#62 Grok Bot, Spotify Xirp, Webflow massive announcement, Manus Splits From Meta, Meta Muse Glimmer

We're recording. We're recording. Awesome. Hello there. [laughter] All right. So, let me Yeah. What? What did you Well, I was going to say this is a bit different. For those uh thinking it might be a bit uncanny, something's a bit odd here. This is a I hate to I hate to tell you this is a pre-recording. We recorded this yesterday and because we're both away, we're both busy and overall this might just be a very very different episode. My theory it's going to be more conversational. I think we're going to be probably a little bit more re relaxed and um slightly less aggressive, but only slightly. Yeah. [laughter] Or more aggressive. I don't know. More. Let's go crazy. All right. Well, should we talk about what we're going to go over first? We'll do we'll do the intro first, right? Ah, we do the intro. Okay. To the show, I think. So, yeah. What are we What are we going to talk about today then, Kawaza? What are we going to talk about? So, uh, this week we have Spotify joining the de AI developer, I don't know, community. We never Yeah, we never thought we would see that. We'll talk about it. There is a really interesting conversation around Web Flow. Like huge news is coming potentially Web Flow 3.0. Uh we have a new product from Grock, from SpaceX AI. Not Yeah. from SpaceX AAI. We have Metam Muse. Um and what is the the one with Manis? I I don't know. Yeah, splitting from conversational. We'll figure it out. We'll we'll talk about this and all of this in today's episode. This is Command AI. I'm Kavarza and I am Sam and this is a week in AI design and dev. So, wait a second. Are we like publishing this whole video on YouTube as well? I think I don't know technically how it's working. We're raw dogging this right now, but I think this is a pre-recording inside of Streamyard and then we'll just schedule it to go out. I don't know how it's going to go on Twitter. I I thought you would just download it. Maybe. We'll see. We'll see what I'm not editing it. I'm not editing it. But um yeah, it would be nice if we could schedule it so we could join the chat. So if we're out and about, do you know what I mean? Because I'm gonna be out and about. I'm gonna be driving actually. So it might just be that I'm on my phone chatting to you guys in the chat. But we'll we'll we'll try and make it as live as possible given it's a pre-recording. Otherwise, it might just be an uploading. Let us know down in the comments. The chat is now the comments section, not the chat basically. Yeah, that makes sense. That makes a new world. We're touching new ground. Now we can go. Well, there is a new product by XAI, by SpaceX AAI. It's a Grock bot, and it looks pretty cute. I mean, look at this nice animations, by the way. And it seems like it's a desktop app. I'll show you in a second. I just installed it. And you have different bots to do different things. Uh, so they have different contexts and they have different probably personalities and so on. And look, look at this like this goofy guy. I don't know. I I like it. I think it's pretty uh funny looking. Anyway, yeah. Oh, you you see it here as well. So, message bots like teammates and apparently you can also add them in different places. And yeah, you you have basically a UI for this agentic type of work and you get the pricing 200 bucks [laughter] a month, but it says try for free and I just downloaded it and we'll see what it can do or like how it looks like. So, apparently I can some Hip Shot responses to it. It feels like there's an app that we I think we demoed on the show called Osaurus and it's a you can create different agents with different personalities. They're different dinosaurs. So there's a cuteness to it like Osaurus again creating different bots, right? Uh different agents that are responsible for different things. And then it seems like it's the co-work section isolated into an app. That's what it feels like. Let's see. Let's see how it goes. Yeah. So, it says it can do chat. It can build stuff. It can generate um images and I assume maybe videos. It has voice basically. It's chat. And then a bot. Maybe it uses the computer. Yeah. Uh a new kind of colleague with its own computer. So, I think this is the point. So, Grock gets its own computer. Um, it can sign into your tools. It has routines and it can work anywhere, desktop, mobile, which which is interesting. Uh, which BT app, but they've almost [snorts] branded it in such a way where it's like this is something different like it's got its own subscription. It's not your Gro subscription. That's that's true because I hope I was hoping that I would get this with the Twitter subscription because I have a we have a Twitter subscription to go live um because of that and I thought because with that I'm getting grock but with that I'm not getting grock bot which is you know um not very streamlined as what Elon says he wants to do, you know, everything in one app. It It's not like that. So, apparently, you can like choose from these. I can negotiate like I have an app that can negotiate and then Yeah, I basically have an app. Uh I a bot. Yeah, right click. Right click on that app. Right click on that bot because that was very simple. Can you customize the bot? Can you go can you give it like its own based prompt? Oh, maybe you develop it in the chat actually. Yeah. So, there is like a little bit of settings here. What are we negotiate? So, it immediately starts doing something and I can have routines per uh bot and they get their own computer. Okay. So, this is interesting. I could potentially tell it to go on eBay and message people. And [laughter] this is so dumb. Yesterday was the solar eclipse and I I went out to buy the the solar eclipse glasses and they were freaking expensive. They were sold out literally everywhere. I spent five hours trying to buy it. They were sold out everywhere. Just look into the sun. They call it sunaxing. Sun maxing like token maxing. Yeah. Yeah, I should have done that. Maybe it would have been less painful. Uh maybe I should have burnt my eyes instead of my wallet. So I [laughter] went on that's a quote. I went on uh we have an app in Germany, eBay client site, and it's like just Craigslist uh something like that. So to get secondhand stuff and there were tons of people reselling these glasses because they bought them ju just to understand how bad it was. I went to um a shop they said they ordered 500 and they sold them in less than half an hour. I don't know how they managed. Anyway, so long story short, I wanted to buy it like from these people and I paid €15 for four pairs. front door. Yeah. I hope that was a business for five for four of them. Yeah. For €25 each kind of like uh I bought two of them for 60 and another two for €4. And if I had how is relevant if I had this spot maybe it could negotiate better than me. I was trying to negotiate with these people and they were like saying bro like there are a hundred people trying to to buy this from me right now. So I wasn't negotiating only on money. I was trying to tell them that I can get there faster so they can get the money faster. I was negotiating on anyway. Um do do you want a pair of glasses [laughter] maybe? No. No. And what was the percentage of coverage that you had? Uh we had 89.6%. Fairly, you know, it's not 100% though. I mean, check my Instagram. Uh check my Instagram. I posted some on my story. They were it was beautiful. And I went there like it's I don't know. It's not going to happen anytime soon again. So with family and it's kind of like okay let's let's I mean you can chase them you can chase it you could go on holidays but it will cost more than a hundred euros. There it is says goodbye there. Oh yeah. Did you put your glasses in front of your lens of the camera? The first story. So if you go back to the first story Oh, look at that one. I just Yeah, that sunsets. That's a unique one. That one. Uh if Yeah. So go back because you kind of like missed it. Yeah. You're asking me now. Oh, see this is the glasses. Yeah, this is the glasses coming up. And these are without the glasses because then the sun was sitting. It was already uh low on the horizon. And these are just some beautiful gradients. Got to love a gradient. Yeah. Uh anyway, so point being, if I had that this bot, I could tell it to go there potentially and negotiate a good price for me. Maybe I would pay less, but then I would have to pay 200 bucks for the bot. I mean, it probably would have told you to walk away from €25 per glasses and that would have been also a negotiation tactic. They would have been desperate. They were they could have been bluffing. They might not have a big list of people. So, okay, whatever. Take it. I'll leave it. And negotiating against me. Exactly. Uh, so I I will make sure to set it up later on and u give it a try. Make sure to to subscribe if you haven't already because uh we will we will make a video about it. We will talk about it. It seems like um to me as you said you you you mentioned something very interesting um chat GPT app claw desktop app now uh the Grobot I think all of these are going in a similar direction it's a one experience app this is where you have automations you have your agents you have your chat you have your apps that you are building with uh chatpt I didn't know is one of my teammates just sent me a like a prototype he built on chat GPT hosted on chat GPT. I think we knew it but I never used it. Yeah. Yeah. Organization only tool so it was only available within organizations. They might have released it or he's just using No, I bought I I bought I bought the organization uh plan for my team because the 20 bucks for uh OpenAI gets you much more than the I think it's kind of to me it feels like roughly like the hundred bucks you get from uh Anthropic. M anyway anyway so all of these apps are going a similar direction and I would suggest you also installing it uh give it a try it it says you can try it for free see what it can do you can create a new bot and wait I just made a new bot what what can it do it doesn't have any yeah okay v1 maybe it's a maybe it's I thought like I would I thought there would be like a more kind of like um setup to it, but it's but it's a it's very instant. So, every bot to me sounds like a new a new instance of um a chat, but it's not just a simple chat because this gets its own computer and it can each can create a project and a website and so on. So, what's on the plugins? What's on the plugins? It's the only other uh I just clicked it. So, there there are a bunch of plugins, but aren't these MCPS? Yeah, exactly. It's a normie way of saying MCP. Yeah. Uh, so like different companies they call them MCPs, plugins and connectors. What else? Yeah. Um, I think Chach plugins too. Yeah. Yeah. Wait, why is Webflow here? Oh, it it might recognize what I already have. Maybe maybe it might just might work off the um agents like global file. Yeah. Or or Claude. Yeah. Or claw. Some of them are picking up can read them because it's strange web flow being next to you know the rest and also not having [clears throat] an icon. But yeah, I I think this is a cool concept. I will give it a try. Uh look with this. You see also the agent has its own computer. So I mean hopefully hopefully my biggest question here those who are paying for Super Grock, they get this. Hopefully they get this cuz that's $200 a month. Imagine paying another $200 a month literally just for an app that runs on your computer. I hope there's some Yeah, here we go. I might have done something really dumb. I might have signed up for Grock and not cancelled. Ah, see, look. It's in between. Oh, no. Look at that. Elon Musk is already getting money. Yeah, Elon Musk. Yeah, I mean, he needs it, so it's fine. But I can't see anywhere in all of these features that come with Grockbot. So, unless you can see it, it's, you know, it's obviously smaller for me than it is for you. I can't see Grockbot on there. So, if you're paying Super Grock Plus, you might have to pay another 200 for Grock. How is this how is this sensible? I don't get this. I I I think you wouldn't get Super Grock then. But but then no that doesn't make sense because you you want spend 100 bucks for chatbt you get essentially what chat GPT is plus you get co-work sorry JGBT work and all the rest of it there's something's not quite adding up here because that just doesn't make any sense well well I think maybe they are trying to I mean they are just trying different things but maybe they are trying to make this to be the main because it does it it does look good to me. It does look like super clean and you know with different bots and it does look useful to me like an all-in-one app especially that let me see this this page clean often to your computer not a lot of features it's got not not got a lot of features to it I mean maybe it needs to not features but like it will grow in time I think it will get bigger in time and more complicated [snorts] so presumably it can code. Am I seeing some code there? But Grock build, this is Grock build. Oh, okay. Sorry. Yeah, which is a terminal app, I think. I don't know. Okay. So, I No, I I I was reading this wrong. So, Grock build is its own thing and then we have Grock bot. But how does the pricing then work? Because no this is super confusing because look if I go to XAI/pricing there is no mention of Grockbot so it seems like bot is completely its own thing. Yeah. So basically what you what we $200 it's a bot app with it. Yeah. But anyway, it doesn't make sense for these to be separate offers. I'm pretty sure sometime soon they will merge them. Look at the bottom below below these two. What's that? Oh. Oh, yes. Already on Cursor Ultra or Super Gro Heavy, Grockbot is included. Yeah, good point. Yeah. Okay. Super Gro Heavy, isn't that $300 a month? Yeah. Yeah. So you get that's but they don't mention it here more powerful model. Maybe Grogbot needs a really powerful model and that's why it's kind of included in that plan or something and Super Grog Plus doesn't give [snorts] Yeah. But also my bet is that something like this needs a lot of tokens because it has it own computer. It looks at the Yeah. needs a lot. It's just not competitive though. You know, you can get Chachi BT on a $20 a month subscription with all of these features plus more probably. You can go for $100 a month and get clawed with co-work like this. Just I I can't see the competitive how this is competing really unless there's something we're missing. It feels like there's something we're missing, which probably is the case because we're trying to understand it live. I'm um uh you got cut off a little bit. Okay. Should we move on anyway? Okay. That's all we can that's all the juice we can squeeze out of that one. And we're back. What? Or maybe it's to come. Yeah, I think I mean maybe it's you, my man. Maybe it's you. Who knows? Should I Should I do um Manis because Manis is a agent? Oh, you're back. Uh I was Yeah. Yeah, we can go Manis. Uh that one, but I I haven't read this article. No. And I was going to go through the article. No, I've I I'll share my screen because I've got the I've got a new article because I noticed that that was a pay perview paper thing. I'm just going to go through the article and I'll I'll tell a little bit of a story. So, Manis, speaking of bots and and agents and things like that, which we always do, Manis was one of the OG agents and when no one really knew or understood what an agent was. We knew what chatbots were, but we didn't know what agents were and they're the doing. And it was Manis and Gen Spark, which were the two kind of OG ones. And I loved it. I actually used it. It was my first experience to like, whoa, I can get an agent to go off and do some research and then I get another agent to go do some wireframes and another, you know, it was a really exciting time when when Manis sort of first came out. And then Meta planned to acquire Manis for 2 billion. Now, this is a Chinese company that Meta planned to buy it for 2 billion. And I'll be quite honest, I thought they had bought it for 2 million because there were some changes that were taking place and all the rest of it. Uh you might recognize this logo here. And then I get an email and this email was a bit weird. Um and maybe I can even get it up actually. Uh excuse I'm going to use my voice because I've only got one hand. gmail.com manus. Cool. That worked. Okay, here we go. I'm going to try and not dox myself and I'm going to share this tab instead. Can you see the email? I need to share it now. Yes. Cool. I got this email and it says uh my subscription's been cancelled and I didn't know I had a subscription to be honest, but it was like this is unavoidable as part of Manis's transition to independent operations and to comply with regulatory requirements in specific jurisdictions. This doesn't make any sense to me. First of all, I didn't have a subscription or at least I had a, you know, an account, but I didn't have a subscription. And then this just told me nothing about what was going on. You continue to use your benefits at no charge during the date of backup period, whatever. And then this article came out and basically this is where we're at right now. This is where my uh my presentation fall short cuz I'm as curious as as you probably are as to what is going on. like it just felt very abrupt, very like kind of fishy. Could it have been Could it be something to do with the fact that they're Chinese and Meta is American? Um here we go. Founded in China in 2022 before relocating to Singapore. So maybe that was part of the agreement or or whatever. Uh the deal quickly attracted scrutiny from Beijing and Washington with Chinese officials investigating whether it violated the country's rules on foreign investment. So I wonder if that's Meta side or their side. Well, I guess Meta invested in them. The country's National Development and Reform Commission issued its decision in April instructing the parties to withdraw the transaction. So okay, it was on Meta's side basically. um kickstarted complicated unwinding process. Beijing has since tightened the tech export controls on crossber deals. So, it's bit of both really by the sounds of things. Um and China to compete for talent, hardware, and data amid an intensifying AI race. We need to back up our data as part of a separation from Meta. looks very political. But yeah, I guess that's that's that's all we have for it really. But it was just it was just really abrupt, very confusing, and I was like, what what's going on here? So [snorts] yeah, and we all know what agents are nowadays. I'd be very interested to know what one why Meta purchased such a thing. Uh two, what angle they have from a competitive standpoint. and three like no I don't have a number three that's it well meta like Zuckerberg has purchased like they they went on a hunt right they he tried to get his hand on the best researchers and he paid for like just some random startups in the AI and I saw like videos on criticizing like if it makes any sense to make to buy these companies And as we see like we have some news about Meta. He he's still trying. He's not giving up. Uh he's trying to be one of the dominant players. It seems like right now we have three American dominant players. Um OpenAI, Anthropic, and I don't know like lately it feels like Grock is becoming more I agree more of a thing. Uh Google kind of like we're still waiting to see what they will do. Yeah. Um and Meta is trying to jam themselves like to to to fit in the club. But it's so harsh that you like imagine they like they are probably spending billions and they come up with a model that is not as good. It's so hard to release such a thing like for all of these labs. Yeah. you are. Yeah. Okay. So, he's muted. I was just using my voice to to go to our next article because, you know, I love a good segue. And that's a perfect segue into Meta releasing their open- source model, Meta Glimmer. Now, we all know that uh Meta was banking on open-source being their entry point, their USP, their unique selling point into the world of AI. And so, they released uh Llama as their open- source model that you could run. It was okay. It was very fast, but it was very small. You're talking, I think it was like between three and nine billion parameters. So, a small, you know, not so intelligent, but fast model. [snorts] given the capitalist society that America and the western world is, they withdrew that they they could not get a a significant amount of performance out of Llama. So then they came out with the the closed source model Spark, Meta Spark, and that got released a couple of weeks ago, not even that far away ago. [sighs] We thought the open- source um ambitions of of Marky Zaki Zaki Marky Marky Zuckerberg were dead. But then two days ago they released Muse Glimmer um which is and if I share my screen now a open-source version I I I don't want to say open source version of Spark but an open-source model and you can see here it's 30 billion parameters and it's dense so it's not a mixture of experts um and what else can we uh good for agent end to end agentic task completion. Obviously, given that it's dense, it means it's got it loads all of its intelligence up front. So, it's quite heavy on your GPU. However, it's got quicker access to a lot of intelligence rather than um mixture of experts, which is I guess more efficient with all of that standpoint. So, reliable tool use, multi-step reasoning, failure recovery, whatever that says. I guess it's when tool uses fail. Multimodal, which is something we all really enjoy. uh especially when you're getting it to check its work and produce artifacts and things like that. So um and they're saying it works well with open claw and Hermes agent. So really quite cool. Oh and multilingual as well. So quite cool, quite nice. I did take a look at this before. I'm in Bulgaria right now and I spent all the morning before my flight trying to get this to work. And the first thing I did was look for uh an MLX port of a Glimmer. MLX is an AI that's more uh it's quantized and and I guess you call it quantized. Um formatted for Mac machines because I have a Mac machine, right? I tried to get it in open code to do some code. Did not work. I tried even I tried this and got it to work. This is just using straight up Python, but it means you have to put the prompt in the in the description here, which isn't really good for coding, but it worked. Wasn't really good for coding. So, I I use this to serve it, use an open code, did not get it to work, and I wasted hours looking into it. What I then just did is I think the if I go files and versions here, no, this is the raw model. If I Sorry, I'm going to use my voice again. Muse Glimmer. GGUF. Oh, there we go. There it is. Um, if we go into here, I don't think this is the one. No, it's not the one I'm after. Muse Glimmer. GGUF. There we go. This is what I want. Go files and versions here. They've got a couple of different ver. You can see that right on my screen sharing. No. Oh, no. No. Okay, cool. Here we go. They've got two different versions here, uh, GGUF versions. This is not optimized to run on Mac. However, it does run on Mac through LM Studio. And lo and behold, I got it working inside of Open Code. And um, the uh, code was It was rubbish. Didn't work at all. So, I don't think this is a very good code uh, very good model for coding. However, that ate up all of my morning and I didn't get a chance to load it into Open Claw or Hermes to see what it agentic capabilities are cuz that's what it seems like this model is um more geared towards more more trying to focus on sort of thing. So yeah, I don't have a real world um use case for it what it's intention, but I can tell you that its coding ability was absolutely nonsense. Although when you're testing local models, there's so much to think about. I might have had some sort of uh configuration not done correctly. But yeah, it built the app, but it just looked like crap and it didn't work. So yeah, that's kind of it. Um on the uh size front I'm still showing this tab. You can see that it's 19.7 and 16.8. This is quantized. I don't know to what level, but the if we go back to meta models here, go back to the raw model. Um and maybe files and versions here. Uh it's hard it's really hard to tell how big the actual model is in this standpoint in this uh format um because it's all just individual files. But the point is 16 and 19 gig is a model that can run on something probably as low as 24 GB of VRAM. You're not going to get a lot of context out of that. You're not going to be able to run it for those long horizon tasks because it's going to quickly just run out. But it's something that can run on a fairly low powered machine. I recommend something, you know, 48 to 96 GB of RAM for meaningful work, but this will run on something quite quite low powered, which is quite cool. So that's Muse Glimmer basically in a nutshell. Any questions? No, not really. I I you basically answered everything and I don't think I have a need [laughter] for it. I I'm using Fable to run the the dev server. So yeah. Yeah. I find these local models and this is something I've been running um kind of on my own channel. I've been using I I did I didn't get a chance to do it, but I I wanted Claude Opus to write me a spec that I would then give to Muse Glimmer. I ran out of time. So, I failed you all. I'm no good at as a YouTuber. I'm no good as a human being. I might as well end it here. So, see you later. I I I was thinking about this like if if for example Fable can put up but you need a fairly detailed um instruction for for something and then by that point you might as you might as well just use Opus to code it. Yeah, that's that's heavily debatable for sure. But yeah. Yeah. Yeah. I mean, if you if you make a plan that Muse could use it locally, you could just use Haiku at that point. I think it's kind of like similar in size. Maybe correct me if I'm wrong. It's hard to know because Hiku is closed source. But I mean, all that to be said, you know, we get onto this idea of like when is the correct time to run local models. I love it as a hobby. I really love it. I don't know why. It doesn't I don't use it dayto-day. I really don't. People keep asking me what my go-to model is. I don't I just like trying these new models. I love the satisfaction of getting it running on my on my local hardware. I can't explain it. But I think when privacy is utmost importance, redundancy, let's say you're um on an airplane and you need to get some AI work done, but it is going to be slower. It's going to be a hell of a lot slower than than these. I mean, we're getting to a point, we might be laughing at the what I've just said right now in in six months, but it's so much slower. And it raises the question as well. We talk about where Grock sits in the market of you know are you we we did we use Grock build on this on the channel. It was so fast. And that feedback, quick feedback cycle is often more important than letting Fable churn out uh an app in an hour and getting some bits wrong. Like what's you need to you need to only you can answer those questions at the end of the day and determine whether local AI is right for your use case. Some people just do not see it and that's fair enough. other people it's all they see because their organization will not allow them to farm off their user data to or it's a cost-saving it could be cost as well you know um so yeah it's all it's all the big debate and I think again only you can answer whether a local model is good I love it just because it's a hobby that's all I have to say I'm out of segways I'm out of segways plus I think well that's kind of it I think we've got web flow now Um, well, we have Spotify. Oh, yeah. Go on. Take us through Spotify. I'm interested in this one. So Spotify just um wait let me um so Spotify just surprised us all especially like the dev community because they came out with a tool for the developers uh know your systems so your agents can too. Um, so they have this new tool that to me from my understanding it's like two things. The how do you pronounce this? Zerp is it like would you say zer whatever? I don't know [snorts] how to pronounce this. Zer. Yeah. And portal. Uh so from what I understand you basically allow any agent that you use um to kind of like gain knowledge on your entire codebase or like business I guess not just codebase. So let's start from here works with claude Gemini codeex knows your organization not theirs. Uh, Zerp is a harness for the model you prefer. Switch between Claw, Gemini, and Codeex without losing context, which is interesting. Run sessions locally or remotely. Stay vendor independent while the context layer compounds with every session. So, it's kind of like a memory independent of which tool you use. Like if you're in codeex and you tell it to add something to its memory, that piece of memory is not visible to Gemini or CLA, unless you could technically set it up all of it that way specifically with MD files, but I think it won't be the same. Anyway, um so here this is maybe something like that, but on a much bigger scale. Uh if I read here, ZERP is different. It's an agentic development environment that understands your code uh connected to portal and it knows your services, ownership, dependencies and architectural decisions all of it in every session. Your agents don't guess, they know. This is interesting. So basically this portal is this a place kind of like a knowledge base for your agent. So from what I understand like you're not your agent connects to this or like searches against this and then understands okay this is this is the organization this is the project uh independent of the agent. So that sounds pretty interesting. [laughter] The comments on Twitter were funny. Like they were like, uh, you you give AI to developers at Spotify and they have like 2,000 developers and they are bored out of their minds and they are just creating other tools for developers. Uh because you know how how much more are you going to do with u uh with Spotify? It's yeah. So yeah uh the agentic development environment with institutional memory. So it's just memory but on a bigger scale with all of your agents. Uh other agents see the file you're working in. Zer sees the system it's part of. Who owns uh the upstream service? What depends on it? Why it was built this way? Every session starts uh grounded and every session makes the next one smarter. All of it sounds just like documentation. This is all documentations for all of your agents. Uh the context layer your agents and engineers share. So I guess again documentation uh and obviously okay documentation living documentation built automatically. So, isn't it just an MD file in the cloud and then kind of like a harness that allows or like tells the agents, hey, before you start any session, look at this or maybe I'm oversimplifying. Well, I mean, you know, this might not become a trend depending on how annoying it is and when I look back at this, but hip shot reaction is this looks like a just a branded um Obsidian uh framework. So, [clears throat] Obsidian is a MD uh markdown file um network, let's call it. is is a docu series of documentation files that you can interlink with each other and see like how all of your documentation here we go there's the nodes how they're all connected to each other how they all relate and it's and people are using this to basically document their lives and their folders maybe you maybe you've got one folder on your business you got one folder on your personal life um maybe within your folder of your business you've got uh onboarding sales whatever are all documented in markdown files and then you can see and interact with the relationships between them and this is simple tagging. It's just it's not complex but this went on steroids as many things did when you stick an agent in front of it because otherwise this is just a note-taking app. stick an agent in front of it and then you can start to really like oh what's our on boarding procedure for new developers or maybe you have meeting and m meeting things in there like what did I speak to um John about in last week's meeting and and [snorts] do we need to do anything on do you know what I mean it's a knowledge base it's a knowledge base exactly that yeah so it's simple one yeah yeah and it and you know maybe it doesn't need to be complex or whatever, but it feels like and this is a very DIY way of doing it. However, the last thing you scrolled to on Zer was this works with CL Claude Gemini. This bit here works with Claude Gemini and Codeex. So again, it just feels like it's [snorts] it's abstracted the agents doing the hard work, the heavy lifting. They've just created a framework behind it and potentially a UI behind it that Obsidian shies away from. Because as you saw, Obsidian is just a note-taking app. It's very, very simple. You You don't really interact with Obsidian. You just interact with your agent. Your front end is just your agent. So maybe that's what they're offering here is more interactive UI. That's what it seems like. I've I I haven't used it. Um Well, it seems like Yeah. [clears throat] Oh, go on. Go on. Go on, go on. I was going to say it's interesting how Spotify have um have done this, but I think you covered it by saying I think they're just bored. [laughter] Maybe this is something they've used internally that like actually you know what I think cuz companies are looking to you know bring a all-encompassing AI solution. They [snorts] they might I don't know whether they use claude internally don't know how sensitive their data is but whatever you know. Yeah. So what I want to say um you can join the beta uh join portal office hours. So every weeks every week they there will be a demo. So they will show it and that that's interesting that they are they are kind of like engaging they're creating a community and we are all you know here about community. So maybe if you're interested just just join the office hour and and and see what's up. M um I just had a conversation with someone um and one of the biggest things with any organization Oh god, I sound like a tech bro, don't I? [laughter] The biggest thing I've seen with many organizations is um is interoperability between their different systems. And I wonder how this and it I I say this because it just mentioned something real quick. you know, maybe the interface is the agent, but you you do you want it to you want it to interface with the different tools that they're using. Um, some of these tools are old and shitty. Some of them don't even have an MCP, let alone, you know, an API, let alone an MCP. So, yeah, I I I'll have to see more about how it integrates with it if they want to this to penetrate organizations because, you know, uh that's where probably the money is going to be in it really. Yeah. Cool. Anything else [snorts] on this? Uh, no, not really. It's mostly a question and I don't think I'm the the right person for this tool. By the way, this this looked uh like harp to me. Yeah. I mean, there's only here's another phrase for you. There's only so many ways you can skin a cat. Yeah. [laughter] Do you know that one? I've heard that one. I've heard that. There's only so many ways you can skin Pedro. Yeah. Uh I the the one that I didn't know was Balls to the Wall. Balls to the Wall. Yeah. Yeah. Um so yeah, Baby Jesus Christ also is [laughter] that's that's a very much a Sam thing. That's not that's not something anyone else says. [laughter] Yeah. Imagine I say this and people are like where did you learn English? [laughter] Yeah. So uh let's me actually uh are we moving to the next one which is web basically this would be the last one. Yeah. [snorts] Uh all right. So, I'm hopefully we're in the comments section and there's a chat to the right or left of this video, right of this video. Um, and this is going out as a quasi live event. So, let us know your questions and hopefully we can answer them in the comments or the chat. Otherwise, like and subscribe. Like and subscribe now. So, we're on a journey to reach a million followers. We're so close. So, just help us get to a million followers, subscribers. Sorry, that's my uh that's my plug for the day. Okay, so [snorts] I will I'm on your Instagram. I'm I'm trying to Have you seen my last video? It's going crazy like Oh, yeah. That be behind the scene. That's behind the scenes. [laughter] That behind the scenes. It It took me 40. Yeah. Yeah. Check the behind the scene. Behind the scenes. Yeah. This one is going crazy. Oh, BTS. The Oh, yeah. Yeah. [laughter] And these are all all like actually trying [laughter] different takes. Jesus Christ. Look, this is what I'm doing for content. Yeah, you got to sub subscribe like look look at the hard work. Um, while doing this, by the way, I had to [snorts] pause twice to catch up my breath because I was running across the room for the slide and I was running sometimes too fast and hitting the table. So, [snorts] wow. There you go. Dedication. Or you just repeated the same five clips over and over again to make it look like it. No, it's [laughter] it was 47, by the way. It's really dumb. Okay. Um, my monitor went off, so I'm I'm waiting on that to uh get back. So, there are a few interesting tweets that I want to bring up and mention because we're going to have a web flow conversation here, man. Uh, Twitter has a new UI and I I'm kind of lost a bit. Anyway, okay, got that. I can't. My UI looks the same. I don't know. It's to to I I saw No, it doesn't look the same. The also like when you write a draft and then you cancel, there is like a delete button that is red and it was never read and it just bothers me. I actually I'm a part of like I I like the algorithm shows me like the the design stuff on on Twitter a lot and people are so annoyed with like bad design decisions that I can understand. So one of them like Twitter is very unc inconsistent uh with with UI patterns within the app itself because everyone is building something different and like everyone is doing it in a different way but the worst is open AI but Google today took over and they are today the worst. Um we could show that as well but yeah anyway uh let me Okay, let me share my screen. Uh, I haven't checked the note by the way regarding web flow. I just went with what I had. Ah, okay. Okay. Okay. All right. So, Web Flow announced something really big potentially. Well, they haven't announced what it is, but they made an announcement about the announcement. It's quite big if you look at the patterns here. So they they shared this website with the assets and you notice something here. There is there's web flow app it app. Yeah, exactly. And there is a folder that should tell you something. And obviously there is the September 2nd. There is I think this is for being in person. I don't know what this is. This is the note. This is Yeah, this is the note. And then we have the drive. And what is this? This is the Web Flow app. You can put it inside of the That's fun. applications. And once you do, it brings you to the Web Flow conf to check it out. So what what does this mean? This means to me at least when I look at something like this a web flow app it means something like cursor to me potentially even lovable which we will talk about in a second or replet something along those lines and all of what I said sounds nothing literally nothing like web flow. Web flow has been this predictable UI with its own CMS, its own JavaScript framework, its own systems and what I just mentioned is completely different from all of that like cursor or or even um the chat GPT app, the the codeex app or the yeah I think this is what they are hinting at and this is what they are going potentially even like claw design a UI mainly driven by an agent and then with some UI for controls on top of it. That's I mean it's a it's it's an app it's an app that works on your machine. That's for my opinion that is a complete departure from something that always existed in your browser. So regardless of what the app looks like, the actual um delivery method of web flow has fundamentally changed, it's an app that's clear from this website, right? Yes. So it's files and this is this is something that kind of like became more clear to me as somebody who came from Webflow because when you work inside of Webflow you're not dealing with files but the all the apps that I mentioned all of them all of them are working with files you are like editing and creating new files and these new files these files that you work with they become your app they they are like pieces of your the the app the website that you are doing like HTML as a file the JavaScript all of these. So why does this matter? It means that web flow is potentially opening up because they know and this this is very smart of them if they actually go this way. They know they can't compete in a closed ecosystem anymore. that era is has gone uh and we we see this so let's talk about this one first so pixelgeeek you know him like I remember I joined I I joined the web flow train by watching a video from from him on Facebook that that's 2018 2019 era and this tweet is I I I think this is is a very interesting tweet. I'm trying really hard to keep it together. Life is just so up and down. Another client leaving Web Flow and will no longer need my services in the near future. I guess this is all part of the hard process of running a business. But I think there is more into it. It's not just about a business. We're specifically talking about a shrinking industry like web flow and no code alto together or maybe a no a new no code emerging. Uh here we have Robert saying, "I've had three le uh leave in the last few months, but then got a new client bringing a ton of WordPress clients to Webflow considering um issue with AI token limits, AI memory issues. I wonder if in 12 months how many will come back to web flow." And I replied to that. My reply was they won't. I don't believe anyone will. We are in a bubble, but there is no going back. There will be just more tokens. We're building infrastructures. We already have so much infrastructure. It might cost more to have to run the the next fable, but the the sonnet is ne will cost next to nothing in 12 months. I also see clients leaving web flow or thinking the whole thing rethinking the whole thing. Uh had four now this year by the way like one of my clients they are they don't want us to build for them they want us to build a to design a brand guideline for them. This is a whole new era new take for us. So that's why I started to pivot last year and now I officially have a client who's paying for a a complete web app I'm building for him and it has nothing to do with web flow. Nothing at all. It's just all Claude and also a little bit of uh Chetchvt and Sonnet and yeah I'm getting paid for it. This is the new reality and I I think like everyone like on the Web Flow bandwagon should actually acknowledge this. This is the reality. Web flow might manage to create a new tool. It it's just not going to be Web Flow as we know it. You might even call it something else. Maybe for the because of the brand and whatnot, it's still going to be called Web Flow, but it's it's a new beast. It's a it's a completely new way of thinking about this. And there is even more about it. Lovable just raised $400 million at a $13.3 billion valuation. That's insane. Web flow the last um valuation we know of was around four billion and I can't believe I I don't believe that they they are valued much more than that because the the market has changed a lot. So anyway, lovable that we were all laughing at last year to be honest is now valued at 13 billion. I'm not going to go through the tweet but there is something interesting about lovable lovable value at 13 billion where Figma the tool that we like if you're a designer and don't know Figma like what have you been doing like this this has been the default right they are now valued less than lovable that came around it like their peak not peak But like last year we heard a lot about them. Anyway, so [snorts] this is this is the crazy new world and my take on it is it would be foolish of framer, Figma and web flow. These three giants of no code and design to [snorts] not pivot to this direction. Agentic first, UI second. That's the future, not the other way around. And whether you like it or not, companies like Lovable are taking over because they are enabling the next line of people, the next generation. The same way that Web Flow allowed me not knowing code still create animations. Lovable is allowing the next generation to create a website even not knowing how to work use an UI not even knowing what a 301 redirect is you can still create a website not knowing what that is. My thoughts on my thoughts on lovable though and we'll get back to I think we we should probably get back to web flow a little bit and discuss all that but like okay [snorts] my thoughts on lovable are that whilst they're being valued more than Figma Figma are a very focused and dedicated group of professionals who require that tool to yes earn a living right whereas is Lovable is existing in a bubble where and impressively so it's introduced as you say a bunch of normies to building apps but are they a dedicated group of people and I can't answer this it's just an open question are they going to be people who yes they've everyone's signed up to lovable because they've heard it and they've paid monthly or even yearly they've committed to a yearly thing to see how it goes but actually by this time next year or the year after all of those people would have churned because they are not they don't require that tool to do the work and honestly my opinion is lovable is a gateway into professional apps like I don't know I'm I'm [clears throat] not I'm not sold on it however I'm 100% on that I'm [snorts] 100% on that lovable and similar tool being tools being a gateway. Uh I I've been seeing a little bit of a new way of thinking about this whole AI era thing that it's not going to destroy jobs. it's going to create like an insane amount of jobs because at the end of the day a business owner might use Lovable to you know get something together but once they start making money with it once they start getting good at you know whatever they are doing they want to outsource it still to somebody else to build on top of it and do more or they want more tool you know that love yeah there will be more jobs Like it is hard to grasp for me, but there are like tons of people making money with Canva. They like profession I I I don't want to get hate for that, but like professionals like Canva, if you're making money selling something, you're a professional, you know? It doesn't matter what tools you use. So, no hate there, but like Canva is extremely simple. And for a long time designers would cringe at other people saying like, "Oh, I designed this with Canva." Like they would cringe at it. They they would look down to at it, right? But still my point being there it there there are always these new upcomers, these new tools that simplify something complex and then they offer it to a wider variety of people and that's how they start. But then they get better and better and more professional and that's also how it happened with Figma. Figma was very simple and now you know as you said it does a bunch of things but I would still argue what's the point of using Figma like it's becoming less and less important as it's becoming easier and easier to just design with uh prompts spec es especially if you have something specific in your mind you can you can get like something really cool on with AI. I truly believe in uh a canvas though. I love a canvas. I show me a canvas and I'm like, "Sign me up." Because I do I I'm one of those people who I can I I can imagine something in 3D in my head. I can render it in like in 4K highest quality, but I still prefer to also have a canvas where I can compare things. And I think this is not going anywhere. We want to have versioning and comp comparison like a visual tool. And I think this is what lovable is lacking. But if I'm not mistaken, uh what other tool did I mention? Uh lovable. And we have the name is on my tongue. On the tip of your tongue. Yeah. The name is on your tongue. You said uh loable uh not replet did you? Replet. Yes. Yeah. So replet I think replet is doing an interesting thing by going after design. M. So in replet you could at the start you could just do like just um do normal like similar to lovable work but now you can do design and I I might be wrong on this but they had some tools to like create designs and compare them next to one another. I don't know if if it was Replet. Give me a second. Um yeah, that that's not it because what did the tool do? Uh you could see like designs next compare designs essentially. Compare designs claw design like a canvas like a canvas. Yeah. No, maybe maybe I'm wrong. But anyway, I would love to see something like this. I would love to see something like cursor or claw design with also canvas where you can explore designs and compare them and then maybe merge them kind of like git but with Figma mindset because git I'm still wrapping my head around the the whole git mindset. It's a very programmer and [snorts] like code merging mindset where with um Figma canvas you just duplicate things and then you mix them and merge them manually. And if we get something like this, I think I personally can get a lot more out of an AI uh in terms of designs. I can create these variations and then sure and see them side by side as opposed to looking at code side by side previous versions. I mean you don't necessarily need to but the workflow is just completely different but this idea of backup. It's it's just backup at the end of the day. It's it's backup. Um yeah, you don't necessarily need that with with design. Um if you create variations, but it's almost like every every node or every frame needs its own version tracking system rather than the file itself, you know, but I mean, you know, saying all that we're talking about web, but what's that? Replet design. Um I I I was not mistaken. So I I I did see that uh I I just never used it, but that's kind of like the idea. I haven't used it because riplet is so expensive, but yeah, this is a canvas and then I just did holding alt and then duplicating. This is what I want. If Webflow can give me this and a UI on top and then allow me to publish anywhere, that would be really cool. It's interesting because, you know, I would argue that Web Flow is not a designer's tool. It's a web development tool, but you can clearly see it would be interesting to see Replet's user base because everyone is throwing everything at the wall and seeing what sticks because the the sky is the limit when it comes to AI, but their user base might be so low for you might be the only person using Replet Design in this current moment right now. It it's it's cool. Listen, it's a nice idea, but it might just be that the user base is not someone that uses them all or whatever. So, it it's going to be really really interesting because this could be Web Flow's last attempt to solve. That's a good point. Yeah. Yeah. If they don't get this right, there's a lot of pressure on them right now. They don't get this right, I think they will lose a lot of people uh literally overnight. Yeah. I hate to say this, but um yeah, you know, we made a video about instatic and it kind of blew up. Uh I continued to use Instatic. I haven't stopped. And in CA, actually, I gave it to one of my teammates and I was like, let's build on top of it. Right now, I'm I'm not ready to share it because we're doing it for a client uh project. We are building plugins for instatic allowing us to migrate away from web flow. I really don't want to like well I want to I guess but I really I I don't love it that I'm migrating away. I really love how familiar and beautiful the web flow UI is. But man, at the end of the day, if I want to do something specific with the CMS, I have to I don't know like it it it becomes, you know, it it becomes so cumbersome like you can't even do like very simple CMS stuff like um reordering stuff. It Yeah. Yeah. I think what's that about? I I think they are way way too stuck in their uh old architecture and they know this probably now. So they they have to make a new thing and they if you've uh paid attention recently they haven't pushed as many updates or as many meaningful updates. Uh well I see like these AI first companies they are like putting a lot of updates there and I think that's partially to do with how they started from a simple UI that was a chat and then a browser and now they are adding on top of it. And my bet I remember like earlier I always said like my bet would be on web flow because they already have figured out the UI and the UX of the editing on canvas but I guess I didn't see um the big issue there being the the architecture that they are already running on. They just can't change it. So they're building something new. I thought about this and you know I reflected on one of your first when we first covered that web flow is behind and and x y and zed. Um and I listened to something you said and I remembered that actually this the the uh proprietary system that web flow was running under it's just JSON. I think I'm 80% sure your web flow website is just a JSON file that that gets translated into you know the HTML website cuz if I remember rightly you can copy and paste between like systems and if you can actually instead of it show as JSON. Yeah. Yes. I've done like I have a component Yeah. thing. Yeah. H, which is funny because you think that would be quite an easy thing for people to be able to edit and and control and whatever, but clearly and JSON is very very very flexible. But one thing I'm thinking about is what's one thing that Web Flow have, one of the biggest things Web Flow have really only just introduced, but they still haven't nailed it. Like they haven't nailed it. They're pretty damn close, but we waiting for it for a while. Do you remember or think or know? I'm not sure what you mean. I'm thinking about like many features that they That's why that was the worry in asking because I bet there's tons of stuff you want or whatever. [snorts] Multiplayer like Yeah. Well, the multiplayer is not bad. Like we are using I know that's what I mean. I think they haven't nailed it, but they are they've they've they've only just really addressed it maybe in the last year, 6 months, something like this. If if we were to say if we were to poly market, put some money down on what the web next web flow app experience will be because it's clearly going to be an app. Um, how will multiplayer um work? Because one that's one of the biggest things of Figma is you've got all these ma mice moving around and whatever and you can all play and d that's the fun of it. It does it does work slightly different for code and the way it code it just does not lend itself to that kind of experience. But if this was a local app, how will synchronization work between because you don't you can you can have plugins that synchronize your cursor environment. So you can code next to each other and you see the you know whatever how will web flow address this cuz I think if they lose that again the collaboration aspect because I think a lot of eyes that we've got a lot of our Twitter feed that we have um with regards to hash webflow are independent designers and developers who are using web flow and we've spoken about it before that is a probably a large portion of but a small portion of web flow's kind of income. So our requirements are very much on the um lower priorities on what Web Flow wants. They need to still maintain everything that these marketing teams want that these companies want. You know they're happy to throw lots of money if security. So let's talk about that that that's I think is a stellar point. Uh so basically what you're saying is what will web flow be able to offer to these companies to these teams uh in in their new form in their new app. So things that they have that are in like incredibly useful today, they they have this concept of components, right? They are extremely useful to to clients because if I build something on web flow, I can basically create a custom template for that client that they can use to build on top of like they can use to build new pages with. That's really really useful. Um, but there are also bigger stuff for like bigger clients, things like optimize and analyze. And this is what your VIP coded app doesn't have by default. Like you could add Google Analytics, you know, with a prompt. Uh, or well, it's it's not as simple. Maybe Web Flow can actually add this regardless of how you build your project. you know at the end of the day it's going to be served on a server somewhere maybe you can add web flow analyze and optimize to it because all of that is also AI and AI can you know change the the copy can change the assets things like this will be like incredibly valuable so maybe web flow will be the final destination maybe you have that final destination maybe you have like 20 vibe coded well you probably have like a hundred uh VIP coded apps, but only you you have like three of them that are on the finish line and you have a website for them and the app and you publish them. And this is where you might still want to choose Web Flow for all of that infrastructure that they have around it to maintain it as well. Yes, maybe it's not for us like you and I as individuals, but maybe it is still very valuable to clients. I have a little bit of a prediction a little bit. Yeah, I think this app will let's call it web flow build is going to be free and web flow are simply going to be a hosting platform for the for the app or website that you build app or website the that will be free. I think they were going to remove the site plan or something like that. But I think the extensions that you just talked spoke about, this is what prompted the prediction, the analytics, the potentially the CMS. I think the CMS might also be a bolt-on, but it's not going to be included. It could be something you add later on. Um but these um I think what would make people really happy is a free experience for uploading and using web flow that's given to developers that's given to hobbyists but the moment you need a professional um uh enterprise level website and you can then buy the I think it's like is it 200 a month their their marketing platform or something like that I think that's where they'll see their revenue I think 500 yeah I think that's where they'll see their revenue or over a certain and bandwidth. That's when they'll start charging you as well. But you can launch unlimited multiple hobby hobby projects just because that is so easy to do nowadays. It's like why would I pay for web flow for my hobby project when I can just clawed up and vibe code a portfolio website that gets maximum a few hundred people visiting my website a month and host it on Versel. They need to compete with that model. And so I think the build, the editor and the hosting will be free and then over a certain point that's when they charge and for those bold. But what do they do with the AI? Because the the build build tool will be will be completely AI. The build tool will be an AI. Oh, what you mean charging for tokens? Yeah, that's a point. That's the point. Okay. Unlimited Web Flow websites. You pay us $30 a month to use the AI build tool. But actually, if you don't use AI build, if you don't use the AI, you can use the build tool for free, you know. So, the build tool allows you to use AI to build. Yeah, but can you also use like just a UI to build? For free. That's my prediction. For free. I don't I don't think there will be a UI at all like to build. I don't I think it will be like So, you just think it's going to be claw code basically? I think it's going to be like claw design. Claw design. Oh, claw design. Yeah, I think that would be silly. I would like them to maintain some sort of UI, but I don't think you're crazy. I don't think you're crazy, but it would be silly. What's the point of me adding a div block? because some people you're you're in a you're in a um pocket of people and there's many of you guys that rightly so are happy for AI to create a div block or whatever. Some people still use some people still use the they want they want to move things around. They want to you know they want to have No, I want that to be clear. I want that. Yeah, I I I would love to have a canvas. Yeah, sure. To add stuff. That's an interesting one. I'm I'm still not sold on that, but I still don't think you're crazy. I don't I don't don't know whether they'll have a canvas in the way that you want it, but whatever. Yeah, but you know you you know why I'm like so like like pushing on that is because I don't see a reason for design to exist so separately from development. I do see a point in exploration being separate from production right and the exploration can be on a canvas. This is where you play with stuff. But it doesn't have to ex that doesn't have to exist in a vibe coded app somewhere else or in Figma somewhere else. It can be all within the same tool and the truth the source of truth can be still code. This is the point. It it is all code anyway. Like even Figma, it's technically still some code. So my my um No, I agree. My my prediction is not what the tool is going to be. My prediction is more the framework in which web flow will operate and how it will earn its money and that is you know it needs to compete with these free tools that people are jumping ship for. H but how can it offer something that it can earn money off and and benefit the people who are willing to spend money? Hobbyists are not willing to spend money. They love the free stuff and they can now achieve free stuff. So they just need to hit that price point that's not a price point. So yeah, they're going to offer you as you say like a cursor ultra sort of subscription to use the AI, but the building of it of of you just using an AI just like maybe you know and love and then the hosting it's that's going to be still so easy and free to do. That's why they still have a UI because it's still free to do. If you're willing to put the effort then and not and if you're tight on cash, web flow will still enable you to build a website just like you've always known web flow to do and then they'll host it for free and then once you certain amount they'll be like cool. Like Versel like Versel. Yeah, I I was thinking Versel. So I take back what I said about the UI not existing but I still think the U I think maybe what I'm imagining is like clot design but where you can start by prompting or you can decide to choose from the web flow framework or whatever framework that they have to start a project where you have pages where you have like component where you have all of these concepts but you might not want to go that way. Maybe you you vibe code something with React with Yeah, you you might even may be able to make a mobile app. I don't know. Similar to uh Replet, I think. So they'll I think they'll merge what do they call it? Web flow cloud or whatever it is. I think that's all going to come in together. Yeah. No separation. you can go that direction of like kind of like linear way of okay you have a UI to add stuff similar to what web flow is today but maybe with a much like closer to the or like the the source of truth that they said to be literally code so maybe it's just astro uh but themed with like UI so you can do stuff maybe I'm not getting that completely right but Then again the main thing would be you add you you build whatever or you import like if it's if it's just files you might uh code it with cloud code and then imported in web flow and all of the sudden you have a bunch of features such as a UI to edit um visually maybe an animation engine to you know animate stuff with an agent. I think that won't be crazy. Like imagine I vibe code something with claude. Then I just take that folder, open it in web flow, and then fire up the Gap um box, like a chat prompt box, and then I tell Gap, hey, the these five things that I selected, animate them with a stagger, and it does like that. I I can't see that uh being a thing and very valuable. Mhm. Um, and for that to coexist with the with the AI, I think that, you know, to plug my own video that I released this week where I, you know, built a framework that we used for the latest project we were on, um, my I experimented with skills to be able to create a Web Flow website on your local machine. you know, I experiment with the skills, but the skills just weren't they weren't authoritative enough, if that's even the correct word to use in this instance. They didn't guarantee that the the agent would create a web flow website as a static HTML website. And so that's why I moved to an agent MD file that had all the parameters and everything. And and there's a video on my channel if you want to go look at the GitHub repo. It's completely open. You can download it. But the agents MD file told it that you always need to create a page wrapper that you create um uh c uh uh combo classes for when you create classes and variations but use mast as a as a base framework. These are the break points that you should use for for web flow because that's the only one that's the only breakpoints Web Flow understands or or works within by default. My point is is when it comes to animation specifically, I can imagine then something being in the web flow harness the the the sort of rules that it might create GAP animations in a kind of semi specific way that enables the UI to then take over. So there's that interoperability where you might then create key frames and still be able to use key frames cuz that's really that's a really nice touch. Web Flow interactions was always a really nice UI experience that I still think people like to have that that that little new that touch. Do you know what I mean? Yes. Yeah. We brought this in the past as well, right? It's the same sort of thing. Key frames, layers, dragging things to offset each other by a certain amount, playing that and just having that loop over and over again. So maybe the harness allows you to just prompt an animation that uses GSAP and then under the hood the UI can interpret that. Or maybe maybe theJS can interpret Gap animations. I don't know. But it's going to have to be a GSAP because they acquired Gap, right? You you can you basically like anything here you define it in the code and then it creates a UI for it. I mean this is not a new concept and we have this is proprietary, right? This isn't Yeah. Yeah. This doesn't use GAP. This is proprietary. Exactly. And we have like dial kit as well which I have mentioned numerous times before. Uh you define these um controls and then you bind it to a specific thing. Like for example, you have a corner radius and then you you connect it to this and you drag it and the corner radius changes. Like my AI does this all the time. Like um there's a what was it? Experiments. Yeah, like I did uh I did this a while ago. See like this is a three 3JS thing and I can uh put this in a web flow project like copy the JSON. By the way, as you mentioned, you can copy the JSON and all of this this UI is defined by my AI. Like I didn't do it myself, but that's kind of like my point. This is a specific type of UI being generated. Well, it's deterministic in how it how a piece of the UI looks, but it's um how many of these controls you need is determined by the agent or like the agent codes adds them, right? These are just props and then we connect them. So, I guess something like this could could also exist. And I I remember um I think it's Josh. Yeah, Josh uh Pocket. He he was working also on something like dial kit but with animations. I don't know if he still has that but Mhm. So you you could do the same thing for Yeah. for Yeah. for Web Flow. Yeah. Well, when is when was it? It's not far. It's like two weeks away. I think it's uh late August, isn't it? The um what uh Webflow Conf uh September 2nd. Okay. Yeah. Yeah. So early early September. Or will be revealed I guess. Yeah. Uh September 1st actually. Um that Yeah. Not long. Two two and a half weeks something like that. Cool. Should we should we end it there? Wrap it up. Well, yes. As I say, very very different recording. It would be good to have some feedback on how this feels, whether it feels a bit better cuz, you know, although I I do like doing the stream, it certainly feels more comfortable doing it uh pre-recording it or doing it behind the scenes and then releasing it. But either way, is that all we've got time for? Yeah, that's all we've got time for according to Kabaza. So yeah, like, subscribe, do all of that stuff. Um, I've been Sam. [laughter] Yes, you remember. And I've been cover. Keep on vibing, baby. Peace out.