#45 Claude Code Leak, Pretext, Cursor 3, Gemma 4

uh, did just cut out, but I managed to get it back on. Just know that if it goes out, it'll probably just take me a second to, you know, just pop back in. Hopefully, my internet's okay, though. As long as you're not lagging, are you Where are you? Um, where am I? What can I say that you're going to understand being in not from the UK? No, I mean, like, is it like your parents' house? Is it like a friend's? Yeah, it's my friend. They live in, like, the middle of the country, and, uh, yeah, just a bit of peace and quiet for the weekend. Although, I am working today, which is very stressful. Um, but, you know, you might see me disappear often, you know, scream and shout at something.

Last week, I got to say, like, you have the best view so far. Like, you should make all your YouTube videos in this video. It's the best. You reckon? Are you saying Are you saying that? Yeah, the composition is just perfect. It's just so artistic. Oh, he's gone. He's gone. Okay. So, now he he's back. I wanted to say comment something, but yeah, he's back. I'm going to just take my pills. Yeah, I can still hear you when all of that goes on, although it seems to be frozen now. Yeah. I'm going to try to connect these babies, and we'll start the show. This is going to be a fun episode, but this camera has, like, I don't know whether this is going to work but it has like a camera see any different magnetic Yeah, yeah, yeah second little such a cool camera but it's so it's a CGI camera but they just doesn't work with Mac and I saw our friend Ryan like streaming with one of these little cameras.

And I even messaged him saying how are you getting that to connect to your laptop because I've not been able to get this to consistently you'll see me drop out you know and he said just try the right cable but I literally stayed up one night trying all of my cables and one of the ones I've got with me one of them seemed to work and it would never disconnect but as we're just finding out now it's not working. Why don't you use your iPhone? the continuity It could that might be a fallback solution but it's also not a microphone. Well the microphone on your Mac is really good also on your iPhone. For anyone watching if you want to make videos and you don't have a camera like this you don't really need it I don't have it for the show I just have it because I used to do photo and videography.

If I would have to start today from scratch and just make videos I would absolutely just use my iPhone. That was just a test shot now. How is the microphone now? How's the microphone now? It's not as good as before but I would say it's acceptable. Let me try something. Let me try and link this bad boy. to my iPhone I've got linking mode. I disconnect that. Okay. He's back. He's gone. Now he's back. Okay. Weirdly, it's not connecting to my my phone anymore, which is very annoying. I connected this Bluetooth this this microphone. Which also, by the way, DJI, if you're listening, doesn't freaking work. Just taking care of the lighting. Well, um I see people are joining.

Uh we We we're trying to get some some of the I I don't know, like the camera setup work working and also the the mic situation. And And we'll be starting in a minute. I I wonder if it's because I linked this to my phone that now my phone isn't being picked up by the Do you want to connect your phone? Is that what you're trying to do? Yeah. Um use continuity, right? Hang on. Hang on. I'll be back. It's uh How funny is that? Just now my phone just isn't being picked up by Continuity. We'll have to figure that out. Uh and that might need a restart or something. Well, that Well, that's why I left and then came back to see if it would need to restart cuz it was just working.

Of course, it was just working. And the moment I connect this this this thing to try and get better audio, you're not picking that up, are you? Uh no. I'm also um setting up a few things. Might just be this camera. I could even just do that. You know what I mean? I could even just do that and that would work, but my phone isn't being picked up. I'm not sure if you heard that, but you can even use your phone as a microphone. Uh You maybe shut off the DJI completely. I have. I have. You have? Mhm. We've got full on AirPods. How was that? Uh no. Is it the the previous gen? I'm going to close Chrome and then come back.

Okay, so Now he's joining again. We have interestingly enough, we have some people joining watching us. Getting things sorted. We decided to go live before we have everything ready, which is more fun. Um yeah, this is how the show is produced. And Sa- Sam is trying to get connected and I'm trying to finish up the the content and I'm just preparing some stuff to show you guys. Um okay, so Can just I reckon we just wing it, mate. I reckon we just wing it. Uh you can't use the previous So, if you use the other view the other camera, it gets disconnected, right? That's the issue. Yeah. Yeah. So, can you for this This view would be fine if you don't move the laptop because when you move it, it creates uh some noise and also the vibration is a bit annoying.

Mhm. The video quality is not nearly as much important as the audio. Yeah. Yeah. I'm saying this and myself I'm making noise there. I wonder why my phone just isn't being picked up now. Like, what What gives, Mac OS? Like, what the Do you know what I mean? Like, I had my I had my phone connected. It was It was working. How do I Can I Let's go into uh photo booth. It's not even picking up in other apps as well. Like Like, it's just not even being picked up in in any any other app. Which is just super annoying. Oh, well. Oh, well. And then the next issue is is probably my battery is going to run out, so I might have to figure that one out.

What about this? What about Actually, I'm going to use the time to see What time? There is no time. I'm waiting for you to go. I'm waiting for you to start. Okay, I can now How is the audio? Is this okay? Yeah. Can you even hear me? Wait, can you? Yeah, it is it is fine. I I can hear you definitely. This is my iPhone audio. My my iPhone camera which is now working. Yeah, why do you need the cable for charging? I hope I won't Yeah, okay. This Okay, let's go. How is this? Is this better? Can you Is this coming out of this microphone or am I still No, this is No, it's not that microphone.

It's another microphone. It It's not that. Yeah, the iPhone is good. The iPhone is good. Let's stick with that then. Yeah, it's good. Cool. Let's see how we get on with this. Thumbnails are going to be fun. I'm going to ask Nano Banana to remove them. Okay. So. Um, how does Google pronounce this by the way? Okay. Should we start? Yes, let's start. All right, so for the audience that has been with us so far, thank you for just chilling there with us. Uh, now we are starting officially and here's a list of things that we will talk about. So, this week we have a huge announcement from Google. Uh, an open source or open It's not open source, it's open weight, I believe, right?

Uh, model has been released from Google, Gemma four, super powerful. Uh, we have Anthropic accidentally leaking their entire source code. Um, and we have cursor design mode. We have pre-text to talk to you about. Uh, it's a nice library that presumably is going to change how uh, we will build layout and uh, UI for the next decade. And what else do we have? That's pretty much it on my list. The You mean the You mean I meant the the the leak of the source code, yes. Yeah, yeah. Yeah, so Yeah. Uh so all of this in today's episode and more. Uh this is Command AI. I'm Kavarza. And I'm Sam and this is Week in AI Design and Dev.

Cool. So should I start with site of the week? Are you excited to see site of the week? I'm I'm very excited about this. I'm very very excited about this. Okay, allow to share screen. Cool. Um All right. So this is site of the week and it's one of the most beautiful websites I've seen in a while. And again, it is from Shopify. As you see, when we load the site, there is this glitchy effect. At the first site, I thought that that's glitching or lagging, but it's not. It's just how the effect is. And as we scroll, we see some really nice imagery and we see how these masonry layouts come together and I'm like, why why did I never think of this like for masonry layouts that I have built in the past?

It's really nice. Then we scroll, we have this clock. And so far you're like, yeah, this is quite nice, but there is nothing crazy about it, right? It's just nice. What is so crazy about it? Well, what is so crazy about it? This pill. You click it and now all the sudden you have this interdimensional view of the site where if I bring the mouse down, it's like the camera is looking down the page and I can see the the rest of the con- content down. And I can scroll. You see the clock? It's there. It's so trippy. It's so freaking cool. It's just such a simple concept concept. If you just think of the idea, it's very simple.

But look, I can look up and I can scroll up among uh like between the layouts and it's just insanely I don't know, it's so creative. It's so beautiful. And as I said prior to the show, I said, "This is a really short scroll. It's not like one of the other projects that they've done that it's like really long. It's super short, but it has some content in it. Uh there is the slider. You can uh watch the content and there is this nice player that you can open close. So, there are more details to it. But yeah. Short and very very punchy taking the the Twitter the designer Twitter by storm because how beautiful and creative it is.

Again, another amazing job from the Shopify design team. shopify.design I was going to say Yeah, I was going to say they always manage to do creative interesting websites, which is really cool. Yeah, if you remember they did this one. The AI the Renaissance. They they were playing with AI and I can't see how they went from that to this. How? Because in that one, they also had some interdimensional thing with the lock and the two browser thing and someone probably said, "Hey, we can do this in the same view and just have a x-ray 3D x-ray view of the uh the site. So, you can see how the ideas are being recycled. Uh I absolutely love this.

I love that I can see that it's the same idea being recycled, but done so well, so creatively. Just to drive tons of engagement, people get to talk about this, and we enjoy the beauty. Ah, this is so freaking good. Yeah. That was it. Short and punchy. what is the website about, really? What is What is the website about? I don't Make the new normal. Uh how we work is changing shape. Um so, is what's possible. Okay, that that's an interesting statement. I think it is about design. And how it is changing um Mhm. what we can do on the web, and especially on e-commerce. Uh let's see what these are. I I do think it is about design, like the possibility of what is possible to do.

Which is changing, right? I'm working With AI, it is changing. Mhm. Um I'm working on an app at the moment, a website and an app at the moment, which uses like video. And it's it's killing the app. Like it's relying massively on video. And this is still, I don't know, correct me if I'm wrong, I could just be trying to convince myself otherwise here, but video is still such a huge problem for the web. Yet, we still we get really with 3D is sort of an easier thing to deal with. The website I'm creating has video transitions between certain scenes. And it just takes while to download, to buffer. It's still such a complicated thing.

Yet, this seems to be running fairly okay on modern machines. Yeah, so video this is what a lot of people get wrong. They see a 3D website and they assume it is the heaviest thing ever. It is not. Like if done well, WebGL can be so much more performant than video. Um and with video you have like real limitations with bandwidth. Uh with aspect ratio you have issues. You have issues with the maximum resolution. Um and aspect ratio, I know I said it, but like I want to say it again. It It is a major thing. But when you render things in a canvas, you you can calculate things and you can scale things depending on the resolution and based on, you know, so many different factors.

Uh and video it is just more strict. And it it requires high bandwidth and it is really easy to drop frames and yeah, it is just I agree. It is more headache to work with video since ever and it is still. Uh not to mention that like WebGL is significantly more difficult. It is difficult, but it is also more versatile, more performant and can do so much more. Mhm. Mhm. Yeah. Yeah. Cool. Let's uh Let's move on. Let's continue. Uh do you want to take the next one then? Jim I'll I'll do that. I'll do that. I've got a window open here. I I have also like windows open if you want. I could also I don't know.

Uh I would I would have to read through. However, you like. Um I will share my screen. I'm dealing with one monitor. I'm I'm I'm dealing with the the difficulties of a single monitor right now. You do have an iPad, right? I have an iPad, yeah. You could use it as the I know you know. I was going to say, do you want me to go set this up now whilst you're sitting here and No, no. No, no. It's fine. It's fine. It's fine. So, Gemma Google dropped Gemma 4 four sizes from 2 billion to 31 billion ranking number three on the open model chart globally. Built from the same tech as Gemini 3 under Apache 2.

So, it's it's free to use. It's free to use. Um there are four sizes apparently. There's 2 billion, 4 billion, 26 billion, um mixture 26 billion mixture experts, sorry, and a 31 billion, obviously big daddy. Um so, a range of different um sizes and all that lot for your needs. Uh and the fact that it's ranking I mean, let's see. Is my screen sharing by the way? I've I can't keep jumping around. Your screen share is working, but yeah. If you go to the one chart Yeah. This one here? Yeah. If you zoom in Model Yeah. That good? Yeah, perfect. Yeah. So, what are we looking at here? It's Qwen 3.5. This is the ELO score.

What is ELO? This is the ELO the This is generally the That's the ELO the I don't know what a measurement they are using, but we see that the it's significantly better than Qwen. Well, not significantly different sizes, right? Well, you got Qwen This is This is 27 billion, and then this is 397 billion. So, the size there The it's beating Qwen in comparison to its size. So, these are 26 billion, that's 27 billion, 31 billion, just a little bit ahead, and they're all roughly you know, well, I mean, this is amazing. This is where you want to be, right? So, really really cool. GLM 5 This is our good friends Minimax M2.7 then no.

But, you know, GLM 5.1's just been released, so it's a shame we don't get to see that one. But, yeah, I mean, for a free model to use on whatever you want, it's it's pretty good. Like I said, built from the first same technology as Gemma 3. Gemini 3, not Gemma 3, correction there. Most capable model family you can run on your own hardware. So, I wonder what the actual I mean, let's open this up over here. And do we see a actual like file size uh files and versions here? And 10 gig by the looks of things, I reckon. I reckon around 10 gig. That's the smallest one. So, it's not tiny tiny. I would expect something around 2 billion to be half that size, to be honest.

But I I guess it's pretty capable for you know it can do a lot for uh for just 2 billion parameters. Um The entire family moves beyond simple chat to handle complex logic and agentic workflows. So, this could be a interesting you know, vibe coding tool. I might have to try this, to be honest. Try and see if I can get it working on my vibe coding setup. And uh yeah. Powerful, accessible, open to power the next generation of pioneering research and products. We've sized the Gemma 4 model specifically to run and fine-tune fine-tune efficiently on hardware from billions of Android devices worldwide. So, not not uh not Apple. Um to laptop GPUs or advisor developer workstations.

So, yeah. I think I think I'll be keen to put this to the test, definitely. Cuz you we we've got all these Chinese models that are making a lot of waves like yeah, Qwen 2.5. Um Mini Max um None of None of which can really run on your hardware. It was just Qwen 3.5, which we did a we did a little thing on a few weeks ago, which could actually run on smaller hardware, um which was okay. Um even going back here, actually. Let's remind ourselves of the Where are we now? Gemma. Let's all Gemma. Let's all Gemma. You know, um Quen 3.5 just not performing as well. And that was the one that could potentially run on your own hardware.

So, yeah, pretty cool. I'm going to try it out. So, So, if I add uh to the conversation, uh I see this yet as another win for Google, uh especially now Google is again picking up steam as we see. Um something that we didn't cover, by the way, another reason why you absolutely need to watch the show Silicon Valley. It's like still relevant. Uh we didn't cover this Google research introducing Turbo Quant, our new comparison uh comparison algorithm that reduces LLM key-value cache memory by at least six times and delivers up to eight times the speed up. Um I'm not sure if this is directly contributing to Gemini 4 or not because this is about the the memory cache, but uh this is another thing that is Google doing for like reducing uh and just making the the model smaller and faster.

Um w- which is even more important if you know I don't know if you know about uh the the new limitations by Claude. So, Claude now uh has introduced more limitations uh during peak hours to users. And it just shows that it is it's very new. It's very new. I think it came out a few days ago that they are introducing new limitations. So, you hit your limits much faster. Why all this matters is because it seems like we're not getting the models much more powerful, at least not order of magnitude more powerful. What we are getting is a lot of um uh efficiency gains and making the models smaller and making the smaller models uh faster and cheaper and things like that.

So, this becomes even more important so we can do maybe swarm of a swarm of agents uh and that kind of things. If I also read here uh this tweet also resonated. Uh it cannot be overstated how big this news is. Being able to run a frontier AI model on your own hardware means token costs are effectively free. Given that you have the computer and the electricity, but yeah, it is effectively free. AI will be uh you I don't know how to pronounce this. Ubiquitous and cheap Ubiquitous and cheap as the internet itself. This is This is like really big. It is becoming uh a commodity like a truly in a way that you can have proper models on your own hardware and for all sakes of um uh privacy and things like this.

Uh this is a dagger in OpenAI and Anthropic. Let Let me Let me just jump in here though as well is that this person What was his name? Sorry. I don't know him, but it Yeah, it makes sense. Nick Nick, I don't think Nick has used this model. He's or they, sorry. Got to be PC. Um they have you know, what he's saying in theory is correct. Like, we're definitely moving in that direction. It's the direction a lot of people want. But, like I don't know if you know this, but my channel has been really peaking testing local models, running local models, and using them to code, using them with open core, doing things with them.

And yes, granted, my usage is primarily for code-focused. I would say that's one of the least demanding things outside of chat that you can do with AI. Running this on an M5, you know, again, a MacBook Air, but I don't think we're going to see leaps and bounds when it comes to the Mac Pro and the Mac Uh sorry, so M5 Pro and the M5 Max. I I still don't think we're there with AI. And what he's saying here seems like sounds like that he's actually run this, and it's actually something usable. But But I've got a hunch that he hasn't. So, I put Basically, what I'm saying is the long short of it is, I want to use this.

I want to actually see it in real-world practice, cuz this sounds like it could apply to anyone. But I reckon what he's saying applies to those few with very, very expensive hardware. Um just being able to run it locally, you know? So, yeah. I'm apprehensive to take his tweet to the everyday person like you and I with our MacBooks and our M1 Pros and our, you know, whatever. I I agree. I personally tend to use the the models in in their absolute maximum setting, unless I definitely think that the task shouldn't be overthinked. Um if it's the case, I will drop to a lower model, but I don't mind um waiting 10 extra seconds or even a minute to get something better.

I don't even use the Haiku. So, I I don't want I don't know where this is standing comparison to Haiku. That that's, uh, what would be interesting to me. I don't know if you know from the top of your mind where Haiku would be here roughly. No. No idea because we don't know how many parameters Haiku is. Yeah. But definitely at the end of the day Yeah. Uh, at the end of the day this this is this is an open model. Um, you can run it on your own chip. May maybe there are people with like lightweight enough, uh, tasks where they can use this. I don't know. May maybe there are tasks like that.

Um, maybe it is going to be useful for bigger models. Uh, maybe you can have Sonnet creating tasks for your local, um, AI. That could be something in the future. Um, yeah. But here in lies the problem though that um, means that people will will never be happy with AI locally right now and that you're comparing it to and many other people will be comparing it to the frontier models, Open AI. Yeah. Like you will just not get that. People were surprised that hang on, how were they model that's running on so hardware that's halfway across the world, how is that still faster than AI running on your actual machine? And the truth of the matter is they've got got such powerful hardware.

You need such powerful hardware to run this stuff. And this people when people first start getting into AI they're like they are comparing it to the frontier models and they are weighing up those options. And ultimately, if you are doing that, you're going to be very, very disappointed with how well AI is running locally for the everyday person right now. Quen 3.5 was a glimmer of that running on that app, but that's just doing chat. And who really cares about AI chat running locally? I mean, it's nice, but people want to get real work done. And I just I just don't think we're there yet. And all that to be said, I want to try it.

I want to I want to pull this down and give it a go. Definitely. It is quite fun. I mean, I'm paying 200 bucks for Claude. Uh I'm under max model again. Uh after downgrading, now I upgraded again. And I'm thinking like I'm paying that and I'm not hitting the limits. So, why not use, you know, the best model for whatever task I have? Why should I go through the headache of doing this unless I find it fun. Uh which I do. And I think I will uh find some time to download this, maybe set up a chat. Because as you know, I have shown it before, I do have my own chat app. Like a complete chat app with web uh search and some some comparison tools and things like that, completely vibe coded.

I'm thinking, why not put this like run this on my computer and make it available to my friends and family? Something like this, just just for fun. Just as an experiment. Maybe I want to do this. Uh that would be really fun and it wouldn't cost me, you know, a fortune letting uh my friends, you know, um run the model from my computer. So, something fun to do. Yeah, exactly. I think yeah, that's what you Well, games. yeah, yeah, yeah. Oh. It's funny that's where I'm sort of starting to land in my content is doing a lot of local AI stuff. Oh, wow. Yeah. me or was that you? You just went crazy. No, I think you're hallucinating.

Okay, fine. No, no, it's just a cable. Okay. Jesus. Yeah, like that's all that's all that's all all us lot watching YouTube can expect from this stuff is just having fun with it and just enjoying it and seeing where it goes. We can probably pretty much ignore. I don't know, I'm speaking too soon. I haven't even tried the thing but we'll probably we can probably safely ignore this sort of stuff for us. You just need I think even the the H100 was even mentioned in the article. Do you know what I mean? That's like that's like a I don't know, do you know how much that GPU is? I think it's tens of thousands. I I think it was 40k.

No, I think Okay, 40 so way off then. H100 I'm I'm searching. Yeah, 40,000 thousand 38,000 but I Yeah. How much VRAM is it? I think it's got like 97 gig VRAM. So you need a few of these. You know, you need a good few of these. Is that it? 80 gigabytes Yeah, exactly right. Yeah. That that thing looks sexy. It looks so good. Look at the insides of that. Oh, yeah. Look at the size of the size of the Look at that. Yeah. Oh, god. Right, we should move on. It's getting a bit too steamy. Okay, I'm I'm adding this Okay, I'll I'll buy it off the screen. Yeah, yeah. I'm putting it in your bookmarks.

Cool. Um Oh, that was embarrassing. That must be awkward. No. I'm also watching two screens, so if I'm getting distracted a little bit, I'm simul- multitasking right now. Uh I have something that I haven't Let me just show it and we'll we'll see what you think. So, I saw this So, I saw this tweet about this tool and it got me I don't know if I'm creeped out or excited. You add this AI tool to your Google Meet and you chat with it. But, they it has a face or like multiple faces and it looks so good. I mean, it's a taste thing like there there are different styles, but Well, hang on. Is this a Google product?

No. No, it's not. It just Oh, Pika. Pika. Con- conversations tend to get better with a face and a voice. Um I don't know. That's why we are thrilled to release the beta version of the first video chat skill for any agent powered by our new real-time model Pika Stream 1.0. Uh the skill preserves memory and personality and enables real-time adaptability uh and it and if you use it with your Pika AI self, they'll be able to execute agentic tasks during the call. So, that's where we are going to spend our GPU uh powers with like, you know, giving agents uh I guess a face and personality. And it's on GitHub. I haven't tested it.

I just saw it and I I don't know. It it it it looks good at least the design-wise. It looks really I don't What's the goal of this thing? I don't understand. I don't understand. You instead of chatting like in a text just with text, you chat with faces and it seems like they are real-time and they just reply to you and you can have multiple ones with multiple I I guess you could have Maybe you could have Anthropic models with OpenAI models with Oh god, you could have Grok. Uh you can have all of them in a chat and then chat with them. Grok had yeah. Yeah. Yeah, that um but yeah. Exactly. So, that's the idea and it seems like that it can also execute tasks.

But what type of task? I mean, I don't think it is just it is like more powerful than anything else we have. It's just that it has a face. A cartoonish face that seems to be real-time, but it's just a video. I don't know how much of that is real. I would have to test it, but I haven't yet. I just saw it prior to the show. Interesting. Was this Was this marked as one of the things you want to cover? Is this uh No, you got to find out what it just Car Simulator It was yeah. Yeah, we can Well, I don't have it um It's not active. Uh it's not working for me, but Where is it?

Yeah, here. Uh I need to Uh should I cover this? Yeah? Can you hear me? is him? Probably me. What do you mean? Yeah, it's probably you. for a moment. I wondered if it was me. Okay. Uh I don't have the design mode. That's the thing. I tried I I tried my best. I tried like for 15 minutes. I couldn't get it active. Uh but we can we can talk about it. Mhm. Well, should we start talking about Car Mechanic Simulator 3 as a whole dig into the design mode. Is this Cursor 3? Interesting. Okay, so it seems like something different then. Is this Cursor 3? It looks like it. I mean that that looks pretty much as I would expect it to look.

Um okay, let's actually let me search that. Um It's a complete overhaul. Cursor version 3. They rebuilt the whole thing from scratch and centered it around AI agents instead of files. So this is a massive paradigm shift, especially for Cursor. Cuz they had this kind of agent and whatever toggle which was just half-baked and you know, I I started to find Cursor to be really clunky, really just full of just awful choices. It started to turn into vibe coded slot, to be honest. But this feels like it's taken the best of everything, particularly Codex, cuz I think Codex well, I'd say Codex set a trend. But realistically it was Anti-Gravity that set this trend. Um and IDEs centered around agents.

And now it's a much cleaner. You still got files and I actually really like this. I do really like it. It's a shame. I want Well, actually I'll speak about that in a second, but yeah. I'm I'm I'm here for this overhaul. Um okay, I have Yeah, I have stuff to show and mention. The thing is I haven't explored the Cursor 3 completely, but I have Look, I have all of these windows to talk about exactly that. Go for it. All right, let's talk about cursor three. So, what you see here is the old cursor and the issue with it is it's crammed with a bunch of like UI choices that are very questionable, very difficult to navigate between and it's really easy to get a completely messed up UI per project, per instance that you open.

But now, there is this new button, agent window, which is cursor three, which looks awfully like looking like what You know what this looks like? This looks like exactly like the codex app, which also looks like the T3 code. Well, code looks like codex app. All of them I have them to show you, but let's talk about this in a second. So, but let's talk about this first. So, this is cursor three and what you see here is a simpler UI where you have your chats and projects. Yes, projects like plural, multiple projects on the side and per project you can have a different chat going on, an agent doing something and then you jump between these projects without looking too much into the code.

If you want, you can you still have the browser, you have the terminal, you have the files, which I really like. This is what I prefer more more compared to something like the codex app. But I'm not sure if it's needed. It's just that it's a nice-to-have. But yeah, so this is the the new cursor with with the focus on the UI being much simpler where you focus just on multiple projects and multiple chats. You you focus on giving away tasks and not looking at the files or the code or anything like that. Uh as you know, I've been using uh Codex app for the majority of my work in the past. Pretty much all my projects I was in the past few weeks, I was running this every day, all day.

But I made a switch. I made a switch recently, like as of maybe two, three days ago to T3 Code. So this is by Theo. And the reason I moved to T3 Code and as you see, all of these have the same exact UI. The UI is almost identical. You have projects and chats. That's it. And then in each you can have a new agent, you you have a new project, you can have a new chat. You tell the agent to do something. Um you have a few buttons here for publishing uh pushing a commit and opening the files, opening the terminal, things like that. Um but that's pretty much it. It's very a very much agentic view.

The reason I I like T3 Code is that I can have Claude or Codex. And these are running from my subscription that I already have, so I don't have to pay extra. And T3 Code itself is free. It's open source and free. So, it just using my already existing Claude and Codex um subscriptions in this much simpler, nicer UI. Uh what Cursor is doing is exactly that. It's the exact same concept, but as you probably have seen here at the bottom, it says free plan. I was paying for Cursor for two accounts actually at at some point, but now I'm on the free plan because I don't know. I I haven't used it as much, but with this I don't know.

I kind of like the idea of having my files just as an option to um to see here with my terminal uh with my browser it's just nice to have the option. But, yeah. Uh if I go into settings here, I see a lot more, like a lot of uh option already exist options, things like agents um that can uh that you can change. You have the model selector. So, it's it seems like the Cursor 3.0 is fairly advanced compared to the other tools. Uh so, it's not a bare-bones. That's what I'm trying to say. Uh I don't know what the beta here is. Uh I access the beta. I don't know what this is, but they have plugins, they have rules and skills.

But, in a way these are not necessarily something that you cannot add to the others um either here in Codex. Uh you can do pretty much the same thing. You you have the MCP servers. You have Here you have environments, uh things like that with T3, uh it's the the least advanced one in a way because it's open source, it's done by Theo and his team, but I do believe he has a good sense of like product design for what this should be. Uh so, I'm giving it a try. At the end of the day, all of these are the same concept. This is Yeah, this this is just a I think at this point this is a matter of what you what you prefer.

You might prefer this, just a good old terminal. And what I really uh miss sometimes about using a terminal is this. Uh I I just you know, having four different windows, four different things going on at the same time, having tabs. I wish I could do the same. I was even thinking like uh T3 code is open source. I was thinking like, can I just do some vibe coding and take this and have like multiple versions of this running on a canvas. And yes, on on a canvas like on a Figma like canvas, so I can zoom and pan and have like new notes open. It just All of this uh is just about how we are trying to do this agentic work that is not completely figured out, but it seems like more and more companies are going this direction, which is also a bit scary in a way, uh which is not looking at the code.

It's mostly looking at the um at the tasks uh or like giving away the task to the agents and letting the agents do the work. Um on a kind of separate news um still related to cursor, they have this design mode apparently that I should have, but I can't access it. So, it says it is with control shift D. And here, if I just let's let me minimize everything. If I go here to keyboard shortcuts, you see I have this toggle design mode shift command D, right? Here, I press shift command D, nothing happens. And you know what's really stupid? They They had this shift command D for two other completely separate things. So, I just by default three different short three the same shortcut existed for three different things, which was probably the issue, but even now that I removed those, I still can't access it.

I have this old uh selector, which is different. So, it seems like this one Let's Let's play the video. It's not just a selector. You could do this before, but there is something new about it. So, it seems like now you can draw um I guess this is just a screenshot that is going to be sent to the to the model. You can absolutely do this manually, but it is just much nicer to have built uh in the tool. But, as I said, I can't access Well, it seems like this is in Maybe I'm wrong. It seems like this is inside of the cursor three. So, but why is it showing in cursor two? Yeah, this is This is where it's becoming kind of a mess.

Yeah, I'll I'll try to get it working. Um yeah, I I like I like having the code open for most of my projects. Like it's just nice to again dip in especially when you're you know, unaware of how it's building the project. You can just sort of familiarize yourself. You can just change one line if you know exactly what needs to be changed. It's nice to have this code open at the same time. And so I like that cursor seem to have brought everything all in one together. However, I'm still undecided and I do hope that these some of these some of these IDEs still remain as pure as VS Code. Because I'm still pretty much undecided.

I use cursor Codex purely because of the double rate limit that we've got and I've got a free account right now. Only for an extra month, I think. And the thing is I find I'm not aware of what's going on half as much as when I'm coding in Claude. in Claude code, sorry. And it's because Claude is very clear about the lines of code that you change as and when it changes them. So, if it's going to do something, it shows you what it's done and then it moves on. The difference with Codex is that you can keep an eye on things on the right-hand side in the code diff tool and you know, show but it doesn't it's not linear.

It's not saying, "I'm going to do this." And then you see the change. And then it's going to do You know, you know what I mean? You're just seeing everything that it's doing. Now, I'm I'm very open to the fact that I mean I can open it in VS Code. I'm very open to the fact that um I could be I'm even looking now. I It's not there. It just it in the feed let me uh I can't really share my screen, but in the Codex feed it just links the file that it's changed and you can't it's just not clear what lines of code it's done. Um I'm even clicking on the diffs right now and I can open them up, but it's sort of again a bulk thing.

All that to be said, I'm still undecided whether I like I I prefer the separation of warp with Claude code and then just the code if I want it to be. I think the ideal situation would be Codex, but be more transparent about the code that it's changed. Just going to get a charger. What do you think about that? Discuss. Um I'm also undecided. I'm changing my mind constantly because I I think it isn't we haven't figured it out completely. And as I said, it it it is shifting. We went from VS Code full control over code to um Copilot kind of way of just changing or like getting AI to help you in line to Cursor to let the agents do everything, but in front of you so you see everything.

You remember the days that we we were like accepting changes like one by one. You would press 10 times accept in just a in the chat in one run. I was doing that in Cursor cuz I was like, wait, where even though I don't code, I was like, where is this going? Like is this the right thing? Are you doing the right thing? Are you editing the right file? To now letting the AI to do the whole thing and we don't even look at the code. Uh so Yeah. It and it is definitely what I try to do with this showing like all of these to say it is a spectrum. Uh you can you can do either of these and I do think what the cursor three is, it seems like now that I also found the design tool by the way.

I do think it is the sweet spot. But being the sweet spot, it doesn't mean it is necessarily for me because you don't want to always be in the sweet spot. Uh I do think like right now the T3 code for me is the winner. Because it is Codex was for me already the winner, but Codex allow you just to use the Codex AI. With T3, I can use my Claude and Codex, whichever I want. It is it is absolutely like a game changer for me. Maybe I only wish to have the browser here just just so it is a bit more clean. I'm a bit I'm a bit undecided on this. Maybe I start uh coding something myself.

T3 code everything in a in a canvas. I'm I'm a fan of like putting things on a canvas and have a like bird's-eye view. If I show here like a canvas like this, even this tool that I'm building, it has uh it has a canvas. Everything is a canvas for me. Uh I just like that view. Yeah. Yeah, to me that's Yeah, that's I'll have to try it. I'll have to try it. I'm open I'm open to it, but yeah, I do I just Yeah, I'm Claude Code I'm a Claude Code sucker to be honest and I just like the way Claude Did you see that Claude Code have introduced a new um What do they call it?

Not fluid, but uh it uh anti-flashing. So, things don't and it jumps more and now it's a lot more fluid. You can kind of scroll up and down, which is quite cool. Um And you have mouse interactions. You can use the cursor to click around. Wow, 2026 you can use your cursor to to Mhm. to move move where you are editing in the text. But, I mean uh it's the terminal and it is Not going to lie, like for a lot of people in our audience probably this looks scary still. I've gotten used to it and I and I like it quite a lot. It's just that um I don't know. This is a bit more persistent in a way.

You know, my chats exist. I can um I can send them to the archive. I can access them in the archive. This persistency is the reason I'm staying with this and the the multi-view is why I also really like this. And I also I mean, I think you mentioned it one time. I think it is it has also some aura or some sort of like, you know, you're using the terminal. You're interacting with your computer and with your AI through the terminal. It's kind of cool. It's it is it is looking cool having all of this. But, by no means I'm not saying this is like uh you know, a measurement of Yeah. being a pro, but it is cool.

There are some people doing some interesting stuff. Tmux is another way to to look at this sort of thing, but again, I don't know. Maybe maybe I I I want the variety. They're all sort of starting to merge into one with subtle differences. I I I quite like the separation of CLI and code. Uh sorry, actually looking at my code. Um but yeah, it's it's interesting enough how far we come and this combined obviously with Composer 2, which we spoke about as well. Like um I'm glad they finally addressed the AI slop Doesn't mean I'm going to be paying an extra year for it, which is coming to I think I will, but it's good.

It's good to see. Yeah. It just felt really sluggish and slow, to be honest. I just didn't like it. Uh cursor This one? Cursor or the previous one? Yeah. Uh to me, it is the UI that is changing all the time and I can't it seems like I can't do what I want with AI and I hate this UI that your your new agent, you know, you have new agent and you have new agent here. It just doesn't make sense. You have duplications everywhere and if I go from here, this I'm right now in the what do they call it? The editor to agent mode. You see how the agent is on the right side.

It goes to the left side. Why? Why? Uh you cut off, so I was just melting down over over this. I was saying that you see you have this agent on the right side. Do you see this? Mhm. You see how I have the agent here and the chat here and I have the agent new agent and also new agent button and if I change the view from editor to agent, they go to the left side and I'm like why cursor? Why? Like it's not and these two it seems like you have two templates that you can freely change and they are just calling it editor or agent but you're almost completely free to change them however you want and it's just not persistent enough um throughout uh the it doesn't give you the the same experience every time and it gets really busy super fast.

Yeah. You're cutting off um I think Yeah. Anything else we can add to the conversation? No. No. I mean I'm just I'm just glad they're addressing the the mess that was this cut the previous to say the least. We have a comment about what cloud code leak. Uh we'll talk about that. We will talk about that. Do you want to talk about pick uh for what it's called uh pretext? Um I since I'm sick I I need to call catch my breath so it would be great if you can go I'll go after. Yeah. Yeah. Yeah. Yeah. Yeah. Let me uh start cloud code. Yeah. Set it up. I also have something cool about it to show from West Boss.

But, first you go on and and then I will show Show what I have. Okay. Well, let's see. Let's see how we go. Again, single monitor. So, ClowdCode just became open source. Unintentionally, but it became open source. Um through some mishap, the source code that creates ClowdCode, which is famously been closed source for, you know, eternity, again intentionally closed source, unintentionally leaked a couple of times now. Someone pushed that to the uh NPM registry, the source map, which is a a file that allows you to decompress all of the all of the jumbled up code that a build process creates, and actually define all of the files that make up ClowdCode, revealing loads of juicy secrets.

Um we'll quickly go through this article here, just to make sure we've covered everything. And Dropped it accidentally leaked uh part of the internal source code for its coding assistant, ClowdCode, according to the spokesperson. Surprised the person in the chat is like, "What ClowdCode leak?" Cuz this was everywhere. Jesus Christ. Uh the leak could help give software developers and Dropped it's competitors insight into how it built viral coding tool, and nonsense That's important. Nonsense tools involved in the exposed. Thing is, I have to be honest. You know, I got sent this by people. I actually do not care, to be honest. It's juicy gossip, and we'll go through everything, but I don't care. Like, I'm not one of those people who is like in a mad thing about open source and Claude code needs to be open source.

Um, but you know, it's there and someone actually has rebuilt it in Python. Which is a which makes it which is completely legal in that you've you've created something you've created something a derivative piece of work from something which means it's not now, you know, it's no longer you know, open Sorry, it's no longer like copyright claimable. So that was the first thing that was done. So go and get that. Obviously they're DMCA'ing everyone who picks up Claude uploads Claude code open source to GitHub. They can DMCA request those and they will and have been doing. Uh, but the Python library and I think there's a rust version as well. I'm not too sure. Those are all, you know, go nuts with those.

They're they're all there. But I think it's what we've learned from the code that's actually uh, where is my window now? I've lost my window. Uh, it's what we've learned from it that's the the most interesting thing. Oh, I've really lost that window. Here we go. Here it is. Um, just going to go through this to make sure we've covered everything. Claude code blah blah blah. Okay, we can we can ignore that now. Um, I I find it interesting how it is for uh, Open AI and other competitors potentially. Are they like sitting down and looking into the code and trying to understand what they can learn from it potentially to make their products better?

Absolutely, for sure. And and and Claude code does have a magic to it. Um, the not only do have they have they got great starting point with the actual model itself with Claude. Like it's like the Apple Apple ecosystem, you know, the software runs great on the hardware because they have that control of the entire pipeline. So there's just a a real nice magic between like synchronicity between the two that may or may not be able to be translated to other tools, you know, it's just it's just how it is how it runs with with Claude. But they'll definitely be looking into some of the stuff. Interestingly though, I saw a video by the Primagen.

And one of the things that someone found was if it's trying to understand if it's a negative like if you're frustrated or you're pissed off with with Claude and whatever. And instead of instead of using their AI to actually query whether it's a negative sentiment, they literally just use search and replace like plain old JavaScript to look through whether it's like piss, off, rubbish, crap, and all whatever. You can see that. Yes. Exactly. And so you'd think that that is some of this stuff would be actually really really complex or they'll be using the most sophisticated thing, but actually they're using basic rejects find and replaces on on the is cheaper, right? And what they do with this to be specific, they are just going to flag the conversation.

They are just going to From what I understand, they do use this for analytics. Now they might be able to go and look at the conversation. I'm not sure about that. Is it uh is it something that they can do, look at your conversation, or is that encrypted? I'm not sure. Mhm. But it is I I know for sure that they are at for the least they are using this for analytics. To see how satisfied or unsatisfied you are with the uh with the session. Exactly, yeah. And also there were some things that um I'm going to take you I'm going to add myself to the stage. Um there are some leaks that happened. One of them has actually already been within like a day they already release it which is this buddy feature.

Tamagotchi virtual pet. And I hatched that as soon as I saw it I was like what was this buddy thing and you type {slash} buddy and it hatches this creature. And there's different rarities. It was like this Tamagotchi {slash} Pokémon hybrid thing. But if we have a look at some of the other stuff here we got Kyros autonomous demon always on agent with GitHub webhooks 5-minute cron cycles {slash} dream has got auto dream. This is kind of creepy actually. Yeah, memory. So obviously things running in the background just understanding what's going on and constantly observing your GitHub and um certain things like that. Ultra plan. So we have ultra think which expands the memory thinking budget that you have.

You can you can just type {slash} ultra think it goes all rainbow and then it just can think for longer. This is ultra plan. Offloads planning to a remote 30-minute opus session 3-second polling teleport sentinel returns results. So that's going to be quite interesting. I am a fan Do you know what 3-second polling means? 3-second polling means whilst it's running for 30 seconds you're going to be checking in every 3 seconds to see what's going on. Ah. Now I'm not sure what I mean Yeah, cuz seemingly you can change that polling number. So, Interesting. I I did use this in I did use this in the T3 code. Yeah, in T3 code I use Ultra plan.

I didn't I don't think it went for 30 minutes, but it did go for I think I saw 18 minutes. And is it remote? Is it done your machine or is it remote? I'm not sure. I I I didn't check. I I I used it and I didn't check too much. Cool. Yeah, so this one's obviously sending off to um you know, remote um remote server doing its thing. Probably going to have to pay extra for that to be honest. And and Uh where am I? Uh Tungsten tool, this is an internal thing, direct keystroke and screen capture of virtual environment. Is that like a you know, a bit for a data collection thing? I'm not too sure, but seems like a internal only thing.

Voice mode, we already had voice mode. So, maybe it's using Deep Ground Nova 3 as its model now instead of something else, I don't know. Um and coordinator mode, multi-agent swarm with workers and scratchpads. Again, something I thought that we already knew about or already have even. Um bunch of other stuff, undercover mode, anti-distillation, interesting. So, they've already started to catch that. Session memory, auto memory, auto dream, and native client. So, yeah, a few things have leaked, particularly that that uh buddy thing. I'm not too I don't really understand it to to honest. It just keeps, you know, winging and whining in the background um and saying things about my code. I don't really know what.

Um yeah. Just check my notes here. That's kind of it, really. I mean, have you heard anything else that's been going on with regards to the leak? Yeah, I have cool stuff. I So, I honestly I I just I'm not I'm not You know the spinner verbs? So, when when you send off Claude to do something, it says thinking and it also uses other words instead of thinking, these verbs. And the list also exists. They have 187 of them and it's just a cute thing to, you know, this is by the way Vis found it. And yeah, it is just fun to look at this. I didn't know most of these words, uh but I'm going to look them up.

It's just kind of fun. Zesting. Yeah, sketching and puzzling and I don't know, gusting and frosting and I don't know, it's just forging and for- forming um grooving and imagining. Yeah, it's just a bunch of b- booping. It's just kind of It's kind of cute, so uh I think they also asked who came up with this and they they did mention uh I I forgot the name, but one of the employees and then the rest of the employees from Anthropic uh came together and just added more uh verbs, so yeah, it is quite funny. That's cool. That's cool. Um So, yeah, it'd be interesting to know what they what actually comes of this. I think it's more out of curiosity than than anything else like what truly makes up the the brains of something that we all think is just this mythical being.

Um but yeah. There you go. So, a few more code leak. Yeah, and a few other notes that I saw from other creatures covering this were like uh there's a tons of comments like more than usual and the comments sound like they were not written for human. They were written for the next agent looking over the code to understand what it is. So, that kind of makes sense and it is the new era that we are in. Uh and also uh I I saw comments on how it does look like a code that was written all by just prompting that meaning that it's not like the highest quality code. Mhm. Yeah. But I I'm not qualified to judge that.

I mean, you know, it is it's it's the way things are now and it really. It's it's quite crazy how um how we've just good is good enough. You know? Yeah. Good enough is good enough. It works. I I'm I'm paying 200 bucks because it it it is the best. Mhm. It is the best and I'm happy with it. I have noticed a few little bugs creeping but they all kind of contain little bugs. You know, uh Codex, it's not checking off the the the list the the task list that it's doing as it's doing it. It's just forever rotating on the first one and then boom it will just like complete them all. Um it takes ages to complete them all but it it will just check them when it's done.

Um on Claude code I've seen it what the numbers or letters from my status bar are infiltrating the text as I scroll up and down which is quite annoying. But, again, it just it gets the It shouldn't accept it, really. I'm not saying that we should accept it, but I'm also saying that um you know, it it it does the It does the work. It does the It does what we need it to, so yeah. Who knows? Um yeah, I don't know. That's the Cloud Code leak. I don't think there's much um much much more to say about it, to be honest. Yeah. Yeah, cool. That's it. Um okay, let's Do you want to Are you good to talk about Pretext?

Yeah, sure. Um uh let's me I'm not sure which one of my screens I should show first. Um that's why I'm um I'm I'm thinking like what Okay, I'll just show the I'll just show the main thing. Um All right, so we have a revolution. Yes, a revolution in front end. Well, maybe. Uh it is a very promising library. My dear front end developers and anyone who's interested in the future of interfaces. This is by Cheng Lou, uh one of the core members of React and an engineer in Midjourney. I have crawled through depth of hell to bring you for the foreseeable years one of the more important foundational pieces of UI engineering, if not in implementation, then certainly at least in concept.

Fast, accurate, and comprehensive um user land text measurement algorithm in pure TypeScript. Usable for laying out entire web pages without CSS, bypassing DOM uh measurements and reflow. So, you know, when you have bodies of text on a web page, the DOM has to measure the length of the text, the width, the height, where the text breaks, and all the measurements for where these pixels should go. So, what he has done is bypassing the DOM and bringing us uh like text layout that you can uh use with this library at 120 frames per second. Uh let's before going into his use cases, uh let's show what people have building with it. So, this is one of them, and we can actually test it.

one I sent it for me. It was quite cool. Yeah. So, none of these are things that we couldn't do before, just to be clear. It's that we can do them now with accuracy and performance without using the DOM calculations. So, right now the DOM doesn't need to calculate where the text goes, it's all done with this library. It is extremely fast and accurate and performance uh performative. So, here we see a masonry layout with hundreds, if not thousands, of probably thousands of text boxes. Again, at 120 FPS. Why does this matter? This matters because when you have a layout such as this, how do you define how big the scroll uh amount is? How do you define where when you scroll, when when does the the the computer starts rendering, the browser starts rendering the next Do you render everything and show it?

That That would mean you need to have a lot of rendering to do, but if you don't do it, then how do you guess the the the the height? You can't. So, you can't unless you do the rendering, which is really performance intensive, even though it is text, but it is performance intensive. Um but now he he does it with something like this, which is so impressive. This is impressive, especially like imagine um with the interfaces we have. I don't know if you have noticed it or not, but even in ChatGPT, if you have a long conversation, it becomes extremely laggy. It becomes so laggy to just scroll. It is partially because of uh the browser is trying to understand the height of this text, and especially if you resize things, the browser has to calculate everything again.

But this bypasses that DOM calculation entirely. Here, there is this really interesting uh example, and this is where my brain broke for a second. Uh I I I I was doing this, and I I I was thinking like, "Why can't you do this with CSS um text pretty?" Sure. Again, I'm I'm not the developer here. So, uh I was looking at precisely, let me show you. Look at this. Oh, wait, it handles uh this, you know? Um this text. This is done with pre-text. This is done with CSS. And I was like, why can't you do this with CSS text, uh, pretty? You know, you you you can you can do this. Uh, but I I was completely forgetting or like I didn't understand before testing it.

I did start testing it. You can break the lines and have the text to look like this with text-wrap pretty. But, you don't know the box size. Like, you cannot make the box size to be what the text inside is. Uh, with CSS it's not aware of that. So, you have a lot of wasted pixels. You have this layout with this wasted pixels. Well, with pre-text it's all calculated. And by the way, he did this by uh, doing tons of, uh, calculations with agents running on multiple browsers for multiple weeks. So, that's also, uh, quite impressive. And here is another use case. And a lot of people, by the way, so far you probably see this and you're like, why are you making text unreadable?

That's not the point. It's not about making it unreadable. It's about creating any layout with like high performance without the need to to do all that calculation, all that measurement. All of this is just a a showcase of what is possible. Uh, we have another cool showcases, uh, before I read that that one. So, here we have a 3D model that is moving and doing all the recalculations automatically. Again, these are not practical things. These are more like the just just to showcase the performance. Uh, Uh maybe I should read this and maybe you can you can include it in the video. Maybe you exclude it. I'm not sure. Maybe I will. Maybe I won't.

We'll see. Yeah. I I I found this tweet also explaining it fairly well. So, let's go through it. Let me explain the importance of this. An engineer solved a problem that's been uh plagued plaguing uh the interface the internet for th- Play How do you pronounce this? Plaguing. Plaguing. I don't know what that means, actually. Uh infiltrating or or Yeah. Okay, makes Yeah. disease type of thing. Like yeah. Yeah. Now now I know. I I kind of Okay, now I know. Now I know for sure. An engineer solved a problem that's been plaguing the internet for three decades. Every website you've ever used relies on a text layout system from the nine- 1990s. The browser loads a font, measures text, figures out where lines break, and positions everything vertically.

Every step depends on the previous one. Every step forces the browser to pause and recalculate. You felt this problem plenty times before even if you didn't know what caused it. Slack's scroll jumping when message heights are wrong. Google Docs getting slow on long documents because every keystroke recalculates everything below your cursor. AI chat apps getting junky when streaming uh because each new token can cause a line wrap that shifts the entire page. Same root cause every damn time. Text measurement is locked inside the browser's DOM. It's slow and and there's been no alternative for 30 damn years. Pretext bypasses all of it. Pure TypeScript text measurement, no DOM, no CSS, no browser reflow. You give it a You give it text, a font, and a width.

It returns exact line breaks with and heights using pure math. Around 500 times faster in many cases than the standard approach. Supports every language, that's very interesting, including mixed bidirectional text. I do this all the time, by the way. I do this all the time in my texting. Um and it's broken. This is broken on every single app ever. Uh CJK, I don't know what that means. No, I don't. I don't know what that is. Yeah, let's Google it. CJK stands for Chinese, Japanese, and Korean. Okay. Japanese, Korean, Arabic, and emojis. The engine is 15 kilobytes, built and validated by running cloud code and codex against browser ground truth for weeks. The demos are vi- wild.

Hundreds of thousands of text boxes virtualized at 120 uh frames per second, no DOM measurement. Shrink wrapped chat uh shrink shrink wrapped chat bubbles with zero wasted pixels, something CSS literally cannot do. Responsive multi-column magazine layout that reflow dynamically uh variable font ASCII art. It's just a byproduct of this. You can do ASCII art with variable font. Uh over the years developers moved rendering to canvas, scrolling to custom uh implementation, positioning to JS, but text was the one thing you couldn't move out of the browser. It was the last piece locked inside of the DOM with no alternative. Now we have a solution. This was built by Changlu, one of the foundational developers behind React, Facebook Messenger, and Midjourney.

He's not just anyone, lol. Uh if you build anything on the web, this now changes what's literally possible. This unlocks new UI patterns, layout interfaces, and experiences like we've never seen before. Go look at the demos in the quote post. It's open source, npm install, um insane. These are all running in a browser. The future of design is still to come. So, yeah, this is very optimistic. This is saying it is going to be huge. And the examples people are making are big. I have one more thing to show. And I want you to be critical. Um I don't know if you know Let let me find it. Uh what did I call it? Mm.

It's an interesting time, and I I'll we'll we'll end on like a light not a light note, but we'll end on something slightly tangential from the normal stuff, but it's interesting when this comes, and that'll make more sense when I talk about what I talk about in a bit. Mhm. I did a Webflow project. I need to find it. Mm. Is it in my playground? Webflow is being slow. Don't show me the spinner. I found it. All right, so one thing to show that might be useful and you tell me I will find some time to go and do this with pre-text. So I did this last year with GSAP and with wipe coding and you see if I load the page we have this container.

You see we have some space around this container with a lot of text and when I I move the cursor, we have some funky interesting animation going on. And if I press start, here we see these canvas windows with text in them. This was quite difficult for me to get right because it was hard to By the way, all of this all of these see them as windows where I have these like text. These are canvas. So I'm rendering text in the canvas. The reason I'm doing this is one is for performance and two is like I wanted to do more stuff, but like mainly for performance because if you do this if you do this this many if you create this many nodes, you you cannot load you cannot do anything.

This can be really fast, really problematic because each one of these would be a separate actually multiple separate div blocks, right? It would be thousand and thousands of div blocks on the page. So, because of that, I made all of these in the canvas. But, there are artifacts. For example, do you see at the top of these, there are these It doesn't show correctly when I zoom. You see, this is another type of artifacts. It's not perfect alignment. It is hard to measure the height of the text and then reflow this in the canvas, especially the with, you know, doing it dynamically depending on the the user's browser window size and the font size and things like that.

It was quite difficult to get it right and it's still not right. So, I think this library can help me. Do you think I'm understanding this right? Do you think this can help me? Instead of doing it in the canvas, I can do it just with pretext? Well, the benefit of pretext, you like the Sorry, you don't This is all graphical. It's all It doesn't mean anything. You don't want this. You actually don't want this to be, you know, screen readable, the the text in the background. So, it might play to your advantage to keep it in the canvas. However, I think the benefits of pretext is that you're keeping you're keeping indexable font text and it being screen readable and accessible and manipu- manipulatable.

Um with, you know, again, it's it's it's text that needs to be selected. It's text that needs to be read and understood by screen readers. It might make it easier from an implementation standpoint to use this library and it does a few of the, you know, positioning things like that, but you would most certainly want to wrap all of that in some sort of um a rear hidden or some way that that a screen reader wasn't going to like I like I mean am I going to find that if I go to Well, I go to this website with a screen reader, will it read all of this text out to me or is it going to be hidden?

No, no. No, I Yeah, but the the idea of this was something very geeky by the way. That that's why it has a purpose. There's a reason why it has this effect. This is like communicating something. Yeah, that's totally fine. But but yeah, I agree. It doesn't have to be readable. It doesn't have to be readable. It shouldn't be. Yeah, it shouldn't be readable to be Yeah, to be specific, it should not be readable. I agree with that. Uh but when I saw the pre-text, the first thing I thought about was this. But even though now I kind of understand that it doesn't need to be pre-text, but still because still with pre-text it means every one of these would be a node in my document.

It still will make it really heavy where doing it in the canvas um has its own benefits. Now, I want to uh like show one last thing with this. I found also this tweet. This is not relevant to pre-text necessarily. I've been playing around with Chrome's experimental HTML in canvas API. Interesting. I use it to create my videos and I wanted to see if I could make text selection work on a curved surface by moving the underlying element around on pointer move. It works pretty well. Um again, this is not necessarily pretext. It is kind of similar to what I have. It is text in a canvas. Um, text in a canvas has its own issues often with fonts and rendering and spacing and sharpness.

Um, and definitely not selectable or so far hasn't been selectable, but now it is because it I think it has a tween. Uh, again, it's not the same thing, but I'm showing this in the same context because it depending on what you want to do with pretext, this might be something more interesting. It might be something that you Uh, and it is open so it it is probably something that you need. So, Uh, anyway, I will probably spend some time going in Yeah, do it. into both of these. Yeah, uh, quite fun. Quite fun. People are People are building cool on the internet and I'm here for it. Yeah, yeah, yeah, for sure. Well, Yeah.

it's in like we'll we'll we'll wrap up there. Um, I'll I'll go through something that I just literally saw this morning and it just got me thinking and I know music isn't like a big part of your life. It's a massive part of my life, but there's a there's a commonality and a reference to AI in this time which I think is totally relevant. So, let me just show you a video that I watched if I can share my screen. Have you heard of a band called um, "Agent de Portein"? I don't know. No. Or seen? I'll play music. You've almost certainly seen this. It's these two guys and they're black and white painted dots.

And the music is very, very trippy. It's all like this guy goes on to explain if you're a music nerd, the reason why they are intriguing, because first of all, it's they're using weird time signatures. And if you're in the if you're in the audience and um you know, you understand music, you know what time signatures are. You can see sometimes it's 6/8, sometimes it's 9/8. And they also use different notes than what we're used to hearing in Western music. They're using notes in between, so you can see this guitar here that's got all of these so many more frets than a normal guitar, which I could show you a normal guitar, but there's almost like a metal bar Yeah.

what they call fret in between each of these things, which gives them access to some of these notes. And then lastly, they talk about the looping. So, he's recording one piece of guitar, having it loop, and then playing another piece of guitar over the top, and it's all very intricate, and it's it's mesmerizing to watch from a musical standpoint. But then he gets into the art side of things and why this is so intriguing right now in the age of AI. And maybe we can listen to what he says. I'm hoping I timed that right that we could just play it. But their appearance is important, too. Often gets described as being Dadaist. And if you don't know, Dadaism was an art movement about 100 years ago, which ultimately led to surrealism, and it was all about incongruity, about things that didn't belong together, things that broke the rules of what the viewer expected.

It was even the birth of the whole what is art debate. And it's interesting that Dadaism back then was a direct response to the machinery of war and of modern life, and that's probably what's happening now. You go back even further, look at the effect that photography had on art and how once that anyone could create images using this simple little machine, art changed radically. Painters stopped trying to depict what was in front of them as perfectly as possible, figurative art, because what's the point? And they got more creative with impressionism, and ultimately that led to Dadaism, expressionism, cubism, because what is the point of trying to paint what's in front of you perfectly when you can just take a photo.

And that freed artists, it actually pushed them to do something more interesting with their paintbrush. Because right now in Google Gemini, as of this month, anyone can create fully produced music tracks with just a text prompt. Modern music is unbelievably formulaic, and AI is trained on those formulas. So, I'll stop it there. Did that come through okay? It did. It did. And it does make sense. So, it basically pushes us from the mundane and the templated and the simple and what is in in the terms of photography, what is, you know, real to to go into abstraction and a level higher and into expression, you know, to express Well, yeah. Yeah. I And I like that.

why I think this is And the reason why I think this is important and he I thought about this before he even brought in the AI conversation. I was thinking about this when he was talking about art and stuff like that. And it's it's so interesting because I've spoken about We've spoken about this before where um I've forgotten the exact word I used or exact phrase I used, but it's like we we are redefining the worldview. We're redefining our worldview. If you go back to some of the earliest episodes of this show, I talk about where we we have no idea the implications of what AI can do cuz we're going to redefine our worldview.

We we we particularly talked spoke about it when we were talking about job loss. And and this idea that we're all becoming redundant. And it's like, well, yes, but our worldview will be repainted and we'll think about things in a very, very different way based on the new norm. And on the photography aspect as well, this is just a pin. I'm just going to put a pin in this. We talk about, you know, ev- everything happens in cycles. And we talk about like and we we're talking as though AI is the first time we've been presented with something that's completely taken job more specifically jobs, you know, but in the context of life and the world and stuff like that.

We think it's this new thing that's going to whatever. No, it's been done many many times before and the the the examples we've given here on this show is about farming when farming used to all be done by hand and then machinery came and you could replace your field of 40 100 farmers all picking seeds and fruits and vegetables out the soil. All of their jobs were wiped when the tractor came along. Okay, so we're just talking about this cyclical um thing where we're we're always presented with a new paradigm that we inevitably adjust to and create new a new worldview for ourselves. Now, that's a bit too intense. Bringing it back to creativity and fun.

Again, it's so interesting that the revolt that happened against well, what's the point? We can take a photo now. They saw that technology had replaced them. This is an interesting use case where designers or artists thought well, instead of letting it join us, how do we revolt against it and become more interesting and become more relevant. And that's what I hope that AI will bring. I hope that AI will push us not I mean we're we're we're in a difficult generation right now. And I watched an amazing video last night when I couldn't sleep about Mr. Beast and his whole thing being awful for children. His whole you know, this this this um uh retention grabbing editing that we're all kind of guilty of, but we're we're losing our attention spans.

Most of you probably lost what I'm trying to talk about here in the attention span nightmare that we're in. But um we are we we've got to we've got to balance that kind of understanding with instead of crossing our arms as creators, as developers, being like, "Oh, no, AI is this, AI is that, like, we can't whatever." Be empowered to do something about it and use AI to your advantage and leverage it and or go completely off the rails and do something very, very different, which is why your retext really had me you know, kind of both sides of the fence here because we're losing the the the reason for a website, not not losing it 100% completely, but definitely it's been, you know, let's say truncated to use a a text-based word talking about text.

The need for beautiful websites. We don't need them to be as beautiful anymore, particularly I mean, they never were, let's be honest, like, recipe websites and things like that. They never were, but because lots more people are sourcing information that they need through AI, they're not going through to the website, thus much more much less websites need to be designed. They still, however, do. I'm never going to to say that a website never needs to not be designed. All that to be said, it's interesting how this text-based thing has come maybe as a revolt against AI. To tie all this back to your text to pretext, maybe that person was determined because it's like, "Fuck this, we need to make websites much more interesting cuz they're all just going to become the same and functional."

So, yeah, a bit of a a roundabout way to tie all that together. I found this video particularly interesting. We'll obviously put his Will Francis, we'll obviously put his thing down below. But, this whole concept of art, culture, and the internet is is kind of it's up for grabs at the moment and I hope we can see the opportunity that's potentially here and artists can yeah, just be artists again. Yeah, but I do like No, I I I very much like the take. AI is just a tool. And I like the idea of using it uh to go higher in the ladder of abstraction, whatever that means for you, however it allows you to express yourself.

Uh it can be strictly for learning. It you can do whatever you are doing and use AI to learn. Let it do your early experiments. Let it do your early um resource gathering. Don't trust it. Don't take its output as final output. It can be just like a tool, an assistant, just like how internet is. Um AI is in is in the same way. And I agree with you. It's another paradigm shift like the example you mentioned with farming. Um the AI companies tend to make it to be much bigger than it is. It is bigger than the previous paradigm shifts, but probably not as big as the AI companies are making it to be.

Uh by the way, I grew up uh as my my dad was a farmer. And I was born in a village and we had a farm and I grew up helping my family on a farm, like actually doing work. And I uh my dad was uh was someone who was always looking forward to to like things in a new way. So, he he got always his hands on like the newest machinery that didn't actually exist pretty much anywhere or like I tried to build something together. I I I did that, so like trying to You know, there is no glorification in suffering. In suffering to like doing things in a manual way. Like if you can imagine, and if you can build tools, what what are we?

We are We are We are animals who can use their brain to build tools, right? Um we have a freaking spaceship on the way to the moon, right? We are building tools, so um I'm really I'm really looking forward to AI in the future with just for building tools. And this pretext is a perfect example, because he used AI to do the dirty um job. The do the calc the the busy calculations of comparing results of different browsers, which would take months, if not probably years, by hand to to try to understand how different browsers do this. And he did the optimization. It is an optimization that he has done by letting the AI to do this crappy job.

And what he did, he was he moved the ladder He moved up in the ladder of abstraction. And um Yeah. He He is the creative one, right? Mhm. Mhm. And And that's it. It's like I I do feel like I mean, obviously we I can only respond to comments that we get on our videos, rather than actually speaking. And And obviously you get the loud minority, right? But the impression we get, ladies and gentlemen, of the of the uh chat, um the impression we get is there's a lot of uproar and not a lot of support for AI where design from designers who can look at it and be like, "You know what? I am going to get it to just build me a freaking layout because be damned if I do that again.

I'm going to go and spend an hour on Pinterest, find those crazy designs and inspiration while that AI is doing all my dirty work for me. You know what? Why why there's a lot of this kind of crossing of arms and and rejection of it when it can be it can play as little role as you want. It could write you an email to your client telling, you know, how you want an extension on your project. It can be very very small or you can take it over. The world is your oyster, you know, grab it by the cojones and and see it for what it is. But, yeah, all that to be said is uh I'm I am intrigued about the whole paradigm we're moving into and how areas of life and and how we do things and how we think of things is going to be shaped in a in a way that you can't even comprehend right now.

We're in the midst of all of that and new and interesting ideas will come of this because it is the new norm and um yeah, we're nerding out over Agent Provocateur. Um, they're a French band they're French-Canadian band, which is why I talk like that. Um it is so fascinating right now. Everyone's mesmerized because we are sick to death of seemingly the same old pop music, you know? So, yeah, that's all I have to say about that. Cool. Yeah, it's a it up. This was a exciting world ahead. We managed to get through it. Yeah. Um, even though this this week we didn't have like tons of research done previously, I had some of these I I have a tag in my Twitter that called for videos.

So, when I find things that are interesting, I put it in for videos. I have one for long reads to read them sometime in the future. Who knows when? So, I gather lists and yeah, I think it was pretty good. We we we managed it pretty well. And it was interesting. I was personally like interested in reading that long tweet. I I skimmed through it, but now I read it and I'm like, okay, nice. Yeah, I understand it now better. You know, we are technically journalists, right? Like all of that we are journalists. We are broadcasting our news, we're broadcasting our opinions, and this is journalism, right? And in especially in the day and above By the way, I'll send you this video, this Mr.

Beast video. It's it's incredible. It can It talks about how this again this retention viewing is just not conducive to the growth of of our little ones, our kids, you know? It's teaching them to just be entertained in that moment and seek nothing more outside of that. And the comparison he makes to a guy I forgotten his name, Mr. Frederick, I think they call him in the show, who complete who granted was from like the 1960s, 1970s, took a complete different approach up against all these modern cartoons that are coming in that are all flashbanged and animals and crazy cartoons and you know how we all know and love. He really revolted against that and took the slow thought-provoking approach and talked about the kids' feelings and how what they want to do with their lives and what they how they should approach difficult situations and stuff like that and that was his whole thing.

Playing more of the long game in the growth of of and the maturity of of the the little ones. I'll send you the video. It blew me away. Yeah. Yeah, that would be amazing. You know, I have a son and he he's in school. He Yeah, he's in school. He's in Montessori by the way. I don't know if you know about the Montessori school. It's I I I don't think it exists only in Germany. It might, but it's it's a private school. It's a private school. They are not very tech not tech, but they they are not happy to see you for example there on your phone, right? When you are in school. It's that type of school that they are encouraging It should be That's a private school thing, right?

It's crazy. Yeah, yeah. It should be and the the way they are They they they have a bunch of like a bunch of good things about them and how they teach kids and it's it's about being independent and doing things you know, on your on your own. But I I do see this issue what you are talking about with with content out there with cartoons with with this like short attention span. The issue with this short form content is not just one of the issues is that the attention span you you watch 30-second long videos. Another issue that I learned about recently is how each video makes you feel different from the previous one and you're just not built for that.

Your brain cannot comprehend this amount of uh feeling changes because in one you're excited, in the other one you're sad, in the other one you're angry, in the other one you feel completely different. Uh and when you watch so many videos and your brain starts to process this many different feelings, uh you just cannot. Your Your brain is not meant for this. And let let's end this by I don't know if you are excited about any movies coming out this year. Well, it's already out. Hail Mary, definitely. You have heard of Dune, for sure. Probably. third one. Yeah, yeah, yeah, yeah. Yeah, yeah, the Dune. So, I've got really fascinated with the idea of Dune.

There are like 23 books. It's really deep, the story. There There is like huge philosophy behind it. Uh I haven't read the books, but I'm reading and watching like long-form videos about the books and about the stories a lot, and it's fascinating. Yeah, video essays. A lot of the the concepts uh in Dune, by the way, something that most people probably don't know, is a lot of these things happening in Dune are not actual conversations between people. Uh a lot of the content of this these books are thoughts. These people like thinking. So, that is uh something very interesting that I learned about it, and a lot of it is about um it goes deep into philosophy.

It goes uh like really deep into society. There There are like the You know, the main thing is thinking machines, right? It's kind of like Mhm. It's not exactly AI, but it's kind of like AI/computers. Um it is when human let the thinking machine do the thinking for them thinking that it will set them free but it just put the power into the hands of the people who owned these machines because they can shape how you think. And yeah, maybe end with that just become aware of what you think and if that thinking is yours or is being injected to you by one of these like short attention grabbing pieces of content. Just be critical of that.

Yeah. Yeah. Yeah. What a philosophical end to the show. Well, that is all we've got time for. So, join us next week on the show. We've got a website command AI show.com. Like if you're on YouTube, we're How many of you are on YouTube? Probably not many. None? Zero on YouTube right now. We're all coming in from Twitter. Jesus. Hello, Twitter. Come and join us on YouTube. Give us a like there. Make a make the algorithm work for us. Um yeah, join us next week for all the news in AI design and web. I've been Sam. Oh, you cut off but I guess it's my turn to say I've been covered Sam. Keep on vibing, baby.

Peace out.