#54 SpaceX Acquire Cursor, Framer 3.0, v0 Design Mode, GLM-5.2, Snap

M. Bada bing, bada boom. There we go. Bada bing, bada boom. We are live. Hello there. Hello there. Hello there. Right, let me get my windows ready. Jac show. I can see we're live on Twitch. Uh Twitter. Sorry, not Twitch. Decidedly not Twitch. Uh let me know if my audio is right. I'm playing around with a bunch of settings at the moment. It's like super loud coming out of OBS, but not as loud coming into uh Streamyard. So, mic mic Oh, I have the mics. Oh, it's doing adjust. It's doing Am I really loud now? Um, not not right now. Right now, you're fine. Okay, me you're fine. Did it change at all? Is it changing now as I toggle on and off adjust mic volume automatically?

No, it's trying to be smart. Okay. Okay. What about if I do this? Is that louder? Is that louder? Yeah. Yeah, that is loud. You don't sound convinced by that. No. Yeah, it's loud. Slightly louder. It's not crazy louder. Weird. Yeah. Streamyard trying to do all sorts of clever stuff with the audio and Okay, let me just double check I have everything. Let me know if I'm too loud. I'll have to revert to my normal setup if I am. Can we go Can we add a new destination like like can we swap a destination midway like right now? No. No. What do you want to swap? I don't know. I was thinking like if we have both Twitters or like like you know we both we have our Twitters and I have like two people from my Twitter but I never see engagement from Twitter so I was thinking about um swapping it for Instagram.

Well that's true if you're faking watching uh it's true but most of our audience is from Twitter. Yeah. Yeah. So engagement. No, but you know when YouTube is good, it's really good. So um you're really dark. Yeah, I would I'm it. It's 33 degrees out there. So I have my shades uh down and somehow my camera doesn't I don't know. Maybe can you play with your ISO? Um, not from here. The the thing is like I feel like it's more the it does it automatically. I have it on auto. Uh, I can I can play with it. Give me a second. So, that might be your problem. But yeah, that's all I play with is my ISO.

Everything else stays the same. Well, let's see if it's better now. I have to Okay, maybe that's too much because it's 400. Yeah, it's too much. No, you see the the color balance is the issue. You see how yellow I am? I'm not this yellow. I I promise not a Simpsons character. Now, you see my Yeah, I'll go back to auto. What What camera do you have? I'm sure we've spoken about this in the past, but what camera do you have? Alpha 7 3. Uh oh, wait. What's Alpha? Yeah, Alpha 7 Mark III. Wow. Didn't realize you had such a fancy camera. It's full frame. Yeah, it's a full frame. Get good good picture from it.

Yeah. Uh I have also the 50 mil uh lens on it. The nifty50. Let's see if it connects. You see the white balance is now different. I I just I have everything on auto. It will adjust though automatic, right? Yeah. It should just adjust in time. But either way, yeah, you look a bit dark, but it's all right. Now, now I have the light in my face. Okay. Do I look presentable? Marginally marginally better. Hello. Marginally better. I'm I look dark because I've I've been on vacation. Okay. Oh, is it? You've been tanning. Yeah. Oh, man. Well, I'm burning. It's uh I've had to close all my windows now. So, I'm the heat fest. I've got the fan that's on a low level so you don't you know hear it.

I'm I'm literally as we speak looking for my fan. I have no idea where my fan is. We have no fans. We have no fans. No fans. Poor us. So, you're entertaining the audience. Uh, while I'm I'm entertaining the audience. While you're in Yeah. Um, I have nothing to say. I'm just going to upload a thumbnail for the video for YouTube. I've never watched h how are people I wonder how people are watching it on Twitter because I've never like watched any I suppose I've watched a few long form videos on I do sometimes like um primagen if you're watching this primagen I'm watching you on Twitter okay well yeah I was I was thinking about this today like most of our audience exist on Twitter but as you They they don't engage because they are bots made by Elon.

They are bots. Yeah, exactly. Oh, you did so well on Twitter. It's probably true, you know. Well, he did want to get out of uh to bail out to what's it called? Like to to not buy Twitter because he was claiming it. What's what's it called? When you buy remorse. He had buyers remorse. Hit him hard. We have it for $44. He had it for4 billion. I bought my Twitter subscription in um in South Africa, which means my paid for Twitter subscription is something like3 pound a month. Something crazy like this. It's nothing. That's crazy. I think it's only marginally more expensive than that. It's not too much more expensive, but either way, I get it really cheap.

I think I'm paying eight bucks for the video upload or something that we need. Anyway, okay. So, let's should we get started? Yeah, let's get started. We talked about Yeah, let's just get started. So, uh I have a I have a long not long well longer than usual intro. So, let's get started. But man, before we start, like I'm still It's still strange like our view count is lower than usual. It's cuz you mentioned bots. Yeah, it's because you mentioned bots. Now we're getting on to us. Yes, basics is just uh allocating the the the resources instead of bots to But the thing is we've mentioned them in our title. You'd think they'd be like, "Oh, wow.

They're going to talk about us. We got to send the bots. We got to spy on them." Probably. Who knows? Okay. So, so this week SpaceX is acquiring cursor for 60 billion for a fork of VS Code which is free by the way. Uh on a similar news uh cursor announced origin uh a git hosting for the age of AI. This is a response to GitHub being constantly down like 60% of the time uh being having an uptown of uh uptime of 60. That that's just insane. Remember we had 99.99. Yeah. Unacceptable. Not anymore. Unacceptable. On a huge news in web design, framer unveiled framer 3.0, skipping framer 2.0. You have AI doing everything from design to exploration to building the CMS everything all in one view.

uh Reloom is launching hosting like becoming a full-on platform to go from research to sitemap for which we were using to wireframes design and now hosting as well like everything uh that's very interesting vers shipped Eve which Sam will talk about I don't have much idea about it uh but we have also V0 shipping design mode, which looks so so damn good. All of this and more in today's episode. This is Command AI. I'm Kawza. And I'm Sam. And this is a week in AI design and dev. I don't know what that is, but I'm on I'm on board. So, you want to take on the the Elon in small the the pocket change. Yeah.

Yeah. Yeah. Let me get the thing. Uh I mean I don't really want to do it to be honest, but I suppose I have to. Where are my windows gone? I can. No, you don't have to. I can I can do it. I don't know what to say for um the the intro. I I can do it, but I don't know much about it. Well, we agreed that Elon has somehow out of nowhere magically, probably maybe highly likely will be the leader in AI very very soon. He snuck up on us a little bit, I think. Yeah. Um, okay. So yeah, a few weeks ago, a few months ago, even was it a few definitely might have been a few weeks ago, you might have heard that um and uh Curser or SpaceX or XAI or whoever they are, Twitter, Twitter was set to have an option to buy uh Cursor, which is, as you said, a fork of VS Code, probably one of the Aentic um idees out there.

there was an option to buy them for60,000 $60 billion or if they didn't exercise the right to buy them then they would uh give cursor 10 billion which is win-win totally for cursor like they they're like yeah and they were left with that option and you know nothing was nothing more was said of it then recently yesterday maybe even they have actively they they've made a decision to buy them for $60 billion which is again insane amounts of money not in stocks in stocks. So and and I was confused about this um SpaceX are giving um no no hang on now I'm confused again for a second time. So SpaceX is acquiring cursor for they're using their stocks they're using giving them literal money.

They are saying, "Hey, you own now I don't know how many percentage of their dollars." Yeah. Worth of their the the SpaceX. So, basically, the people at Cursor now own a a portion of SpaceX stocks, which I want to show you a few like short things about. Okay. Like it's fun stuff. I wonder. So, we don't know at what price they purchased the stock. This is the problem. They didn't they wouldn't because they're now worth how much they worth now? 2.4 trillion. That's the uh By the way, can we have this insider um link maybe? Yeah. Okay. So, the question is 60 billion for how how much of SpaceX? We don't know. And when when is the question because they they started hot and right now they are valued as at 2.4 trillion with a T like trillion not billion trillion dollars making them uh the sixth um most valuable company in the world.

Right before the show they were the fifth like they were before like like they were above Amazon. That makes it really really really crazy. SpaceX being above Amazon. Now with this market cap you would think that they are making a lot of money right I mean these companies are making a lot of money. We've heard Apple is making a lot of money. SpaceX is being in the same zone must be making a lot of money. So we can actually go to earnings, right? And then look if SpaceX is anywhere here and we scroll all the way to the top 100 and SpaceX is SpaceX is not there. Why? Because if we search them SpaceX and look at their earnings, their earning is minus3.68 billion.

So they are not making any money but they are among the giants. This is the crazy thing. They are literally among the giants making in market cap while being in the negatives deep in the negatives. This is the stock wizardry that Elon Musk can do. You could say this is a bubble. You could say this is over inflated. Whatever it is what it is. We see they are not making much money. they're losing money. But yeah, crazy times. Crazy crazy crazy times. A comparison that I like to to say and then we're done with this is compare it to Amazon, which SpaceX was for a short amount of time above Amazon, meaning that they SpaceX was valued more than Amazon.

What's crazy about that is that Amazon is literally everywhere. like you cannot like societies, countries do not function without Amazon because we're talking not just about the Amazon shop which is fairly important to the society but also the AWS which is the backbone of the whole internet like without that pretty much every website goes down like everything will collapse without Amazon. If Amazon ceases to exist, bad news. If SpaceX ceases to exist, um, not much will happen. Yeah. I mean, yeah, they have like important business. You know what I'm saying? Like it's not like a crucial part of the society or like Yeah. Any I don't know. I don't know. It's just like valued a lot.

And also you could say at least they are making money that's why they are valued uh so much but no they are not making money uh either. So so people including cursor are banking on that this valuation is there because they're banking on the fact that SpaceX will be making lots of money. These are people who believe that SpaceX are going to be, you know, earning the money even though them as a company are losing right now. Like you say, Amazon were losing money for a certain amount of time. These are called loss leaders, right? So Elon Musk must have put one hell of a pitch forward. You know his was it TAM total addressable market is like wild and exactly all those pictures of all those spaceships must have woo SpaceX even though by the way this isn't their pitch deck to you know whatever uh to investors um like the first like six pages is just pictures of rockets and the rest is all about their AI ambitions.

Yes. So the the whole reason they are so high up here is because they are saying to the investors now the public you and I they are saying we have this huge market that we we are in we're losing money in it but the market is huge the addressable market is huge we will make money sometime in the future but I I don't see a plan for that like not just SpaceX Nobody has a plan how to make money with AI. Not not really, not long term. So that's an issue. And also like cursor joining SpaceX doesn't mean that cursor is not thinking that SpaceX will make money someday. It's just that they hope that the overall basically size of the company will stay about the same amount or preferably grow preferably for them to grow.

So sometime in the future they can sell their shares. I I bet when they join uh and they get you know 60 billion dollars uh worth of stocks those stocks will be um locked. they can't sell them right away. Typically, it's probably over a year or a few years. This is where they can begin to sell those stocks because if they allow them to cursor can immediately sell those shares and with that they can tank the whole value of the company which SpaceX obviously not just SpaceX no company wants this. So that's just a standard acquisition uh term. The cursor is locked there for a set amount of time and they hope the value will grow or or stay roughly the same.

Um but I don't see SpaceX becoming profitable anytime soon. Is Tesla profitable? Yeah. Well, you know, so obviously the other side of this coin is cursor and like the it it's interesting because like it's not well first of all actually just to finish on that it's not 100% clear what the what the details are of the stock. It's not like does this give cursor priority to their GPUs? Does this involve any other sort of deal uh other than here's $60 billion worth of stock? You know, keep going, keep doing it as you're doing it and that's it. We we just own you just have to report to us, you know, as normal investors do. So, I we don't know what the what the ins and outs of it are.

But what this does do is SpaceX, XAI, whoever the hell uh this is now. I guess it's all SpaceX. Elon Musk essentially owns, you know, quite a lot of this the what do they call it? vertical um uh integration of a lot of AI right now down to the fact that he wants to ship the GPUs to space. So even the space hardware he's able to to you know uh narate yeah that's why SpaceX is so so much valued because people believe in this vision apparently you know he owns the whole thing. Yeah, and we didn't really report on it, but like you know, Grock builds was released um like last week or something like that.

And this is a again an ecosystem of you know really putting forward Grock into the coders of you know tomorrow as a model. I haven't really tried the Grock model really and I really want to but just doesn't interest me to be honest. Um, so I'd love to hear from people who have used Grock and what what it's like. But anyway, I'm getting off track. They own the data centers. They own the the the actual AI model itself. They own the software that uses the models that they've got so much of this pipeline um that they rival Google in that respect that they they own a lot of it, you know. They have Twitter and they got Twitter.

They have the rockets to send so they they for the full vertical they have also the rockets theoretically we don't know if this will happen anytime soon the rockets to send the data centers to space and then the models to train on those data centers and then um the like the Twitter the data how valuable that is I don't know but arguably with um cursor now they have valuable data probably much more valuable than again I would argue a lot of people uh like how many people are still using cursor you know and how many how many how much how many of those users are valuable users you know are they people who don't want to pay for anything we'll get into that a little bit later on I don't know it's is it is data but it's whether it's valuable data time will tell they built the I guess they built the cursor compon compose a model from it, you know, and it's pretty good like from what I've heard.

Uh that that's really good. That means they have good data. Uh but also this is interesting because um Google is paying SpaceX to rent their their servers. Um even though they are rivals in you know having both AI models but they are also they have this partnership. Same for um Anthropic. They are hosting their models on SpaceX hardware. Now by acquiring cursor, SpaceX has insight into what are the answers coming out of these models because people use them in cursor and then how people integrate them in in their code. And with a separate video that you can watch separately, we will talk about origin um cursor's version of GitHub. All of this coming together means that SpaceX will own a good amount of data, probably more valuable than you know what other companies have.

So, it's probably a good business decision for them. And with their with SpaceX being valued at this like insane number, buying cursor for them is just it's just so easy. It's just so as you say like peanuts. It's nothing. It's pocket change. Well, yeah, exactly. And this will this will get raised a few times in this in this show like I watched the um interview with Dario Armade and he he similar to the question we spoke about before the show like um he she asked him like what's what's the key to success or what you know what have you and he basically comes down to you know not creating the foundations are creating a good model that's that's what it all comes down to.

So, it's yet to be seen. You know, they might have all this data, this and that and all the rest of it and access to hardware and and Twitter. Um, but unless they produce a good model, it might all just been for nothing. And that's the problem with Grock. It's like people are people adopting it? Is it a good enough model? Like, so are you saying you don't like Grock? You don't like I I saw a video. Who Who was it from? like Brock playing um a game uh one of like those shooter games I believe and Grock winning by far because apparently um I think it was from Mo, you know, Mo is making like like I've recently started watching his videos.

Uh, apparently like other models like the ones from Anthropic, they they think about morality and like teaming up and stuff and Grock. Oh wow. figured out how to run people over with with vehicles just developing a strategy and not caring about That's crazy. That it's really fun. And uh even though the other models are smarter in a way, they are not good at playing these games. And Grock is just I mean, you could have seen that coming if I would ask you which one wins the games more. Probably Grock if it's especially a game where you don't think about morality too much. You you you need to be selfish. You need to be objective and uh kind of like reckless in a way.

And that's what comes to my mind when I think of uh Grock. It's a bit reckless. Winner takes all, baby. Uh last thing to last to say is that uh SpaceX will pay a 10 $10 billion termination fee if Tuesday's deal, so next week, if it falls through, Yeah. because this was yesterday. This they announced it. If anything falls through, they'll have to pay 10 billion to uh SpaceX uh to to Kursa. Wonder if that'll be in stock or whether that needs to be in cash. Yeah. So anyway, anyway, should we move on? I don't have much more to say on that one. Yeah, I we got we got it really covered. Yeah. Yeah. Yeah.

Yeah. Um this will be a quick one. I won't bore you to death with it, but let me let me get my email up. You might have seen a little email from Anthropic uh about two no on the 15th. So about a week ago now saying that uh the p the the programmatic usage pool of cache that you can use to use the agent SDK and the um clawed-P that they're no longer rolling it out anymore. They're pausing it, which to me sounded like good news. I thought, "Oh, yeah, that means I get to use my my usage." Um, sorry, I'm just trying to find my email as we as we uh talk here. But of course, um iCloud search is utter trash.

Where is it? I don't use iCloud. Maybe that's why I'm too Google brain there. No, it must be in uh must be in my other one cuz I thought I deleted it. Okay. Well, let's do this. First of all, this is the original email that you would have seen. Okay. Hi Sam, starting June 15th, Max 5X plans claim. Can you zoom in please a lot? Uh, and then start again. Yeah. Hi, Sam. Starting June 15th, Max 5X plans subscribers claim a $100 monthly credit for using Claude SDK and Claude-P including third party tools. These are things like Open Claw. These are things like um Hermes or any tool um T3 code. Any tool that was using clawed subscriptions under the hood would get given a pot of cash to to to to use that sort of stuff cuz of course we all know that the claw subscriptions massively subsidized massively subsidized.

Uh but on the day of that was supposed to take uh take it take shape. It was supposed to do we get another email which I'm looking for right now. Hi Sam in can you see my can you see my email page? Yeah. Um, in in May, we sent you an email announcing that starting today, claude agent SDK, cord-p and third party apps built on the agent SDK would stop drawing from subscription subscription rate limits and move on to a dedicated monthly credit. We're writing to let you know we're not making this change today. We're working to update the plan to better support how builders um build the Claude subscriptions. Nothing changes for now. agent SDK and claw-p for third party usage continues to work with your subscription exactly as it did before and there's no credit to claim.

Your subscription limits are unchanged when we have an update. We'll share it with you. Now, I'm I was like, "Yes, yes." And I was like going to go and update a bunch of my stuff to use Claude so I don't have to, you know, I could use Claude with my subscription that I'm paying for. And funny enough, I recorded a video on Hermes, which I haven't released yet, but I recorded a video on Hermes, and it was automatically linking my Claw subscription, and my usage credits were going down. I was like, what? I I just did not did not understand uh what was going on at all. And I thought, there must be something going on or that something's broken or that it's taking its time to roll out.

Then I check this article here and they quote they've got a screenshot here which has slightly different wording. Um uh cord agent SDK called P and third party usage still draw from your subscriptions usage limits. So the extra usage that you get or that you can put money into. So that's that explains everything, but it's not stated in this bloody email. It definitely doesn't state that. So it it makes all sense, but it's just really annoying. And I think I sent a tweet about the fact that Antropica are just so at their communication. Like I think they could be a lot lot better at their communication. I'm wondering where this was taken from. Um is their publish post.

Oh no, that's conductor. Sorry, I'm looking off screen. I'm just wondering if if this screenshot was taken from um the page that that announced it. Let me just double check one thing. Let's have a little look. Starting June get a budget how it works. This is the original post. There's nothing for you to do today. An email. Here we go. Yeah. So, it must have been taken from this. This is a page on there. So, they've they're able to update this and gain a bit of clarity, but um either way, point is still not ideal. And long story short, sorry, I'm not sharing my screen. It's this page here. Yeah. So, you can see the change.

That's where they took the screenshot from. So, what I once thought I mean, this is still bad. This is still bad. like that pot is was better than taking from the money that we've we've we have to top up that pot. We have to top up that monthly uh usage credit. Now it's taken from that. Whereas they were previously just going to give us a chunk of money that it would draw from. And that was actually a good thing. It wasn't the best thing, but it wasn't a good thing either. Um now this is even worse, which is quite annoying. And I think people have missed that. I don't I don't see a lot of people like angry like you you you were going to give you you you gave us something then you took it away then you were going to give us something and now you've taken it away again.

A little bit annoyed about that. I I don't know honestly why they are doing this. They want users for example to not use um cloth their cloth subscription with tools like Hermes and um like programmatic uses and also other types of like agents like uh those but why don't they just ban those in a way or like just restrict well we've gone we've gone over this right I it's because they can't figure out a reliable way to know if someone's using Hermes or Open Claw or whatever. They just can't because you're essentially just calling the API and they've just there's no indicators that accurately tell them. There was that one guy who got charged a bunch of money because uh he had a line a comment in a file that said Hermes in it, you know.

Um what else happened? Um, I mean, I think that was that was kind of it to be honest. So, they can't reliably do it. So, they just had to like make a decision. It's all or nothing. And I think it just goes to show how much they're actually subsidizing because basically the these these tools mean that you can build something that can actually use the co the tokens that you buy, actually use the subscription that you pay for. We all like this is a matter of this is this is um a well-known fact right the there is huge sub subsidization is happening all over all over the industry the people who pay for the higher tiers that don't use all of their tokens credits um subscription whatever it is whether it's uh Netflix whether it's uh Amazon like there are people that do not use.

And that's what you bank on. You bank on people not using what you give them. And you and you've got to draw you've got to find as a product owner, you've got to find that balance between well, what looks really good, but we know that no one will go over it. Or how much are you willing to sacrifice? Are you willing to like make it the offer sounds so good it's hard to refuse, but actually people end up not using enough anyway? So it's do you know what I it's a very it's a balancing act there. It is a balancing act but overall from what I understand there are arguments that uh Anthropic is not making much money uh if not anything on these subscriptions they are probably losing money on these.

Where they are making money is on enterprise. So they are trying to lure you in. You basically subscribe for 200 bucks. You get tons of usage and once you like it enough, you probably will continue using that in your company and you you know you ask your company to buy this and then they buy the enterprise and this is where they're making a lot of money. So why would they limit you? because it's a losing money system anyway. That starting point and using these uh agents kind of like guarantees like users that use these like um programmatic users and agents will draw much more from the tokens and they want just to make sure they have you know you know they are not giving away too much of free stuff but they're managing it pretty poorly.

There is no doubt in that and that they are communication just doesn't exist. It's it's that bad. The naive the naive people will say and and I understand because people don't understand how business works. But the I think the naive people will say that hang on I've paid my subscription. Um in fact let me find it actually. Um where was it? Uh it was on a Twitter post is here. This person is a naive person in my eyes. So this is on this is this is something we'll go over later. So don't worry too much about it here. But this person instead of limits there should be a fixed number of tokens for subscriptions. There should be there should be prepaid rollover tokens when you barely use any in a month.

I shouldn't need to wait 5 hours every 10 minutes. Right. That's that that's something that's something separate. But rollover tokens when you don't use any like this you don't understand how business works like not not that I'm saying it's a good thing or a bad thing. This is just the way that it is. I don't I don't run any subscription services so I don't benefit from this at all. But they they subsidize. They they make sure that that you pay and that you don't they want you not to use your your your maximum amount or whatever, you know, they they they benefit from you not using it. So why they why would they just let you have a rollover and why should there be a fixed number of tokens?

It makes sense that one would think and feel like that. I totally understand that but it's just not the way it works and it is just not the reality. The reality is these tools are expensive. Like the amount of money that these companies are burning is insane. So yes, we are paying 200 bucks and it's a lot of money, but it's all relative. Like we like it or not, that's the reality. that $200 that I'm paying is not enough for that company to make money with because they are burning crazy amount of money on everything that we are talking about. So like the training the the servers the the graphic cards everything the the electricity it just too expensive and the reality is we've been kind of lied to uh like a sweet lie and we're just waking up.

We're just waking up to the fact that oh that you know model that we've been using to generate images with and you know what what else it just burning money for the business and they cannot continue this forever. There was this funny view video uh from Big A uh he it was about the SpaceX uh whole thing but it's it's the same thing. It's he uh he generated an image with Grock. Uh the image was Elon Musk putting lipstick on a pig. And he just to make his point, he made it 20 times for free. And the whole point was like, look, I can do this. There is no world that I can generate an image.

Elon putting lipstick on a pig 20 times for free and that business will somehow make money with it. No, they're they're losing money with it and that that's that's the reality. So, we'll see how how far this can go. I hope more companies will sign up. They will pay the enterprise so we can continue enjoying our not, you know, too cheap but also not as expensive as what reality is. Uh the the $200 subscription. Yeah. Yeah. Um I forgot I was going to bring it back to the point I was trying to make, but I've totally forgotten the point I was trying to make. The point is is like no, no, it's all good. We we we go off track sometimes.

But uh yeah uh the point is this the the agent the uh credits that we were going to get given um the subsidization I think I I think it was something to do with the subsidization that takes place like it's just it can't go on forever. Absolutely can't go on forever. And uh there this is their way of just trying to manage that. Oh no that's where I was going with it. And all that to say is OpenClaw, Hermes, all of these um platforms, they make damn well sure that you maximize your token usage. They like there's there's no world if you're using these tools every month you for sure get that and get your tokens.

And that goes against every business decision ever made when it comes to this subscription uh all you can eat buffet uh option that that we're so accustomed to. You know, it's like those restaurants, it's not in their uh interest to have um you know, people who who eat too much to go there, right? Those restaurants and it's not in uh you know, anthropics business before I sleep. Sometimes I do ultra ultra think workflow and then just something. So I sleep and claude works. You should you should look into loops. Um loops basically set off what two agents. One is constantly reviewing. It's on a loop just reviewing the code and then it's passing the prompt to its other like they're just a little team that works whilst you're sleeping.

Could you do that with workflows? Um well you can one agent is there to prompt the other agent. So you could prompt that a that uh let's call it the monitoring agent. You could prompt the monitor agent to create workflows if it wants to. Yeah. Yeah. But like those two agents exist in two separate terminal windows essentially. Yes. Yeah. Yeah. Yeah. So how do they interact with one another? Through files. Okay. So I need to I I get the idea but I need to see how that is. Yeah. Yeah. Yeah. Yeah. So um the the reviewer agent because the the implementation agent it it can it selfdestructs like it just removes itself. So then the reviewer agent can spin up other agents like another agent.

How how do they ping each other? like the reviewer has to get pinged that hey I'm done review my work how how will that happen um it it can check git it can check git diff differences in the in like here's the main branch that uh that agent has been working in another branch or another work tree are there differences uh what are the dates of those things are have there been any changes so um yeah something to look into. Yeah. Yeah, I have to look it up. You can set goals in clawed code like slashgoal and that's one way of doing it, but it's not a particularly good way because it's I mean you could in theory run a um a loop on your uh codeex agent or vice versa.

Opus could be monitoring the codeex agent. Do you know what I mean? Like you can have two different agents. That's also interesting. Yeah. M I I was thinking like if I should do computer use uh but that will burn even more. Yeah. Um honestly I would love to do this for what I mentioned before like I I was trying to create an automation tool to uh to to have many chat and like a publishing tool all in one place. So for example, make a video, publish it to every platform and also monitor the comments to automate replies. I made the software overnight. I just ran it literally overnight workflow ultra uh think it was a working not just prototype.

was a working project, but connecting the figuring out the APIs and like setting up the developer account. I'm I'm stuck like in the Facebook loop. It's not it's not recognizing it. Well, I don't know. I would love to set up an agent who can like, you know, try that until until they get it. I don't care if they hack Mark Zuckerberg's own account until they get the API access. Can't you can't you give Claude the Chrome MC uh Chrome I did. Yeah, the Chrome extension and it navigate and use Facebook. It could it couldn't figure it out. It couldn't. No. Yeah. like it has ideas for why it doesn't work and like it kind of like goes in a loop of oh it doesn't work because of this and then we change it like oh well we have to change it back.

I've always had trouble with Facebook to be honest creating an app creating a dev account and and all this that and the other. Um but yeah and they all do things differently as well. They're not always the same way of doing stuff. So it's like okay what have I got to do to create this account? Do you know? Yeah. Yeah. Anyway, this was supposed to be a quick one. Yeah, it was supposed to be a quick one. What were you going to say? Um, separately that it's over. Yeah, if you are interested in this one, um, maybe you can help me out on it because you're posting stuff to LinkedIn like you could automate posting to everywhere.

I I made it like uh it works. the the posting part works, the automation doesn't. To LinkedIn. Um uh yeah, I did post to LinkedIn. Yeah, that was the first one. I even posted like test app and people liked it. I was like, what thing is Yeah, cuz mine was winging that it needs a business account so I didn't bother. Yeah, I so I have my Kabarza agency account which is a business. It's not a personal account. Through that I authenticated my own account. So it when it posted it posted with my personal account but the business account was just for the API setup. So that worked and I posted also to Instagram. My video is still there.

I posted a video saying like if you see this video it means like I posted through my app and it worked. Um okay so that works but I don't get a web hook when somebody comments because apparently like I have to have a different um a different account. That's fair enough. That's when you know it's 6 p.m. Yeah. Yeah. Yeah. 6 p.m. Yeah. 6 p.m. Pedro gets food. Okay, so we should move to the next one. Curs cursor ships origin to take on the behemoth that is GitHub which has been going down down downhill for the last 3 months. But it's been built apparently it's been built for for the agent the agentic future. I don't know really know what that is.

There's not a lot on it, but I guess we could have a quick look, right? Yeah. So, we are launching code storage. Launching. Launching. We'll see about this. Origin gives teams and agents a place to host, review, and collaborate on code. Available this fall. Join the wait list. So, it's not a launch. It's not sh it will come later this year. The website though, I'm a big fan of the website. It It's just like a single section. I was gonna say what? Yeah. This is nice. This is nice. Yeah. So, anyway, you can sign up, but it's an interesting idea. We had uh this conversation with GitHub not being uh their uptime being like really really bad.

Like they're I think they're like at 60% or something like this. So when you pin, I'm not talking about the GitHub homepage by the way. It's like the git like the the functionality doesn't work when you try to um uh update or like use git. So this was uh something this is something that cursor is trying to do and this is very interesting because it goes hand inhand with the conversation around hosting on one side because they just they they shipped uh their editor so you can like build a marketing website you can visually edit stuff now you can also host your code there this can become like a full-on platform for marketing stuff not just for app, but also you know they have their own AI and now they will see very clearly what code you are shipping, what code you are like pushing to production, what prompts you use, what answers did you get from each one of these like different models.

So it seems like a very strong position for cursor that they are building for themselves because they will own the whole loop of you know you you prompt them they see what works they see what you ship. Yeah. Um pretty interesting. I signed up. We'll see what it is all about. I also signed up. There was a bit of tiff like to and froing in the comments section of this Twitter because you could only sign up for your work with your work email and uh you say you sign up with a personal email so they must have fixed it. Yeah. Yeah. I used my Gmail just to test it and they fixed it. But there are like too many people unhappy about um not you know please use your work email as you said earlier cursor sees you if you are using your Gmail sees you as not worthy maybe you don't pay so if you use your Gmail you're not going to pay 200 bucks a month for you know a cloud subscription so probably also you just want you just want it for free you just you're probably someone who's wanting it not to not be a work address is probably someone who wants it to wants to use this for free which you know um arguably should you get early access I don't know um the other thing is is well what do we think what do we think what is git for the AI for the AI ages because I'm I've got the Claude plugin for GitHub and I tag Claude in my GitHub issues and things like that it's quite a nice way to work Um, could that be it?

Could it be um a a self like improvement? Could there be bot like agents that just keep addressing code reviews and, you know, um, yeah, just improve the code, whether it's bug fixes or whatever. Could it be a way for non- techy people to also interact with the code that you're writing so you can combine, as you say, with marketing people or something like that? What What are you What are your um predictions? Yeah, this is what I'm also like thinking because you asked who's using cursor and I would say like from the webflow community that we know like most of like the people that I talked to they use cursor for their programming stuff um when when it comes to web flow it's somehow cursor found its place um like where designers and like kind of like not super professional like hardcore developers work.

Um they they found a place there and a lot of like not super techy developers started using it and including I think a lot of uh like you know hardcore developers are using cursor. So it you know how GitHub has uh issues and how you can like do all sorts of stuff there. I feel like they are tr they probably will try to make it super easy to host your code. So it like just using cursor like automatically like save your code kind of like and it's in the cloud or something. uh maybe like a also CDN to integrate your code somewhere else and something like issues will be like comments that you make and maybe other people can make and agents will start working uh uh on them.

So all of this I can see it happening somehow. Um at the same time what will stop a GitHub you know from just copying their uh featur uh features once they release them and if they are good only time can tell I I tell you who will stop GitHub copying these features GitHub because they're Yeah, but why why GitHub is so bad? Like what's the problem? Why why is their uptime so bad? I don't know. same I I think just engineering as well some would say it's the there's no CEO there's no lead leadership at GitHub um that just from a administration slash kind of leadership perspective they just they've just lost they've just lost it they haven't they've lost their way you know um you know Nokia did the same thing you know they were the top of the game they were amazing but they just lost their way and they just fell behind they weren't keeping up.

So, um, slight tangent, but I was looking at, uh, there's a supermarket called Sainsburries in the UK, and they have a like a a card system for earning points called Nectar. And I was trying to link my Nectar account because you can link it with, you know, do you know what AIOS is? Do you know what AIOS is? No. It's points that you could you can get for flights, right? For you can you can spend them on air miles, right? You can spend them on um plane tickets and and whatever. You can link Nectar and AOS together. And every single corner I hit a hurdle on trying to link these frigin these these two accounts.

Like I I just wanted to change my email. Didn't work. Well, actually, sorry. I wanted to link my accounts. Didn't work. Probably cuz the the the in the the email on one account was different from the email of another account. So I was like, okay, I'll change my email address. The change of email didn't work. The iOS every it was like how how can something be so wrong? I mean, so many things can be so wrong in an organization. It's crazy. And I know nectar aren't the, you know, the king, cream of the crop when it comes to technology, but it was just an example where like just a company can just be poisonous through and through every single corner you turn.

And that's a culture issue. That's a culture issue, right? You find one bug here, one bug there, fine. But every step of the journey, I was encountering problems. And and that's probably what's happening at GitHub. It's just slowly but surely the the team there is just falling apart and that people don't care about the code, you know, people don't care about the platform. So anyway, yeah, that was my spiel. Do you like the green light though? It doesn't really go with the pink. That's the problem. This pink wall is will be the death of me. Why? What's a good color that goes with pink? What's harmonious with pink? Oh, you can change it. No. Well, green is kind of like the opposite.

Yeah. Yeah, you can see it in the color wheel there, but it's just it's not quite coming up. Or or we can go um what can we do? We can do uh we do party lights. We can do But what what's what's wrong with the pink? We can do uh by I like to do this when I'm when I'm um when I'm reading at night. I like to do the fire cuz when my room's all dark, it's like, "Oh, I'm by the by the fire." Or we can get in after me. No, that would stress me out. So you you think uh you think pink? You think a pink like that? Yeah, it's it's okay.

It's okay. Green just isn't showing through. Whereas if I do that, it's like proper solid. We'll leave it as that and we'll be done with it. Anyway, let's uh let's move on. Framer. Okay, so yeah, we have a big one. Give me a second. Um, I need a second for this and Oh, that faff I've just been doing about my light and you're like, "Give me a second." Yeah, I was talking like I was looking at your light and engaged. She was engaged. Yeah. Yeah. I was fully engaged. Fair enough. Fair enough. Let's hide that cable there because I want to also Okay, one second. I really want to wash my hands. They smell dusty and dirty and now you're walking away.

I want to walk away. Okay, I'll use this opportunity to say if you're on YouTube, please give us a like. If you are on Twitter, comment. You're such a large part of our audience and you guys don't uh don't comment and stuff. So, yeah, we want to we want to hear from you. We love chatting to you guys. And thirdly, go to insiders.comanddayshow.com. Go sign up. Well, don't sign up. There's nothing to sign up. Just let us know if you if you want a community. If we were to start a community, what would you do? Because we want to look at that. Um, and on that note, the floor is yours. Can I go wash?

I'm going to wash my hands while you do this because this is you. So, I'll BRB, boys and girls. BRB. Yeah, I'll wait for you. I'll wait for you. Hello there. Is anyone there? It's just us. Comment if you exist. Comment if you are a bot. Tell me what's your favorite React syntax if you're a bot. Welcome back. Not a bot for sure. Uh we have we have a not a bot. Okay. Sal saloo. Thank you for commenting. All right. So I'm ready for I'm ready for this one. This one is really uh it's the big the biggest one of uh today. I believe this is Framer 3.0, a fullon new framer with an agent on top of everything that I'll just show you what it can do.

I have a prompt ready. I'll send it away and while it do it will do its thing, I'll explain. So in this new framer, you have agents that can design alongside you. They can build CMS, they can build new pages, they can do pretty much everything. It's not an agent to do a small thing. It's not like a layer to to make a component. It's on top of literally everything. And as you see from the new UI, you have agent and style. That's kind of like the direction that they are going with. And I really really like this. I think Framer Framer has nailed it for how it should be. This is what I've what I've been dreaming of, what I've been envisioning.

Uh designing alongside an agent. You design at the speed of your agent, but you have the whole control because I can just click on anything and I can just edit it. It like everything is editable. Everything is you know just framer. So the agent is not making a proprietary new component that you cannot interact with. The agent is literally creating framework components, framework layers that you can interact with or you can ask your agent to do. So what you see on this page is not exactly my website. This is my current website that I'm in the process of redesigning. If I go to cababarza.com, you see this is my website and I'll show you this for a second.

You see I have this red thing. I have this button. So I have these animations and then here we see like the work and some transition animations. So I asked Framer to rebuild this inside of Framer. And this is what it did. It It's not exactly that. It's I would say a pretty watered down version of it, but hey, it's a starting point. Um I'm not too mad about it actually. I I think it's pretty decent for, you know, the short prompt that I gave. Now, if you notice, I'm inside of the design area. This is what's really really cool about agents in in this new framer. We already had design. So design is like a canvas free form canvas where you can have like multiple things uh next to one another.

You can just just like Figma. Now, you can use an agent uh ask an agent to do stuff like just right now I asked an agent to create a blog post page based on my designs. I was hoping that it would show it right away. It's not. I guess it's it's analyzing everything. It's creating um a design plan. Okay, it it's still working. Uh the when I asked it to copy my website, it was quite fast. Anyway, so as you saw, it was also trying to do something with the CMS. Yeah, I think it's trying to do something with the CMS as well. You also see these like PlayStation looking buttons. Uh they they showed these in the kind of like uh presentation and before they started.

This is kind of like the new design. We'll talk about the new other new feature as well. But yeah, this was like the marketing uh the marketing way of showing it. And here you see we have these buttons that I have no idea what they are before like hover. Okay, maybe the plus is yeah insert something. Then we have the the layouts. We have text and the last one is what is this? Shapes vector. Yeah. So we have that and then we have everything in the menu I would say pretty well uh and very like clean uh setup and as you see yeah so my page is being created post and if I click on it it's not showing up.

Yeah I it's not showing up and I don't know why I have no idea. It's sayinging down the bottom right. Yeah. So where your your agent? Yeah, my agent is still working. So it it's still working. Uh anyway, so while this is working, actually I can have another chat. And this one I'm going to uh I'm going to ask it to create five variation of my hero section with cool um web gel graphics on the canvas. I don't know how to that off. Have a look there. Just click around like I want to I want to know what models it uses as well. So I can see it's using GPT 5.5. Yeah. So you can go with Sonnet 4.6.

You can go with Opus 4.8. Let's just go with 5.5. Um or you want to use another another one? No, no, no, no. What What do those other buttons do as well? Uh so this is select I think I can yeah select this uh plus is adding like I can add something like an image or something or a file and then this one is just send. So, I send this and I hope it does what I'm asking because I want to use as I mentioned I want to use the the design canvas thing and not the pages because I think this is where framer is not doing a great job at separating these. These are not the same thing.

These are like your final pages. Oh, this was a folder and oh yeah, we see the also the article coming together. Anyway, so these are your pages and this is like your design area. Oh, see ah that was cool. You saw how how it selected uh a section and then did this whole uh duplication and now it's going to add webgl effects I guess. Oh yeah. Okay. So yeah, see these are these are pretty bad. Maybe it will it will be better. We'll we'll see. We'll see how uh we'll see how it will be. Let's give it a second. Um give me a second. All right. So while this is doing its thing, let's check uh our post.

Uh, and we can actually go to the blog post. We see this agent is still thinking. So, going back here, this one, we see it's also probably still thinking. Yes. So, you can have these going at the same time. Now, now, while these do these um work, I will tell you about the agent a little bit more. What's really cool as you saw you can choose your uh model but not just that you can actually connect connect to it via their new MCP and use your own environment you use your own you can use cloud codeex app whatever you use uh open code I guess you just connect to your project and then you're prompted away from there now we see the work with uh the hero is really really bad.

Like from the marketing videos that I saw from them, it was much better. Maybe my prompt was not as good. I had high hopes. This is pretty this is pretty bad. Anyway, so yeah, that that's really unfortunate. And this one is still working. The the blog post is still working. It's it's decent. Um, and I don't have a like proper setup. I literally asked it to copy my uh website and I didn't give it like a visual to to see. So yeah, it's it's it's working with limited amount of data. But um, honestly, it's it's pretty hard to defend defend this. It's it's it's pretty bad. It's pretty bad. But the idea, yeah, I just going to say we can get into the actual design stuff in a bit, but the idea is really cool.

So, can you um two questions I have. One is about pricing, how all that works, and two, so can you just create a CMS collection right now? No pages, no nothing. Can you control from like an administration level to ask the agent to do it? Yeah. Create me a CMS of of team members from the agent. create a CMS of team members just the CMS. We'll just do the CMS first and let's see how it will be. So yeah, from from the updates that I've seen from them, the agent is on top of everything and it it's one play place to interact with and it does everything. And yeah, you see it's it's doing that.

you see or lose it. It feels like this is what web web flow was supposed to do months ago and it's it's so true. So you see we have the posts, we have three posts, we have team members um and yeah, it did it. Now I can ask it to add 10 team members. I want to see what is the So we have ro by email, LinkedIn status. Yeah. Do they have an image? Yeah, they have a photo. I want to ask to add team members but add like another um field. Another field. So what should be that field be? Add um catchphrase field uh members each with their own catch. How do you write phrase phrase?

Yeah, it will understand. So, I I want just to test. I'm pretty sure it can do this, but let's see if it will add. There you go. This has added the field. Well, it's adding the members and the catch members. Yeah. Nice. So, this is Webflow. If you're watching, this is what I've been asking you. And Framer, if you're watching, like, you earned my respect. Actually, Framer earned my respect a few weeks ago when they updated their comparison pages. They finally did it. Um, and they are they are being finally truthful. It's a different story. I've had my own issues with them not being truthful with the community, but now it looks like very fair how they are comparing themsel uh against Web Flow.

And with this update, this is so damn good like in in in the way that they have been thinking about the UX. I have this agent window and it can it can interact with everything. I don't know if it can do like analytics, localization, and settings. Um uh what's what is in setting for example? Can it do redirects for example? That would be cool. Yeah, go on. Ask it to do a redirect. But I I don't have anything for redirect. That's That's the issue. Yeah, just say it. Just say redirect um redirect uh about to about us. Add uh redirect page. Uh a redirect. Redirect. Just say read. Yeah. Slashabout to slashabout us. Let's see.

Uh, it will probably say I don't have we don't have a configure URL redirect to add a redirect and verify the project state. Yeah. So, it seems like it Yeah, it can do this. Yeah. Added. Go check it out. Let's check it out. There it is. Yeah, there it is. Framework. Because the way we described this, there you go. We described this like web flow are trying to uh what did you say that they're separating the agents? You've got an agent that helps build your website in the flow there. You've got an assistant here, whereas this is wrapping the entire plat is a platformwide and a platform all-encompassing agent that can do loads of different things.

the um you know even though web flow uh designers gravitated to web flow funnly enough it's not a designer platform as we can see because uh framer have adopted the design mode and I think that's where they could probably do it a little bit better is actually having design mode and pages mode just to again have that distinction once you know they're there it's it's fine But I think yeah that clear separation of design mode versus pages mode um and having some clear differentiation there. But this is this is great. This is this is amazing. Like imagine uh as a as a framer designer slashdeveloper setting up a whole a whole design system building everything and then you you have all the all the small like detailed controls that you usually have.

This is what web flow is totally lacking. Like web flow if you see this like this is this is what you need to copy. Like we are at a point like to put to put it bluntly there is no way around it. Web flow should copy this like framer has been copying web flow like the whole idea of this UI came from web flow. Web flow nailed the UI for CSS like that's like super clear. They nailed how you can interact with CSS from a UI. But this implementation of how you have agents on top of everything, Framer has nailed it and it's it it's really really looking good. Um, now the results that I'm seeing are not really even close.

It's they are not really even close to what I've been getting from other tools. And just just to make a com like I showed this last week as well. This is what Lovable has gotten a client of mine or like a lead. Uh this is Lovable Design. This is Opus 4.7. It's It's not bad, right? This design, it's it's just not there. This is Figma agent, which is like horrible. Like this is horseshit. It's that bad. And this is like Fable Five. RIP Fable Five. But like I'm I really miss you dearly. Uh but this is like Fable Five. Like this is this is darn good. This is as good as it gets when it comes to agents doing design.

Why am I mentioning this? Because what I'm getting here in Framer is not nearly as good, but my prompts are not specific and I'm using 4.8. I I'm not sure if it's like on the max use whatever like ideally I would not use this interface. I would just use the MCP maybe because with the MCP I can, you know, have it in a different window and still do everything. As you saw, we changed something in the settings like that's that just talks about someone who doesn't know, you know, about this. I mean, let's we called it a redirect. We told it what the slash pages were. Imagine someone who doesn't know what they're doing and don't know about redirects.

They'll say when someone goes to about us, I actually want them to go to the we've renamed it to about us or we've do you know what I mean? They'll use natural language and frame will figure out oh you want a redirect you know. So this is why it's like super powerful to have that you know whereas web flow is so techy. You don't you have to know that redirects are even a thing or you have to know that um well yeah you just have to know the interface and then you it's pot luck whether you'll be whether you'll get an agent that they've built in to be actually able to manage it. Um how much how much are we talking what what's the prices cuz I saw next to the the models they have times one time 1.3 like what what are we talking about here with the price?

Uh we'll look at the price in a moment. Okay, cool. Because uh let's see. So just before we go here, I started uh another agent session. I just asked it to make a responsive uh layout out of this. So you start with something in desktop. It can the agent will take care of responsive. And as you said for someone not techy, this is amazing. And imagine for like even me setting things up for my clients and then my clients ask me to do something. I can tell the agent much better and make sure that I get you know what they need. U so that that's that's a literal gamecher and I think this is where also not just web flow figma has to go.

This this is the way design is going to happen in code. Design and code are becoming the same thing. They they always been the same thing but we kind of like separated them because of you know accessibility like who can access it. Um now that's changing again like design is becoming we design in code we and we have all the access here like this is just this this is just too good. This is this is the perfect uh I I I don't have much to add like I could ask for an MCP and they have that. I could ask for like I don't know what to ask for. Yes, the results are not good, but it's not their fault exactly.

It's also partially the models. Yeah. Let's dig into that a little bit because it's sort of their fault. I don't know what their system prompt is, right? But it's not big deal, right? Let's let's talk into let's talk about this overall design AI design world that we live in. Like I do think this is really good, but I I always think that AI will always struggle with design for those that really care and know about design. Like it it will do some great like you showed us the stuff that you did on lovable or the website that's on Lovable and you just so happen to be really happy with the designs that that came out of it, right?

But the undeterministic nature of AI means that you might do the same thing today and it won't look the same. Yes. Or you've just done it right now and it just doesn't quite live up. You use the same models and this and that, but it just doesn't quite do it. And the problem is is that it's again it's pot luck whether it's created something that was in your mind to begin with. It it's it's it's playing Russian roulette with with whether it's going to satisfy you or not. So this is for me this is why AI will never be good at design. I think it's good and I've said this I I said this in a tweet and I've said it a couple of times on the show already that where AI will be really good for is the exploration.

It's the research experimentation. has given you ideas, but it will never be it will never be good at the final project product purely because it can't do that 100% of the time. You're always going to be disappointed. Probably probably nine times out of 10, you're probably going to be disappointed with the I if you're looking for perfection. If you're looking for the final design, nine times out of 10, it won't it will never be good. So that's what that's kind of why I say design AI is just not good for design. Like you might you might strike gold. You might get it. You might win, but it's a lottery. Yeah, it it is it is a lot.

You said it perfectly, fine. Yeah, perfectly. It's it is true. And that's why if you are a web designer, a de web developer watching this using web framework, I don't think AI is coming for your job. uh if you are a template maker maybe maybe AI is like good at that sort of stuff but even if you like make really creative template AI is not good at creativity uh either it's it's good at exploration and that's my point with framer getting it right when it comes to how they have implemented the UX side of things um and it will just get better from here like we can we know that that yes these results are bad it by the way workspace used over 80% of its li oh my god um um command click on view usage just so we can see just so you have that ready for um pricing and yeah so let's talk about pricing in a bit but yeah so the whole point is it will get better from here their expl their implementation in terms of UX X is great.

So now let's look at the I wouldn't recommend you going in with the expectation that this is going to be your design tool. As a designer as a as a designer I'm doing that as a as a I'm not a designer but as a designer you shouldn't look at this as being a helpful design tool. Should be an a helpful administrative tool. When you said templates, I thought, well, you design the first template and then you'd say, right, take this design and then go create a blog page, go create whatever, you know, tell it to go and use your design to to do whatever. That's as a designer. If you're a marketer, if you're a founder, if you're, you know, someone who isn't precious about the design, then use this for design, but also use it to help build your website, you know.

So that's why I love this implementation because it's so versatile that it really should it really suits a lot of a lot of people. But you know again you might get lucky and get a good design but most of the time you're probably not going to be satisfied just because you have such a precise idea in your head. The reason why well coders developers who say AI is not good at code is because they have a very specific idea of how they want to architect something or how they want to do something. And it's and and and design AI is the same for code as it is for design. If you're very prescriptive, if you say make the border radius of this card 18 pixels instead of 24 and you get you're going to get the exact, then you're going to be very happy.

Whereas if you're not precious, the more broad you can be. There's a graph somewhere that's that's like pressure and uh preciousness and satisfaction with AI. The more precious you are, uh the more unhappy with AI you're going to get. the less precious you are. If you can sit in the middle, I think it's great. Like, I'm not I'm not precious about how a certain feature has been implemented. I'm just like, do it, do it however you want it. It's, you know, it's you got to take the rough with the smooth. If it does it in a terrible way, is it really that big a deal? Probably not. But we'll talk about Ship Studio in a second, Lou.

By the way, we haven't ignored it. We just we'll just talk about it in a second. Yeah. So, now let's talk about the usage. So that this is crazy. So this workspace that I'm on has 3,000 credit units and I'm already basically using all of it uh with just a few prompts by the way. Not a website done, not really not much of design done and I'm already already there. Now you can add more and as just just to be clear I'm I have burned around 3k in usage right just just now you burned two you burned 2.5 out of your 3k but it says like you're yeah um I'm almost hitting the 3k that's what I'm almost hitting the 3k one more prompt I'm hitting the 3k now how much is 3k of Yeah, we don't have 3k, but we can do 2,000 credits.

This is insane. That's yearly, mate. That's yearly. Wait, that's yearly. Monthly credits. If you just wanted to top up for the month, 2,000 for the month makes it better. Makes it easier to swallow. 50. No. So if you go to€10,000 €200 €260 per month and€ 10,000 just like I don't know what exactly that means but like from the feeling of it like I've been using Claude uh the the 20 bucks subscription the $100 and the 200. I' I've I've tried all three and I'm like constantly switching between the 100 and the 200 depending on my workload on in that month. This 10,000 feels like much less than what you get even with the hundred bucks.

It's it's that little. So if I go with something like I don't know the maximum 10k per month for just credits this is this is what feels like the 200 bucks by the way from um from Anthropic and if you go to yearly yeah 50k just on AI credits for a website this is crazy but this is where you know having your own MCP comes to play but I don't know how they because that is technically using their API but you use your own AI so I don't know if they will char they will require you to do something about that they will charge you for that I don't know uh I would love actually I will tweet at them and ask this um I'm just looking at what the frame MCP allows you to do agent skills MCP server.

Yeah, they're not not very clear about what it allows you to do. Um the pricing is free. Yeah, the pricing would be free. Yeah, for MTP server. Yeah, because you're you're the one using your credits, right? So, it's just you're just triggering API. You're just triggering API. Yeah. But like Figma, for example, doesn't let you to use their MCP with the free tier of Figma or if even if they do to a certain degree, I believe Um, so what I'm doing right now is I'm just looking on the API docs to look at what sort of things you can do. Reference and we looked at the API. It's not actually that it's not actually that fleshed out to be fair, but we'll have to look into it.

I think we'll have to look into it. Um, so it's not cheap. If you can do all that with your own if you can do all of that with your own M your own agent, that would be amazing. That would be amazing. But again, you know, use your use your tokens wisely. You get three. Oh, by the way, how much what plan are you on right now to get this AI agent? I'm on the pro one. This is the one. I didn't buy a framer to just have it parked here. Uh I'm a part of this Lenny's newsletter thing and I'm on the pro one. The 30. So $30. The $30 plan gets you the basic plan doesn't get you the AI agent.

I don't know. 1,000 monthly credits. 1,000 monthly credits it gets you. Can you see that? You got 3,000 monthly credits for the 30 th000 monthly credits for the thing. So, um I I I'll chime in here. Lou, if you haven't checked out, we we covered uh Ship Studio last week, but I think it's just I think it's just totally different from from Ship Studio. Like Ship Studio allows you to edit code that you've like created. I think it's just a different like it's a complete different thing altogether and plus ship studio doesn't contain hosting. Ship Studio doesn't host your website for you. You can't set up redirects. Well, technically you could because you could create a versel um uh configuration file and set up redirects like that.

So in a way that you could do it like that. But if you're a marketing person, if you're a founder of a business and you have no interest in design, no interest in coding, like Framer is absolutely going to be your go-to. Whereas Ship Studio is like a good thing in the right direction. Um, and I fully support it, but it's um it's just for a different market. It's just a different person. So, if the ship studio suits you, that's all well and good, but it's not going to suit people who just don't care about either design or or dev or they care about design, but they don't care about dev, you know. Yeah. And they want to tinker with things.

Yeah, I think I will give it a try. But the whole point for me for this video is that this interface, this UX is uh what I find to be really, really good. I asked it to to create a mentorship page based on the mentorship page that I have on my website and for this one it got it super super close to what I already have. So um it's much better in that and it we see it already also made a component. We can ask the agent to make components and uh things like this. We also have branching by the way. How do I do branching? I think branching might be a like a business feature to be fair.

Well, they said they are interfa introducing branching, but I don't know how to That's what I mean. It might not be on your plan. It might be on an enterprise plan. I would say I might it might be that it's a branch of your whole site. So it might be if you go back to design mode and you the PlayStation icons. What do the PlayStation icons do? So this one is insert. Okay. So that's a new element. Yeah. This is just layout. Okay. This is text. I don't know why they do this, but yeah, whatever. Vector. Okay. Yeah, they are technically all this. Yeah. Yeah. Um, it's cool, man. It's cool. I think I think this I I'll always praise Framer on doing things really cleanly.

They're not always ahead. They're the apple of the no code world. They're they're a little bit behind. Doesn't do everything you want it to do, but they just they roll things out when they've got a really good solution for it. And that's what I'll always praise Framer on, if a little bit expensive. I feel like they are quite fast actually. I feel like they are the first in many thing. They are fast at pivoting. They did pivot uh once when they became a website builder and now they I feel like they pivoted into this AI agent flow. They didn't faster than anyone else. They didn't have rims for font sizes for like years, you know?

Like it's those sorts of things. Yeah. It's still kind of like framer. It's kind of still limited. They don't have classes. It's it's a mindset. I would say it's a mindset of how they are. It's a mindset. Yeah. And and it's not my favorite mindset. Uh when it comes to building, what I really like about this direction is the idea of a design canvas being integrated here. Yeah. And then once you have a winner, you explore. Once you have a winner, you set it to be your, you know, main project, whatever. So, that that's that's really really good. So, Webflow, um, I want this. Take note. Um, let's move on because I'm just looking at the time.

Yeah, the time is Yeah, let's let's quickly talk about Eve. you you'll be happy to know I don't massively um haven't gone too deep into Eve but Versel seem to have they want to introduce their own agentic framework a framework for building agents that you can deploy presumably on Versel and I've gone as far as looking at the documentation sorry it's going to be a bit of a boring read but essentially it is a directory It's this is your agent and this could you know you you define your agent whatever you want it to do is it something well yeah you you know how flexible an agent could be right so and then you you just design you give it the the tools it's able to run you give it the skills it's able to have access to sub aents that it has access to and it's a folder structure that defines and and it's this its own encapsulation of an agent which you can then deploy.

So you start writing some code here. You set what the model is. This is this is all this is the bare minimum that you need. Now my understanding is presumably this will sit on Versel once you've deployed it. You just basically give it a task you know um and it will revolve around a set instructions. So in this example you're a senior data analyst. Answer questions about the team's data. Prefer exact numbers. uh if you can compute it, compute it, state assumptions, use tools, and then you just basically get an API that you can just send requests to this agent. If it if you ask it what the weather is, yeah, it might tell you what the weather is, but it's not really like a good use case for it.

So, why they built it? we've had and I was confused because I was like, well, what really like we've got managed agents with uh with Claude, but what what's what problem is this solving? Um, and essentially it comes it basically comes down to just every installation or every instance of an agent being different. They want to standardize something. So that's kind of as far as I've got in understanding this. I think I need to dig in and use it. There's some there does seem to be something quite interesting about it and probably something that um you know especially companies maybe they just want to standardize a lot of their processes uh and all the rest of it.

I don't know. Uh will you use but that's as far as I got. Yeah, I want to try and use it. I do want to try and use it. Um, and I think I'll have more to say once I've used it, but this was this was brand new. This is brand new information to me, so I I'm sorry I haven't I haven't had much uh experience with it. But if you do, let us know uh if you've had experience with it, then yeah, that's that's kind of it. But I think the next one is Versel too, isn't it? because Versel have been busy with the V design mode. Yeah. So, did you want to take this one?

Yeah. Uh, I want to show the video. It's kind of like uh not as good after that framer one because it's still in the same direction. Well, you know, we spoke about ship studio. There's just different implementations, isn't there? Yep. Oh, all right. I will add I will just add one thing about that agent because I'm looking at this page here. Mhm. So the the channel connect is an interesting one. So you can actually just connect it let's say in Slack. So this this Slack agent will forever you could have it on Discord. Maybe it's your documentation for your company. Maybe it's you know uh something but like you can at@mentntion this agent or chat or or you know chat to this agent and it's just a specialist in doing in doing that one thing.

So imagine these planted throughout your organization throughout your company and they are super good at that one thing and in this instance what was it? This is get weather. So whenever someone asks the weather or whenever someone asks weatherbot in your discord, it will just tell you the weather. Obviously, you can think of clever, more clever use cases for it. But yeah, I think it's just a way of getting agents dotted around your organization and uh providing insights that are relevant specifically to your organization. So yeah, that's all I'll add. Yeah. Okay. So um All right. So Vzero has now a design mode just like every other vibe coding and agentic tool. You can talk to the agent but now you can also directly interact with what the agent creates.

So introducing design mode. In design mode you can comment on something and basically stack a list of comments to your agent so the agent can take care of it or you can start directly interacting with the canvas as you see here. you start interacting with it and you just change color, sizing, whatever directly on the canvas. That's really cool about it. Um, it's not just there. I will show you in a second. You have a whole panel similar in a way to no code tools like web flow and framer where you can, as you see also here, you you see the layers, you can you can change stuff. I I also really really like this.

So this is a real really clean implementation of a design mode inside of an agentic tool. Mhm. But one thing is missing and this is what we covered in the framer video and that is a canvas, a free form canvas where you can explore ideas, put them next to one another, compare them. That is missing for me. But the rest of it is really really nice. Uh I checked also you know you can go to their docs and you see the design mode. Something that is a bit odd. Uh so you have the chat and design and this is the design panel. What's a bit odd to me? This is a bit ugly. Uh like we have Figma, you know, the the the style panel.

We have the web flow style panel which is my favorite. We have framer which is also really clean and they are so user friendly but I feel like this is just a you know a a bit too cluttered or like not very easy to read. I mean you can focus on on it if you want to but it's yeah uh I it just I find it odd in the world where we have uh web flow perfecting that uh style panel we still see new tools not really having having that level. I mean with web flow you know with the padding and the margin controls um and also with the even with the uh border things it's it's just things are just so so clear.

Anyway uh this is an interesting update. I did use Vzero for some time. Um I had it with the Lenny's newsletter but now I'm I'm not a paying customer but this looks really nice. uh you ask your agent something and you go to design mode as you see here and you just click stuff and and and change them. Uh as we said before this is the future like the no code tools are adding aentic first uh UX and the agentic first uh tools are adding a design mode. So both are becoming essentially the same thing. Yeah. Uh just to comment on your you know we've got web flow and this and that and whilst I agree with you I think you're coming at it from like um like a design first perspective whereas like the dev first perspective is like you know we were happy with WordPress for ages just bunch of fields that we just type in what we want like with there's devs who are more technically minded that just can't think outside of that like if I want I want this to be that then it's like you numbers will be perfectly fine rather than a UI.

Um so hopefully they'll get there. Hopefully they'll look at other no code tools as inspiration and start improving it. But for sure like I I think that uh it's not as important for some people. You know we we're we're really entering a world where it's like one size could fit all but it's just never going to satisfy anyone. And similarly, you know, the uh canvas thing, I wouldn't want that in V 0. I don't want canvas in V 0. Do you know what I mean? And also, my question to you is like, why wouldn't you why wouldn't you just keep Figma for that? Like, why why couldn't you be would you be happy with a Figma import?

I'm I'm using Figma right now exactly for that. And I'm super unhappy because I don't see a point in doing design and development in in a separated um fashion. They are the same thing. Like the design if I'm creating a design in Figma, right? So um like if I'm creating this design in Figma, this is claude code literally thinking in HTML and in code and implementing that in like translating that to Figma and then I translate that back to HTML like there is no point at that anymore. What I want is a canvas. What I want is like essentially framer a canvas where you can explore, you can duplicate, you can play around with things also manually, not fully agentic but also manually.

Once you're happy with it, it's already code. That's my point. It's you don't do a translation again. It's already in code. You shouldn't worry what's happening under the hood, right? Translation. It could translate back and forth, back and forth a billion times. like it shouldn't worry you that that's that's what's happening but it's scattered. It's not in a seamless flow. So I would it takes a lot of time. I would argue I would argue and say actually I quite like the separation. Like if I was using V0 like I don't want this design mode interface. I wouldn't I wouldn't ever use that. I wouldn't care to. I guess it would be nice, but like this is where you start to get like really cluttered applications and frame is a great example of like that it's really good what you can do.

You can have you have got that design mode and you have got the the implementation but it is a fairly basic tool for all intents and purposes. it sacrifices in its how far it can actually go for that, you know, universal sort of thing. But I'm not gonna I'm not going to say you're you're right or wrong. I'm just saying like it's just it's an example of like how this what you're never going to get this oneizefits all thing. So it's you're going to have to just gravitate towards the tool that solves your your needs cuz your needs are going to upset a bunch of other people. Yeah. But you can have it as an option.

That's my point. My point is just like as you said like with framer the the canvas is a separate thing. What I'm imagining is exactly that you have you can have your own workflow. You can just work from the terminal and create the thing on you know whatever platform that is. another person can use the platform directly with AI and another person can do it on the canvas um and manually. So it's not one, you know, just one linear way of doing it. Um it's just you have these options and Framer has done a great job at making it still look like it's a very minimal UI. Yeah, but it's a simple tool. That's that's the thing.

It's like you know um what master of all trades master of all trades what the jack of all trades master of none of comes to mind in that you know if you try and do everything they're not all going to be great you know and I do want those very dedicated tools and I want them to be great so that's that's all I'd challenge that with and again you're not going to you got to find a solution. Every tool can't do everything. I don't know. I feel like uh if any of them add a canvas uh and a CMS, it's kind of like very close. Um Lou is saying these tools need to add a new thing like a throwaway section just to play around and experiment.

And I feel like this is what framer has uh in the design canvas. These are just experimentations. You can literally drag things around and just play with them and do whatever you want with them here. And they are not going to, you know, harm your website. They are not like going to break your website. These are just experimentation here in the canvas. And then once you are ready, you can put it in the final uh page. Yeah. But cool. Anyway, uh V0ero design mode like every other tool. Uh this is really cool. Um we had cursor as a note here. Um obviously I can talk about claw design but what was cursor? Um I'm not sure why cursor is here.

Did cursor has a design mode as well. Yeah, they've been Was that something you wanted to show or mention or whatever? No, not really. Okay, fair enough. We'll move on. Claude Design have also improved their design mode, but we couldn't really figure out what they've actually done. So, yeah. Um, let me talk about GLM 5.2 might actually be the first. Well, actually, let me think about this cuz Kimmy was darn good as well. So, I've been playing with GLM 5.2 2 and Kimmy K2.7 code. And I have to say they've both impressed me so much. And I think I think we had to put this to the test. We've reached a point where Chinese open source models that are cheap are actually able to do real work.

Um, so yeah, GLM 5, Fable 5 level, not Fable 5 level. And this is what's this is what's crucial about that. This is what's crucial about this. Um let me get up GLM uh 5.2. So just to um yeah just to let you know where am I? Screen studio. Sorry. There we go. Um this is the sort of prices we're Hang on, hang on, hang on. I'm working with one monitor right now. Okay, here we go. Uh, cool. So, this is the sorts of prices we're looking at. $12 a month. This is you're not really going to use this. This is for anyone doing any real work. You're not going to really want to you you'll reach your daily limit like or at least your hour like 4 hourly, 5 hourly limit in in an hour.

Um, realistically, we're talking about the pro or the max. And if we're going to go for it monthly, it's $64. Quarterly 57 and yearly is is $50 a month sort of thing. So really really quite a cheap alternative to something like Claude. And I played around with it. Let me actually can I show you? I got it to redesign an app for me because apparently it's it when you say Fable, it's apparently doing better than Fable in design benchmarks. I need to design benchmark. Show me. I I want to see that. Let me show you that design benchmarks. We're not all about benchmarks on this channel. Um, sign benchmarks. Oh, my video just comes up at the top of the search list.

That's cool. Um, here we go. Visit. Oh, it's my own post again. I didn't plan for this, but my own post is coming up in Google when I'm trying to search this bloody benchmarks. So, it's coming up top. Can you see this? Um, again, I'm single paging right now, so I can't see. It is coming above Fable on the uh design arena. So, I put this to the test. And if I get it up for you now, um, what was it called? Um, email outreach. Here we go. Um uh bun run dev. So this is the GLM 5.2. This is GLM 5.2. And where do you access it from? What's that? Like um I use it from open code.

I use it from open code. So, I didn't run this locally. It's like a 1.5 trillion parameter model. Um, and it's interesting you say that. I think someone someone's quantized it down to three bits and is actually still really good, which gets it to around 250 gig. I can't run it on my Mac, but 250 gig can run on a Mac Studio, you know. Um, and the performance is is quite good. I can't show you the leads page, but it's not it's not really all that bad. You know, it did have it started with a really clawed design first. So, there was a few parameters it sort of worked within, but I but I gave it a I gave it a um prompt that I have given Gemini in the past, which is here's an app that I've got.

Don't change anything except just redesign it. Don't change any of the layouts, though. And I I have to say I was more impressed with Gemini than I was with this. It It's arguable because this is such a boring app. It's it's an outreach tool, you know, it's a basic dashboard, but um all that to be said, I added some features. I uh you know I changed some code. It got it done really quite fast. It wasn't like it didn't offend me whatsoever. It was it was it did what I asked it to do. And like for the first time I was using this a Chinese model and it just seemed to get done. And so where I think this is where I think this is a real turning point is opus fable you know for real research and creating clear implementation plans for a lesser model and then you've got the Chinese models working away getting getting the real work done that I want to put this workflow to the test and see how useful it is.

But it I was like it was just really satisfying knowing that it was barely costing me anything. And I was like, "Oh, I think we've reached Opus 4.5 territory with Chinese open source models now." And I think they're really usable. But as with all of this stuff, I got, you know, people slating me in the comments, these are rubbish prompts and this is not good or whatever. You all have your own use cases. You all have your own styles. I strongly suggest putting these to the to the test. Um, I've bought myself a year's subscription because I've got loads of referrals. Actually, I'll put I'll put my referral link in the in the description. Can you send some of those to me?

Tokens. What? Credits. Tokens. Yeah. Like gifting each other tokens. So, no way have I earned $445 in the last few days. 52 people 52 people have been invited through my video. This is crazy. Um, anyway, I'm going to give you 10% off. I'm going to give everyone 10% off uh off GLM. So, this is through GLM coding plan. This is through. So, yes. So, let's let's break this down. I use this through open code. Um uh you can just you can just open up open code and just type slash slash login and scroll down to uh ZAI coding plan or GLM coding plan, one of the two and you just you use it like that and it was really clean, really simple, really fast and uh I I just got to work and for Zen AI.

What is Zen AI? Zed AI. Zed AI. Z AI. Um, it's just the name of the the company who made Zen. Ah, okay. Because we have Zen uh on Open Code and I'm like, "Yeah, Z." So, you know what I would suggest is buy the light plan for 16 bucks. Give it Yeah. Give it a light bit of usage over the course of a week. Don't push it. What? Don't go too hard with it. Just give it a whatever. because I've now bought the light pro for a year and I was like, "Ah, I should have bought the light pro for a month, tested it out, been satisfied, and then gone for the light pro."

But then I have all the these referral credits, so I'm I'm privileged here. But, um, yeah, definitely, you know, it's cheap enough to try out, you know, it's cheap enough to try it out. So, I would, uh, I would recommend. All in all, really happy with both GLM and Kimmy. Both of those both of those new models are doing really really well and uh I I had a lot of fun using them. Cool. I will give it a try. Yeah. Um you better. You bloody are better. Right. Are we going to do snap or are we calling it quits there? Not really. I don't know. Okay. Not mine. We'll wrap it up there because I wouldn't mind I wouldn't mind bouncing as well because I got um I got DM for me.

Cool. All right. Well, saying all that, Lou, big shout out to Lou. I'm just showing a random comment here. You don't have to read it. There you go. A man I love Vzero. Shout out to Lou. He's been been cool on Twitter, no less, as well after saying that we've uh we've haven't had much um Twitter engagement. You've been chiming in today. If you are on YouTube, give us a like. If you go to commandiohow.com, you can watch everything we've done, all of our previous episodes and stuff like that. Um, and insiders.comow.com. Register for a makebelieve community that we're thinking of spinning up. Let us know what you'd want to do uh if you were to join such a community.

Join us next week for all AI design and dev news. We're live every Friday. Hey, if you're new here, love to have you on. And if you're on YouTube, give us a like, leave us a comment, something like that. I think that's it. Yeah. Yeah. You've been Sam, I've been. Keep on vibing, baby. Peace out.