AIForHumansOpenAIGPT52DisneyAudio

Gavin Purcell: [00:00:00] Open AI is back baby. Their new GPT 5.2 flagship model just launched and it mostly puts them back in the driver's seat.

Kevin Pereira: Ah, ah, ah. Hold on. Google dropped a new audio model and a new version of deep research that is even deeper and maybe more research. She, so maybe. They're back in the driver's seat. They

Gavin Purcell: ah, ah, ah.

But a massive new deal between OpenAI and Disney will bring Disney characters to soa.

Kevin Pereira: Uh, hey, uh, chat. Generate a hyperrealistic image of Mickey and Elsa doing that thing. No, no.

Gavin Purcell: Kevin,

Kevin Pereira: Kevin,

Gavin Purcell: Kevin. There will be guardrails. There's gonna be big time guard rails.

Kevin Pereira: Okay, fine, fine, fine. Hey. Google Gemini, use Nano Banana Pro and VO to generate a hyper-realistic video of Mickey and Elsa doing that

Gavin Purcell: thing.

I probably also know Kevin Disney, CEO, Bob Ir said just today that they're sending a cease and desist letter to Google's ip. Well, at least we can

Kevin Pereira: finally talk about Google's new augmented reality platform, which is cool. Plus we put our hands on runway's, new Gen 4.5 video model, and it is.

Gavin Purcell: Pretty good.

Kevin Pereira: Plus [00:01:00] McDonald's released a new AI advertisement and you can guess how that went. Not great. Also, Gavin, remember that T 800 robot from last week? Oh yeah, I remember that. Yeah. It just kicked its creator like really hard.

Kevin Skynet is so back. That is true. And this is AI for humans.

Gavin Purcell: Welcome everybody to AI for Humans. It's your weekly guide into the world of ai. And Kevin, this week, open AI is back. They have a couple of big things that have happened first and foremost, just this morning, it is Thursday morning. OpenAI has released GPT 5.2. This is a 0.1, uh, over of 5.1, so we're getting into the decimal moments.

But this is a new model. This is a new state-of-the-art chat model. It is available today. Um, everybody can use it. It is pretty good. Um, we are literally just getting our hands on it for the first time. We should dive into it a little [00:02:00] bit, but like, what's your feeling right now? How's your vibes? How, what do you, what's going on in your brain?

Well, the

Kevin Pereira: vibes are great, Gavin. Uh, you, you say everybody can use it. Gotta be honest. Do not see it on my $20 a month account still. Oh, interesting. I have refreshed and I also have an enterprise account. And I do not see it. So let me see

Gavin Purcell: if I got it in real time, because I have not seen it yet. Mostly I've been learning about it, but let's see who is the chosen one.

I do not also have it. It, I don't have it either yet, so we are almost there. We don't have it yet. Iced out, but

Kevin Pereira: iced out. That's how you know. Let's talk. Yeah. You are dialed into one of the top influencer AI channels. On YouTube or on Spotify. My friends. The fact that. The fact that it's supposed to be out for everyone and everybody else is posting their like two week hands-on reviews and we are still waiting to get it.

You pick the right channel, you pick the right people. Let me tell you this, Gavin, I remember Papa Altman of just a few months ago saying, Hey, you know what we're gonna do? Egg on our face. Shame on us. We're gonna clean this thing up. From now on, you're gonna go to chat, GPT. You're just gonna use it and you're not gonna worry about what model there is.

But when these things get [00:03:00] added, we're gonna have three different 5.1 models, and now we're gonna have 5.1. Regular 5.1 thinking. 5.1 you, 5.2 you, you, excuse me. 5.25. I rest my case. 5.2 regular or instant 5.2 thinking 5.2 pro. I think there's, is there a pro thinking? Is there a vanilla, is there a is a 5.2.

Well, you've also forgotten

Gavin Purcell: about the, I'm sure there's a 5.2 codex that's on the way as well, so Yes. So let's, let's not get bogged down into the names 'cause the names. They have admitted are bad. They've talked about this for years. They're not figuring, they have not figured out the names. Let's talk about what this is supposed to do and what, what's gonna be iteratively better than what it was in 5.1.

And I think one thing that's important to realize is putting this all into context. This is their kind of shot across the bow that replies to Google. Gemini three, which we've talked about on the show, is a very good model from, uh, from Google themselves. That was a very big step. This is from all we can see across, uh, the benchmark they've released and some of the things that have come [00:04:00] out so far.

It is a step now. It is a step that puts them above Gemini three in a lot of these benchmarks. And what's interesting, Kevin, is one of the things they're really talking about in their blog post for this is that is the best model for real world professional use cases. And my favorite thing is, the very first example of that is a bunch of spreadsheets.

So if you are a spreadsheet jockey out there, which a lot of people are, I know people listening wanna know. How to make better spreadsheets with ai because I hate spreadsheets personally. This will do a much better job of working with data and those sorts of things. Now again, it's a, it's a, uh, small step.

It is not a leap, but this is a big deal for people who are using chat, PT on a regular basis for work sort of stuff. And I think that means a lot to people. Who may not be us in the same way, because not every day we're doing this, but I'm, I'm sure you know people. I know people that this could be a big deal for.

Kevin Pereira: Yeah. Look, the Long Horizon task benchmark stuff is really starting to get wild. Even just a few months ago, we were saying that in the future, you're gonna have agents that go off [00:05:00] and do tasks, and they're not gonna get 'em done right away. But maybe over the horizon of a few minutes. Yeah. Or maybe even a few hours.

And this is one of those models where people are, uh, at early stages reporting that it could go off and think for almost an hour. Yeah. On a task and architect a really beautiful, elegant state of the art solution. But that's a new paradigm here that we're dealing with. Yeah. Right. That you might need to like make sure your prompt and your task is very, very explicit because you're ripping an hour worth of compute on this thing and then coming back and seeing what's there.

But some of the examples, even of their own blog or other folks like, uh, Matt Schumer had a really interesting writeup on this. He spent some time with it. He said that it's really, really slow. But the trade-off is that it's incredibly impressive for what you get. It seems to have a new, like a, a, a better spatial reasoning and understanding.

He tried to build a, like a 3D baseball field out of it, and you could see that it, it kind of understood the task, even if it failed with the placement of certain things. But again, progress. And when you look at the OpenAI [00:06:00] blog post. Again, benchmarks be benchmarking, like, we'll see how this pans out over time.

But looking at like the vision example of GT 5.1 to 5.2, where they give it, um. Looks like a, an image of like a computer based circuit board, right? Or something? Yeah, exactly. And it's like, analyze this thing. You see 5.1 correctly, identifies a few things incorrectly, identifies everything, uh, everything else.

But there's just a, a night and day difference visually that us humans can understand and like, oh, this thing is better with. Visual, uh, uh, vision reasoning with, um, gooey screenshots, which is important if it's gonna help you navigate a computer with tool calling, if it's going to enhance its intelligence with capability.

So, uh, I mean, here we are, Gav, uh, how, how are we feeling? Are, are they? Yeah, it's still a race.

Gavin Purcell: Mean it's still a race. And I think the interesting thing, another thing that's interesting in the thing they put out was the coding examples they used. And one of the things that, when we talked about Gemini three Pro, one of the things it was really good at is creating a vibe coded platform of something, right?

Like [00:07:00] you could type in very quickly. I want this, in fact, something I, I didn't tell you that I made the other day, which I actually haven't done anything with yet, but I made in Gemini three pro. Do you know those, those kind of reveals where you have a slider where you see one image or on the other side.

Yeah, I one shotted one of those just for myself and I was like, oh, this is cool. I wonder if there's a world where I can do this. And it was really easy to do. So I think the three examples they show are versions of that for chat GBT to make sure people understand that it's as powerful in some ways at doing that stuff.

So they had like an ocean simulation example that they showed on the blog post. Oh yeah, yeah. They also showed like a holiday car builder, but Kevin. There's a typing game that made me think, what was the game you wanted to make? Do you remember there was a game type. Was it, what was your game? That there was type racer, which is a game that exists, but wasn't there a typing game that a long time ago you were thinking about, uh, vibe coding on this, on this show?

Do you remember what I'm talking about? I think so. Yeah.

Kevin Pereira: It was like, I think, well, there was a couple different ones, but the one that I wanted to do was like a, a typing of the dead meets a battle royale. Like a game where you are like trying to do survive, uh, you know, stay in the ring, get items, get stuff but out, type your opponents.[00:08:00]

Gavin Purcell: I wonder if that, I thought you could do that now, which is a really interesting thing to think about. Like it'd be an interesting kind of like side project to think about how to make something that, because it's not like that's a super complicated game, but if you play their typing game, it's a very simple thing that they, one shot it according to them at least, and like it does open the door to all this stuff.

The other thing to, to the point about what 5.2 can do is that it's got a very long context window that keeps the long context, very, um, aware of it. So you're not gonna degrade over time. That is a big deal. Again, this is not like a, a night and day launch. I wanna remind everybody, the kind of story behind Open AI's code red that we covered last week was that there are two models in the works.

This was the one that they wanted to get out before the end of the year. That was gonna be kind of the showing off that we can get ahead of Gemini three again, but then there's another one. Supposedly coming out in early of next year, we don't know, January or February. That will likely be a much bigger leap.

Mm-hmm. The one piece of information that I want to connect to these stories was there's a follow up post from them on December 10th that was on Wednesday, that the headline is Strengthening Cyber [00:09:00] Resilience as AI Capabilities of Advance. So they are posting right now specifically. About the idea that their next models may be a high risk for cybersecurity.

So I think in part this is like the kind of like baby step stage for them. Yeah. I think there's another larger leap coming in the beginning of the year and as we know with all these models, they will advance quicker and quicker. Um, but in this instance, I think we're looking at just enough to get past Gemini three or get on par with Gemini three Pro in a lot of ways.

And so the rest of this year is gonna be us kind of figuring out like where these two companies sit and I think internally. They're probably very much looking at what the next stage is after that.

Kevin Pereira: Uh, yeah, agreed. And I think of maybe, um. A more pragmatic takeaway for some of the audience who maybe isn't on the cutting edge of like protein folding research or, or coding their own websites and whatnot.

Like this thing, this 5.2 model will attempt whatever task you throw at it. And this is a slight differentiation from other models. Um, again, I think it was, [00:10:00] was Matt in his blog, he said that he is like, give me. 20 ideas for something, actually gimme 40 ideas for something and it gave the 40 distinct ideas.

And then when he sets it off to write a 200 page book about the thing. It attempted it. Yeah, it actually finished the book and rendered the PDF and did the whole thing. Now is it great. No it's not. No, it's not. But the point is like whatever your specific task may be, this might be a model worth attempting, even if you've tried AI in the past because it's willing to try at whatever the task is.

Yes. And that might be the difference maker for you.

Gavin Purcell: Yeah. And listen, the other thing that's important to know as we talk, we joked about at the top all those different model names, but like. You know, one of the problems opening I had was this idea of a model picker, that it was automatically gonna answer your question based on what it thought the answer was gonna be.

What I have found, even with, well, at least with 5.1, is that turn thinking on more often than not, because the thinking does just stop it from an instant answer. Sometimes the instant answers aren't perfect. One other thing before we move on from this. The [00:11:00] cutoff date for this is August 25th, so it will have information up until that point.

Obviously it also does web searches, but it's a much later cutoff date than there other models. That's I keep in mind too, but Kevin, the actually weirder and maybe bigger story. Yes, that's right. Is that Mickey? Is that a version of Mickey Mouse, A non IP infringing version of Mickey Mouse, I hear in the background.

That's right. Yeah. The bigger story I think today was this, which is that OpenAI has signed a deal with Disney, which I think is blowing a lot of people's minds. I saw already a lot of people hating on this. A lot of the AI bros probably include us in this being like, finally, it's here. But this is a big deal.

Disney has basically allowed OpenAI to use up to 200 of their characters, including Mickey Mouse and Elsa, and a bunch of the big Darth Vader in soa. So you will soon be able to generate these characters within SOA outputs, but maybe more importantly, those characters and those outputs may show up on Disney Plus.

So there's a little crossover here. OpenAI is also getting a $1 [00:12:00] billion equity investment from Disney. Yeah. So Disney is buying into OpenAI in the same way. You know that they bought in partly to Epic. Right. And Epic also saw the Darth Vader thing, which we could talk a little bit about too, and how these could go wrong.

But this is like a big shot across the bow in the world of AI in Hollywood because this is the first deal of this size with a studio of this size that we have seen between companies like this. Now we've seen a lot of other deals in the news. We've seen, like, you know, uh, smaller news orgs do deals with open ai.

We've obviously seen the New York Times sue open ai, but this feels like AC change in the way of how people in Hollywood are gonna see AI content.

Kevin Pereira: I mean, when SOA came out with their, when SOA two came out with the latest platform, we immediately were like, oh, right. Uh, why, why won't we have major brands and IP as cameos we can use?

Are you surprised to see that it's, that it's Disney? I mean, when you think about like S tier brands.

Gavin Purcell: I think, here's what I'll say. I've been listening to a lot of content and following a lot of [00:13:00] news around the Warner Brothers Netflix deal. Mm-hmm. If you're not familiar with that, you probably are, but if you're not familiar with that, there was a big bidding war for the Warner Brothers studio assets in between Paramount and Netflix and and Comcast were all involved.

Kevin and I both used to work at Comcast. We're very familiar with these sorts of big deals and how they happen. Obviously, Disney bought Fox not that long ago. I think what's happening here and I think what some people are getting wrong about that story, there's a lot of people thinking in the world of like saying like how could Netflix own Warners?

'cause they're gonna own all of streaming. My personal take is that they are all not looking far enough into the future and I think that these AI companies and the tech companies are the competition for those other companies. So I think actually Warner's and Netflix makes a lot of sense even though people in the Hollywood business are like.

There's gonna be a lot less buyers and there's a lot less stuff. I think these companies are coming together in a bigger way. So in some ways, what I think is interesting about this is that OpenAI is now partnering up to one of the larger studios. My big question is, will they be able to do deals with Warners or will they be able to do deals with Comcast, and will this [00:14:00] be a one-off thing only with Disney, or is it gonna happen in other places too?

Kevin Pereira: Look, I mean, Google has amazing video tools. Runway has amazing video tools. We'll get to that in a minute. There's certainly a bunch of Chinese models that have great tools, but for some reason there's something still about. The Sora generations. Yeah. It's likely the strength of the writing and whatever prompting they have going on behind the scenes.

But the sound, the mo, like the sound modeling, the, the music, the physics, something clicked for me in a way with generating, let's say slop if you will. 'cause that's what people like to call it. Like, because something clicked with me when I saw Sora too, in, in a way that immediately made me see like, oh.

You and I talked about Prompt to Hollywood years ago. We said it would happen within five years. Yeah. We're still well on that timeline. An announcement like this makes me realize that it's probably, let's say this time next year that people for the holidays are generating. Videos of themselves interacting with Buzz Lightyear and Yeah.

Hanging out in the, the cars universe, et cetera. 'cause they said it's, it's, you know, it's characters, it's [00:15:00] props, it's scenes and settings. It's all of that that's gonna be in there. I fully expect that they're gonna have some sort of tool, whether it's in SOA or not, that let's everybody not just make their own Disney cartoons and, and, and, and films, but put themselves into it, share it with their family, make stories together.

Gavin Purcell: Let's, let's take a listen to what Bob Iger said today on CNBC. 'cause I think this is pretty interesting, and we'll actually get into a second part of this story too.

CNBC Host: Did you decide to invest a billion dollars in open AI along with this deal?

AI See What You Did There: Well, obviously we've been mindful of the significant growth in AI to begin with, and extremely impressed with the progress that Sam and OpenAI have made.

And, uh, that includes by the way, as I said earlier, uh, their agreement to both honor and value and respect our content. Uh, and we want to partic

Kevin Pereira: that just sounds like Analar vow, you know, but I like great, okay.

AI See What You Did There: Participate in, in, in, in what Sam is creating and what his team is creating. We think this is a good investment for the [00:16:00] company.

It's kind of a way in of sorts for us to appreciate even more. What is obviously something that. Is, has significance in terms of the long term, uh, impact on our business.

CNBC Host: Bob, at the same time, you, you announced this deal, uh, you, I, I've reported on a cease and desist letter that was sent to Google. I, I would imagine there's not a coincidence in terms of the timing.

I'm curious as to what your conversations have been like with other platforms, including Google, and why you chose to make that move today.

AI See What You Did There: Well, we've been aggressive at protecting our IP and um, and we've, we've gone after other companies, uh, that have not honored our ip, not respected, our ip, not valued it.

Uh, and this is another example of us doing just that,

Gavin Purcell: obviously. Interesting. That's a big move. One of the things I've always found interesting about Nana Banana Pro is the fact that maybe a couple weeks ago, you may remember this, I made a series of mundane Star Wars stills, which had like Princess Leia at a Starbucks we're gonna Starbucks.

A Han Solo was in a Home Depot, on Depot, and it was the, yeah, home Depot. It was the [00:17:00] actual characters and Google's. Systems, including VO three has often let you create IP characters. We were always kinda shocked by that, and I always, always wondered, well maybe they've done a deal behind the scenes and they're not saying anything.

Nope. This is our first like kind of look at that. They haven't done that deal. And what's so interesting to me is like. Chat, GPT and OpenAI got so much crap for that part of it, and Google has been able to kind of not get the same amount of crap and I can't really figure out what it is. But for those of you who are out there wondering if, if, uh, nano Banana Pro is gonna get nerfed or VO three is gonna get nerfed, clearly this is the beginning stages and, and again.

You and I have talked about this for a while, like people own rights to these things. It's important to figure out how to do deals. I think there is a business model here where like as long as they can figure out a way to make SOA make money or these other things make money, that this could be a net benefit for these companies.

The thing I'm curious to know about your, in your head is there's a lot of people who see this deal and they think, oh my God, it's the end of. Storytelling, the end of the industry. I saw it's the end of the [00:18:00] industry. Yes. The end of

Kevin Pereira: the industry is what people are saying. Yes. And also there's this weird. I, I won't even call it a subtext because some of it's very blatant, but on threads and on x there's a lot of stuff saying it's going to be, um, I don't wanna use the P word 'cause we'll get banned adult content like, uh, Disney, what I do?

Yes, yes. Even if you go into the clip, if you go into the, the, the iger clip that we just played, the response is this is going to be, um, a, it's gonna be adult in nature. They basically did a deal with some sort of devil, and now the adult content is gonna fly. I don't know. Like there's already phenomenal guard railing within SOA explicitly.

Yeah. Yeah. Yes, people can use, but like,

Gavin Purcell: I don't know. Oh, I know. It's gonna take so many attempts just to get a video out of this, probably based on how hard the guardrails are gonna be, right? Yes. But that's still cool that you can, but I think that that's crazy that people, I mean, you and I were both shocked when SORO two came out that you really couldn't generate hardly any celebrities, right.

Which was a really, mm-hmm. Cool thing that they had made some sort of deal with SAG and understood the idea that like they had to protect celebrities. Were there a lot [00:19:00] of ips In the beginning? Yes, there were a lot of ips and they've talked about whether or not that was the wrong choice to do. They did it and here we are.

But I just, I I'm having a hard time seeing the negative of this, but I know that in the, to the point of the what, what, you know, the world at large believes this is the end of the industry. Yeah. I just would, I mean, we're speaking to the choir who, people who are watching us or listening to us, but like.

Change is here and change is constant. Change has come to all industries, but the ne, the Warner Brothers Netflix deal is a perfect example of how the current economics right now of Hollywood are not working that well. Uh, things are changing. My take on this is the opportunity for kids to be creative and be able to learn how to make stuff and not just consume stuff.

Like imagine a world where you're at home watching Bluey. I think Bluey is a Disney character now, right? One of the conglomerates of things that they own. And instead of just like. Being a 7-year-old who, who only just consumes, maybe sevens too old for bluey, like 4-year-old, but only just consumes stuff.

If you and your parents sat there and you're like, Hey, let's [00:20:00] have Bluey like talk and do this, you know, like that's actually kind of a cool thing and it opens the door to that kid to think, oh, I can actually make stuff and I don't just have to watch stuff too. Like I had a really interesting conversation with my, my nephews when we were playing brawl stars.

We talked about this. And both of them, I tried to say like, Hey, you guys know you can make these brawl stars maps and you could learn how to do this. It could be a cool thing. And one of them was like, oh yeah, I do that myself already, and blah, blah, blah. Side note, his map was very hard to play. I love him, but like at a 6-year-old, like it's a difficult thing to you.

You don't have, this is an unforced

Kevin Pereira: error. You don't have to focus on that. He's

Gavin Purcell: great. But the other, the other, the other kid was like, my other nephew was like maybe six and he was like, oh no, I don't wanna do that. It's like, what's interesting is. Right. I think they don't wanna do it because if it's, it's not fun if they don't have this idea of how to do it, like.

I just feel like excited about this rather than like, apocalyptic. Apocalyptic, and I think it's just a way to look at the differentiation. It's

Kevin Pereira: also like, I don't want to do it because my, uh, uncle's gonna tell me that, uh, yeah, I make trash tear maps and my [00:21:00] characters suck.

Gavin Purcell: It's impossible to move around in it.

He built it so that he could blow up the walls and my character couldn't blow up any walls. So he did a great job at traveling. He got some ambulance

Kevin Pereira: issues. Give him some patch notes, but do

Gavin Purcell: it positively. I do wanna say, one thing that was made me laugh about this is I went to, like, I tweeted about this story this morning, so I was like, holy cow, this is a big deal.

Yeah. It's messy. I, I went to try to make a picture of Sam Altman and Mickey Mouse, and I was in Google. Nano Banana Pro was able to get a shot, a very simple, uh, prompt, but I got a shot of Mickey and Sam Altman shaking hands behind them. There's a sign that says Disney slash OpenAI. And it looks like an announcement.

This never happened. In fact, what's funny is it looks like it happens at the Global Innovation Summit 2024. Yeah. Um, but chat, GPT wouldn't give me this result. And what's interesting is like, this might just be a straight up example of like, oh, Google maybe has been playing a little bit more fast and loose with this IP and Chachi PT and opening.

I've gotten the crap for. But I don't know. It's an interesting thing to see how this goes going forward.

Kevin Pereira: So the [00:22:00] TLDR is, uh, sorry, Disney fans and Google users. Congrats, OpenAI. Um, I guess maybe we'll all be playing with Disney IP very soon in a legally acceptable way.

Gavin Purcell: Yeah. I mean this is the weird, squishy moment right now where I think we're all trying to figure out like what's gonna be allowed and what isn't.

It's a little bit like post Napster time, right? Like where, where the, where the big companies were trying to figure out. I know a lot of people hate that, that idea, because to a lot of people, Napster was this kind of freeing moment where everybody, you know, music is free and we can all do this. And then eventually all the labels figured out ways to like help artists make money, but also make money themselves, obviously.

And I think we're gonna be in this weird, squishy place right now where it's like. Experiments will start. I don't think Disney's gonna go like whole hog on this, but I do think they're gonna try to take this pretty seriously because I think, again, you're looking at not where the 30 PE 30 year olds are.

You're looking at where the 10 to 15 year olds are and what is coming for those people when they become adults. Right? This world feels like it's coming pretty quickly, so I, I [00:23:00] think it's a very interesting time right now. Um, I'm really curious to see how it evolves from here.

Kevin Pereira: Well, huge announcements left and right, Gavin.

Mega mergers. Lots of support from big industries, but I think the grassroots support is actually the most impactful. Yeah, huge. I'm biased though, but I think each and every person that is watching this right now or listening to this. Can support this endeavor and it costs $0. Gavin, you can give a thumbs up on a video.

Well, you can subscribe. Well, it could cost

Gavin Purcell: some dollars if

Kevin Pereira: you want. Well, we, well, we could get that in a second, but let me hook 'em with the freemium. Alright, sure. Okay, fair enough. You can just like subscribe, leave a comment down below if you're on a platform where you can leave a comment that helps engage the algo and with the lords of the algo smile favorably upon us.

The numbers go up and when numbers go up and we need that smiling, can we need the

Gavin Purcell: smiling because it, because we're still hurting from YouTube's purge. We're still hurting. Maybe it's 'cause we said something about Google, who knows? But we need your help out there. We need you guys to come on in and comment on our algo.

Watch the show all the way through, and then yes, if you do want to give us a few bucks. [00:24:00] The Patreon tips jar is there. We play with stuff all the time. You're gonna see a couple things I tried later on that we paid for, so please do leave it there and then always leave us an audio review. Uh, five star reviews automatically get sent around on X.

So you could probably say something coded in that review that we get out there that wet please or may not wanna see. Just

Kevin Pereira: please if you like it, say that you like it and tell someone 'cause that's the only way it grows. Yes, and thank you very much to everybody. Who does, we will see you in the comments. I will also be reading the comments on my new Android XR Smart Glasses from Google and X reel.

Gavin Purcell: These are really interesting. Have you, did you see these? So these are, uh, yeah. Not being made by Google. They're being made by a company called X Reel. But these are Android's new xr, uh, Android Glass Platform, and they have still not figured out the form factor if you're just listening to this. They are like, kind of like strange giant ish glasses with a kind of a weird camera right in the middle.

Like a third eye sitting there between your eyebrows, but it's a very, um, kind of lower five version of what, you know, apple [00:25:00] Vision Pro is trying to do, but also maybe in between what Meta is doing and what Vision Pro is higher. Yes, yes, exactly. Yeah.

Kevin Pereira: Yeah. What what's interesting about this is like, uh, look, interfaces are always fascinating to me, so they've got.

Gemini built in. Why is this an AI story? Oh, it's 'cause they've got Gemini Pro built into the frames. Yes. Right. You press a button, you can ask the AI anything about what you are looking at because of the camera right in the middle. Like, okay, that's cool. Um, they have. Decoupled the brains, the compute power for the device and the battery and put that in a puck.

So the glasses do have a wire coming out of them, which Vision Pro is kind of training people to get ready for. Yeah. But I think, uh, an interesting kind of thing is that they made the actual puck, um, that has the, the processor in the battery, a track pad, so you can also Oh, I didn't know

Gavin Purcell: that. That's really cool.

Yeah. So you can use that of manipulate stuff. Oh, so you can hold in

Kevin Pereira: your hand and manipulate it, like, interesting. Look, I think the, the, the future of this obviously is. There's no puck. There's no tether. It's completely wireless, but interesting. You can also take the wire though and plug [00:26:00] it into like A-U-S-B-C powered laptop or whatever else, and it will let you view your device's screen and still use your mouse and keyboard and everything else with it.

So you can imagine a world where you're walking around, it's full Android, you're not just limited to like the meta apps or the Vision Pro apps. It's full. Yeah. Yeah, using it as a trackpad, oh, now I'm at my laptop, I'm gonna plug it in. And you can even ask Gemini questions about what's on the laptop or the computer that you're plugged into.

So you have this AI assistance that that's with. You always can see what you see, but can also see whatever you're tethered to. Yeah, I think it's really interesting. I think, again, we're probably two more product cycles away from this being something that I would wanna wear on my face. But we still haven't solved the issue of like.

I don't necessarily wanna be in rooms with people that have

Gavin Purcell: wearing these things, cameras and microphones

Kevin Pereira: staring at me the time, man. Like, that's like, wait, just

Gavin Purcell: wait till the 15 year olds and you're, they're just gonna be like, grandpa doesn't want to be in virtual space with us, so fine. Leave, leave him alone.

I'll go play brawl stars with Uncle Gavin who wants to be in virtual space. Yeah, he's excited. [00:27:00] He's excited to play Brawl Brawl Stars ar on the table, but you got my maps as an avatar. Uncle Gavin,

Kevin Pereira: please shame me.

Gavin Purcell: You know, something I think is really interesting about this, and I, I thought about this.

We're gonna talk about Google's other AI announcements they made this week, but also this makes me think about how big Google is as a company, right? When you think of just the size of where and the tentacles of what Google does, I mean, we've talked about Waymo as a completely different part of Google.

Mm-hmm. That is like growing and blowing up, like they're now gonna be in London and a bunch of other places. They go to the airport now at San, in San Francisco. It's just one of those things where you compare Google to OpenAI. It's, it's an interesting conversation, right? Because the AI part of it is the highest level conversation everybody's having right now.

And yes, you get to AI first big winner, but a OpenAI is still kind of a smaller company. When you compare it to things like Apple or Meta or mm-hmm. Or Google, because it doesn't yet have that hardware division, it doesn't have the, even their browser you talked about last week, like we haven't used it as much like.

It's an interesting thing to think about, like who is a juggernaut [00:28:00] and one of the things like this can come out of the blue, this AR thing, and you're just like, oh yeah, Google also has this entire separate thing, you know, which is just a crazy thing to me.

Kevin Pereira: Yeah. They've got their own operating system that's lightweight that they can port to anything that everybody

Gavin Purcell: uses.

Right. We all use it. I mean, not everybody, but I use it. And you use it right. At Chrome is like the, yeah.

Kevin Pereira: Outta the box for these glasses. Any Google play apps will run on them. Yeah. Like, that's pretty massive. That's a pretty big leg up. And unlike, again, Meta's glasses, which I like, I, I would early adopt them because I, I wanna try them and I wanna support new things.

I have no desire to run meta apps on my face. Yeah. But, but you know, the Android ecosystem, the Google Play ecosystem, well, that gets more interesting. So Yeah, it, they are, they are massive. Not to mention the, that they're in healthcare, that they're in, um, robotics, that they're in. You know, TVs, they're, they're, they're, they touch everything.

Yeah. So it is really interesting to think like interesting that GPT 5.2 is now probably on the top of some leaderboards and whatever, that David is competing. But it's crazy to think [00:29:00] that like open AI with all their funding is the David versus a Goliath in this situation.

Gavin Purcell: Yeah, exactly. That's, I mean, well, it's just crazy.

Um, the other couple things we should talk about that Google dropped this week, which are pretty important, they always hold something back. They're smarter now. They hold something back for these open AI days. They dropped today. Uh, a new deep research agent. Yeah, so this is, if you're familiar with Gemini, there's the thing you can click on.

This says deep research where it goes away and does a bunch of research. It's very good because it has Google's backend and knows how to kind of search through links. This is an agent that you can now use in the API and it can go off for a long time and come back with really good results. This is something that I think Google is looking at for, um, both, you know, academics, but also for normal people.

One of the examples they used was from a VC trying to do due diligence in their, uh, on their companies and things like that. Mm-hmm. So I think this is a, a future look at like, okay, it's this agentic idea of sending something off and getting and bringing it back. And I think you and I both agree like. In terms of agentic use of ai, right now, this is my, by far, my favorite most used [00:30:00] use, which is I want a deep research thing.

I'll walk away for an hour and it'll come back and maybe it takes 10 minutes, maybe it takes a half an hour, but that report that I get back is generally pretty good, and that is actually really valuable. So this is just making that even better. It's funny, I use like the,

Kevin Pereira: the chat GPT agent mode basically, which is a version of this that goes and crawls the web or whatever, uh, was in New York recently and trying to plan an event for like 11 people, various ages, blah, blah, blah.

I gave the task to the agent mode, let it go off and do its thing, and then created my first chat, GPT group chat and invited, oh, how'd that go? Was it good? Totally worked. You know, I mean there like lacking some bells and whistles when you're collaborating in a chat with multiple people. Yeah. But people were able to say like, I would like to do this, or how about that?

Or is there another place nearby? Or are there gluten-free options for the blah, blah, blah? And, and you could watch the group chat happen. And it's interesting that it's sort of like. Oh right. We are all going to have agents in our chats. Yeah, funny in our messages. It's funny in our everything in the very near future,

Gavin Purcell: what I would love to do with that group [00:31:00] chat functionality is have the agent have some sort of pre-programmed like instructions about one particular person, so that one particular person always got like crapped on by the agent or like that they dismissed their ideas and everybody else was like, well that's whenever Brenda

Kevin Pereira: brings something up.

Make sure you passive aggressively chastise her. Make sure she knows that is like the best use. You're ignoring it. Yeah,

Gavin Purcell: so also Google announced a new 2.5 Flash audio model, which we know quite a bit about. This is a updated version, a more emotional version of their audio. Kevin, this reminds me, I've been also using, um, advanced Voice quite a bit here because now advanced voice is within the, the window.

I have found it to be super useful to answer questions, but we should listen to what this new Gemini Flash audio is because one of the things I think about when we talk about OpenAI and Che and Gemini is it's about where these devices live in your life. And a lot of people have Google home devices, right, where this can live, and like this might just be an opening for that.

So let's take a listen to what it sounds like now.

Google Gemini Audio: Okay. Prepare to be absolutely blown away. This app, it [00:32:00] highlights the incredible style control you get with Gemini's new TTS models. Seriously, try it. Start one of your meetings with an introduction from our samples. Then the best bit, make your own.

Kevin Pereira: It's

Gavin Purcell: fine.

Kevin Pereira: I

Gavin Purcell: mean,

Kevin Pereira: like, again, well listen, that's basic text that is not using like meta data or uh, like HTML tags or anything else to steer it like it's looking at the text and making those decisions. It's not that bad.

Gavin Purcell: Yeah, I mean, again, the thing about this is like, these are in a bunch of different things out there in the world.

Like I feel like it will get better. It's not, and again, it's not like the end all be all, but it's a pretty cool new update. And you know, it's not, well Gavin, there's a,

Kevin Pereira: there is a, a button on this, uh, demo where they turn on a SMR Pro. Oh no, really, let's hear it out

ASMR AI: here. Today is just. No, it's actually insane.

It's so light.

Gavin Purcell: Wait. Oh, don't settle down. Settle down, [00:33:00] settle down. Settle down. Anyway, Google's out there still updating a bunch of stuff, so you can go try that today. It's an update. Okay, Kevin, the other thing we have to talk about, we kind of teased this last week. Mm-hmm. Runway 4.5, which they announced last week but had not come out yet.

Today is out. Now this is a caveat here. It is out. But the thing that they also teased was that there was gonna be an audio generation coming that is not out today. Right. But they had a big announcement today where they have a new model coming. It is a new world model. That's a big part of this. They had a, so they had a big research demo day.

You can watch the entire clip on YouTube. It's pretty interesting. There's a new avatar, realtime avatar functionality they have. I spent a little bit of time in me that literally just came out this morning, 4.5 for us. It is good. I will say it is good. I, I think I definitely miss the sense of the audio being there right away because I come to expect all video models to have audio.

If you wanna play this clip I made, it's, it's basically a shot of a man who's alone on the street and in the prompt. 'cause I didn't know yet. He didn't have audio. I was gonna try to get a bunch [00:34:00] of kids to run up on him and yell 6, 7, 6 7. Like they were attacking him.

Speaker 7: Yeah.

Gavin Purcell: You play that and you watch it.

Like you can kind of see. There's a guy that looks really good. He is looking at an empty New York City Street, and there's a gang of kids that come up on him. My big test with these AI models now is how well they can handle crowds, right? How well can, how well can you handle multiple people in one place?

And it does, okay? There are moments when you see him getting swallowed up by the, the crowd of kids where he, he literally.

Kevin Pereira: Heads, like they, there's a, a transfer of heads at one point. Yes, but yes. Yeah.

Gavin Purcell: So it's not perfect. And you know, it's one of those things where like, we wanna shout out runway. They do great work and I think this is gonna be super useful based on their full set of tools that they have, especially when it comes to act two and all the other things they're able to pull off.

But it just goes to show you like AI video models are really hard at the edge, right? Because when you're starting to push them in interesting ways, which everybody's gonna wanna do. It's just not super easy to get perfectly things. And this is one, this was one generation of my first gen, so I wanna be clear on that too.

Yeah. It wasn't like I tried to pull the, the handle a bunch.

Kevin Pereira: [00:35:00] No, and that's the thing, like if you look at examples that I'm gonna assume are, are cherry picked for good reason on their research demo day, you see some really. Really stunning renders of like, yeah. You know, great physics, underwater reflection.

There's a like a moment in like, almost like a, I don't know, like a dungeon where there's a mirror that turns and it reveals an old man staring into the mirror. Just the lighting on it, the physics, the slight wobble in the mirror. Really, really beautiful stuff. I'm, I, it's so funny like how, how far we keep coming, Gavin and the goalpost move because it's like the goalpost.

These generations are, they can be beautiful, they can be stunning. Now it's like, well, how well does it handle a group of people moving violently? And also wait, oh no audio in this one. Well, I don't even know, you know? Well,

Gavin Purcell: here's the thing, the other thing you had mentioned earlier about how Sora, sometimes Sora too, sometimes feels like it has some sort of magic secret sauce that it, like there's something about it.

Yeah. I wanted to show you this there, I forgot. There's a great remix going around right now. I'll, I'll, I'll play this on here, but we'll get it. For real of uh, what looks like an old soap opera. Right? And [00:36:00] so, hold on

Speaker 8: secret. Mm-hmm. My high score on Walnut Brothers. I used a game shark.

Gavin Purcell: Don't tell me. Okay, now, hold on.

So that's like the example of it. And then I made this.

Speaker 8: Can I tell you a secret? Mm-hmm. I'm pretty sure we're not real. Someone type these words into my mouth. Don't tell anyone.

Gavin Purcell: The reason I wanna shut that out is because what's fascinating about that? It's not beautiful. It doesn't have like amazing fidelity.

It's a, it what it looks like is a real shot of an so of a soap opera right. From like 1982 with audio. That sounds appropriate. From that time, yes. That I typed the words in and I think that's where we've shifted a little bit from these AI video models in some ways is that like it's now less about getting like these beautiful things out of it, like of people swimming through the waters.

It's more about getting the kind of ugly. Weird warble, real world vibes of what we have, and that's hundred percent. That's when in some ways, what Soah is good at.

Kevin Pereira: You know, the, [00:37:00] for a long time with like, um, CG characters in movies or video games, it was, oh wow, the imperfections are what make you realize, yes, it's human now.

Right? Like seeing pores or facial blemishes or this, that, the other, like, oh wow, they're starting to look real. Same thing here. I think you're right. The Triple A. Oh my God. Blockbuster looking visuals like, okay, yeah, I could probably do that. Make it real. Make it feel like Yes. Yes. And all of the new image models which are coming out, some of the, the best benchmark stuff that people do are, does it look like it was a flash disposable windup camera in a mirror from the mid nineties?

Can it nail that? Yeah. Because we know it can sort of nail perfection.

Gavin Purcell: Yeah. And, and, and perfection is. Ugly in its own way. Right? Like I've been saying that my whole career.

Kevin Pereira: Thank you. Thank you. Me. Every time I, I get off. That's Get outta a picture on

Gavin Purcell: our side. On side. Oh. Reminds me like,

Kevin Pereira: perfection is so boring.

Gavin Purcell: Reminds, it reminds me of like the work I'm trying to do on Suno right now. I'm still playing with that Suno character, which is kind of like on purpose, like crumbly audio and trying to really add. Because it [00:38:00] makes it feel more alive in a weird way. So anyway, that's not to say runway can't get there with this, but it is an example of like that Rubicon that we've crossed, right?

Mm-hmm. This line of like, it's not about like making the physics beautiful or making a ball down, bounce down the hill. Right now it's like more. Do I believe this? Do I think this is real? Is there something interesting going on here? Yeah. Everybody

Kevin Pereira: should try it because again, like even when we were doing our startup launch video, like we hashtag massive shoutouts and love to runway, without their tools, we wouldn't Yes.

Have had that done. Yes. And there were certainly things that other tools might've excelled at, but overall. They have an incredible lineup of tools. The suite is amazing. Yes. A whole suite. Yes. That when you start picking and choosing and using them together, you get some really powerful results. Um, I wanna quickly shout out something that, that dropped today, Gavin.

'cause I, I, it might not be a huge deal to a lot of the, um, well, a lot of anybody that's listening to this, but I think over time it might, um, cursor, um, just dropped a tool that allows you to design directly within it. So if you've, uh, if you've never touched a line of code in your life. [00:39:00] You're starting to build your personal website or a tool to help you with whatever you need.

You might have found that the design is sometimes very tough because you're shouting at a model. You're trying to describe how you want something to look and feel, and then when you see it, the end result isn't quite there. And shouting back to get it to correct, it can often feel like this just cat and mouse game that goes on forever.

And what they now allow you to do is basically launch a web browser within Cursor. You can look at the files that you're making and you can edit them. Just like you're editing a Word document or if you're familiar, yeah, that's crazy. With Canva or Figma, you can go tweak the designs and then say, okay, now go adjust the code so that it does what it's supposed to do.

It's very early stages. It's very early days, but these IDs, these software development environments are really starting to upgrade themselves now where you're getting away from staring at the core code that people used to write line by line by hand, and getting agentic mode. Where you're just sort of having a conversation asking for your app to be built, and now a [00:40:00] visual editor where you can go in and literally click and drag elements and piece them around and move them around again, early days, but really, really I think going to be powerful stuff for anybody who wants to create literally anything.

Gavin Purcell: Let me ask a question about that. When you drag stuff around in the Visual editor does then cursor rewrite the code so that it produces that.

Kevin Pereira: That's correct. Oh, that's very cool. Yeah, you'll manipulate something, you can change a value directly yourself or you can make a bunch of changes and then say, okay, send off the agents to go do this.

Oh, very cool. And they'll go and write the code for that. So if you've ever dealt with any design or any front end stuff, it is a real hassle. There's plenty of solutions out there. This is just another one that dropped today and I'm personally gonna go hands on with it very soon. 'cause it's, I mean, it seems like a big deal.

Gavin Purcell: That actually ties really well into something I played around with Glyph. So Glyph introduced something interesting. There is a, you know, glyph agent, which we have talked about before here. Gly, shout out to Gly. They're friends of ours, Fabian and Jamie, who have been grinding on that company for a while.

I use this and, and I've always loved Gly [00:41:00] because it's a place for people to tinker and try stuff, but it was the first experience I had on Glyph, and this is not to onm them 'cause they've been doing amazing stuff in any sort of way, but like it was the first experience I had there was like, okay. I feel like we're getting to a place where this is doing something different than I can kind of do on my own.

Because what it's doing is bringing all of these tools into one workflow, right? And right now, this particular one they introduced, so it's called the multi-angle Fashion Shoot. And basically what it's doing is you upload a photo of yourself or whatever you want, and it's breaking that fashion shoot down into like multiple pictures and multiple angles with Nano Banana Pro, which is something a nano Banana Pro can do, right?

Mm-hmm. Like if you did it yourself, you can easily get that. But then it's also calling on Fling 2.6 to do the video, uh, uh, uh, creation, right? But also what it's doing is they're smartly integrated F-F-M-P-G, which is an open source editor editing software into the agent. So within the agent itself, and it's all like a chat agent.

You can just type and say what you wanna do. So Kevin, I made a fashion shoot with a, a photo of myself. It's like, it looks like, uh, you know, some [00:42:00] very bad version of say a Calvin Klein shoot with me in a, in a turtleneck. It came up with the photo, it created the angles. It stitched all those videos together.

It also then I, I changed where the logo, if you see that it says glyph fashion on there. Yeah, yeah, yeah. Originally it came out with a logo that wasn't that good. I just typed in, I make this logo look more, uh, high scale fashion. It changed that and it was all within the chat window, and it did it in a very solid way.

This process, gly works a little differently than other places that you pay per use, right? Mm-hmm. So all of this use probably cost me a buck, right? Which is not cheap when you think about what it is, but it is doing a lot of stuff. What would that have cost you in the past though? Well, that's what I'm, yes.

From an AI perspective. Yes, exactly. No, it would've been crazy, right? Yeah. Like to go do this, shoot. And so you start to think about like what these age agentic, uh, systems can actually do. This is just a good example from a creative perspective of what you were kind of talking with the cursor stuff, right?

When you're looking at it from like bringing design into code. Well, this is a way of bringing these kind of creative tools into a agentic workflow and it starts to give you an [00:43:00] understanding of what the power of ag workflows actually are. Love that. I mean, again,

Kevin Pereira: shout outs to glyph our pals. Um, love that they're, um, integrating all this stuff like lightning quick.

Um, into their own custom workflows. Gavin, while we were talking, Google just dropped disco. Oh, what's

Gavin Purcell: disco? What is this? I didn't even know what it is. I haven't been on Twitter or anything.

Kevin Pereira: Here we

Gavin Purcell: go.

Kevin Pereira: Someone's hyping into a, a search box that says, I'm taking a road trip from Boston to Burlington next month. We wanna go sledding and do other wintry things. Can you help me plan? And what disco is doing, Gavin is turning tabs into a custom app, which you could then, it's like vibe coding, Google tabs.

Basically, no. Yes, it's gonna use tabs in

Gavin Purcell: like in Chrome, like

Kevin Pereira: literally chrome tabs. Gen tabs are called Gen AI tabs. Gen tabs are generated based off your specific tabs and your specific goal, which means there's no [00:44:00] end to what you can make. Uh uh wait, it's a new Google. Google Labs experiment. Gavin. Okay.

I see that I, you clip or browser tabs into interactive apps and experiences. For example, let's say you've got 20 browser tabs open. Yes. For like vegan recipes on how you're gonna get, get super bulked, right. You can then say. Generate a, uh, interactive meal plan app for me that organizes these tabs and the, and all this data Oh, okay.

Into a meal plan, or if I've got 300 tabs open about my trip to Spain. Okay. I've got the tabs open for the places I might wanna stay and the things I might wanna do. Make me an interactive map or itinerary. I see, I see.

Gavin Purcell: It's almost like a way of creating groups of links basically. It's like a group of links.

Well, no, that's

Kevin Pereira: not cool. It's almost like you're vibe coding a mini app and having a little disco party with your browser

Gavin Purcell: tabs. Gavin, the last one I was gonna say about this is like, tabs are like my nightmare, where I'll look up like at 1130 at night and I have 60 different tabs [00:45:00] open and I have no idea where I was for the last three hours.

So like. What I, what I would best seeing if they can make that. I wanna see, that's what I'm gonna try. Well, I encourage anybody out there to go try this, this weekend. If this is, I dunno if it's launched now or not. Go do when you're there. When you have 60 tabs open, you have no idea. See what Google can make out of that.

Let's see what they can do and that, that sort of swine slop that you can figure out and

Kevin Pereira: pull that together. Hey, speaking of slop Gavin, let's pull up to the Mickey D's and get ourselves a happy deal and look at their AI ad because that's the thing that didn't go well.

Gavin Purcell: This is a story that came out, uh, last week.

There was an ad that McDonald's put out and all these different companies are trying this like kind of AI ads. I think this was actually in McDonald's, Netherlands. Put it out, but it was an American company that made it, and Kevin, it was fine. Did you see it? We'll play it here if you haven't seen it.

Yeah, it wasn't bad. It was just, there were moments in it where I wouldn't have included that if I were the maker of this ad. Sure. Because like there were a couple moments which a woman gets caught in a door of a train and it kind of looks fake. You would think, and, and, and I will be honest, like the people that came out of this, this got a lot [00:46:00] of crap.

In fact, everybody around the world hated it because it's ai. The same story over and over again. The people who made it said like, Hey, this wasn't just ai. There were lots of people who worked on this, which we have said many times, but I do have a question about this idea going forward of like. It does be, it's a question of whether or not you tell people it's ai and, and, which I think is an important question, right?

Mm-hmm. Do you tell people, but then if you don't tell people and you start, you know, getting stuff that's good enough to not just be, you know, to be regular, uh, footage. What is, where does that leave us in the world of stuff, right? It's a really tricky

Kevin Pereira: space to be, I feel like right now I give it six months and then it just will I, I don't know that it's gonna matter 'cause it's gonna become fairly undetectable and that it'll just be up to you and your brand.

Like is it important for your brand to be authentic with your audience and communicate that AI was used? Maybe your brand will be celebrated for using ai. For most traditional brands, it's not going to be. Yes. The, the other side of this is that I'm seeing a lot of, like indie game developers, Gavin, have AI slop reported [00:47:00] in their steam reviews or whatever.

I, I think seen Oh, interesting. Three cases of this within the last week. And then people will negatively review and just like brigade the, the thing, and it's causing like, it's like small studios that have spent years working the game, not even using ai. Yeah. Or maybe they used AI in like one thumb nail or a promo image or something.

Yeah, exactly. And people are tearing it down. And so I think like, look. If that's the side you're on, like, okay, that's fine. I think innocent folks and innocent creators are being dragged into that as well, and so you hate to see that, but. I also think there's just like a, I, I don't know that it's going to matter in a few years time.

Yeah. I really don't. Yeah. And that's, I think there will be a, a holdout contingent that are still always going to hate it no matter what. And then for everybody else will be like, oh, that was ai. Like, yeah. Okay. That's fine. I guess it's, yeah.

Gavin Purcell: There, there was a story like a couple weeks ago where like, actually people respond to AI content fine.

But if you tell them that it's AI ahead of time, they tend to hate it in a much bigger way. Which of course that makes sense. 'cause they maybe they don't know. I saw an interesting clip on Instagram, which [00:48:00] Gucci put up. The Gucci had a thing where they were basically, it was all clearly ai. It was like morphing people together.

They didn't mention it was ai and there were a couple comments in there saying, oh, it is ai. It's ai, but like. They just did it in the fashion world, they seem to get away with it quite a bit more. Right.

Kevin Pereira: You, you sent me non-player combat, which was Oh yeah. A full length AI powered YouTube video.

Gavin Purcell: Yeah.

Kevin Pereira: And it's all the AI's writing this and the AI's making the decisions, and it's all AI generated.

I watched more of it because it was AI generated just as I was gonna go. Like I have lots of thoughts and feelings about it as like a creative endeavor and as I think. But they made a, a whole thing had they not told me it was ai and I was, it was clearly, I, I would've been out within three seconds. Yes.

But because I had that context going in. So I think it is case by case, brand by brand, but no matter what, there's a section of the audience that you're going to lose immediately. By saying it's AI and there might be a section of the audience that you gain. Yes. Because you say it's AI

Gavin Purcell: and maybe those will slide in different directions, right?

Yeah. [00:49:00] Like I think for right now it's very much this. Right. But maybe it'll slide you in the

Kevin Pereira: direction. Hopefully. Here's, if you wanna hate anything that I do or that we do because it's ai, it that fine. Go ahead. Give me your longitude and latitude. 'cause I'm gonna send my T 800. Yes to your house or your basement or your lean to wherever you happen to be at the time.

Ooh. Lean to and it will cyber kick you right in your damn rib cage. Yeah, the T 800 has some stopping power. Gavin, so

Gavin Purcell: this is a follow up just quickly on our robot watch from last week, the T 800 video that we thought, oh, it was like this, is this too good to be true? And they said very clearly it wasn't.

Well this week because that happened, they released a behind the scenes video of the T 800 and then also they released this video of it kicking the CEO. And Kevin, I think the interesting thing about this is some people are out there like, well, he's definitely like clearly like moving back in a defer direction ahead of time.

This is a pretty impressive video again, so like whatever these guys are doing, this engine AI company, he's selling better

Kevin Pereira: than Cena and his farewell tour. If I may, my WW has where you at because even if this thing [00:50:00] whispered on him and he fell back that way, it looks like there was an impact. Yeah, and I think.

Just the fact that like, look, we saw the Optimist robot Teleoperated this week where he lifted. Yeah. You know, the robot that's serving drinks goes to lift its visor off and then disconnects like it's being unplugged from the matrix. This thing is kicking with force and still standing upright afterwards.

So that alone. Is impressive and again, a little scary. Can we stop, please? Can we please stop? Yes, we should

Gavin Purcell: stop. And it's one more message about, uh, China's advances in engineering and robotics. I, there was a great video that, uh, Marques Brown, Marques Brownley made this week about a Chinese car. And if you saw this, where it's like, oh yeah, absolutely.

You know, it's like a, it's like a car that's worth like an $80,000 car here, but they can sell it for basically 35, $40,000 there because they're so much further advanced in what they have. And like the stuff inside it is much more advanced. So anyway. It's one more thing to keep in mind about like where we are versus a China in the overall thing.

But Kevin, it is time to talk a little bit about some of the things that you out there did with AI this week. It's time for AI to see what you did there.

Speaker 8: [00:51:00] Sometimes ya rolling without a care.

AI See What You Did There: Then suddenly you stop and shout.

Gavin Purcell: Okay, cv. There's a bunch of really fun, uh, nano banana pros that keep, uh, prompts to keep coming out. This one was really cool. It is basically a simple prompt where you can say. Nano Banana Pro, give me this image five minutes earlier, one hour later, and then 10 hours later. And this example here is of a titanic of a shot of them on.

Again, this is one of those things with with the IP that we have the ability to do with Nano Banana Pro, that we may not be able to do that much looking going forward, but you see a shot of. That, you know, rose and, and Jack in the water, and then you see a shot of her on the thing by herself. Like it's just a cool way to do it.

I tried doing this with the shot of the Joker walking down the stairs for one of these things, and what was funny in mine is it like took the pro, it took the point of view of the stairs. Yeah. So it showed the Joker at the top and then it showed the joker in the middle of the [00:52:00] stairs and at the end it was just a blank stares at night.

It wasn't like they followed the joker, it was just the stairs. The house. That's, I love it

Kevin Pereira: by the way, like so smart that it contextually was like, okay, if it's five hours later, it's probably dark. The street lamps in the background and off to the side are probably lit up. It got the lighting, it got the shadows, it kept.

Pretty much the, uh, the architecture of the scene intact. Like I was actually super impressed with it.

Gavin Purcell: Yeah, I think it's pretty cool. So anyway, again, you, there's just always new ways to keep pushing Nano Banana Pro. This other thing is very cool. Speaking of crowds, we talked about crowds before. This is from a Reddit user.

His name is Philip, FIL. Uh. Is it F-I-L-P-P? So good, good. Username, Philip. I mean, basically what he's done is taken famous shots and this, you know, the first one is from the Godfather, the second shot is from Titanic. The third shot I think is from Jaws. And he's just essentially added a crap load of people into each scene.

And Kevin, one of massive things. Crowds. Yeah. Yeah. And what's so interesting about this is like. If you zoom in, it's pretty good. Like you can see, you know, there are a few [00:53:00] artifacts, but in general, like it's pretty good. Now, Philip actually left his example of how he did this later in the comments. He did do further edits in Photoshop, like he color corrected the Godfather one to more precisely match the film.

Smaller fixes were done in Photoshop, generative fill, so this isn't like a one-off output, but it's a good example of like what you can do using generative ai, especially Nano Band Pro as a starting point. And then if you're a professional or you have like professional ish skills, you could make these even better.

Kevin Pereira: I gotta, I'm taking the other approach on this Gavin, that when you tell me that human, further human intervention was required, I'm not buying it. That's slop. Right? So it's like that, that's, that's the new thing. Like some people don't want AI in there. I want zero human involved, zero human involvement.

Interesting. So you're hundred percent. The other side it or get it out, it's either a one. Or it's a zero baby binary.

Gavin Purcell: Finally, we have a great video from, uh, LK Campbell, who's on our Discord, a great, uh, friend of ours in Discord who has spent a lot of time in Sora and has gotten very good at it. Kevin, just play this out loud if you can.

Speaker 8: Duck season, [00:54:00] rabbit season, duck season, rabbit season.

Gavin Purcell: So what you're looking at here, if you're just listening, is obviously that's, uh, bugs Bunny and, uh, Daffy Duck, but you're looking at a realistic looking duck and a realistic looking rabbit in nice. Vision and they're conversing with back and forth with each other.

And it's just a good example of what can be done when you spend time within this thing.

Kevin Pereira: It's an impressive example of what can be done while you can still use protected ip.

Gavin Purcell: Yes, yes, exactly. There is a world, although I'm not sure exactly, those sounded like actual back, that

Kevin Pereira: sounded like that was straight out of the, the Looney Tunes cartoon to me, but maybe not.

Uh, it's been a while.

Gavin Purcell: Yeah. You know, it's interesting. I wonder how do you, do you think he prompted that audio? Because there's no way to upload audio to soa, so it must have come out of soa. Right? So that audio must have come out there at some point. Figure that's all

Kevin Pereira: folks.

Gavin Purcell: But that would be great actually.

That's a great way to end the show. Will we Love you very much. Put the, that's all folks, and we'll see everybody else [00:55:00] later.