#21 Sonnet 4.5, Sora, Lovable Cloud, Cursor Foundations, Vercel Situation

Oh, it's loading. It's loading. It's loading. And we're live. Very good. Awesome. Welcome. Waiting for Instagram to show. Okay. Okay. Yes, it is. So, working on Instagram. You see my hat made? You don't see me wearing a hat. I've gone I've gone formal this week, you know, and I want to try something. Should I take it off? No, you can you can do what you want. I want to um see if I because I've set Look at my glasses. I'll have to either take them off or And your green chair. And my green chair. Yeah. No green screen today. No green screen. Yeah, you should green screen your room. Well, that's what I normally do.

Or I should I should give you a snapshot of like my room. I've told you to do that to me, but you haven't given it to me. You did. I should I should just do that. That would be really funny. Then it looks like we're in the same room and looks like we're in the studio, you know. Yeah, I will just change the angle a little bit. And what I'm trying to do I'm trying to take my um my what's it called? Uh polarizing lens uh polarizing filter off this lens, but it's on so tight and it's so thin that I can't grip it. Plus, I'm not strong enough. Uh, so what will happen if people scan that QR code?

I should do it myself. Oh, we should remove it. It just goes through to YouTube. Oh, okay. So, yeah, if you scan that QR code, you'll go through to the YouTube where you can hit subscribe because most of our live audience comes from um other means. But then throughout the week, we release Yes, finally got it. We release shorts and shorter clips so you can kind of catch up on on all the junk we spoke about. Right, this won't work. I have to balance. I'm balancing it. Hang on. Yeah, take me darker. Hello everybody joining especially the ones on Instagram because I can see you joining. Um so hello to all of you. Uh, we will start very soon.

I don't understand polarizing lenses. I can see that in in my right in my left eye here. If I remove that, it's there, but then you polarize it and it's if I rotate rotate rotate, it disappears there. Like there. I wouldn't worry about it. Like look at the in my that that won't disappear because that's just I think that's just too harsh. It kills it a little bit, Mhm. but it's just too harsh. So, really, what is green? Is it like What is that reflection that I see? The green. Mhm. Oh, it's the it's the special coating on my glasses. Okay. So that I don't um die. All right. So um we got 13 people.

Should we Yeah, we have we got people we should start. Uh so yeah, what's been going on this week? You have um you have Sam's what should we call it? Sam's uh doctor's appointment room or something like that, right? Yeah. I I'm just asking like client questions and I'm using this to uh to get some answers. Free consultancy. That's what you consult. Yeah. So, all right. So, this week I got a client and I think this is true about most businesses. Like what do businesses want? They want a website that attracts visitors. Uh and they want to convert. uh how do they attract visitors? They want to do maybe something on social media um and you know other things.

So the client asked me among other things. So we will do design development you know the normal stuff but they also said what can we do with AI for content generation something that they don't have to involve in so much something that they don't have to you know do much work or give much input and automate and just have some content so people read it watch it whatever and then they go to their website and also So like how what are other means of you know uh especially in today's age getting more visitors on their website. So for content I told them about the quality with AI that is not you know top-notch it's not there yet.

uh it's very distinguishable and easy to see that it's AI maybe sometimes even low quality and they were like yeah we know but it doesn't matter that much because having some content is better than no content and our audience doesn't really care all that much they are not AI savvy and if it's AI generated so be it if it interests them it interests them the question is like what are tools and ways and what are things that we can do for them? Um they would be literally paying us to use AI. This is what I've been saying by the way for ages that the AI itself is not taking away your job. If you refuse to use AI, your job is gone.

But even the client is coming to me saying like willingly to pay me to use AI to do something for them. They don't they don't want to sit and you know with AI. They want us to do that for them. So now I got to go learn. Uh but first picking up your brain, what do you think we should do to get some traffic on their website and yeah all that? Well, so just on the topic and I know we've discussed that like something is better than nothing, right? Just to touch on the point that you do run the risk of just like um destroying your reputation a little bit just by putting out shitty content like you ruin trust.

But end of the day with time it will disappear. And that's the other thing just to bear in mind that that marketing strategies just and sorry social media strategy is a long long slog. And I wonder if they understand that to build up an audience like unless you just like very rarely does someone just explode onto the scene. They've either built up the the chops in other channels and then they launch a new one and then they've apply everything they learned to that new one and they're really really good and they're able to drag all of their previous audience over. But realistically, you start, you know, I I I think you and I are still on that journey of building up a social media strategy that converts into customers.

Ignoring all that, um it depends how light touch you want AI to be because on the one hand, you know, you can use it to develop the strategy, right? Uh, you can use it to write the script, you can use it to generate ideas, but then you still put in the manh hours to actually handcraft the video, right? But you use it to Yeah. fast forward through the creation process where you don't have to do any thinking and you're saying to them, "Look, all I'm doing is getting Claude to come up with a script and a plan or whatever and um that beats us handwriting it." what what we've got a script in a day whereas before it would have taken us a week as an example so they're they are getting b they're benefiting like that right yeah and they don't necessarily need video so maybe I should have clarified so for example let's say post that they can use uh on LinkedIn so for a post on LinkedIn we need a text and we need also preferably an image it could be also a video.

Um, but it doesn't have to be like a headshot video like this. And for the post, h how can we like generate these posts? Uh, they have a huge website. We made the website with them for them with tons of pages and products. So maybe we pull in that data and create Yeah. something for them. Yeah. Like so to plug my latest video that I just released yesterday, uh you'll see that I use Manis and I send it I plug in all of my websites, all of my properties, all of my information and say go off to all of these websites and generate a website. Whereas in your instance, you're saying go after all of these websites, pull this together, and generate a series of posts, LinkedIn posts, and this, that, and the other to to to release, right?

So, yeah, you can get an agent to go off and do that research, put it all in. Um, what I do in that video then is that because it's done all of that research, I'm in the same feed then and I say, "Well, now develop some content for my website because it's got that context." So, um, there's it's just that's the thing with AI. It's just so limitless. You you can use as much or as little as you want. An interesting one is um, Synthesia. Have you seen Cynthsia? They're the they're the talking head. You can develop an avatar cheaper than ever now. It used to be like three grand and you used to have to like take your video and this and that and well lit room and you know, right?

You that could be a way to then plug those scripts in because you need a script to make some video um and pull together something like that. Um this just question sorry the automation part of publishing it. Yeah. So they don't maybe so maybe they don't want to publish something bad. They want to have some control um but they don't want to do too much. I assume they want to have a say in you know what ideas but then the AI should do the most job most part of the work to find the ideas list them create the post everything and set them as sort of like draft instead of like publishing and they can like read it and be like okay that's fine publish.

Mhm. Yeah. I mean that could be manual or you can um set up an automation in N. Let's say for example they've decided every Monday at 9:00 a.m. they want to publish content or whatever. Or maybe you mess that around. Maybe you just get an AI to like generate generate a schedule every single week fresh every single week and then you automate against that schedule to say pull in from a notion database which has either you know the main bulk of that LinkedIn post uh maybe it's destination so maybe it says you either have LinkedIn or YouTube or Twitter or Twitter will be Twitter's very expensive so you can't use API for Twitter but point is yeah you you could figure out some sort of crazy automated workflow based off of a a database, you know, and again, that database could be um populated with content that you created through AI.

So, yeah, lots of ifels's. If there's a, you know, if destination equals LinkedIn, then take this content. If it if it has a video link, then upload the media. If it has a do you know what I mean? like you could you could automate it. Um I wonder how much you're actually saving or what you're actually gaining from that to be honest. Um, but because you need to you need to somehow put it into a notion database anyway. And it's arguable that you're trying to you're just putting effort into other areas, but you know, um, do you do you suggest me using Manace to start with? What does Manis do? Um, it's an agent. So, uh, I mean, all of the AIs have the ability to go off to the web now.

As long as you put an AI, uh, sorry, as long as you put a a URL in there, they've all got the ability to go off and do research and whatever. Um, and create stuff. So, maybe start with some of the AIs. What do you subscribe to? What AI do you actually Open AI? Open AI. Yeah. So give it to OpenAI and see if that does anything. Didn't they didn't they release an agents thing? Didn't they have an agents thing? I don't know. But just give it to OpenAI say, "Hey, this is the company. This is conversation in a different way last week." You know the the thing that you cannot make the any of these AIs to count from one to a million.

I had I thought about this because I re I rewatch our stuff, right? Um got to boost those view counts, you know, always give it a like. Um and yeah, and and we'll talk about Sora today and there is a limit on the amount of time a GPU can run for. And I mean this is getting into a big old discussion now because our the the length of time um who's the J David Shapiro I've told you about him time and time again. His late one of his latest videos he's tal in GPT2 I think we could get about 90 seconds worth of time that an AI can run for. Interesting. Yeah. And now we're at a point where we'll talk about it today.

Sonet 4.5 can run for 30 hours or 30 minutes. I forget which one it is. It's in our notes. So I can just look here. Either one is kind of like crazy. Yeah. 30 hours. Now this is text. This is code, right? Video um and audio presumably there is a limit that AI just can't run for that. So how you might do how you might get it to count to a million is say count to 15 and then like store that context or do this of something and then count to the next 15 count to 15 in intervals basically in 15 you know whatever until you get to about some way to like I don't know whether it can do that I don't know whether you can command AI to command AI to do that but um but yeah there is a limit on on how long these these models can run for.

And it's just so that text we've got it to 30 hours with 4.5 audio speech is is going to have its limits. So maybe it's just trying to cap those limits. Yeah. So you need to figure out a way to work around. You talking about World War I, you're saying, "Well, tell me why World War I started." rather than just give me a documentary on World War I, you say, "Yeah, tell me how it started." So what I should do, for example, in this case, I should create maybe a sort of a table, but start with one column and be like, can you add these things? And maybe it does like 20 and then maybe I add another 20.

And then once I have the list of a 100 let's say post then we can like add another column and then enrich 10 of them. I think you can you can ask the agent to break that down even further. So, why not start with topics as opposed to like write me all the posts when you've got that create 20 topics for me to talk about or 10 topics for me to talk about. And and just as a pin, this is what separates an expert SEO marketing person from us who don't know the first thing about it. 20 topics, some a professional might turn around and say 20 topics is far too watered down. you really you just want three topics that you want to um approach or or your topics want to um be separated into broader topics such as howto's listifies do you know what I mean like may an expert will be able to tell you those things but you could get AI to figure that out for you but an expert will be able to just touch on those nuances but anyway us dumb asses b create 10 topics and Now work through each of those topics and create 30 posts for each of those topics and then create and then from those 30 posts you know and get it to output these to markdown files.

And this is why agents are great is because they can keep referring back to the stuff and they can work sequentially as opposed to open AI which is boom boom boom like it's just one and then one and then one and then one. You constantly need to you know work with it. So yeah, like there'll be ways I think there'll be ways to cheat the system and and again start with the topics broad and then start narrowing and then broaden out and get it to write those docks. I think I think I you know you'll have to let us know how you get on. I I think at the end of today I should try and see what works.

Yeah. Yeah. I I did actually in the while you were talking um signed up for Manis. This is my first time using it. Hey. Hey. Did you use my referral code? Well, I I'm not a paid user. So, once I get I'll ask use my referral code. I'm going to put my referral code right now in the chat so that people don't just kind of funny like seeing how it goes through the the page and it kind of struggles. Uh it it struggles quite a bit and I did ask chat GPT at the same time. Uh and they they are both still working. So, it's been minutes. Well, let's see. Uh, can I see when I did it?

So, it's been four minutes. Okay, four minutes. Both of them have been doing this. Uh, doing what? I can't show it unfortunately because it's client website, but I'm basically giving the client website link and saying, give me a full list of their products. I know the list of the products, but I want to see if it can if it can do the list of products. I don't see anything on the QR code manusc if you got an iPhone, take a snapshot. If you got an iPhone, take a snapshot because you can now read QR codes and do this, that, and the other from the from the snapshot. Um, so it's difficulty. Well, well, I I should see.

I don't know. It shows some errors at the same time, but it's funny because it does leave it do its thing. Don't Don't Yeah, I'm not I'm not I'm not canceling it, but I see running JavaScript document document element.out out outer HTML. So, it's trying to browse it, but also run JavaScript and see the entire HTML structure and read it, which is yeah, trying trying to do everything. It's quite a lot. It's insane to think about it like as we human we just go through the website and just visually take it all in and we understand you know what's there but this AI is trying to get the code look at the site and I see it on the site that it's scrolling the page and trying to press tab and see the links by the way yeah I know yeah I in my video as well yeah it's cool Anyway, I told the client uh about they they said how how can they increase traffic and I don't know how much this is going to increase it but you know this ML is it ML file that you can add to to your website kind of like the site map is it I I'll check uh you know oh robots txt or or a sitemap XML very like XML but for for agents.

Oh, for agents. I don't know actually. I haven't looked into that. Uh, Web Flow has this. So, maybe I'll find it. I don't know where it is. It should be under SEO maybe. Oh, llm.ext. So, llms.txt. Um, something that you can add to your site plan. So, this is for LLMs. to hit this file essentially and read the list of page. I don't know what should what should be inside of that. I don't know. But what I mean I don't again like what do you I don't think this is going to increase say traffic right they want people to go to their website right they don't want LLMs to well you kind of want to because people are asking LLMs and if LLM can read your website like if you they're not going to get that traffic though so if you ask in Farsy eventually yes it is traffic it's a brand a awareness like if you ask in Farsy what is the best source to learn uh web flow.

my name comes up which is great you know probably in English somewhere down below will come up as well but this is like how people interact with the web with internet essentially uh and you got to place yourself where people are like we are doing this with this we are we are live streaming now to Twitch because like our audience is not on Twitch we are live streaming to Instagram and YouTube because we and Twitter because they are there and I think it's essentially the same thing. You want to put yourself where your um leads are or prospects. Anyway, quite a long topic. Does that I mean does that kind of help? I mean, you know, I can I can send you a video where someone uses one tool to like, you know, create video to script the video, then to generate video and and add voice over and it works through all of these different tools.

It's impressive, but I don't think I don't I couldn't can't speak to it making a meaningful difference. All all I can all I can help you do is speed up your workflow or lift you up where your skill set doesn't immediately allow you know such as script writing such as strategy such as um you know or even yeah that's that's all I'm helping I should try it I should try it and maybe even report back make a on it. Yeah, that that would be actually I think I think that's a great idea because the client is willing to pay me to use AI. I mean, I'm the same with my I I'm telling my team to use AI, you know, because uh if it can speed up, why not?

Anyway, uh speaking of my team, I see you guys in the chat. Hello. Right. Should we Should we get started with the Yeah. the meat and potatoes of it. Yeah. We've got a lot to get through. Yes. All right. So guys, here comes the We have a comment. Uh so anyway, ah that's okay. Here comes the intro for this week. We have like some really cool news to talk about. So this week, a self-driving car tried to run away from the police. Open AI released an impressive video model that opens the gateway to the hell prompting like brain rot videos. Uh and we are going to talk about 11,000 jobs being lost to AI. Uh all of this and more in today's episode.

This is Command AI. I'm Karza. And I'm Sam Gregory. And this is this week in AI design and dev. Cool. Smooth. Okay. So these are things uh that we are going to talk about and this for the start. You you want to start or Well, we've hit a snag already. Um comments asking me to restart which is wonderful. What happens if I I can disappear for a second, can't I? This ain't gonna Y I'm here. Okay. Right. Let's restart. I'm gonna I'm gonna My worry is about what it's recorded to my local drive at the moment and it's gonna freaking um Let me think. Let me think. Maybe we just abandon you need to restart.

Can you just share my to share my screen? Ah, okay. Yeah. So, I'm going to leave. See what happens with my file upload. Do some like juggling or something. Entertain the crowd. Yes. And I'll see what I can come up with. Right. speech of sex. Pick a card. Yeah. Well, I'm uh I'm practicing, guys. I'm I'm practicing some card tricks. I know some card tricks, but I've never done any card trick for the camera. So, uh I don't know. I don't know. Let's not talk about card tricks, even though I I could do it. I mean, it's kind of it would be cheating in a way because I could see it, but uh anyway. So, let's talk about Manace.

I I used it to I gave it my client's website to look at the product list and just return a list of products that they have and it's not it it's it's quite wrong in places. It's not complete. Uh it I I can't show it on yet, but it's it's quite it's quite bad. Like if I would take a screenshot of the footer and all the links that there are in the footer and would give it to any AI, it could just look at the list of the products and be like, "Ah, these are the products and services that they offer." But yeah, anyway, man failed at making the the list. Do you know I've never had it fail but it didn't it it's not a clean complete list.

It's not complete. That's my I've never had it fail on me. Must be must failing. It's it finished. It went through but the list is not complete. Then I'm surprised because it took what like 12 minutes 12 minutes to look at the one page where you see clearly all of the products in the footer. But yeah, it is where it is. I I should be smart about using it because it's not Yeah, you did it whilst we're on the call. So I'd be, you know, I'll be interested to know what your prompt looks like whilst we're It was very simple. Yeah, I did the same with CHP and CHP failed the same way pretty much.

Um I I CHP and Manis go to this site and give me a list of all their products uh with the page URLs. Anyway, let's uh let's continue. Let's go. Right. So, um this was wild to me. It's uh Whimo, which is the it's a driverless car. Um, a driver should try with the intro. Sorry to interrupt. Yeah. Yeah. Yeah. Yeah. A driver's Whimo car made an illegal Uturn to avoid the police, which was there to control drivers for consumption of alcohol or drugs. Yeah. So, basically, it was pulled over by police in San Bruno, which I think is San Francisco. I think the Whimo cars only exist in San Francisco. um which made an illegal U-turn.

Now, it's suspected, it's not we we can't find this out, but it's suspected that there was a DUI checkpoint, which is where police just stop and just check that you're not high or drunk. Um that that it picked it up on the GPS or something like that and it just made a U-turn to kind of avoid it. But that's all speculation at this point. The issue drivers driverless car. Exactly. Right. So the issue is is that when they eventually caught up because they saw it making this Uturn up with a car, there's no one to to to to find to to you know it was just a I don't know whether the car was empty actually.

It might have been empty. I'm not too sure. But the point is the driver's seat was empty. No one no person made that U-turn. So they've got no one to find. And it sort of revealed this loophole in the law, a regulatory loophole that prevents officers from actually like citing a driverless car for for traffic violations, which is bonkers. And it's, you know, with with in this instance, we're talking about just an illegal U-turn. But like in instances where it's murdered someone, let's say, or if it's hurt someone, who are you who are we to blame? This is just this is just the bottom of the or top of iceberg. Do you know what I mean?

When it comes to this these problems that might encounter with these AI powered cars. Pretty um pretty wild. And yeah, like I say, leaves this gray area leaving officers powerless because they just can't do anything. You just got to let it do its thing. I wonder how they stopped it. Uh I think they have like these driverless cars have protocols when they see police to to stop or like to to know and stop. I don't know. Maybe I don't know how they but as you said it might be because if the AI is trained on human and human try to kind of like avoid police if they see them. Maybe the AI just kind of like learns this by default to try to avoid police if they can even take cover.

They slow down when you see the police. Yeah. Whatever. Yeah. I don't know. It's crazy but quite funny either way. Um, something that's uh not quite as funny is I read on the news that Entia have killed um more than 11,000 jobs worldwide over the last three months uh due to AI obviously which leaves us wondering like who's next. So this company is firing 11,000 people because of AI. Because of AI. Yeah, it um that she said something quite crazy here. Uh we are exiting people on a compressed timeline where res-killing based on our experience is not a viable path for the skills that we need. Um where is the other one? uh she she specifically um says something to do with AI which I'm not too sure where it is but like yeah I think it's basically if you can't upskill on AI if you can't bring AI in and this is the thing it's it's not necessarily about like replacing because I mean you know you were talking about other companies who are trying to like literally replace people with AI.

I don't necessarily I mean technically they are because they've got rid of people because of AI but ultimately the people who are left are expected to deliver in accompaniment with with AI you know yeah a longer discussion around this I believe uh was very well covered in that video if you bring it up uh it's in the docs this video from cold fusion and I highly highly recommend everybody to um watch this video and also subscribe to Cold Fusion. It's like one of the best YouTube channels out there. Uh and he talks about as you said these are like some other instances of other companies trying to replace um you know human with AI and they failed.

Uh Clara was one of them. With Clara, they tried to do customer support with AI and it was so bad that they had to uh issue an apology and like completely roll back. Uh Taco Bell and McDonald did also try apparently with um like drive-thrus and things like this where some people apparently like uh they they they went there to buy something and by accident they bought like a thousand or like 100,000 uh other types of like I don't know finger food or whatever it was added to their it was added uh to their bill automatically. So that was a really bad fail in this case. I agree. It doesn't say that it say in this way that it's replacement in this way but it's nonetheless with AI.

It's important to also mention this company they have like close to 800,000 staff. It's a lot of people working there. And 11,000 is yes it's a lot but it's not you know a big portion of the company. and restructures used to happen be before AI as well but with AI it's a bit more aggressive and it's AI related obviously but it might be also like a restructure due to AI but not to replace people with AI we don't know but that's that's what I would also think it might be um might need to restructure the company so so they can rehire potentially in the future for pe people who can use AI, people who can leverage it and you know have the higher output.

I I've said this in the past. I uh if you're a company, your objective is to make money, right? Uh that's the first objective. One way you make money is by spending less money. You know, how do you spend less money? By firing people. But if you fire people, you are not really making money. So you fire people who don't make you money. But if you can hire people who can leverage AI or any other thing, it doesn't matter anything. If you can hire these people to leverage something to make you more money, you will do so. It's in your best interest. So yeah, I don't think it's that the AI is replacing in a way.

It is uh saving them some money. But if you if you give two options to a company, which one do you think they they choose? You give them the option to save some money by firing people or you give them the option to make more money like significantly more money by even hiring more more people, you know, having like more costs potentially, they will choose the second, right? Yeah. Yeah, it's a really good point. And just to sort of smuggly look at those who are full-time and you know career career jobs and full-time jobs used to be about security and this is clear as day how just a bit of a kick in the face.

So your job is never secure. The moment they find an excuse to get rid of you, they will. That aside, really good point about the fact that restructures always happen and we are we are still figuring out AI and the best thing we can do to figure out the AI is to to commit to it and see how it goes. They may as they may likely hire 20,000 back in this ongoing restructural thing as we learn about AI. Like we're I don't know what it is about the human psyche, but we always just we we We know everything. We know everything that's got nuance. We know everything has context that once you start to understand the context, you you you begin to like see through it.

But for whatever reason, you look at this and you think it's finite. You think it's done. 11,000 jobs gone. you know, um bit more optimism, I guess, and that yeah, 11,000 jobs may have gone and it's probably going to be a very tough time to um to those families, obviously, but it's not it's not for it's not for certain. It's not it's not indefinite, you know. Yeah. And touch on what you Yeah, I agree. uh touch on what you said regarding careers like even without AI and before AI you were what were you optimizing for? You would go to university or learn something and become valuable. You would quite literally be valuable for a company because you would make them money by having a specific type of skill.

Um it's still the same thing. It's still the same thing, but it's with AI. So now you have to know that skill which is AI to be still valuable and relevant and it is harsher than before. The curve is much sharper. Uh the cur like the turning curve the way that I imagine it is much sharper and you have to kind of like learn a lot of new stuff. But just as we had this early conversation in today's call that I said, "Oh, this client wants to pay me to use AI, but I don't, you know, I I I I don't know like exactly I I've heard you can do content and this and that."

And now I'm in the position where I see, okay, there is money to be made using this specific skill that I still don't have, but what I have is, you know, interest and curiosity. So I'll just I'm asking you and I will next ask chatpt and go on YouTube and see how I can make content and I will do probably like one test content for them and be like would you like to have more of this? Should I build an automation around it and there I have a new job in a sense. Does that make sense? Yeah. And also you can tell them that this is our first stab and over time we'll get better.

We'll refine it. We'll learn about we'll get data on statistics we get back. So, there's a million and one ways why this might be the worst, most likely will be the worst bit of content that you ever produce. So, you're going to get better. You're going to realize, oh, if I add this to my prompt, then it 10xes the the quality of the script, you know. So, yeah, you've you've basically just been hired as a media prompt engineer, which I know a lot of people hate that word um or that that terminology, but yeah. Um, we can talk about this for frigin ages. So, should we move on? I think we we we've Yes, we talk about this quite a lot and it's so important, don't get me wrong, but like, yeah, I think we're definitely it's definitely we're in a transitional period on all levels and nothing is nothing is for certain.

We're just a new world is ahead of us, which is scary, but it doesn't have to be. You know, look, we've we've got so we've come so far, guys. We've come so far with with many many shakeups. So right something more optimistic is cursor um brings a free AI uh course that you should watch. Uh this course co covers coding fundamentals from beginners uh in a YouTube playlist. I'll just share my screen with you now which I thought was just really nice like I was keen on watching you've got introduction here obviously how AI models work. I was keen on watching the hallucinations one. Mhm. Um tokens and pricing context. So, it's definitely geared more towards, you know, coders and things like that.

But I think everyone can learn something from this. And it's really nice. These are like roughly five minutes each. These um these videos, just something to get your teeth into. I thought this is quite cool. They just released them all in one. Yeah, I'm I'm looking forward to watching them. I will probably tweet about, you know, what uh I think about them. Yeah. you'd be done in an hour. Watch it on 2x speed, you'd be done in half an hour. You know, they're like five minutes. Yeah, they're like five minutes. And I think that um and and what I will say is like like I'm such a nerd for fundamental stuff. Like even HTML fundamental, like I I literally can I I can reel off what a what a HTML tag does like it's just there in my brain just because I've watched fundamental stuff over and over again just because I'm yeah I'm a nerd for it.

The amount of little tricks and things I learn from courses like this when you learn the fundamentals you think right well okay let's for example little little quick um lesson on AI here. AI is inherently stateless. So everything you send it a brand new chat. It's just one one. It's not there's no there doesn't nothing um if you ask a follow-up question, it's sending the whole conversation again. It doesn't remember what you've been talking about. You have to send it the whole conversation again. And that just fundamentally changes your approach to AI. and and you know whether it's building features whether it's just how you talk to AI and things like this like understanding the fundamentals I think really give you the most amount of this is obvious I'm biased obviously the most amount of flexibility and most amount of power to use something because if you're just learning how how one person you know does a clever thing you're never learning you're never able to completely freestyle and be um what's the word I'm looking for?

independent um creative with that subject matter because you're always taking an abstracted layer which is kind of goes back to my whole YouTube channel and web flow and code because I fundamentally not want to take the power away from web flow but so much so much magic happens in web flow and it's like whoa it does this thing and it's like I I I remember when I first opened up a web flow I'm like oh it's just a gooey for writing HTML like it doesn't add much. So, yeah, understanding fundamentals gets a gets a win in my in my books. Yeah. Yeah. Um, anything to add? No, that I'm just going to watch it. It's super short.

So, yeah. Yeah. Yeah, why not? So, um, we should link it, by the way, in the in the video description. Yeah, absolutely. Yeah. YouTube description, right? So, head on over to YouTube. Where are we now? Media assets. There you go. Head on over to YouTube. Over there. Come on over. Come on over where the where the water's warm. It's all toasty over here. It's getting winter time now. Move on. Somebody in the comments uh Mahi I believe uh he mentioned that uh that show mentioned what sorry that show that you have not seen that word that show is banned on this show the s word if you say it I'm gonna bleep it out even our audience know yeah so good how How many seasons is it?

I think it's like five or six. And there is the there are like pieces of it that is like universally kind of like known as like kind of memes and jokes. So if I even say it, it's not truly spoiling it. So there is this part where they have an AI and they put the AI in charge of the code and they ask AI to we talked about this in one of the episodes. They ask the AI to to solve the bugs and the pieces of the entire product starts to disappear because the AI starts deleting the codes because that is solving the bug. U but it's it's quite funny and the show is crazy accurate.

The accuracy of the show is insane. You have to this stuff would have happened. They're not like they're not telling the future. This stuff would have happened. They've like capt. Yeah. But they captured they you know everything from Elon Musk to Bill Gates to Microsoft and some Altman is sort of in the show and the characters are so well written that you see the resemblance of what is really happening today in the real world in the show. And it's good because it's funny. Like they made it to be funny and accurate and relevant after a decade or two years. Um, Bill Gates has written a book that I would like to read and and something makes me very uneasy about Bill Gates, but basically he's written a book about the future and what he thinks the future.

And so much of that book, I think, has come true. And he's written about AI and he's written about a bunch of things that forgot what it's called. I think it's got some weird name to the book, but I definitely want to might might get that an audio book and uh listen to obviously. Anyway, yeah, let's uh nice little announcement. I'm not sure how many vibe coders we have in the audience. I certainly am a vibe coder, but I sort of know my way around it. So, Lovable, which by far and away, you said it yourself, like they have the best best faking branding. Um, but they may have just took the lead on Vibe coding apps by introducing authentication, database, storage as well as an easy way to implement AI features like AI chat bots and things.

Uh, they call it lovable cloud and AI. So, it's quite confusing. They they they even self-deprecated uh self depre. Yeah. self-deprecated themselves on the live stream about the naming of this. But ultimately, well, here's the thing. I'm a big fan of Replet and the reason why I like Replet is that all of these other coding platforms relied on superbased for your your databases and your authentication and all of the other stuff and all they really did was take your take your input throw it to the AI and take the output. That's why you've got so many vibe code apps because it is just it's it was literally the lowest hanging fruit out there, right? Replet were one of the only tools that actually had some sort of I call it IP some sort of thing because they did they gave you the database.

Yes, it relied on um neon which is a postgress database but they integrated really well first party. It was done in the code. It wasn't like a oh I'll add neon database. No no no they included it in the prompt. So this is where lovable cloud kind of does that thing replaces that. So you've got uh data, you got your files, object storage, things like that. Uh users, so authentication and stuff like this. So really, they've they're on par. Well, I say they're on par feature- wise, they're on par with with Replet. Now, I still think Replet is is ahead. They've been doing this for so long now, but it's just good to see that Lovable is really there.

And that there's no denying they have a good clean interface. I do like their interface. is is is definitely a favorite for designs. Their marketing is great. It's a very approachable brand. Let's have a little look at this. Oh, it's just a just a little Yeah, I wanted us to watch this video because I saw the looked like the video has a lot of views. So, I I was like, let's watch it together. I don't have sound, though. Well, yeah, we don't have sound. That's the only problem. Uh but I'm I'm guessing Sorry. Should we watch it? We are, but we don't have sound. Oh, okay. Um, we should. Yeah. But I'm guessing he's given this to a 10-year-old and even a 10-year-old can build an app.

So, I think that's the that's the Sorry. Uh, I would like to still see like what they build. I'll watch the video later. So, Lovable doesn't look like this. I'm just going to Yeah. Right. But it looks like he's building like a like a learning app which is a really really nice way to promote these tools because I think kids and Replet did this actually. Replet spoke a lot about um children using it to build apps to help them learn stuff. And uh it looks like that's what they're doing here. So publish. So yeah, it's very very welcome. The AI aspect is a little bit confusing as they said on the on the live stream last night because this was launched last night.

Um, it's basically a shortcut way to add AI features into your app. So, maybe they've got something around here. Um, oh, and they're they're doing a lovable challenge, so get in on that. And I think there's some cash prizes and stuff like this, so get on there. Get on their challenge. What I find interesting that nowhere on the lovable website uh is there a tag or something that shows that the their own website was created with lovable. Do we know that if if it's creatively lovable? Maybe they show it in the page source. Uh analytics. It looks pretty. No, it's a next.js website which could have still been created to be fair. It still could by the way.

Yeah. Well, everyone's off Netflixify. We'll get into that. Everyone's on Yeah. Oh yeah. Oh yeah. Yeah, we'll now can you like open now that we are on this topic um kind of a small tangent but open like webflow.com please and just right click and go not to the page um like not inspect but to the view page source. You see that uh HTML comment the Webflow logo. Yeah. Kind of like nerdy way of like old way of doing it. Yeah. Just I'm just saying that the lovable website is pretty nice. It's very minimal. Nothing crazy going on. And they are a tool for creating websites, but I don't see many like amazing websites created with it.

And by the way, this is a piece of content that I will make. So, this is on my list to make a a video for which I have to do quite kind of like a lot of research to find the best websites made with AI. Well, check my video out. See if I make the list. Um, remember, do remember these tools. Yes, you can build a website with them. Like, there's nothing stopping you, but I would strongly recommend against building a website with Lovable. These are good for apps. Is it good for your applications which don't need any sort of um SEO optimizations or you know because they're slow like replet apps are pretty slow to be honest and lovable apps are going to do the same like because there's just so much in there so you know you mentioned that this website isn't necessarily built on lovable yeah good it's not meant to it shouldn't be built in lovable but I I I would say is just to round off what we were saying the pricing of them as well is here so it's not I think you get basically 25 bucks of cloud usage.

What that means, I don't know, you know, whether it's um uh storage, bandwidth, um database, you know, maybe maybe in obviously in totality, but this could be a only a few files according to their pricing. So, this doesn't really mean much. And then a pound of AI usage. Sorry, sorry, a dollar of AI usage. Like, that's like nothing. So yeah, I pricing is pretty unclear, but it is in addition to your normal 25 bucks per month minimum kind of subscription. Yeah. So, it's separate from the the main loable. It's separate billing, which again, you know, we're kind of coming back around like so much stuff was been giving out for free, but AI is really um flipped the script a little bit.

It's like stuff can't be free anymore. We can't It's unsustainable to so expensive. Yeah. Yeah. Um, speaking of expensive, Sonnet 4.5. Yeah. Is it good? Our next one, Anthropic Unleashes Claude Sonet 4.5 with 30hour coding sessions. Uh, $3 pricing, $15, $35 pricing, uh, and their most aligned model yet. So, I have been using it. I haven't been massively impressed. However, AIS have personalities. We have personalities. We have off days, on days, all the rest of it. I'm persistent. But to say that I have cancelled my Claude Code subscription as of next month, and we'll get into why in just a second, but this is still an amazing exciting release to be fair. It came with a whole plethora of like tools and things like that.

So basically um just to summarize Sonet 4.5, it is now better and recommended over Opus which was the leading Frontier model 4.1. Um it's kind of better than both. And in fact, let's just quickly show you the D. So yeah, we're up here. Opus 4.1, it's um 79.4. So it's funny. Sonic was the best and Opus was down here. Then Opus took the lead and now Sonnet's taking the lead. I wonder if they have different teams working on this. I don't know what goes through Anthropic's brain, but the point is they're all $3 in, $15 out, which is um just Yeah, you you would think that they would give you the older models for cheaper, but they don't.

Anyway, um so yeah, uh bigger context awareness, memory management as well, and the ability to maintain focus for up to 30 hours, which is incredible. They've also released the cloud uh the Claude agent SDK, so you can build your own agents, which is really cool. I've forgotten where that Claude agent SDK here. Um yeah, building your own agents is is amazing. that can integrate obviously with with chord code. There's a beautiful new VS code extension. Um I can't really show you. Maybe it's um in the article. We can show you that. Um, no they don't show, but it's a really nice um, you know, instead of using it in the terminal, it's just a window inside of cursor or VS code or whatever, which is really cool.

Really nice. Um, and imagine with uh, Claude as well, which I to be honest, I haven't taken a look at. Where's Imagine? Let's just search for it here. So, lots lots to get excited about with with Claude code. Um, oh, and the Claude extension, which is adds AI to uh Chrome. That is still coming. It's still coming. They've released it. I've got it. I've installed it, but it's only to Max users. and that they're the only and we mentioned this last week, they're the only um provider to really be being very careful about AI agents working in the browser. Um prompt injection I think they call it or something like that. So, uh but yeah, just a big big release from wild.

It will be wild with the crazy things people will come up with to hack each other. Yeah, for sure. For sure. It might not, but you just got you got to be you got to be uh careful here. But they're saying also it's their most aligned model yet as well, which is good good to be aligned. Good to um be aligned with humanity's best interest. But yeah, looking at this graph here, it just it just kills everything, you know, on paper. But it's marginally better. And this is what gets me. I don't know like it's like these AI companies and models they are getting better every week. Every week we have a new winner that is a little bit better than the other.

But looking at what has happened in the past one and a half years, not much has changed. We have bigger context. We have better models. But yeah, we still if you are a developer, you can use AI much better than a non-developer. M and as a non-developer you can now build more like better MVPs but they are still not MVPs like prototypes maybe like early MVPs because yeah because here's the thing like this might be part of anthropic strategy on why they don't lower the price of Haiku and and Sonet 3.5 and 3.7 is because they know that you'll use them and they're just as good. you know, they're probably not actually that much the latest and greatest are not that much better.

But when you're presented with a a choice of models and they're all the same price, it's like, well, surely I'll just use the latest if I'm paying the same. If I'm paying that much, I might as well use the latest. Um because yeah, sometimes I use Warp, which is a terminal, and you can choose the model there, and it does coding and stuff like that. I've I've sent it coding tasks on like I I won't won't have noticed the model that I've chosen and then see that it's like their basic unlimited free model and like oh did the job like I didn't have to choose GBC5 thinking or or whatever it did the job you know so I think yeah it's definitely worth definitely worth um playing around with what's out there and and not necess necessarily paying over the moon, you know, because everyone's just going to have the different ways of working.

Maybe maybe the apps that you're building aren't hardcore and you just don't need this, that, or the other. So, because saying that um I mentioned I canled my subscription to Claw Code because I want to commit to GLM 4.6. Now this is an opensource coding model which you know it beats out um it beats out um 4.5 which is where am I looking at this one here y in a lot of them um it's it's not on the sweet bench verified there so and the terminal bench but it be and this is just nothing compared to the pricing now is nothing compared to Claude and I want to commit to doing it and and I have seen the reviews and I have seen people use it.

It doesn't quite produce like the best or whatever. But the problem with a lot of these influencers and things like that is they're looking for the one shot because they know that your attention is going to be lost if you if you say, "Look, you've got to work at it." Because surprise, surprise, guys, you have to work at success. It doesn't just happen, right? You've got to work at a codebase. You've got to give it follow-up prompts. You've got to break it down into smaller chunks. You can't just oneshot an entire website um with just one prompt. Do you know what I mean? Yeah. And so the price on this, I'm just trying to look for it here.

Well, actually, let's go through the features. It's got a longer context window, jumping up to 200K, which context size, I think, is a good parameter to kind of, you know, make sure you're being um efficient with your with your prompting and your tasks that you're giving it. good coding performance. Uh, and you could actually trick claw code into using this by the way because they use the same API as anthropic, but I'm using it in kilo code. Um, reasoning more capable agents and uh, refined write it refined better aligns on human premises. So, I think that's Oh, no, that's that's actually so it can write and do your things. So, you might consider this for your script writing of your um of your marketing tasks.

Yeah. But yeah, um, So I can't see the API price. Do you know where I can see the API pricing in open router but I want to show you also that their subscription fee is very tempting. So I want to do API pricing first just to see if you know I I want to give it uh 50 cents in $1.75 out down from three to 15 you know. So you might find that this model it works for you. Yeah. Um go back back really cheap. Yeah. And all of this stuff is uh z.ai uh pricing. See if this is the right website. Um no it's not. Where's the pricing? What is this tool? This is this is the company who make GLM um Z uh GLM uh Z AI pricing.

Let's see if Google will come to the rescue here. Here we go. Oh no, that's the API. They have a subscription which basically starts at Here we go. Subscribe coding plan. Um this is quarterly. So look at monthly $3 a month, you know, down from a hund I don't think this is more comparable to the $20 a month, but um this is this is more compared to the $100 a month. So it's 15 bucks per month. But if you go, you know, quarterly or even yearly, this works out about $30 per month to get the absolute max from this model. But I'm going to try in the API uh pricing inside a kilo code first.

See how I get on with it. And I might just get on fine with it. So yeah. Are you changing models a lot? Um well now that uh Sonnet 4.5 is kind of the allin-one. I used to swap between Opus and Sonnet I sometimes I've set um cursor to be GPC 5. When when it's kind of struggling I do I do just change the model and just say look can you sort this out and see if it figures it out. Um, so yeah, not not a tremendous amount. Um, perplexity. I just kind of use the default to be honest. So I, yeah, I've sort of found my rhythm a little bit. I just basically use chat GPT for almost everything.

And when I'm inside of like cursor, which I I I spend quite a good amount of time in cursor, um, I use auto auto mode. But when I get stuck, then I try to see how different models compare. Well, you got 20, you got 20 bucks a month. You might as well use it because unlimit auto is unlimited. You get free. I have two cursor cursor uh subscriptions. One for myself and one for my developer. Oh, I see. I got you. Yeah, but I'm not on a team plan. Maybe I should change. Well, I I don't know. I'm not sure the pricing stuff, but you might um you might just give him an open router API key and put it through open router.

That might end up being cheaper than just paying a flat subscription. Well, one of the subscriptions I'm just I I used to be paying actually no, I paid it yearly because somebody asked me to build them something and I used the money to pay for the subscription for the whole year and then uh and then that deal came up and we bought basically every AI tool for the price of one. So yeah. Yeah, that's why I had no Yeah. Yeah. Yeah, you might consider because in cursor you can give it an API key to open router and then you can select any model and that just might end up being cheap. I don't know how often they're developing.

I don't know how heavy they are but maybe you cancel their subscription for one month trial open router and say just stick this API key in. Then you can manage their usage. Then you could put a cap of $20, right? Um and see if they come close to it. They're like coming in at $5 then it's like cool use open router. I didn't know but I could do it that way but that is probably smart to do it for next year. Yeah. Why they have you bought him a year's one then? Yeah. Yeah. Anyway, I love all you know I love trying new tools. I love I love um seeing what it can do for me and stuff like that.

As I say, 4.5. I had a rough first day. Uh 4.5. Just nothing was getting done. There's no surprise that today wasn't a great day for me either. And I did try a few other models. So, again, it's probably just tiredness, whatever it is. So, I'm not I'm not going to outwardly say 4.5 is is bad. And people have had success with it. Um, I just implore people to make up their own mind on some of this stuff as opposed to just listening to dumbass influencers like me. Um, should we move on? Yeah. We're going to warn you guys now. This is not a political discussion. This is not a political podcast. Yeah. This is just an interesting observation on the tech industry at the moment.

Well, this is something that happened. Yeah. and it got super viral on Twitter. So, uh I don't know how Gee, I think you shorten that to Gee. CEO Versel was here photographing himself with the prime minister of Israel. Yes. And you might support, you might be pro Israel, you might be pro Palestine. That's not the point here. that the onslaught, the the exodus of Versel from this tweet was wild. I mean, Jesus Christ. Yeah. Um, uh, look at D. Uh, you know, Danny, if you scroll up a little bit, you see Yeah, this one. Yeah. So, he's replying saying, "Not the flex that you think it is." How can you be sorry this is going this you know company owner to company owner here how could you not know I don't know because the CEO of um I'm sure he knew that yeah this is going to be controversial I'm like you cannot tell me otherwise he knew that but why did he do it you see there Lionus we have uh he's also kind of like an I know I I follow him he's like yeah so people are not happy.

So basically here's what happened. CE of Versal takes a picture uh with Netanyahu and people are not happy with it. So what do they start doing? They start cancelling Versal uh subscriptions and Versel is known for being quite expensive by the way. Um like when you get in it can get really expensive but anyway people are cancelling subscription and I'm sure he saw that coming. just cannot do this and not see it coming. Yeah. Uh we don't know how many people are unsubscribing. Yeah. You know, uh it's one thing to say that you are, another thing to be like, well, Vale is a very very simple platform which is why it got so popular. So, but a lot of people moving on to replet though.

Here's the thing, you know. Yeah. for some reason replet I don't I think it's because to be honest I think it's because Amjad is very vocal about his political views and he's very pro Palestine I think and so yes you know where one is there the other's there and so a lot of people and and he's making tweets on how to move next year's projects onto replet and making it easier for those people so he's he's definitely fighting the Palestine corner here Um but yeah, you were saying you don't know how many people were moving off vers. Yeah, we don't know. Um but it's it's interesting because I saw somebody saying like hundreds of thousands of users moving away.

No way. That's that's not even close. I don't think it is. But people are um people are and I don't think like there are enough people who would who would like move to Versel because of this photo. I think the net is quite negative for him. I don't know why he did it. I I don't know like why would you do this as a company owner if you try to I don't know like it's it's kind of like mixing the your company with uh with the with politics especially this one is very sensitive at this time and it's not just customers moving away I bet there are some resignations happening as well people might not continue working in the company and people might not want to join the company and this is something you don't want to do especially like there are we know we know of Microsoft for example they had deals with the the US military right uh that caused some people not wanting to work there or you know with Facebook with everything that Facebook Meta does some people might not want to work there but when you look at basics if you are a top engineer and I know Elon Musket he also became very polarizing recently uh in the past year or so.

Um but regardless of Elon Musk, if you look at what SpaceX is doing and what you know, you could be working at SpaceX or I don't know like in a finance company as an engineer, you'd probably choose SpaceX because of the cool things that they are doing and the cool brand that they are doing. The my point being the branding of the company really matters for the people who will support it including people who want to work at your company and this is a net negative for Bersel I believe I mean you know it's funny you mentioned Elon Musk because and I know he's not quite tech but he is tech like and his chumminess with uh Donald Trump right and again love or love Donald Trump doesn't matter.

point is when did when did tech merge or get so political or obviously everyone a lot of people going to have their own thoughts and feelings on this but you know I think of artists I think of musicians and things like that where you you know you take John Lennon you know or or Paul McCartney being wife beaters and things they're hor they they've been horrendous people maybe not so much now because John Lennon's dead But they you separate the art the artist from the person in some ways. I don't know. I I guess I'm just saying it's interesting how um Ghee felt the need to merge his political beliefs with with his work. It's it's messy because you could get into social media, you know, revealing too much about people's lives.

You could get into I don't know. There's just so many avenues you can go down. But long story short is that it's it's left him in a bit of a mess. And I wonder if he's there thinking, was it was it worth it? Because people feel very passionate about their political beliefs. So maybe he's like, you know what, I don't want these people on my platform if they're anti-Israel or whatever it is. Because it's something to do with Argentina there as well. I think he had an Argentina flag. Um just at the top there. I think that's Argentina, isn't it? Yeah, it is. I don't know what's up. Either way, I mean, yeah, we I'm not not too well verssed on this sort of stuff, but yeah, what can you do?

What can you say? Uh, we have Chris here. He's saying like the autocorrect is bad. I think he means on his phone. It's funny, you know, because it's an AI show. It's everything AI and we are talking about AI taking over the world and the autocorrect still doesn't work properly. But yeah, hey Chris. So anyway, um yeah, people finish off. Should we finish this bad boy off? Let us know what like you should let us know what you think in the comments below. I'm like genuinely no politics. No politics. Like we don't need to know what side you support. No, not not that not that like if a comp if it's fine for a company to to do this.

This is like what I'm interested in. Not what side you are in um on that's a different discussion. Mhm. But yeah, I don't get it why a company CEO would do that. Let's talk Sora. Oh no. Jesus Christ. This gives me nightmares. I was fixed on this video for far too long and I don't really understand why. We haven't really written an interview intro for this but like I mean first of all let's look at the tech here like that looks spoton. There's a little bit of compression but that looks very much like Sam Alman. So this is Sam Altman and this is um how how do you say this in English? This is like um this is what he made essentially.

He didn't make this video but Theo made this. Yeah. Yeah. But he he this is like his his product. His uh Sora did this, right? Sor is the video model that's generated this this video. This is Yeah. This is Sam Alman and the CEO of OpenAI and this video is made by by Theo. It's quite funny. It's quite funny. I wish we could. Well, I don't wish because it just I don't know. Just it does something to my brain. It's brain rot, right? It's brain rot. Anyway, we're talking about two things here. We're making it very confusing. Sora is the Sora 2, sorry, is the new video model from Open AI and it's it's very impressive.

Uh the sort of stuff. Have you got any examples of anything decent in here in these links? Try try the second one. No, no, I'm not. No, I want to show the model. The model was very very good. I mean, you can kind of see. Show the first one. I'm sure the first one. Yeah. Uh also, but uh the this is pretty good. No, the first Yeah, that too. This is their own examples. I don't like eyes that the eyes look so dead in some of the videos, but whatever. I mean, this is Sora. This is all Sora. And yeah, this is just incredible. I don't know. There's been a few video models. We we don't really follow video models too too much on this channel, but like there have been a few interesting ones that have come up recently, but this you see like crazy camera movements and the characters stay consistent.

Uh the scenes changes things. There is by the way sound. We cannot show it over the video but there is sound in the videos and it's very very very impressive what you can do with it. Yeah. Um and then saying that alongside the release of um Sora too as if these videos these high these very impressive videos wasn't enough they've also released the Sora app which I can't download. you can't download, but it's a Tik Tok like experience for Sora created content. Sora created brain rock content and which is why you saw this horrendous video. So all of this comes with this app and inside of this app you essentially have only AI created video.

So it's not like you know on Instagram where you second guess if it's AI or not. On this app, it's only AI. And what you can do, you can allow your friends to use your face to make a video. And so you allow people to do that and you use them. Yeah. You make a video with their face and make something funny. And people take it and share it on other uh social media. And there are also CCTV videos like you know the style of CCTVs that one of them went viral. Let me see if I can actually find it. While while you find that, like just to just to be clear about this as well, the the um process of allowing yourself to be used your likeness and whatever.

It's a simple you read three words and you just move your face. And from that information, you know, I'm maybe everyone knows what Theo looks like. That looks very very That looks spot on. And this very much looks like Sam Alman. Yeah. Um, from, you know, nothing a few seconds. Yeah. If you open the last link that I added, this is where it gets scary because even this toilet thing is not as scary as the next. I suppose it lends itself because of the quality of CCTV. I suppose it lends itself to to this. Well, it's still kind of you see that's Sam Alman stealing something, isn't it? Yeah. Sam Altman is stealing a graphics card because it, you know, the need I mean the fact that the graphic card just floats into his hand is a bit of a Yeah, of course it is.

It is easy for us to to watch this and be like, "Oh, that's AI." But not most people are just flicking, right? Yeah. Like imagine you take your friends pictures or whatever and you use it to frame them to to do something funny but it I don't know cause some trouble. Uh yeah because where's the Sora logo? That's the thing because what he's done he's generated this video he's playing it on a TV and he's filming the video. Yeah there's a Sora logo but there are AI tools that you can use to remove I see something here. So I think they've cropped it very I don't know but anyway doesn't matter you can still show watermarks uh that are not visible to human eye but software can read them or it can be like between the frames so that your eyes cannot see them but you can like yeah you use some software to to recognize it.

that is possible but it's crazy um where it's going. Yeah. So what does it mean for the society? Does it mean like can can we trust video again? Like we should not we shouldn't already be trusting video. We already we already know that it's we should we could in the past. Yeah. If you would see a video of me stealing something 10 years ago, if it was clearly me, you would pretty much, you know, you'd be sure that it's me. Yeah. I I don't know. It's It's the societal impact of this. I don't know. I It goes back to me telling you the very first question you asked me about, you know, yes, you're producing content, but if it's low quality that damages your reputation.

Yeah. This could on the sideline be slowly damaging the reputation of the internet. You know this idea that people are releasing video footage that you know because before Sora by the way we're just using this is just Sora is just our launching pad to this discussion. But yeah, I saw a killer whale eat someone and it was so clear AI because it was really bad. But the comment section did not see that it was AI and stuff like that. And I think the more and more people will learn what AI is and that um more and more stuff that gets created, it just either will people just will skip past it. They won't get the views that they're after.

These are the people that are making it. Or the consumers will be like, do you know what? I just don't go on the internet anymore because I literally can't um can't trust anything I see. Yeah, I watched a video today or something uh talking about headlines. Um, it's to do with it was no ordinary. Is it No ordinary gamer? Um, it was it was it's Penguins Penguins Zero. That guy. And he he spoke about a headline that was released about a guy who held someone at gunpoint for for killing his dog. So, he ran over his the a guy ran over some other person's dog. the owner of the dog ran over, obviously took him at gunpoint, made him dig the the the dog a grave and made him do this, that, and the other.

The comment section, the headline obviously summarized this as headlines always have done for for century, well, maybe not centuries, but you know what I mean, for hundreds of years. The comment section is like, yes, obviously America agreeing with some taking someone at gunpoint. Yes, this is so he deserved it. He killed a dog. this is disgusting. But actually, you go in to read the article and you you you begin to realize or you you understand that the guy who took the person at gunpoint, the owner of the dog, is actually just some psycho. Now, I don't know the specifics of all of that sort of stuff, but it's it's a kind of chicken and egg thing that was going through my mind, which was like people learned that humans are just lazy, inherently lazy.

And they've they've almost gified or they've almost kind of leveraging that by uh knowing that people don't read past the headline. And that's where they make their, you know, um, snap decisions. And, and so marketers or whoever you want to call them, they'll they'll craft their headlines in such a way where they can shift narrative either intentionally or unintentionally. All that to be said is that we've just we're just forever terrible at not doing our due diligence as much as we should. You know, we'll see a video now. Will there'll be in 20 years there'll be there'll still be people who don't know let's say this video is the headline who don't just who who don't go beyond the headline but there'll be a large portion of people who will be like just got to make sure that's AI just got to make sure like I don't yeah that yeah that that box moves weirdly on the screen there that's that's definitely AI.

So the damage is done I think and yeah we'll just have lazy people who just believe everything that that's put in front of them. I don't know if they are lazy but it's lazy people I don't mean lazy to be a bad word. I mean lazy to be I I mean lazy to be human nature. We are all inherently lazy. We are all electricity. But yeah, I don't necessarily I don't I don't mean to shun people who are lazy. I actually, you know, in some ways admire them that they don't go down rabbit holes. They just take things at face value and move on, you know, whatever. Anyway, uh like I'm I'm interested to also know like our audience if you watch any AI generated video.

I know I've asked this in the past, but now it's You mean actively look for it and actively want to like Okay, so to be more exact, you start watching a video and you notice it's AI generated. Will you continue watching or will you quit? Like that's more question. That's more a question. Yeah. Yeah. Do you instantly switch off? Yeah. when you see that it's a race to quit because you might have been interested in the thumbnail or in the hook and you like the idea of the story but as it goes on either because of some quality drop in the storytelling or a glitch in the video or anything that shows that this is an AI generated video.

Will you continue watching or rage quit? I rage quit. I was thinking about this today. It actually frustrates me that you rage quit AI video. I don't I don't watch it, you know, because just because it's AI does not make it inherently bad. I I'm not going to lie with you. Probably will be, but you know, AI just doesn't categorically make something bad. So, I was thinking about that today because again, like, you know, we produce content on YouTube and this that and the other. I bet there's plenty of people who just hate dislike our content just because it's AI and it's like well hang on we talk about all sorts. Do you know what I mean?

Like so it frustrates me that again. Yeah. The the the rage quitting just because it's AI is is quite me. Yeah. Funny uh enough. No to me when I see it's it's AI and if I don't know the creator instantly right why is that? Let me let teach me teach me why why is that low value low it's it's a very low value mass produ produced thing that I so is that because it's AI though or because you've seen that there's low connect with it at all so for example the video from uh the video that made with some altman in it I watched it I watched it multiple times just like you did I think there there there is a re relevancy there and a connection because I know the creator of that video even though he did it with AI I know the the story the context I it kind of like connects me with the idea of like watching that video but if a video is mass- prodduced by a prompt that somebody made and they don't even know that particular video exists.

they just mass- prodduced it and it's just like mass stories. There are like really good videos on this on uh on YouTube that you can watch how it is if you sign up uh for Tik Tok for example and start watching AI generated videos how quickly the algorithm will show more of them and you will see it's like a sea of these videos and they are really really low quality. low quality in the sense that they they won't teach you anything. They won't add any like real entertainment. They are kind of like brain rot just just so they take away one minute from you and they hook you in a way to to take that one minute away.

When I watch a video that I feel like okay it's AI generated it doesn't give me the thing that I want and I want that human connection I rage quit. I will not watch it. But if it's from a creator that I know, if it's, you know, within the, you know, for now, for, for example, with the Sora thing, these things that happened and people were making jokes about it and they were making like essentially memes and I knew what it is about. I'm a part of the the whole story, then I will watch them. But I don't watch them because they are AI generated. I watch them because I'm following the the bigger story.

What where's the where does the line draw? Because what if I was to tell you or what if you were to find out some of the people you watch are their craft and their script with AI. So I think there there are already like there are uh we we have cold fusion. Um he used to not show his face now he shows his face because people are saying are you AI? He's he has such a deep beautiful voice that sounds like AI but it's not. I've been watching his videos for ages. I've been a subscriber like I don't know like six years maybe. Um, so if he uses to answer your question, if he uses AI to enhance his videos, I don't mind because that's the guy behind it and I I watch the video.

If I notice a drop in the quality of the content, regardless of it's AI or not, even if it's a real human talking to a real camera, I still won't watch it, right? It's then is is AI the question then? Is AI really the question? Yes. Are you just it's statistically proven like you cannot tell me otherwise. Like statistically, if you look at AI generated videos, especially like faceless and mass generated from accounts that you don't know, the chances are that the video is absolute trash. So, I won't watch it. I won't spend the one minute watching it. I just skip it. I might Yeah, I might even tell the algorithm I'm not interested. Mhm.

If the if the video is good, I might, you know, watch the creator if I to see if it's like a even if it's AI. Yeah. If it's if it's AI, I might check out I do this. I when I like a video, when I like a video today, I I watched u somebody on Instagram. I found like this this um person, she's making she said she made like a 100K from social media. Mh. and she broke it down to affiliate and videos and platforms, etc., etc. And I went on her page and I checked other content and I was like looking at some tech reviews she had. You see, like this is how the connection works for me with the AI thing.

I just don't have that. It feels very random and out of place. Um well I mean interesting I it's interesting that you might stick around if it's good even if it's mentioned even if it's AI but then on the other hand you're saying you'll just quit if it's AI. So re for me it sounds like if it's bad quality is more so I said this I said if it's bad quality but if it's AI it's almost guaranteed that it's bad quality and I said I would stay if it's good quality that to be seen. So, I would correct myself to say that to be because I haven't I have I or maybe I haven't noticed um I haven't seen a video where I notice that it's AI, but at the same time I see that it's actually good quality.

I haven't seen that. Well, should we move on? Yeah. This is your this is your uh your one. Oh gosh. pencil. This is apparently a tool that sits on top of cursor and it's like Figma inside of your cursor. You can prompt um you can write a prompt to make a design for you. You can change the design in a free form canvas just like Figma. You can move things around and then you can write a prompt to your cursor to develop what you've been designing. Uh, and I genuinely find this interesting because you you have your designs there. It's not for every part of your product, but let's say you want to play with a few different versions of something.

A canvas like Figma is amazing in that it can like multiple versions and you can play around with the designs manually. You can compare them at in one view and then continue with developing one of them. Uh, and I really love this idea. I I I wish Webflow would also adopt something like this. It's really interesting. So, what I'm seeing is these are pen files. So, there you go. Oh, no. Pencil here. So, these are pen files. And what this looks like to me is that ultimately, I'm just really digesting it right now. Ultimately, it's a plugin. Like this. This popup looks like it's inside of the pen. So it looks like it's a for lack of a better word a I if frame into the pen tool and you can see it sounds like yeah and you can see here it's got very basic tooling it it's based on uh an open-source tool by the way you mentioned it on Twitter so that is also really cool and we have like Figma competitors that are open source so if you can take them and kind of like mix them with any IDE that you have.

I do like the idea. So, we'll get into who is this for, but like this makes me think now as I'm digesting it and I'm picking it apart. So, a couple of weeks ago, we didn't really cover it, but a couple of weeks ago, Figma made their MCP server remote, which means that uh you you used to have to run it locally on your machine and then you can MCP from cursor into the local instance of uh Figma. Now, with the remote one, you don't doesn't need to be local. Saying that, I wonder how long it will take Figma to build this tool. and easy, right? Because I think the chat is just hooking into an MCP up to the remote and then this is this frame is displaying.

I mean, this is a proper huge oversimplification, but I don't think it'll be long before Figma's done something like this. However, the page Yeah. and and see the features there. Yeah. Design right where you code, normal design, infinite canvas design. as you probably come to expect from design tools. Um, uh, multiplayer. So, look, you've got people floating around doing stuff. I'm presuming I'm presuming you can use this in the browser, though. Maybe you can't. Have you Have you requested access, by the way? No. No. Get on that. Get on that. I might as well. Might as well do it. I'll do a bit. I'll do a bit. also really new five designing with precision. Yeah.

Yeah. So I do I do wonder who's sitting in this in this world of I I I understand accessing the design, you know, converting the design into dev, but I don't quite understand who's who who wants an interface where they're designing very basic interfaces and then refining that interface with with AI prompts. Um or you can start with a basic interface that is created by AI and then you add your own touch of your design. You can as you see bring the designs from Figma. Mh. So but you're right Figma might just do this and I I see this clearly that a lot of companies are trying to own the whole vertical. You know we've seen this with uh Figma.

They have everything from ideiation that you you do all the way to the design and now Figma sites and the whatever that is with their vibe coding that you can create an app. They want to own the whole vertical of create for the web. They just want to blur the line between all the different disciplines and say look we help you build this thing a product a website. Yes. and also not have competitors like they don't want you to design inside of Figma and then move it over to Webflow. They want you to do everything inside of Figma. Framer does the same. They have like their design now and you know with the development part.

Yeah. I I you know andable lovable is doing the same thing. Yeah. I there will be there will be areas that it makes sense right but I said it last week interoperability the choice to have to be able to like well you know what yes it's good that lovable have their own database now but I actually really like I already pay for superbase I like being able to do this I like being able to integrate it with my n thing I like to you know what I mean like the it goes back to and I said this before the call it goes back to trying to be a designer and a developer like trying to sit in the middle.

You you're going to water down those experiences because it takes a lifetime or many many years to really become a good designer, it takes many many years to become a developer to and to solely commit to those things. And then once you've done that, those industries are moving so fast anyway that the the moment you stop developing, the moment you become a dinosaur. Yeah. Um saying that tools are the same. you start you try you know you you try and build everything into one tool it's it it starts to not become the best at the you know let's say for as a design tool Figma will always be such an incredible design tool because they're laser focused on that well they used to be they they've spent a lifetime refining Figma as a design tool and only now they're sort of branching out into other tools but the core foundations looking at the design tool of this oh I've just uh removed it Looking at the design tool of this, as I said, it's a very basic design tool.

So, what designers are going to sacrifice auto layout, um, effects and things? Do you know what I mean? Like, are designers going to sacrifice a dedicated design tool for that? So yes, these these companies are going to capture a large portion of the audience because it is allin-one and some people like that and I get a certain satisfaction of things being maybe it's an OCD thing. I get a certain satisfaction there but I wonder if they are also simultaneously risking diluting a core product because then they're all just offering everything. Do you know what I mean? Yes. Yes, I do. Um, but I'm not happy with the current state of things either because it's scattered like um I'm working with my designer inside of Figma and then my team inside of Webflow.

We quite literally have like Discord, Figma and Web Flow open and we try to coordinate things. We do this all the time. all the time. uh and it's not you know the worst but hey we could we could have something like this what I see on the screen where we are doing the same thing we have still different roles you know I'm the one who make the decision or like coordinates the entire thing uh the designer does you know the design the developer does the development but we are in the same place and we are talking about the same thing and changing something is not as big of you know changing it in Figma and then changing it in web flow but maybe even if it's still changing it in two places in this case it is changing it in two places but it's still kind of closer in a tool and a lot of this I guess it can be automated it can be get much easier um I see also like some time lost in translation and coordination like you see files as they get big and you have like many projects people are people start asking what is that thing that I'm going to work on it where is it you know having it every having everything in one place that I agree that might not be good for everything but it probably will be good for some things yeah for sure if you can more power to you if you can find if that's enough for you and that solves your problems Then absolutely, like I say, this is going to be the perfect product for some people.

Just as long as you're not you're not diluting the quality for convenience because it's like like I said earlier, you might be taking from one area and then losing it from another area. So, it's like yeah, you might have that convenience, but the output is just not quite there. So, it would be it'd be be an interesting um uh experiment to try and see if it does work for your people and and whatever. And it goes goes back to um Framer, you know, and their infinite design canvas and the fact that all of their tools, all of the design tools that they provide, you know, will work on a website because it's it is watered down by design.

you. Well, maybe not by design, but like you know, long story short is this might be the perfect tool for you and and absolutely if you can make this work. Goes back to me downgrading to GLM 4.5 or 4.6, sorry, like downgrading, simplifying might actually work for me. You don't always need the best for everything, you know? So, consolidation is um it's worth experimenting with. Cool. Yeah, I agree. Uh, we are at the end of the topics. Uh, yeah, we are. I was thinking we need to write an outro. Just like we write the intro, we need to write an outro covering things such as like, subscribe, let us know what you found particularly interesting or what let us know what you want us to cover next in next week's Command AI.

Yeah. If you have specific questions, just these will be released as shorts as well. So, um, yeah, let us know any thoughts or feelings on the shorts as well. Yeah, but we can, uh, Right, you have a meeting. I've just seen a message, so I start stop ying. I was just being intentionally slow. I have a meeting with a lead and yeah, cool. I have to prepare for that. All right, guys. Well, yeah, thanks for tuning in. Let us know what you liked, what you disliked about the show. And uh I guess we'll see you next week. Yeah. Thank you for hanging around and peace out. Keep on vibing.