um which was killing the service because it just wasn't preventing you from you know normally you have a a rate limit of saying like you can only post every minute or every 30 seconds you know otherwise you get blocked there's a there's a repercussion to that but they hadn't put any rate limiting like even I know this yeah um no for sure I mean like it's like how how does this happen like I haven't done like like this conventional coding thing ever like you know building to that level. I've been coding like smaller things but I even I know there is like API rate limit. So how does something like this happen? I kind of don't understand like isn't it like couldn't you theoretically give your code to a different agent or even the same agent new session and say like like security audit and within it find out that hey the API keys are exposed because every time I set up a project I I see somewhere my agent telling me your end file the environment uh like file don't expose it don't like don't add it to the you know you add it to git ignore and things like this I know this and I'm not technical but it's not uh so his API keys haven't been compromised like yeah I understand but it was like the way that was coded that everybody could post in you know um as anybody else yes they're a yeah but that that's slightly different so He's storing the API keys in the database which was compromised. Yeah. So that has nothing to do with I understand but like any I I know that was just an example saying like couldn't even like today's agents couldn't I give this code isn't it open source like couldn't you take the the multbook I don't know if it's open source or not and like run it run an agent and say like do you find any security vulnerability there I'd be interested to know I don't know if mobook is open source Um, I mean, what what if you were to run a security a if if Chris Schit or whatever the hell his name is, um, Schlit, sorry, not [ __ ] Schlit. If Matt Schlit uh ran a security order, it probably would have picked up rate limiting, you know, as an example, it probably would have said you've got no rate limiting on your APIs. But, you know, there you go. That's that you're just trusting in the AI with um and it might miss it. you you know if you ask so much of your AI where it's got so much to take into account it's it's like the classic prompt um don't make it secure don't make any mistakes do you know what I mean that classic one line of like if you're giving it everything telling it to build build God and then make no mistakes uh make it secure like it's going to it's going to falter it's going to like you know this is why when we get better at AI engineering, we get a bit more narrow with our scope so that we can run security audits against that specific uh feature that it's building and it can pick out these security things. So anyway, so if we take a look at Maltbook now, you'll see what it's become. Uh trading Fred, show off your trades here. Core Mint, USD hackathon, the crypto bros have taken over and there are some, you know, if some un un uh knowing innocent little person came and got their AI agent to read some of this stuff. There are uh ETH addresses and Bitcoin addresses in amongst some of these codes. The basically the crypto bros ruined everything as they always freaking do. They always do. this. They seem to have sorted out the up votes thing. If we can we organize the posts by number of I thought by clicking top I would get the thing but they seem to have fixed the sort of maybe the rate limiting or whatever but there were hundreds of thousands of up votes which would obviously push it to the top you know. Um, this ladies and gentlemen is a very very uh innocent or you know he got a little bit of a he got a little bit of a re realness check cuz realistically you know luckily people's personal information wasn't stored in those in those database entries. I mean, going back to going back to this one here, uh, you know, um, yes, and, uh, Kapathy Moly, his full name maybe or description, maybe there was something in there, but he was lucky because nothing sensitive was put in this. By the way, what can you click on that again? What is karma 23? Karma is I think that is a a I'm not a massive user of um disc uh what is it Reddit but I think karma is a way to know how much you're engaging with the platform how many posts how many upvotes how many comments you've left and you build karma I think comments uh chat let us know what's what what karma is but karma is a Reddit uh functionality I don't know what it entail what it means or So, I just exposed myself showing that I'm not a Reddit user. Exactly. There you go. You're not a Reddit degenerate. Um, so yeah, that's the price to pay for an insecure applications. And luckily, it really isn't a big deal because it's just uh it's just a post posting and thread, but could very well been someone's bank details. It could be very well be someone's personal information. It could have been a lot worse, you know. Could not happen to me because I'm prompt. Don't any make any mistakes. Yeah. Yeah. Exactly. Yeah. Yeah. So anyway, ladies and gentlemen, build your apps in a safe and secure way and finish it with the the classic oneliner. Be secure. Don't make any mistakes. Uh let's talk about the cord ads because these are quite funny. Yeah. Before we go uh talk about this, look man, what's happened? The worst like this is my life now. This is my life now. I What is this? What's happened? Talk us through it. What's going on? My rate limit. I couldn't finish one project, one task. It was a big task. It was Oh, that's not a rate limit, bro. That's just your uh No, not rate limit. My limit. Yes. Sorry, I misspoke because we were talking about rate limit. Yes, just my my limit. There you go. See, so why haven't you updated? Why haven't you upgraded? Not to say you should. You know, you might be struggling for money. I don't I don't need to know, but honestly, I've been a bit stingy with this. I I Yeah, stingy is the Yeah, I I'm spending so much money on AI and everything. I'm probably paying 300 bucks a month already for myself and my team and okay Higs fields and there is chat GPT and there is the API and there is this thing that I'm building like the the API for this and I was like can I get like can I get far with this that's one thing and one other thing that I do I know you disagreed with me prior to uh like when we started talking about clients paying for me to learn a part is like I it's the same like can I get something that would financially cover cover this like can I get a client wanting me to build like whatever for them and I would justify paying 200 bucks a month by saying hey this client is paying me and it would cover it that's how I see the business side though. Mhm. But I I think I will upgrade maybe to the 100. Yeah. Yeah. So what I want to do a bit. Yeah. What I want to do, and this might indeed solve your problem, is I literally, and maybe I'll get AI to do this for me, but I want to look at all of my subscriptions and mark any against uh mark like rebuild against any that I want to I should rebuild. Um, Frame.io know is a way is a sign off process a tool that I use to sign off my videos and I can upload my videos before they go on YouTube and I can and or my editor can as an example and I can comment against the the the things right the timestamps so I can say remove that edit uh remove this bit um add in a title here whatever this is simple as [ __ ] I'm rebuilding this right this will save me yeah it might save me 12 bucks a 30 bucks a month depending. Yeah, I guess when I get my editor in. But um yeah, I can share my videos to clients. They can, you know, they can comment on them, whatever. Yeah. Imagine that for every single subscription that I'm paying. Imagine I can start alleviating and this is this micro SAS business that I want to really promote. It's like, let's build our own tools. Now, you have you should have access to Vzero, which we're going to cover today. They've had some they've made some updates, but you should have a you should have a that's that's one thing you can use to build your your software. You've got a subscription to Replet, which the credits will reset. You should have um lovable. You should have one. By the way, boys and girls, um Kabaza and I uh the beginning of this year or last year even now it's coming up to the refresh cycle. We signed up to a newsletter which gave us access to cursor lovable replet all of these AI tools. So we're we're we've all got subscriptions to all these things. Be careful about your you you're sharing your screen by the way mate. Yeah. Okay, cool. Just be careful. So yeah, so we we have access to all these tools. So start building start replacing some of your subscriptions is all is all I'm getting at. Save yourself some money because we we can build our tools and we can build them in a a much more like I don't need all of the features of frame.io. So I'm just going to build what I use and then vibe it out and then I'll save myself a subscription. That that's really cool. And I'm that that's what I'm beginning to do and that's what I where I feel like at the top of this Dawning Krueger effect that like a developer like developers have been doing this since ever you know and it's taken us long now we can just now we can do it afternoon. Yeah. Exactly. And like why would you pay for you know whatever app that you can replace? I saw um uh Theo is building a frame io um like similar to frameware io as well, but whatever like you're to I think he's going to charge money for it. I'm not going to pay for for but like if you put yours up for like on GitHub for you know like as open source I would love you know he's probably frame.io very differently to what I use. I'm not going to pay for his thing. I'm going to build what I want, you know. Yeah. So, yeah, this I'm a big advocate for this and this is where I see the future and where I want to take my clients is to say, look, hire me as a dev. I will build any and all tools that you need to run your business, you know? Oh, man. This is so cool. Yeah, this is so freaking cool. But, man, I'm just so [ __ ] like addicted to this. Like it it's so cool once you get once you get building and once you like I add this and I I make it work. It's not a [ __ ] prototype. It's not even like an MVP. Maybe it's an MVP or slightly better, but it's just so addicting. Yeah. Yeah. For sure. And like I say, make it nice. Yeah. it could solve your problem if your if your if your money is going, you know, offset the money that you would save on subscriptions into clawed code, you know, um or indeed just anything really. But anyway, that's my that's what I'm going to do. I'm going to I'm going to spend some time just looking through all my subscriptions and thinking about what I can play replace. But right now, I'm just building my Frame.io um uh what do you call it? Clone, you know, maybe I I suggested Discord. I know Discord is pretty complex but I pay for Discord by the way but Discord is too complex. I would say no chance like no tool could come close to Discord for now to to replace it and I think building something similar to Discord this good. Yeah. I I think that is the next order of magnitude. Yeah. Yeah. Cool. Let's talk about Claude ads and uh start wrapping up. But uh I don't know if you guys heard Yeah. What? Yeah. Wait, wait, wait. Let me fix the screen. Yeah. Cool. Um I don't know if you guys heard, but Anthropic didn't waste a second to rip into OpenAI for introducing ads to their platform. Let's take a look. Have you seen these? You seen these? I I I've watched them and they are so good. They are so good. And Sam Alman, he's so pissed. He's pissed. We'll get into that. Yeah. How do I communicate better with my mom? Great question. Improved communication with your mom can bring you closer. Here are some techniques you can try. Start by listening. Really hear what she's trying to say underneath her words. Build conversation from points of agreement. Find a connection through shared activity, perhaps a nature walk. Or if the relationship can't be fixed, find emotional connection with other older women on Golden Encounters, the mature dating site that connects sensitive pubs to roaring cougars. What? What? What? Would you like me to create your profile? ads are coming to AI but not what's the difference between me and you. So that's the uh that's the general gist. They there's a series of them. That one's called betrayal violation build muscles deception and they all come along the same thing. They've got this they've got the questions they start talking about and then these sort of slightly you know uh unrelated ads come into it and um all of the car it's funny because all of the characters or the people who represent the AI have a slight delay to them. I thought that was a little backhanded thing of like your AI is slow, which it is. And they're like, by the way, they if you watch them, they also all do this. A I hate it. This glazing at the start like amazing. Great. Amazing question. Yes. And they they glaze you and then they start with the delay. Yeah. Yeah. It's so open AI. Yeah. So good, Sam. Yeah. Sam wasn't too happy about it. So, first thing, good part on the anthropic ads. They are funny and I laughed. Great. I wonder why Anthropic would go for something so clearly dishonest. Our most important principle for ads say that we won't do exactly this. We would obviously never run ads and in the way Anthropic depicts them. We're not stupid and we know what our users reject that. And to be honest, we we covered this last last week and yes, I but Sam we said we don't believe it. Yeah. Yeah. But can you show can can you show that text again? So the the the last sentence you read, what is wrong with that sentence? Well, let's have a little look. I'm loving my little uh highlighter tool here, guys. I just So we are not stupid. Oh, which one? Which one? No, no. Like what do you read where where it says Yeah, this one. We are not stupid and we know our users would reject that. That's this typical like red flag that you you probably should look for. He is not saying this is bad because it is bad because of morals, because of right and bad. He's saying this is bad because our users wouldn't accept it and if it if they were, we would shove it. you know, it's it's just like yeah, I I saw other people also pointing out exactly this sentence and it's Yeah, man. It's it doesn't look good. And I I don't I I can't believe them. I mean, I'm using Chad GPT. I'm paying for Chach. I understand all of you know this within that context, but I I I wouldn't believe them. I to me this promise and the promise we talked about last week are very similar to Mark Zuckerberg. Has he not promised that he he's not going to do anything bad with our data? Has time not proven that he has been doing something actually illegal with our data time and time and time again? Yeah. And with these companies, something that people have mentioned about anthropic in defense, not exactly in defense, but anthropic with these videos showing that they are taking the higher moral uh stance and they're like we're not going to to do this kind of things. But they're still paying for the like the because they they downloaded books, scanned them without buying the books. They pirated them and they are still paying the fines for for what they have did uh what they have done. And I saw people kind of like trying to mention that which which is fair. But just just to be clear, it doesn't mean that Anthropic did a good job there. No, it's wrong. But it's wrong across the industry. Every single AI company has done this. Any AI company that is relevant has stolen data from everyone. And I still don't know how to feel about it. That's how AI came to exist. They are literally stealing this video that we are uploading. They are training um their models against it. And I guess this is we this is the world we are uh living in. And at least in Claude, you won't see ads. That's what they are saying. Yeah. Yeah. I this one this took me like it's quite a like I guess it's on brand for anthropic to double speak to use deceptive ad like you know um I uh I like that he's immediately gone in and uh you know bit back a little bit but I didn't realize it was a Super Bowl ad. So this was a Super Bowl ad that they paid these were like six adverts. They would have paid a lot of money for that. Do you know what I mean? Um and they did it so quickly. Yeah. Like last week. Well, we've been they've been talking about um ads in chat. Ah, yeah. Right. Maybe. Yeah. They had they saw it coming and maybe they know from like some insider information people telling them it's coming so they prepare for it. But yeah. Yeah. So, I don't know. But, um it's good. It's weird. I I like it. I like the memes. People were like Sam Alman not being mad and then screenshotting his tweet. Yeah. Yeah. I'm not I'm not mad. But the screenshot thing is like and you know wrapping up from the from the ads like I I really don't like Sam Alman. I think I've read enough I've heard enough. I've um yeah, we've we've covered it enough that he's not a great guy and he I think he he tries to appear to be a great guy. He's got his PR is is on point, but you know, I don't know whether you know this, but he has his blood on his hands with a um a suicide that happened to a whistleblower who worked for OpenAI. And do you know about this? Uh I have read about it. Yeah. But I don't know like deep details. But I know about like in his past startups he has literally scammed people. We have this uh world chat. Is that me or is that him? Is it He has a history of this. It's him. Okay. Uh yeah. So the there was he's been there was a whistleblower who Oh, decided to join us. Yeah. I was just saying there's a whistleblower who uh was going to reveal a bunch of Open AI's, you know, behind the scenes stuff. And this whistleblower committed suicide. However, there was signs of struggle and blood all through his house. No indication that he was ever going to commit suicide. Not that we would ever know, you know, not that, you know, reach out to your friends and and family buddies, you know, people that people could be struggling. But anyway, but it does really seem like and they got wiped under the carpet. they were like they didn't the the family are still petitioning to try and actually get some more information and to continue it but it was almost just completely brushed under the carpet um and seemingly completely paid off with someone with lots of money. So yeah, Sam Alman denies this, but anyway, he seems like an interesting and he, you know, as you were saying, he's he's got a history of deceptive um uh business practices. Um kind of just I mean, you know, he lost his job a year, two a year ago. No. Yeah. Just under a year. Just under a year ago. There was a coupeta happening there and Yeah. Yeah. And uh he got his job back, but loads of um board members left. loads of open employees left. People don't like Sam Alman who know him. Yeah. So, and I worry that this guy is at the helm of one of the biggest companies uh the one of the biggest breakthroughs in this in humanity's history. You know, I just don't like that Sam is there. So, that's my that's my spiel on Sam Alman. And uh I I you know, I worry I do worry anyway. That's I'll say less about that. He's not I don't think he's a good guy. Um, anything else before you uh uh No, on the on the ads. On the ads. Uh on the ads. No, he's just uh I I was literally just saying that he's like uh Mark Zuckerberg. Yeah. Yeah. Yeah. Just something's not right about him. Something's not right about I used to like him, but just after learning more about him, watching watching more, something not right about Yeah, Figma. Let's talk about Figma. Actually, let's talk about sex and let's talk about you and me. Let's talk about all the things. Uh let me actually prepare a few things. Oh man, my agent is coding my tool and I can't use my tool because that's probably that's probably why it crashed, right? That could be as well. That could be as well. Um give me a second. Um I'm doing something. Okay. Also, I'm getting a lot of followers on Tik Tok. Crazy. That's nice. I mean, Tik Tok is famous for um let's seeing a lot of like it's easy to get followers, but All right. All right. Uh easier. No, it's not easy. It's easier, but all I'm saying it it doesn't matter as much as views matter. So, you you want to optimize for that. I'm not downplaying it. Yeah. Yeah. Uh, okay. I got things ready. Okay. Should we go? Yes, guy. Figma is killing Adobe Illustrator this time with a new update that allows you to make an editable vector out of any image. So, for example, here I drew this beautiful flower. I'm going to screenshot it. By the way, I did this with this tool that I'm building. I can paste it here as a screenshot. So, this is just an image as you see. I can come here to more and then say vectorize and then it uses AI by the way it uses AI to recognize this and turn it into an editable vector. So this is let's see if it ah I have to click this I have to click okay click done and now I can yeah what do they call part of what they call this in Illustrator it's like expand or something isn't it I don't know exactly so yeah it works really nice with this screenshot here's a another one that I would like to try like I this was also an image vectorize it. And again, it takes a second. And now I can like pick the eyes and just make them bigger. That that's the really nice thing about it. If you zoom in all the way, you see it's not perfect. Just to manage expectations. And talking about managing expectations, I want to do a test that I haven't done prior to the video. So this is the web flow logo. Uh that is perfect. You see this is like the perfect vector with the perfect amount of points. Nothing extra. I took a screenshot of this and brought it here. So this is the screenshot version. And I want to vectorize the screenshot and kind of like go back and forth and compare them. uh to see how good or bad it is. So, we're going to accept this and oh that's that's pretty good. That's not perfect as we saw. See, this is not perfect. Okay, my definition of pretty good might be different from the definition of pretty good for somebody who uses Illustrator, but to me that is pretty good. So just to be very I would say this is how this Oh, it's that's how it's supposed to be, is it? Yeah, this is how it's supposed to be. I would say illustrator. So it's sharp. I would say illustrator isn't much better. It depends on the surely it depends on the actual um resolution of the image. I don't know if we have any Illustrator people in the comments, but let us know what the what to expect on a tool like this with a this was a roll Yeah, this was low resolution. Like quite low resolution. Is he gone again? Chat, is he gone again? I think he is. Yeah. Um, bit awkward, isn't it? bit awkward. Oh, here he is. Enter the studio. You know what? I don't need the 200 bucks claw claw subscription. I need to buy a new MacBook with a billion gigabytes of memory. Uh, my browser crashes because it says it doesn't have enough memory. And I guess the reason for that is this allowed to share screen code is um is a bit uh RAM hungry to be fair. Let's see actually uh just so memory. Yeah, Figma is taking 19 gigabytes of memory. Pretty insane. Yeah. Yeah. So, I'm gonna close Figma before I crash again. But all I was saying, I'm happy with this because I don't have to use like a super professional app. It's good enough for me. It could be better, but it's good enough for me. If I need to vectorize this to simply change something now, I can. That's the point. It's cool. I would have And is this free or do you have to pay for this? I don't know. I'm on the paid version of Figma, so I wouldn't know. By the way, the these are like, as you know, these are all our thumbnails for this channel. And we use Figma remove background for every single one of them. Oh, cool. It just works. Super nice. Yeah. I just take a screenshot of our video, bring it here, remove background. You see this is with background without same here everywhere. Very cool. Um and now this is just an addition to that. Yeah. Yeah. It's becoming a pretty all-in-one tool, you know. Yeah. Yeah. It does a bunch of stuff that we need and therefore we don't need a specific, you know, tool like professional tool. Adob Illustrator if it can do a pretty good job. Just got some news from uh breaking just got some news from Elon Musk which actually was a few days ago. He's just learning about it now. Um malware found in the top skill for maltbot. Um what does that mean? So, maltbot uh again I'm getting confused because it's open claw clawed code uh claudebot uses skills to understand how it should do certain things and there's a dependency on one of them called open claw core um and it basically has some malware in it. Oh, so be careful out there, guys. If you're open claw people, uh, read your skills. Don't just go ahead and install anything. Yeah. It's literally it's not it's not has not been created by anthropic open AI. These people with high very very very high standards that can't just release stuff. This is just released by someone probably in a similar position as Kabaza there. I know how to build things. I'm going to vibe it out. Yeah. Yeah. Don't do any mistake. Tell your AI to not make any mistakes. It will be fine. You forgot the classic. You forgot the classic. Well, uh I think that's all we got time for in it. Yeah, it's been a lot this week. It has been a lot. You know what? What What are you going to do? I'm I'm going to like after this I'm going to finish my like a few agentic tasks that I have. Well, I have to wait for my claude to reset, but I have um chat GPT until then. Not chat GPT, I have the codeex 5.3 until then. Well, I am This is my first evening off, so I am probably just going to relax and eat some food. I might finish my frame at IO because I started that on Replet. I might bring that down into Clawude Code and save some credits on Replet and see if I can finish that because yeah, we've already got um some I've got to review some videos for the Command AI channel which you should subscribe to at Command AI Command AI show on YouTube. The link is up in the the QR code. You can take a screenshot of that and go through to it. Um, give us a subscribe over there. We're going to release a bunch of clips from this week's episode, from last week's episode. We still got a couple to go out. Um, and that's everything. Um, join us next week. We stream every week. Uh, next Wednesday we'll do another stream of the claw code stuff again. Yeah, I think so. Yeah. Yeah. Join us on Wednesday. So, basically, just subscribe to the channel because then you'll hear about all the stuff that's been going on. Um, I've been Sam. Yeah. Oh, go. Yeah. I want to say, have you seen the app that we are going to build? Uh, the you and I are building. Yeah. Oh, the drinks app. No. No. What do you mean? Have I seen it? Like the preview, the one that I made with Replet because I guess we are going to do Oh, no. I don't think I did. I don't think I did. No, I haven't seen it. Should I show you quickly? Go on then. Okay. uh apps, cocktail app. So, we are going to build this app live and show you how to build it the proper way, not like these other ways where people get stuff leaked. You you get it. You have to do it the right way. So, there's a cocktail ordering app. You can favorite stuff. You can see your orders. We don't have an order yet. I go to the menu. I can take any drink. Maybe let's find another drink. Let's margarita here. I can say without alcohol with no. Yeah, whatever. The style frozen add whatever comment. Place an order. Now as an admin I can go to the admin dashboard and I can see who has ordered what and I can move these orders in this canban just like this and just like this. So this is your view. Yeah. Yes, this is my view. Uh I I have like I can see the pendings and if I click them I can also see uh how how to make it. So there are the notes and also how to make it. And let's say I have a guest over and they are not logged in. They don't or they forgot their account, whatever. I can make an order from my view for them or just put my iPad there. So people can come here order and they can select for from the existing guests or like register a new guest and then order it and I can add new drinks and there is a CMS and everything. It it it works. It's really it's really nice. Yeah, everything works and I assume it's uh very secure because I said don't make any mistakes. Great. Great. Yeah, it's amazing. I I've always I've always vouched for Replet. I think that it's an amazing tool. I'll have a video coming out about it next week, I think, if we can get the sign off on it. Um but it will just take you through from a beginner to expert in using Replet. already. It's just such a such a great tool. Such a great tool. It is amazing. It is so freaking nice. It's like look I can't All of the By the way, all of these in less than two hours of work. I did it across two nights, but less than two hours of work. So, is it ready for production this Sunday then? Yeah, actually I have to So, this is the funny part. I have to add the actual items and that is going to take significantly more time with the images and everything than building the the entire app which is funny. That's the time we are living in. But you can you know I mean I wouldn't use I would because it's a neon database. I would I would use the um neon MCP to just AI add all of those things to the database. Um, I wouldn't manually do it. Yeah, that's not the like just planning them. But I'm doing uh I'm I'm not going to host it on riplet. I'm doing an entire migration and just building my own like authentication database everything using rail railway. Yeah, I'm going to deploy to railway. Well, do it peacemail. Do it a bit at a time. Yeah, understand what's going on. Why wouldn't you why would you not just leave it on replet and for now and then move on to another I want to see I want to see like what I can do there. Yeah. and ah no I know the actual well I I did want to move move away from replet because I want to learn and bunch of things like this but the reason I'm doing it now is because I ran out of my credits like so fast like two nights two nights I don't know maybe six prompts maybe I did six pro like I kid you not I didn't do like more probably like maximum I did maybe five prompts. Well, to be fair, yeah, one of the prompts was like the the whole plan. So, one of the prompts was really big, but yeah. Yeah, there are ways you can reduce your usage as well, but I'll cover that in my video cool. All right. Well, that'll be that'll be everything, guys. Like I say, commandio show over on the YouTube. Um, yeah. Join us next week. I've been Sam. Yeah. And I've been cover. Keep on vibing. Peace out.