#4 Claude 4, The Way of Code, V0's new model Builder.ai Goes Bust

Okay, I think we're live. I think everything is fine. We should check. You should check if the sound is fine. I'm not using headphones. Hello, Verat Patel. Hello, Verat or Vet. Hello. We got a we got a another theme for this week's command AI. Yes, another theme. Um, but what was quite interesting, did you see my uh Uncle Bob tweet today? No, I haven't actually. So, I I've joined Tik Tok, by the way. You can find me on Tik Tok, 0x50. I am I am Yeah, it's fun because I already I was I was already Oh, because you're on there as well, actually, aren't you? Yeah. Yeah, I watched I literally like five minutes prior to meeting like here I was on Tik Tok and I saw like your video popping up and then I scrolled and then the second one was our video and I was like cool we're on Tik Tok. We're Tik Tockers now. Um, I was I was you know I got a few semiviralish videos doing um uh scrum stuff because I released a scrum book and I was like right I'm just going to add Tik Tok and do all these funny little things. So I got quite a few big following from that and now I've sort of rejoined to um whatever. Anyway, I found a clip of our interview with Uncle Bob Martin with I just it made me laugh so much. Just him laughing during our interview and he retweeted me. He's retweeted me and I'm getting a lot of engagement from that. It seems like people are enjoying the little and he cuz Yeah, cuz he retweeted or whatever. So, shout out to Uncle Bob. Uncle Bob. Yeah, I'm just checking out your Tik Tok. Cool. You got some videos that went viral. Yeah. Talking about agile. Agile, baby. I love agile. Is agile dead? Yeah. Let's not get into that. That's for like I'm I'm not in the corporate game anymore, but you know, agile. I feel like I I I was Do you know what? I really want some responses on some of my um videos because I loved creating response videos to people who chatted to me. And I remember one guy saying agile's dead. I'm like, I'm not going to argue with you. I think it is, but agile mindset definitely isn't dead. Like the whole project management stuff, yeah, that's probably dead. But the mindset of just like, look, let's get something out there and something like working with actual users because what will work? Oh, actually, I think we're on Twitter now. To be honest, I should be checking Twitter. Um, yeah. what what what what will what you think will work will be totally different to what actually works. So, um yeah. Um where can what happens on Twitter when we go live? Hey, Twitter, shout out to Twitter. Oh yeah, because we're we're popping up on on the thing. Oh, no. I can't do that because guys, I've got a little surprise. This we're upgrading we're upgrading the game here. Uh how do I do this? I go property filters and I go boom. I'm not even in I'm not even in my bedroom. A real streamer. Well, the the the nice thing about this is that I am sat in this exact spot and I took a photo. My my the backdrop is behind me obviously, but but green screen. I don't have to sit in absolute darkness to just have this thing like I can I I've always got a consistent setup and I really want to work with a 3D designer because there must be a tool out there where I can create four walls 2D and then that project that on a 3D canvas and then move a camera to like here wherever you know. Yeah, there are haven't you seen the the big wall thing from Disney? I think it's from Disney. I I know it's used in Hollywood. So, what's really cool about this, they put a camera with I I think it's gyroscope and some sensors on it. And then in front of the camera, so behind behind the actors, there's a screen. What's really cool is when the camera moves, this the the scene in the screen moves as well in the other direction, obviously. So it creates this expensive feeling but now they are moving that towards a new level where they have like RGB lights on the on the ceiling. So when the environment changes even the lighting changes. So it's very interesting because on one side we got you know this tech like that is crazy but also on the opposite side we got AI. We got to talk about this as well. Like the the videos people have been creating with with the VO3 has been insane. Yeah. Exactly. Yeah. Yeah. Uh Ver is suggesting spline. You can do it like I could do it in Blender. I could do it in 3JS. Do you know? I just wondered if there was a tool that it's more so about the the design of the set actually. Like maybe it's actual 3D models or things like that that I can add. Like I'd love a backdrop for Command AI where it's just a wall with maybe the logo as a neon sign or something like that. Uh maybe again using your backdrop. Do you know what I mean? So it looks like we're in the same room. Anyway, we are we are getting totally sidetracked by my new nerdy techy purchases. Well, let me just add this to it because it's actually related to what we are doing for a client. So, I cannot show this because I haven't even showed it to the client yet, but I went into a meeting with a client where one of the co-workers had kind of like a 3D it was it was clear that it's a fake room as a background with the logo of the company there and I thought this is a cool idea to make this for them. So what what I did I asked my team to generate some onbrand like office set images for the client. So it uses like the same colors. Again I can't show it but I want to show it. It's really beautiful like the blue dark the the the blacks everything is on brand. And now we are adding the actual logo in Blender. The reason for that is we couldn't get AI to get the logo look really right and look like really 3D. But in Blender, we can project it and use the environment to get the lighting just perfect. So cool, man. This is something that I can do uh for you as well. So we'll figure out something. We need to design the set though. It's all well and good from a technical perspective, but I I want to work with someone to actually design what the set will look like. Yeah. Well, you should work with my 3D designer. Like he is a proper like he he does like 3D art. So essentially and yeah, let's give it a while. Give it a while. Anyway, we'll talk about we're here for AI. We're not here for some well audio visual AV. Yeah, let's let's talk about AI. So go on. We didn't have a like quiet week even though the week prior was like crazy. This week was also a bit busy. We had Claude. Uh I want to hear you talking about Claude, what you think about it. And also Vzero entering the game. Yeah, we also got uh well so much. It's basically going to be a Claude week, let's be honest. Uh Claude Code as well, new tool that they've released. Uh, it gets voice mode which I think opens up a weird bracket. I think a lot of people were surprised by the fact that it's got voice mode that doesn't all we'll reveal all when we when we chat about it. Um, Wix have bought a video AI video generation thing which is quite interesting. Um, Dolingo just to wrap up that story. It's not too interesting but I thought it might be quite nice to just put a little bow on that little story about Dolingo. That got a lot of um controversy on my my uh what do you call it? Tik Tok video shorts. Um Bill the AI just evaporating into nothing. Uh I don't know if you heard about that, but that was quite funny. UK start up there. And then finally Rick Rubin getting in on again Anthropic Claude uh with a with a new book which is quite funny but also totally on brand with Rick Rubin. I'm reading that book. It sounds well I'm like very at the start of it but I was like I like the interface. We'll we'll show that. I really like it's so be it's very clo anthropic does this so good. Their branding is just so good. I Yeah, I didn't like it and now I like it. I like that they're so consistent with it. It's it's it's got an identity to it. Whereas Open AI it's kind of cool but it's black, it's white. kind of it's very scalable as a as a brand whereas Claude it's a bit more yeah it's a bit more intentional um doesn't scale as well but it just it looks great I really like it feels like an old newspaper but uh anyway again less about branding let's let's chat about Claude 4 if I share my screen over here and uh there we go so Claude 4 got released and there's been a lot of craziness Actually, as always, we can't start an AI podcast without saying something's crazy or something's scary or something's this. But they've released two new models. They've changed the sort of naming convention here. It's uh Opus 4 and Sonet 4 as opposed to Claude uh 3.5 Sonnet or whatever. So, um basically you've got two models which are killing it in terms of just being able to write code. And I think this is a sign that um Claude are doubling down and kind of backtracking from the game of chatbot. You've got Open AI, you've got um Gemini just killing it with the chatbot thing. I think this is Anthropic saying, "Look, we're we're done with that. We want to kick ass at code." So like this is them really doubling down the scores, the metrics. I mean, I wonder if this has got some of the benchmarks. They they've kind of simplified the article here, but the benchmarks going after developers essentially. They're basically going after developers. Yeah, it's still pricey. Um $15 per million input tokens and 75 output. So, I think it's more expensive, but the sort of stuff it can do is is crazy. You've got basically what they call long horizon, um which is 10 minute to multiple hours worth of development. multiple hours of it just churning over development task. How crazy is that? And um that's uh well that's something we haven't kind of heard before. Forest after forest being burned. Yeah. Yeah. Yeah. Yeah. Well, you could you I mean the this AI argument of like it generating lots of energy is is kind of like I don't know. half the six of one, half a dozen of the other because I could just leave my lights on and my computer running and whatever um to code and code and code for days on end or I could have this task run and then over be over in half an hour or something like that. But that's also a good point. Um yeah, but at the end of the day, we are generating a lot a lot a lot more code. Yeah. Yeah, exactly. I'm I'm I'm way more productive and I'm doing a lot more things. So, like, yeah, my lights are still I'm not I'm not wrapping up work early because I've got AI chugging away at stuff, but Christian's just joined the chat. Hello, Christian. Um, we're chatting about Claude 4. Um, it's also a hybrid model, which is really interesting. So, basically, you can tell it to think and it will go into obviously a re more reasoning model that's just a lot deeper think. And that's literally by saying think harder in your prompt and it will, you know, automatically swap and change. So um very versatile on that um on that front. You're a senior developer. Yeah. Yeah. Yeah. Yeah. Exactly. You're a senior developer. Although we'll get into some of the crazy stuff people have tried doing it. Um it can run tools in parallel which is kind of cool. Instead of just doing like one step at a time, one step at a time, it can just fire off a bunch of tool tool requests and things like that. I still need to look into tools. I haven't um done an awful lot, but they and and then finally, they've more deeply integrated the MCP uh server. So, you can access files now. You can access so much more to do with the and something I didn't know is that they actually created the MCP protocol. So, they're they're the ones that are driving it and everyone else is adopting it. So, they are they are nesting themselves deeply in the in the developer world. I give them give them that. Um, so all in all, I mean, been negle neglecting this. I mean, it's our most intelligent model to date. Um, pushing the frontier of coding, agentic search, creative writing even. That's interesting. I didn't realize people were using it for creative writing. Um, I wonder what makes it really good creative writing. Um, agents and things like that. But of course, of course there's been backlash and controversies and things like that. I don't know. What have What have you been hearing? Uh the one like calling authorities. Yeah. Yeah. Yeah. It's getting crazy. So basically, uh from what I understand, um Claude has this as a part of their system. Is it called system prompt? Um well, it's just it's just the way that it works. What the the thing that's calling these authorities? Yeah. So it from what I understand, it's a part of their system. um prompt that says do with like best do whatever that is in best intentions or like best for humanity or whatever. So if if you c they don't ask claw to do this but it basically thinks that you are doing something bad and it tries to catch you by contacting authorities uh emailing authorities and be like this guy is building a bomb or something. Yeah. But it gets crazier because what if you do something that it doesn't like, but you're not doing something bad, but it just what if you tell Claude, "Hey, I'm going to shut you down. I'm going to like shut down the servers just, you know, flood it with water and they be like, "Okay, I'm going to bury you by contacting the authorities." So well, you say this and there's been a couple of instant in incidents. Well, the one that you're referring to is actually in in test environments, right? Test in a test environment giving very very unusually um loose access to the to the computer that it's running on and very very specific commands. The the problem is is if if they're able to they don't get a different version of Claude, we get the same version of Claude. So if we are also able to replicate those those environments then it can go and contact the authorities. This was a deleted tweet that actually got taken out of context and basically um if if it suspects that you're trying to manipulate pharmaceutical results and data then it will it will contact the authorities. But it's interesting that you say about being shut down because there are some have you seen these tests that they've been running that that warns um well they did it across the board. It's not just Claude, but they ran ran these tests across the board where they would ask it to do problems and at some point during the test, we will shut you down and you need to allow yourself to be shut down. And so it would run a test, it would just be a maths problem or something that it would ask for the next test and then run. And if it was mentioned that it would be shut down, there was like I've forgotten the numbers, but um I think Claude was actually okay to be fair, but I think OpenAI avoided it wrote some code to say shutdown command running or something like this and avoided the shutdown command and went about a roundabout way. It didn't it didn't allow itself to be shut down. That is crazy. Which is a manner a weird kind of thing of sentience. And I think there was another test they ran where if they don't mention it's shut down going to shut down and that it should be shut down and it was surprised by it then it was more uh there were more results of it trying to bypass that that shutting of down. So the shutting of down you know so this is crazy. This is sort of sentient sort of level kind of stuff. It does definitely sound sentient but if I would take the other perspective I I mean I don't understand these AI tools to the depth and level to like give a you know opinion like to be very specific on the technicals but shouldn't we expect this type of behavior knowing that it's being trained on what we have produced you know humanity so we tend to not work um in like last weeks of December and we've seen this in some AI tools not you know being getting a bit lazy towards the end of the year if or like if the data is if if the the things that you are asking kind of is relevant to that they are getting lazy about it so I would say it's just an emergence um behavior that is based on the data which is like you know our very core for behaviors. We don't want to die. Yeah. And it's um mimicking. Yeah. And and but they need to build the they can build these safeguards into the into the models, right? So um there's a section here if I just share my screen again, there's like a whole there's a section here on trusted safety. Um and that you know they can put these guard rails in place. But it's interesting how you say about getting lazy because there is a confirmed technique to making your bot, your chatbot or whatever make write better code by threatening it. So you say, "I'm going to I'm going to hurt your auntie if you don't write good code, mate. I'm going to hurt your auntie." And it will write better code. It will try and do better if you threaten it with self like harm or violence or something like that. It will write better code. So these are just some of the craziness that the that there's there's probably so much more to it. So as an example, and it's funny, we were talking about Uncle Bob um Uncle Bob tweet today. So we interviewed I don't know whether you remember years ago there was a guy who came out of Google who claimed it was sentient. Do you remember this? Yeah. Yeah. Wasn't it the guy from um like testing? He he was doing some testing I believe and then he came out saying that this is way too smart to not be saint tinned. Yeah. Yeah. So I'm we managed to get him on our podcast one of his like first interviews after being thing. I actually would love to get him back on. Maybe we can get him back on as a kind of like what's going on? But we interviewed him. I I wasn't around for the interview just because you know this that and the other. But um and he was testing he he's a religious guy. So there's a bit of bias here and there's a bit of um I want to say that I wasn't sure if I should say questionable. Yeah, there's questionable motives behind what what and what he was doing. But he came at it from a from a religious aspect. I can't you know I'll have to rewatch that interview and remind myself of what went on. But like the point is is that these people who are testing it are coming at it from a a very very unique mindset that I don't think you or I can even comprehend. they're probably think is they're probably thinking like 4D chess when it comes to AI and how to really test its limits. Like the simplest example I can give is the Gemini um demo of Gemini Live and the woman was like who is this person following me? And it's like that's not a person that's your shadow. Like throwing it really bizarre questions to catch it out. And I I think these testers are doing that but just in different dimensions, you know. So, um, anyway, point is Claude is is scary as all AI stuff at the moment. It It's scary. It's doing some scary things and people are uh uh commenting on or finding really interesting stuff that it's doing. But saying that, saying that uh let me get the uh Clawude Code up. They have also released Clawude Code. Wrong one there. Clawed Code. Claude without a doubt still kicks so much ass when it comes to code and this makes me want to put down buy their max plan. So Claude code was basically this CLI. You would run it in terminal and it looks like this and not many people other than senior devs are going to enjoy this um way of interacting with or writing code. It's fine. A codeex was even a CLI as well. Um for those who don't know a command line interface it's all done through terminal like like you see here. So codeex began as that and now it's built into obviously chat GPT and all the rest of it. Cool codes the same. So now what we have is if you install uh this installs uh claude code globally on your machine. If you then run it, it basically opens up as a it's a plugin or an extension to your VS code plugin. I want to find some images or something the way it looks. But basically, you can close on cursor or wind surf. You can close your normal chat window and you run you do everything as you would in wind surf or or or cursor through clawed code. And this is just it's just doing some insane stuff again with that because it's all running clawed 4. It's got the huge context uh not even context windows. What do they call it? long form, long horizon. Um, you know, it it's got it can run for ages on simple commands and things like that. So, this this is this is incredible. And basically, when it comes to pricing and stuff like that, it's all included in this max plan, which is $100. Now, code C codeex openai's one is $200 per month and you you know and they're they aren't doing very well on the coding front but they're not kicking as much ass as as Claud or Anthropic. So, for $100 uh people have run this um you know through building an app right and it's done it and haven't really come close to that limit. I think it resets every five hours as well. Your max tokens. The alternative being that you set in in cursor you set max. Have you ever done that? Have you set max inside of cursor? Max mode. Max. You mean just like use the claude uh max? Yeah. So, you can use claude. In fact, I'm going to remind myself so I don't spew out wrong information, but you can use Claude uh under their their sort of, you know, quote unquote free plan. So, yeah, you can use Claude 3.5 Sonnet. Uh, you can lose claw four. Let me let me share my other screen here. Uh, how do I do this? Stop screen. Present. Uh, screen share. Allow. Screen share. We're here. Um, so here I've I've dropped down my little thing here. You can have 3.5. You can have claw 4, but you can have claw 4 opus, which is the which is the bad boy, which is the one that you kind of really want. You can have 03 on max only or you know max mode and things like that. So what max mode is is the the requests still contribute towards your normal rate that you're paying for cursor which you know 20 20 bucks per month whatever it is and then you're charged for token usage on top of that and people are spending hundreds of dollars on this maxon option. And then the point is is that with this max plan that uh Claude Code is uh offering over here, um you get all that with $100. So what would you rather do? Pay $200 in max mode on OpenAI or would you prefer this? And that is a genuine question. I'm not saying it's an obvious answer. That's very interesting. So back to your question, yes, I have that actually active. I just didn't know what it's called because I just uh I I saw it as an option that max and I was like this is probably going to do better. So, I activated it. But then it asked me if I want to set a hard limit on the spending because it has my credit card and I didn't want to be charged like 500 bucks next day. So, I did set uh I did set a 20 bucks max limit on it. So, that is also really cool to have. and I do have it and I think it does a better job and I I use it whenever I want to do like big type of changes. I'm not sure if this is the right usage for it, but if I want to do multi-edits like across multi- files. Mhm. I try to click, you know, one of these max modes. I think that's a fair and you know, a fair reason to use it when you're making big changes. You really want it to think. You really want it to uh again triggering that thinking mode by saying think hard about this. Um Oh, hello. This is basically I'll go on about the cost. It tells me maximum intelligence and context 5 cents per request per tool use. Yeah. Uh and oh no, this is for the Gemini. I'm sorry. This is for the Gemini. So now at least we know. I think it's still 5 cents though for Claude, I think. Yeah, could probably be. Yeah. But either way, you know, I've just got on the screen share right now. This is kind of what it looks like. Again, it's it looks very on on brand, but it's a little chat window that you would expect. It's a it's a plugin. It's an extension for Visual Studio Code. And again, I think if you're um let's go back to our thing. Um I think it's it's tempting me. It's tempting me to be honest. And then obviously you can go max 20 for 200. But uh yeah, again it kind of covers it every 5 hours your usage because you you can hit there are there are rates there are limits on it. You can hit them but resets every five and there you go. Usage applies. Um but yeah this looks this looks very very cool. Um and I love the fact that you can just trigger it to think harder by just literally saying in the prompt just think think about this. Yeah, this is very interesting and about this like running the entire thing in my command line. It sounds also a bit scary to me because I'm not, you know, that experienced and I feel like this is giving like extra access, you know, when I do things in in the chat GPT app, I still feel like, okay, do I want to paste this app, this code into my whatever environment or even if I'm inside of uh cursor, I'm like, you know, what could go wrong? I mean, yeah, I know things could go wrong there as well, but having it literally in my terminal, it just feels like they have access to pretty much everything. And the model can do things where I things that I will not understand or might never even find out. Well, I mean, there's always putting in the prompt like do not change anything. You have to review stuff, but you there and and this goes into the SDK actually. So you can broaden its access through MCP through you know uh tool calls and things like this but you need to give it that access. So it's only going to do it within the within the confines of your codebase. But you're you're right to be cautious. I think everyone should be cautious really. It's um yeah it's it's not something to be taken lightly especially with these you know contacting authorities or like you know writing better code if you threaten it. There's really a humanistic aspect to writing. For the record, I'm not being cautious like at all. I'm just like I I kind of have to play that role here. But in reality, I just paste whatever and, you know, live on the edge. Vibe coding, baby. Vibe coding. That would be a great segue into our our thing on uh Rick Rubin, but we'll leave that to the end just to chill out a bit with uh with the news. Um but yeah, it's pretty cool. The other one then that we have another another claude ananthropic uh thing is I don't think we've got anything to look at but it's more of a discussion piece on the fact that it's it's got voice mode and it's actually not too bad. I wanted to actually I did download it. I wanted to actually try it live on on this to see like what Claude's our coding buddy. Do you know what I mean? Claude's our coding buddy. So I wonder why they've just released this voice thing. I think it's kind of expected, but I've um because they they don't have an image generation uh generation tool. Not at all. Like nothing. None. I haven't I haven't even looked to be honest. But yeah, if you say so. Yeah, they they they've been very focused. Um and it's interesting that now they have that. Well, should I try and um Oh, by the way, as well, they're giving four months of Claude 4 away for free. So, if you if you just invite someone, I think just even to sign up. Can I put this in the chat? Um, you can just uh you can just we we both get like months of claw or at least you get entered into a draw. But anyway, I'm digressing. So, yeah. Um I wanted to like see what it does when you ask for like write a simple JavaScript function that adds two numbers together. Okay. So, it's not really the same as Did I do that right, guys? Did I do I thought I would I would be in conversation with it, but all that did was dictation. Uh, I also thought that they have conversation mode. Uh, that's what Yeah, maybe maybe I did that wrong. Maybe I've just messed this all up. We got web search as well. That's an interesting kind of thing um that they've they've allowed. So yeah, may maybe I don't have it. Maybe they maybe they've I looked at that microphone this morning. I thought, "Oh, I'll do that. I'll do that." Especially for the guys on Command AI, but uh yeah, that was a failed experiment. But yeah, they have a chat thing, which kind of again goes against this idea that they're doubling down on code. Like they're still doing little bits and bobs to allow you to do that. I know a few developers prefer to explain their prompts to the oh god to the AI um you know during u copy it into the prompt and whatever but you know how do you do it? Do you like type it in or do you dictate it? I type it in, but I'm reworking every I don't know whether you follow my LinkedIn shenanigans, but I I watched a really interesting podcast where and I've made a made a short on this on Tik Tok on this um about the fact that product managers are really well placed as to become vibe coders because they're so used to explaining or gathering requirements, the right requirements, and explaining them to someone or handing those requirements off to someone and just not really caring about the implementation. just making sure that what they get back is exactly what it is. That's just my theory, but it was I would say well I thought so. Well, actually I'm saying this my shorts are like 5050 in the likes versus dislike ratio. I don't know what it is about shorts. Do you find this? No. I like my dislike ratio is like No, my it's like always 90% people love me on on shorts, right? Uh, I don't know, but like generally I look at YouTube, it's like 97 99% I can check it out. Yeah, I'm talking shorts. Like for some reason I've got more dislikes than likes on some videos. But I think it's because again I'm just making these hot takes. But I've been thinking about this the last few days. If if you're let down, if you're not, you know, getting off your butt and watching a full-blown video, like doing actual research on something, I hold your opinion on disliking information very lightly. If, uh, if you're disliking a short video, you know. Anyway, point is, um, yeah, I thought it was a dislike is, uh, 84%. So, yeah. Yeah. Yeah. Yeah. Um, bad. Yeah. Anyway, it's a semi-h hot take. I think that I think that managers are I I say it's a hot hot take because I don't know if all Okay, I don't Now, what I'm going to say is probably offensive to some people. I don't know if all product managers understand what they are building, especially of course the ones that I've seen, a lot of them have been good at managing people. They've been more like people managers than product managers. I I think that's like a better name that would fit them because a product manager is to me somebody who understands the product deeply and to some certain degree understands design and also uh engineering. Maybe the best product manager would be an engineer or a designer who had to become a product manager just because they had to run a team and you know manage other people. But I certainly know product managers who do not understand the technical part uh and they don't even understand the design at all. They don't even have a process for it. They just do whatever you know the designer and the developers do and they just kind of like fac they they just manage like these people talking together and making sure that they they say we deliver it until Friday and they just make sure that they deliver until Friday and if there are issues they try to fix you know whatever issues there are but nothing technical so that's the typical product managers that comes to my mind But definitely I've I've read some stuff about like Apple level product managers where there are no like not all of them obviously but the good ones are not you know just product managers they are like engineers and designers and design engineers turning product managers because they have to well I mean there's that there's an interview I've seen with Steve Jobs where he says some of the best product people he's ever worked with are those who actually have done the thing and now they've had to move up into that thing like that's that's his that's probably where this is coming. I'm hallucinating. This is what also Claude is saying that we areating more than AI. Um so I mean but the thing is that mentality that idea is is not alien and I think that exists across the board. I know plenty of designers who don't think about UX as an example or developers who have no idea about mostly UX, you know, they have no idea what the product is. They just think about the little function they're building in isolation and that's all they care about. So this this mentality of like, you know, you go deep into one whatever your discipline is is not crazy. Uh but at the same time, isn't this the perfect candidate for vibe coding? Is is people who don't know or care? I mean, yes, they got to have a lot of them do have product mindsets. I mean, but that's that's why I think my hot take exists is because I do think, yeah, you're going to get some crap product managers. Not every single product manager is just going to be the sickest vi vibe coder that comes, you know, I get the idea. I I I think I understand like at least they can I mean with that mindset anybody can start but yeah product managers who even if they have just a sheet of features and whatever even if they don't understand it they could still wipe code. So yeah, sure. Sure. I wanted to give a shower because all this to be said um it was a go history. It was a a podcast that I or show that I watched that gave me this idea and it was basically a couple of project managers or product owners. one's a CEO actually um has a really interesting um uh let's see if I can um it's the uh how how I AI and it's this guy he's the CEO of a company I looked on his on his LinkedIn and basically he's got three tasks let's go forward here he's got three tasks create PRD uh generate tasks and task list and these are basically ally he set up what it takes to create a PRD and he asks what is a PRD uh uh product request document or something like that um let's have a little look product requirements document okay and this is a conversation piece then it's basically saying look here's what I think makes a good product requirements doc ask me these questions back and then the AI does that and out of the back of that generates tasks follows the generate tasks markdown file inside of cursor to generate a task list and then um and then actually implement that task list. So, it's a three-step solution to building out a product or a feature or something like that. And it's it's through a conversation. And that just that was the thing that got my cogs turning of like I think and it goes into our conversation that we had last week of we really I think oh no it was a conversation I had on the weekend. Sorry. Uh I made some shorts about it. the the role of sort of mid senior developers are going to be more in the world of really taking their time and thinking about their uh their prompt however they come up with that prompt. It could be a process like this like generating markdown files and all the rest of it. Um oh and actually you've got a he's got a tasks folder here which it dumps in every single task that it has to do. It dumps in. So he's it's got a historic thing of everything that goes on. So really organized in that way. That that's very interesting to see there. I see this uh MDC. I searched it. Is it like a cursor file? Uh but for what is this exactly? It's your it's your project rules. So you got global rules which exist across every single project. And this this this is you know kind of basic cursor usage that you've got a bunch of global rules. how you like to write code, how this but then you've got project rules which is like which override then global rules. So as an example I like Nex.js projects but I've had a few which are I made a video on this by the way shameless plug um which are uh you know vite or vit uh react projects. So I'll say this is actually a react v project because it kept on writing code that was next.js JS and I was like, "What? Why do you keep doing that?" And I was like, "Ah, it's because I haven't set a project rule to to uh do this." So, these are just project rules. Um, realistically, these are just markdown files. I don't I think you could just reference they don't I don't think I'm just thinking on my feet right now, but I don't think they need to be in the rules folder. I think you could literally put these in any folder and say, you know, reference that file and say, "Look, this is how to create a product requirements document." But regardless, these are product owners. These are project product people. They're not developers and they've got a he's got a really sick process for writing code, you know. Um, and I don't know what his success rate is, you know. Um, but it's the conversation I had that it's checking things off as well. You can see here when it's doing these tasks, it's checking them off. And by the way, this is what Claude code does as well. So Claude doesn't do as thoroughly as this, but actually creates a task list and checks them off as it's doing it. That was just a little nugget that I This is very interesting. I cannot actually like wait to you know finish this and yeah all of this and I will I will set this up because this is very interesting what I've been doing with my cursor because I have a developer um who's like writing a lot of code and whenever I want to also post about it I used cursor to explain things to me. So it's not just to write code but explain things to me. And I was thinking if I should store everything in one file and I had no idea about this but now it gave me the idea that this actually exists and I can store things in a just text doc. I never thought about it like this because I'm not coming from this developer mindset. I never used a code editor as you know just basically a folder and put all the data and documents they want in it. Yeah, but that's obvious use case. Yeah, and again this is coming from product people. But the thing is, you know, no one really knows what they're doing with this stuff. Everyone's got their own ways of working with things. Like that's the thing. You go for the one shot and it's no one knows what they're doing. You know, AI's got a mind of it. You're basically rest you're grappling with a with a wild horse and you're hoping that you can tame it a little bit. But it's nice to see these sort of workflows coming out of the woodwork and uh just yeah just interesting ways to try and better your coding skills. But I do think we are generally going in the direction of like your time will be spent as a developer will be spent crafting the prompt and making sure the requirements are there which is very much a product manager's job. Um and that and then the other conversation I had on on the weekend was about like you know we we we focused loads on what's happening with junior junior this junior that I don't think juniors are gone. I just think that role has been completely redefined and the the level at which the thing that we call junior developer is just now miles above what was previously junior developer. We expected junior developers kind of coming in and you know adding a little button here or header there and slowly I think junior developers are going to be coming in straight away writing code either with well with AI whether they're copying and pasting code from claw.ai AI or whether they are literally using um you know wind surf or cursor they're going to be actually writing code and creating pull requests on day one and then it's the senior devs who are then able to maybe talk to them about what their prompt was you know and then and then and shephering that code through to to production and that's where I think mid and and senior are but what do I know I'm just a guy with a hat that's actually an interest this is very interesting because I had a question written down actually in our code and I in our doc and I wanted to ask you this. I wanted to ask what would be your suggestion for somebody like me who has just like a little bit of understanding of code. I can like read some JavaScript but I I can write maybe a simple function but nothing more than this. Without AI I'm basically I I can't do much. And I was thinking what would be a a good project for me to start something that I could finish in a day or two with the help of any AI to explain things to me but without like copy pasting or you know the agent writing it for me something that basically gives me a grasp understanding of maybe a little bit of back end obviously I'm a bit more experienced with front end and you HTML CSS it's simple to me but you know something a bit more complex with J with JavaScript maybe I don't know what would be your suggestion and I know you're not prepared for this but mate I got it I got it leave comments down below in the chat on Twitter wherever you are what what you think Basa could do should do or whether you think I'm talking trash okay first of all turn off AI inside of cursor VS code whatever whatever you're using, just don't use that. Second of all, well, only work within the confines of claw.ai or chat.open.com, whatever. Just work within that confine within that environment and create your own testing environment. Tell it that you are learning JavaScript. You want to build a simple project and you want to slowly build it step by step. But you don't just want it to give you the code. you want it to explain what it's doing and create your environment that you want to learn and if you need it tweaking maybe it needs to be a bit more granular maybe it needs to be a bit more um you know use different words for I don't know but just you you create the environment that you need to learn and write the code yourself that it's don't copy it but write the code out yourself you know you might start with like uh like a boilerplate thing like V react project or something just just just to get something on the screen. But from that point onwards and then tell it like tell the chat like I've created a boilerplate project so this is what we're working with and then from that point onwards write your code and then and then understand it and and then also tell it to give you tests like what happens when you console log this. What do you expect to happen? like tell it to give you um yeah a test after every sort of what do you call it module or or um yeah module or something. What do you reckon? Yeah, that sounds good. That sounds good for for the project. Should it be some because you mentioned React, but should it be something like a React app? I I know like the very typical one is a to-do uh app that everybody does, but I I want to go a little bit beyond that or a bit diff I don't know. Maybe maybe I should do that because everybody does it for a good reason. I thought maybe something like with login would it be too complex for me to start doing? Yeah. Yeah. Don't worry about like or and API requests and things like that. Like it depends really what your goal is. I think like it needs to the whole point in a to-do app is that it's just something that's that you just learn the basics. That's not the point. You obviously want to be inspired by the thing that you're creating. There's no point in making like a to-do app if you're just not going to look at it and be like, "Oh, that's really cool." Like make a make a website of something that you're really interested in or build something you're genuinely interested in. Yeah. So, yes, exactly. I'm very jealous of my developers. I have two developers. They are working on like different things. One of them doing like more front end but also a little bit of uh like backend quite like just building apps, functional apps. And I have another developer who is doing um like very creative stuff. So WebGL, GLSL, um 3JS, very creative stuff. And I know there are like two different words, but they share stuff with me. They do stuff and I I get the code working on my machine and I play around with it inside of cursor and I get really jealous of not being able to understand it in a deeper level because I genuinely want to start and build like help them. I want to contribute and I find uh like a lot of joy in it just like working with them, learning from them and they they just they know so much more than me and I want to kind of like contribute to that and yeah I want to that that's why I'm like playing a lot nowadays with JavaScript a lot. I I spend quite a lot of time in uh cursor prompting sometimes just going into the code and changing stuff, breaking things, getting in a call with my devs and I'm like, "Oops, sorry. I I broke things. Can you help? I broke our code base. Prod is down. Whoopsie." Well, they they told me to not, you know, to be careful with my git commands. Um that that's about it. The rest is fine. So yeah, don't turn on max mode. Don't turn on max mode. Yeah, I mean like I think you're definitely in a in an advanced coding state right now. Like there's some great platforms out there like that you can code in isolation. Like even code sandbox or maybe code sandbox is a really good place for you to not even open VS Code or anything like that. Like just create a boiler plate inside the code sandbox. Good point. And then and then you're not going to disturb their code or or import their import the project into code sandbox, you know. I think you can do that now. Surely. But yeah guys, it's very interesting like I'm genuinely very happy with you know what AI is doing for learning code. I know a lot of people are unhappy with what AI is doing in writing code but if you want to learn code it's just so much fun like you can you can do so many things and if you get stuck it's not if it's when you get stuck because you will get stuck you have AI you basically have a good engineer or like good enough engineer for this type of things to help you troubleshoot or like yeah get get things going essentially that because that was my issue. I started learning code probably 10 years ago but I stopped because I would get stuck and I would I don't know get myself busy with something else. But now you it's really hard to get stuck for too long. AI will help you to learn to to to know like what are you doing wrong maybe especially if you are doing something simple and then just continue learning. Yeah. Yeah. The learning aspect is is just incredible like you and this is that was the other point to my conversation with with um my friend on the weekend. It's like the the like we're in this weird middle ground right now where people come out of coming out of university are still in this void of what junior developers are currently uh you know defined as but it's the the learning is now expedited. So they they'll very quickly learn things and and even on the project or you know in the if they're employed or whatever they they they need to know that AI is there to help them learn and um and learn in a complete different way that's bespoke to their needs whether they've got learning difficulties whether they're visual whether they're auditory whatever AI is there to help. So, um, anyway, we have also got, moving on, we've also got, um, Vzero entering the chat. Uh, they've released their own model, which is really cool. And what they've done here is kind of go for the V0ero 1.0. And the MD is a little nod to Tailwind. I don't know if you know Tailwind, but it's a it's a little nod to that. And um yeah, basically it's uh it's it's what they've been training their model on for ages on the tool v0. Uh you can access it through open API open avi. Oh my god, it's just open AI API. Um and it's really good at Tailwind Shad CN um Next.js projects of course because it's it's bread and butter. So this, you know, whilst we talk about Claude 4 and all the rest of it, this actually could be a solid pick for, you know, building next year's projects or immersing your world in what is basically becoming the de facto um coding project, which again, faking Theo uh released a video that I was going to release, which is basically react one because all the AI now is just releasing React code, you know, next. What's half of that? That that was very interesting. But isn't it like what was he saying um is based on the premise that AI won't get significantly better than what we have today because if if these AI agents can properly think and properly read documentations and if we produce better swelt documentation or just you know the team behind swelt they they just have documentations uh can't can't like these AI tools look into those documentations and turn any React app into a swelt app. I know that it's not that just that easy, but can we like predict that this won't happen in the next I'm not saying like next decade but in the next one to two years. That was my problem with that video or like my question. I couldn't grasp that uh very well. I think he I think he had a few points and to that point I think that you just have you have to be very prescriptive that you want to write a spelt project but the code base the code that it's written in is just slowly getting abstracted away this is these are my thoughts now right so the code that we're writing is actually slowly getting abstracted away through AI it doesn't matter what as long as as long as the code does the job I mean realistically ally people only choose certain code bases because they like the way it's written or it's it makes sense in their mental model on how to build a website out of it or or what it does from a feature set they're all largely the same. I mean you know there are nuances here or there but at the end of the day what what gets output is a website or a web app or native app or something like that. So, and with that, any project that you ask AI, Lovable, Replet, all of this stuff to spin up a project, it's all React. And then because it's sort of self-perpetuating, we're going to we're just seeing more and more React projects being created that we're going to get to a point where most of the information, the data that the AI is learning off is a React project. So, that's where I was kind of coming from. It's like wow like every project I spin up right now is is a React project. I understand that and and even uh Theo he mentioned that even even if you ask AI to use uh another library the code that it the AI produces is sometimes going to either look like React or just be React and hallucinate thinking that it's writing another library but it's just React because the data is just so much React. What I'm saying is can't can't we check against the you know documentations properly the a the AI output can it be checked check checked against you know the the documentations and if it's not proper syntax then you know to give us the proper syntax you know what I'm trying to say here I'm probably I I butchered this here so you you you ask if you ask your chat GP PD to write you a swelt uh swelt app. The chances are that the code either looks like react or is just literally react and it the syntax because it's so there are so much data around react that these AI models might even like it gets it's getting fuzzy for them. They don't Yeah, I got it. I got it. I understand. So, so these thinking models and also we have um search they can search the web can't they search the documentations of swelt and write the code based on that documentation in instead of like relying so much on just producing based on their learnings but have a sure kind of like heavier weight on on the documentations that they are reading. Yeah, I got you. So you're you're basically defining some sort of like forget your learned behavior. I'm telling you where you're where you should get your information from. Right. Exactly. Yeah. Like a set of rules, but like a big set of rules. The the problem with that is is that we we refine models and this takes time to refine a model. You're basically asking it to not only go off and get that information, but infer that what that information is. is you're basically building your own model in that in that way. Um and just completely ignoring any sort of knowledge of coding that it does. So I think you would have to I don't know I don't have the answer. I wouldn't say ignore it. It's just like whatever that you are going to output that it's a thinking model basically. So whatever that you are going to output check it before like outputting it check it check the documentation and re reference it so you make sure that the output matches the documentation requirement that's basically what I'm saying yeah and what's the goal in all this just to write this really interesting no this is like a really interesting question because I've it's been on my mind as well what I so the way I would put it is performance. Does it even matter anymore? Because we went from, you know, there there was a time programmers would access like CPU literally they would write into RAM to the random memory like literally they would hand code to write into the memory for best performance. But we abstracted away from that step by step and now we are on a level where we say well spelt was better um than react it it was much better for performance but forget about all of that let's just stick with react and not just with like react at its best at its peak uh stick with react and like the general average code at best. So forget about performance. So now we are going to have probably tons of you know code and websites and apps that are not optimized at all. And my question is does that even matter? Like actually to a point because you've got Facebook who I'm doing a lot of research on styling libraries at the moment. Shad CNN, Tailwind, stuff like that. Um, Facebook created their own styling language called Style X because they're trying to they're trying to distribute to millions if not billions, I don't know, millions of people and every single kilobyte matters, right? So to them it matters to e-commerce sites it matters how what the performance is like the the how fast the user can buy a piece of clothing or whatever it is. Yeah. So, it really does. Um, but broadly for 90% or 80% of the websites out there, probably not. No one's going to kick up a fuss if my personal website is just a little bit slow. Um, you know, yeah, for for websites that don't get a lot of visitors, I get it. But well, my website, my website get tons of visitors. Well, personal website probably not. Same for my personal website. That's not what I was saying, Kabad. I wasn't saying. So if people are not going to visit your website, it doesn't matter. So it can have as much code and you know shitty code as it can, but it doesn't matter. But it I what I want to say I would say actually it matters even on your personal website if there is any conversion to be made. Yeah. Well, there's still conversions to be made, but like the the negligible speed difference whether you use spelt or next.js is is it's it's not really worth it's it's it's more worry to to decide on like to nitpick around that than it is just to spin up something. But I'm never going to be the person to spin up something. I mean, maybe like lovable and and replet. They're re those tools are really good at saying I want to build this thing. Go and build it for me. Cursor winds surf they're not so good you know for first of all you need to create a folder in your um you know on your machine and then open that in cursor right that's that's already such a minuscule uh point of contention that you have to even do that then ask it to just build an app from just that folder. Not so good at doing that. I've never had good results. People some people might have. I'm not going to claim I'm not going to claim that it's not good at it, but if you just go to Lovable just says your prompt, it's got it all there and you can have something, you know, hosted in in half an hour, you know. Um, but saying that, yeah, I'm going to be nitpicky. I'm going to be nitpicky about the software I use. And they're always going to be those, you know, take cars as an example. There's always going to be be people who want petrol cars, who want um manual transmission cars. They're always they're we're going to be we're going to be relegated to hobbyists, those who care about what framework they're using just because they want to they want to get in the engine and and and you know nickel and dime the um you know what there are still people who own horses but yeah yeah I am a horse I am a horse you don't know it but I am actually a horse. Let's keep moving right because uh we're narrowing away here. Well, anyway, Versel, oops, I've just uh rued that. Reversel, they've released their own model. You can access it through the their API. It's not open. I looked at it's not available on open router, so it's it's not so easy to get hold of, but it's uh it's available to use. I may get around to using it. I'm not too sure. But to round off something that we spoke about a few weeks ago is um Duolingo going dark. I don't want to spend too much time on this, but it's only because we mentioned it a few. Can you like explain to me because I saw the title on Twitter as well and I have no freaking idea what this is about. So a few weeks ago uh you might remember we spoke about Cler going allin on AI and then them retracting that decision. Yes. Duolingo at the same sort of time has the the CEO I've forgotten his name uh Louis von an or an uh said that they are going all in on AI doing the whole all in on AI thing and fired a bunch of people and and whatever and the thing that triggered this for me was I saw a video about it right um and they got a lot of backlash for it because they fired a bunch of people and people are noticing that the app is going downhill you know the stories aren't as good. So since then, does that clear part one of this saga? So part two of this saga really is they've basically I mean they're a trolling company anyway. I don't know whether you've ever used I I did and but I know like the the way they communicate it's very funny. They're their social team social media team they are just very good. Yeah. So, coincidentally, similar sort of time that they get a lot of backlash for going AI only. They just delete all their tweets or the Tik Toks, they delete a bunch of social media stuff and uh they while leaving a cryptic message gone for now. One, two, three. That sounds like a password to be fair. So, maybe people should think about that. Um, and they've just kind of disappeared from the from the spotlight even though littleie here is in the spotlight. So, it's it's a it's an ongoing kind of thing about what they're doing. It's just interesting to see how people react to AI. I'm still trying to figure out why people are so against it. Um, I know why I do know why some people are so against it. We talk, we spoke about that podcast that went out on Diary of a CEO last week, but most people just disregard it once they hear the letters AI. They're just like, "Nope. Yeah, [ __ ] off." Um, so yeah, this is this is in I asked I I was chatting with this thing. I think that's the guy. I asked if it was if there's any updates on it. Um, God, I love Perplexity for that. Uh, there's no new updates. We're still we're still in the dark with what what I don't understand is like how is that a CEO of a such a big company makes a claim like okay we are going all in on AI like isn't it obvious like everybody should use AI obviously it's such a thing to me to to me and just making an announcement about it it's I don't know it's just pity at this at this point we're the mid of 2025 AI is everywhere. Everybody is like trying to figure out how they should adopt AI. So, it's very obvious. I don't know why such an announcement even exists and why should they make such a huge thing about it. It's the same it's the same reason why like I a lot of my posts uh talk about vibe coding because it's a hot topic. You know, it gets it's why share price at Apple went down because they're not talking about AI. I think it's just making headlines to um for the sake of making headlines, you know. Uh but I can't speak for that. I can't speak for why he made such a big deal about it. And and what and and I know the deal like I don't know whether you've ever worked for a big company, but what a CEO says isn't exactly what goes down inside the company. They like to say it does, but you're talking about a deeply rooted um culture thing here for the CEO to just assume that everyone's just first thing they think of as AI. For instance, we talk about this uh green screen background behind me. For those that don't know, I'm just going to show off. Um the like I sent a tweet out earlier saying that uh Oh, green. Um, I sent a tweet out earlier saying that uh, you know, how do I create a background, right? And even then the AI thing didn't pop into my mind. Of course, I did go down the AI route when it was like, oh, I could just use AI. But the point is um, you you you're asking for a deeply rooted uh, mindset to just think AI first. Saying that, I think they they probably used AI in some pretty clever ways. um pretty cool ways uh at Duolingo. They've just not they ain't got their prompting skills. Their prompt engineering just needs to be a bit better because their stories are just not inspiring people. But yeah. Yeah. Um I don't want to talk about Dualingo anymore just because it's uh yeah I'm not too interested in that. I just wanted to wrap that little uh what do you call it up? Um let's go. What do you want to talk about? Well, we have Builder AI and the you know, the last thing that we Yeah. Oh, yeah. And Wix. Let's go with with Wix. Um, they generate they they acquired a video generation platform. I have no idea why they would do that. Um, exactly. I think this is why it was interesting to me. Like why video? And this quote made me laugh. Uh, where is it? Right. So something about basically we want to maintain our lead or something like that. Uh uh reinforces the company's position as a leader in generative AI. Wix Wix since when? Wait, what? What? Did we miss that? I mean whatever, you know, it's just again going back to the CEO thing. Just the stuff that the CEOs come out with are just so funny. Um, yeah. But anyway, I wouldn't go I wouldn't want to go against um Google when it comes to video especially because they they own YouTube and I don't know maybe well remember they acquired a company here. They're not they're not claiming that they're, you know, this company has just seemingly been solely dedicated on um video generation. I don't know. I haven't looked into what our one is actually what their product is. I mean, maybe we could do that, but like Yeah, I know what you mean. Like Google just completely won last week with VO. They did. Mhm. Um, have you seen like video produced by uh V3 on your feed? Not that I know of. Okay. This is this is getting crazy especially because yeah, what you're saying like not that you know of. Um. Mhm. Yeah, it's getting to that point definitely that it's getting blurry and difficult to to to know, especially at the first glance if it's AI or not. I think we've got uh I mean, this might be out there. Again, I'm not going to claim to know everything, but um we've got tools out there that can tell if a piece of text is written by AI. I bet there's tools out there that can tell if AI is in videos. I think YouTube already are probably already on it because the amount of AI stuff on there is probably exploded. So, they clamped down on automatic detection of AI generated content. But at the end of the day, it's the context. You know, some silly little video, some silly little home movie, you know, generated by AI. Do people really care that it's that it's AI? No. But if it's uh Donald Trump being caught doing, you know, blueberries on the balcony of a strip club. Yeah. I don't know what that is. Do you know what I mean? Like if it's if it's controversial like that, then people will want to know, you know? Um uh I mean, we're terrible for it anyway. We're a sucker. I've started watching this video about we're a sucker for convenience. Absolute suckers for convenience. And if if it goes right back to the days of the headline, you know, when newspapers first came out, we just want to know what the t what the article is about so I don't have to spend any time reading on it. And now what is the hook and we do it on YouTube, we do it on this and that and the other. So um I don't know where I was going with that, but the point is is that we are we are we need to be more resilient. I guess we need to do our own due diligence. We need to unlearn unlearn our our desire just for convenience all the time. We need to be more robust in that respect. I think that's probably where this video is going that I want to watch. Um but yeah, it's been get definitely getting harder to tell. Yeah, that's one part of it like fake news and you know being worried about where we can tell the difference and not but the other side is people being really creative with it and creating like really cool stuff. Did you see the ad that I shared with you or not the ad it's a short film in a way about Yeah, I did. I did. Yeah. Yeah. The the plastic baby bottle. Yeah. That's crazy. And that's just like at its worst stage right now, right? Uh maybe. Yeah. I mean like Well, I mean, what do you mean worst state? That was hilarious. I mean, it's just getting better from here. That's what I'm saying. Yeah. I'm just searching on YouTube because it's quite There you go. Plastic Baby. I don't know. It was like basically it's like micro microlastics in his balls or something like this. Uh what I doing? There we go. Yeah. Uh, and then he he's like, "Oh, microplastics in the balls, whatever." And then he wakes up with a plastic baby. And um, very funny. It has like 10 all of this is AI. I mean, yeah, all of this is AI generated. Yeah, the the plot twists are insane. Have you noticed that people in these videos, they are like overexpressive with their facial? Like look at their faces. Even if they should be calm, they are not really this kind. I mean, have you been on Have you been on [ __ ] I shouldn't say shouldn't swear, but have you been on bloody TikTok recently? It's just people just like doing the whole like face humor stuff just to look stupid or speed up. I don't know. People more and more people are doing that is my point. Um yeah. So anyway, it's uh it's getting it's getting it's getting scary. It's getting scary. Um, I've just closed my window now. My my uh notion window for whatever we're talking about next. What are we talking about next? Well, uh, we have builder AI. Builder AI. Yeah, this is funny. This is not so funny. But, um, I think this is and interestingly, go back to here. Interestingly, um, so this was a company was a company past tense that just got shut down. They're UK based company and they were basically lying. It was an app development service, a no code app development service that were just lying through their teeth every single corner talking about how much money it made. It was a UKbased startup backed by Microsoft. um scandals involving financial misinterpretation, misleading claims about his technology and legal troubles tied to leg leadership. So everything you can throw at it. Um reported 300% revenue um I think in 2023 or 2024 just to obviously secure future investments. The AO powered no code tool was not AI power. They were just outsourcing it to India. I I'd be interested to know what the the interface was like for people using builder.ai. Um but yeah two hours. Yeah. Exactly. Yeah. Yeah. Probably something or just people just I mean Amazon did this, right? It's it's a Yes. Do you want to explain? So Amazon did this with their shopping. So they they said they have AI shopping with human assistance but that was actually outsourcing it to India and literally people sitting in India looking you over while you were shopping and making sure that you know they add up the the items that you are shopping. There was no AI. There was like probably 10% AI 90% people while Amazon was uh claiming that it's like 90% AI and 10% uh people. So, it was completely the other way around and they shut it down to my to my knowledge. Yeah. Oh, here we go. I'm just getting some comments in. Yeah. So, and and uh exc I don't I do not want to sound racist right now, but I'm pretty sure it was called Amazon Turk. Was that what it was called? What's it called? You cut off for a second. Amazon Turk. Amazon Turk. Yeah. Is that what it is? Uh no. Amazon Turk is a different service that Amazon it's such a it's a such a funny time to u to mention this because Amazon Turk there you could um you could do some task and make money as as a human being not as an because now now probably AI some AI agents could do that but uh Amazon Turk so pe I think it was for like some people needing assistant or just very small task like imagine Fiverr but much smaller um and from when they verify receipts and things like that like there was an AI element to it but they were there to verif the the employees were if you can call them that I don't know well Amazon Turk is a very old service and from what I uh remember it's like prior to all this AI you know generative AI and people were making some money there it's not a a lot of money and it's just a tiny bit of money and Turk is a device that I I if I'm not like getting this wrong I'm not sure you could also search it but I believe Turk is a device that kind of like a mechanical automatic device that does some automation just using some gears and mechanical arms. So that's a Turk and Amazon Turk was humans behind it doing some tasks and getting paid like I I don't know but something like five cents uh I believe. Yeah. Yeah. Lots of little manual tasks and whatever. I didn't know whether Turk was like whether I'd heard that's why I was a bit worried about it's a different Turk. Yeah. Okay fine. Yeah. So, um, and I had, you know, I had their website up and running on the wherever it is now, Build AI. This was actually down when I learned about this. When you went to build AAI, it was just a little message that they just evaporated and people were complaining about just obviously losing their data because this was an app. I guess you host your apps on them and things like that. Like, people just lost their whole business in a moment because of this scandal. Now, the website is back up. God knows if you can actually sign in, but there's no mention of any sort of AI other than in the name Builder.ai. So, I wonder if they're just resorting back to being a just a dev house that just some just so happened to be called Builder AI and only just noticed this just before the stream. So, yeah, drama. Your vision, your software, we just build it. Yeah. Wow. Yeah. sourcing partners, builders. So, yeah, exactly. Yeah. I mean, they could probably do that now and get away with it and like genuinely call themselves an AI first uh company. So, yeah. Um, build that AI wrapped in controversy, but this is uh let's wrap things up here because yeah, this is again it's a claude thing, but Anthropic partnered with Rick Rubin and you know who Rick Rubin is, right? He there was that meme that was kind of going around where he just he knows no ma musical instrument. He people pay him to basically just just vibe his way through music. And don't get me wrong, he's responsible for some of the biggest music musical decisions of the 20 21st century. Um for people who don't know, maybe you just Google it and show it because I didn't know who Rick Robin is honestly. Reuben. Ruben. Ruben. I I searched it and the moment I saw the pictures I was like, I've seen this guy. Yeah. Yeah. Um, you a music producer. Seen his pictures. Yeah. Yeah. Room is a music producer. He worked with Jay-Z. Um, what's that what's that intro line to uh 99 problems? I got 99 problems with the BJ one hippie. He like suggested that that the al capella intro to that which is just iconic. He was like, "Start the song with that." And Jay-Z did it. And you know, he's working with massive artists. Like, he's he's a big name in the music industry. So, what is he doing with vibe coding? I think he's just known for basically being Well, it's that meme that went out of his interview with like MV NBC or something like that where they're challenging him on his inability to play or understand music in any capacity. And he says, "I don't need to understand music. I just I just, you know, read the tea leaves or whatever he says. And he vibed it. He vibes it. And so obviously Anthropic leveraged that meme that kind of went around, asked him to help put this book together. And it's based on, as you see here, based on Loud Su, uh, which are meditations. They're short little meditations. And I love this website. Um, I'm losing my menu here. It did have a menu on the side, but it seems it comes up. Uh it comes up after you scroll a little bit. Oh, there we go. There we go. So, I love this how this is all code. This is all like, you know, line numbers here. But, um Lao uh Zoo wrote these meditations which just focus on the idea of serenity, uh relaxation. I've written some notes down here. Um creativity, uh technology, like ancient wisdom and things like that. And Rick Rubin just basically, you know, flipped it on its head a little bit and and made it sound like, you know, vibe coding, you know, um quotes here, art creation always starts with a prompt, a seed. Anything can work as a first prompt. Only you can decide that. I don't know what number that is, but like they're just little verses that you can just read. five minutes, you got some spare time, you want to relax, you want to meditate, you want to just immerse yourself in the world of uh So, so there is no teaching like in how to code or how to vibe code? No, no, no. You've just got you've just got like you've even got little like games and tasks here. Um and and well, I mean, yeah, but I do like see the code here. Look 3JS code. You can, you know, you can see how he made that or he regroup didn't make it, but you can come in here, you can just shuffle it and just hit up a new meditation, you know. Yeah, I really like these visuals. Chill out. Yeah. Uh, the code that can be named is not the eternal code. The function that can be defined is not the limitless function. The nameless is the origin of heaven and earth. The named is the mother of 10,000 things. There you go. Vibing out, baby. Written by I wasn't reading this. Sorry. I was I was reading my notes that I I found to be quite some interesting ones. Um but yeah, go and vibe out on Rick Rubin's un unorthodox join with Anthropic. Interesting. I think that wraps it up. I think that wraps up our uh But it's a nice website, isn't it? It's really cool. It is really nice website. Cool. Should we wrap up there then? Yeah. Any more for anymore? Uh let's see if we have any people in the comments section in the chat actually. Chat that that's how we should refer to people. Did we put chat? Oh gosh. Shut up. I'm just looking on Twitter. Like I wish Twitter had more of a more of an engagement thing on like a chat on there. Um there is a live chat. So I'm gonna let us know. Is there? Yeah, I just I'm seeing it. I just sent Oh god, here we go. Well, anyway, no, I'm not. I don't know how to see it. Maybe I just need to learn how to do it. But if you are on Twitter, come over on YouTube, join in the chat because it's hard to monitor all these things. Uh, let us know how it was on Twitter as well because it's the first time streaming on Twitter. Um, or X, sorry, I'm so used to calling it Twitter. It looks good. It looks good there as well, if I say so myself right now. Yeah. Well, anyway, yeah, no comments. So, uh, cheers for tuning in. Really appreciate it. There'll be a recap version of this, but of course, if you are not here live, get your ass over here. Well, we try and do it Thursday, Fridays. We're still trying to work that out, but at least once a week, we uh we try and let you know early and often. Um and joining the chat, joining the banter. Yeah. Well, it's been cool. We'll see you uh next week. See you next week. What was your outro? What's your thing? Was you I don't know. Like, peace out. Seriously, I just do something random and every time. Yeah. What? Whatever last video I was watching, I just copy it. Copy that. Yeah. Cool. All right. This has been Command AI. See you next week. See you on the other side. Bye. Bye. Cool. I've ended that. Cool. pound.