Creative Flux

Pierson Marks (00:00)
we're live. Episode 23, I believe. Getting up there, it's cool. I think this last week, yeah.

Bilal Tahir (00:04)
Getting up there, yeah. Now it's routine. We did

this last week, but it was Thanksgiving, so, you know, happy Thanksgiving, belated Thanksgiving to all our regular users, know. All eight

Pierson Marks (00:11)
Yeah, totally.

Happy Thanksgiving. All of our regular listeners.

Well for everybody watching Creative Flex for the first time, I'm Pearson. This is Belol.

And we just talk about generative media and AI. Sometimes more AI, sometimes more generative media, sometimes other things that we get off topic on, but it's always fun. And this week there's a lot of stuff. mean, we got Runway, Gen 4.5, we got all this Google news, we got the Code Red from OpenAI. Where do want to start?

Bilal Tahir (00:44)
Oh yeah,

yeah, yeah. I think let's start with video because I mean there's been a couple of cool video models I think, you know, it's good to lead with that. I think, and it's kind of annoying, and I think it was very hypey because everyone starts wake posting around like over the weekend about next week, everything changes, everything changes. Like, you know, as always, the video world is just gonna be...

Pierson Marks (00:49)
Okay, let's do it.

Bilal Tahir (01:09)
new paradigm blah blah blah and then Monday the end what happened was there was a model called whisper turbo whisper garden or some some something that code word on code arena which is this leaderboard where a lot of these companies will put their API anonymously just to get feedback on how good it is before they actually publicly you know announced their model and this model had was ranking number one it had beaten Google's view 3.1

And so.

Everyone was like, oh, is it a new view? Is it a runway? it, you know, a gling, whatever? It turned out to be runway. Runway released their 4.5 model, which, you know, really good, amazing. Did it change everything? No. Which is why I was like, you know, it's all, hate this highway posting, but really cool, like amazing cinematic, you know, scenes. Like runway, can see, is kind of like the major new video. So they really lean on the aesthetics and stuff. And so you can create amazing storytelling.

For developers like me, I'm not as excited about Runway just because I don't want to pay like $5 a minute over video and I want an API, which they don't have. So this was cool out there. But the second model, which I was more excited for, was Kling. And Kling released their Omni model. an Omni model basically...

Pierson Marks (02:06)
Right.

Right.

Bilal Tahir (02:27)
They have a regular line of model which they also raise. So they released Kling v2.6, was an update, minor upgrade on 2.5, so much better there. But the Omni model is supposed to be combined editability and generation. So I haven't gotten a chance to play with it yet. I'm going to do it over the weekend. But they're kind of calling it the Nano Banana for video, which we'll see if that stands to be true, because Nano Banana, we talked about it.

last episode is really good but you can get really granular updates edits to your video using O1 which is the only model short form.

Pierson Marks (02:59)
All right.

It's

super interesting on video editing. We haven't seen a lot of video editing models and I think the spatial awareness. So, you know, we'll tie this back together. I want to deep dive into both runway and cling a little bit more in depth before we break off.

Bilal Tahir (03:06)
Yes.

Pierson Marks (03:15)
But one of the key things that video editing models need is spatial awareness. In images, you have a 2D projection of some environment onto an image. It's 2D. There's no motion of an image. And so it's much more easy. you see a tree in the background. You select that tree. You edit that tree. You add apples to it. You remove the tree because it's just like a 2D thing. However, in video, when you have 3D,

It's now the tree where it lives over time if the camera is panning or the actor is moving You know those apples need to move with that tree and and stay consistent. So the spatial awareness is key, but this is why I think It's a data question

And so when we get over to world models and we get over to, have like YouTube data and all this like really spatially aware information, I could see Vio4 being like nano banana for video where it's just like amazing. Cause if they have the best image editing model, why could they not have the best video editing model too?

Bilal Tahir (04:11)
Right. Right. And like you

said, have the data. The YouTube data is going to be key for that because it has all these training data points about what happens when an apple falls versus a mug falls, which shatters or whatever.

Pierson Marks (04:26)
Alright.

Bilal Tahir (04:26)
So these cues, but to all the other labs who are doing this, to their credit, they're getting better at this too. So I saw a couple of examples of one where some guy put, like one of the examples was you put someone in an image. So there was an image of a character in a bathroom and she's washing her face and she looks up and there's like another guy that appears, but it's in the reflection. And so all these subtle like, the reflection has to be like, know.

different than the actual image you pan to the person, but a reflection looks slightly different. The model was getting those subtle things right, which is, I guess, part of just collecting the better data and making sure the training is understood. And when you train the model, you're actually understanding the physics world, as you're saying, behind it. So yeah, really interesting stuff.

Pierson Marks (04:51)
Hmm.

Totally. That's super interesting. And

you know what sucks is that they named this, I mean, why would you go name O-1? When you said O-1, I'm like, wait, I just got confused. That was like Open AIs thing, you know? ⁓

Bilal Tahir (05:18)
Yeah.

Yeah, it's weird.

One, think they're trying to call it omnib, which I think they were trying to call it some sort of consolidated model, like one world, one world to rule them all.

Pierson Marks (05:26)
Alright.

Bilal Tahir (05:33)
just weird because they also have the standard image to video and text to video model. But they have this other ability which I like is reference to video. And reference to video is different than image to video. And image to video is like you start with, like you take a frame and then that's your start frame and then you generate a five to 10 second video from that. Reference to video gives you more flexibility. You just take a, let's say an image of me or you and you say I want the character walking through the garden, you know, but the first

shot is actually you start from the tree and then you zoom out and you see the path and then the character appears. So you can do more stuff with it right the character doesn't necessarily need to be in the first frame. So I like the concept of reference to video a lot more and I imagine we've talked about the UX behind how you build let's say a 30 minute video. I think what you're gonna one way I've thought about is like you're just gonna have these characters seen so you're gonna have maybe you take shots of a character you have a frontal face side face it's almost like a police

you know shot like scared like a 3d scan ish type of a you know a bunch of photos of the character from every angle and then basically and then maybe the scene maybe hogwarts right this is the door this is the scene so you have all these like shots of the scenes and then you can feed it into the video model which is it's almost like building the world you know

for the video model and then you can say, know, Harry walks into, you know, the Hogwarts main hall, right? And you already have the scenes in there and you can have a consistent video using these atomic units. So I think it's pretty cool.

Pierson Marks (06:50)
Totally.

Right.

Well, it's very cool. that reminds me of like the other day when we saw like the image to video game type of perspective, like the first person POV where you had a character you went to. I think it was they took an image, they put like it as.

Bilal Tahir (07:11)
Yes, yeah.

Pierson Marks (07:20)
you are in the camera with the wand and you're like have the wand and you're like walking around and so it really looked like you were playing a video game like a Hogwarts video game or Zootopia because Zootopia 2 just came out and that's gonna be really cool. I know you like Zootopia, I like Zootopia. That was an awesome movie and yeah I mean it crushed in China too so it's like I think it beat Avengers in terms of like the highest grossing US movie in China so that's pretty crazy.

Bilal Tahir (07:34)
One of ⁓

Yeah, I was surprised by that.

did not know China had such demand for Zootopia.

Pierson Marks (07:50)
But I wonder if we'll see some stuff come out because those characters are pretty cool too. So maybe there's something there to look into. But.

Bilal Tahir (07:57)
like you mean using those characters as a storytelling? Like, yeah.

Pierson Marks (08:00)
I don't know. don't know.

The concept though of Zootopia is pretty cool. Where you have all the characters and you have character creation and you have the Sly Fox.

Bilal Tahir (08:04)
Yeah, yeah.

Yeah, I think, and it's blowing up in

Sora and stuff too, right? Because you can be some guys like the Sloth or the Badger or whatever. You can get character and just explore the world. I think one, lots of reasons Zootopia is amazing. like you said, one of the reasons was the rich cast of characters. Because it's like this whole anthropomorphic world where every kind of predator or animal is just like, has their own function.

Pierson Marks (08:14)
Mm.

Right.

Bilal Tahir (08:33)
and world and society, many societies within the big society and obviously I mean the parallels are to races you know in humans but still you know it's just such a very interesting way a fun way to explore it you know like a world so yeah.

Pierson Marks (08:49)
Totally, totally.

That's super interesting. Well, okay, so we had Zootopia that, and there's another thing that I want to touch on too before we get to stuff. There was a cool theme that at least on my Twitter was showing up is like these 2.5D isomorphic renderings of cityscapes. Did you see these?

Bilal Tahir (09:05)
yeah,

I think was it generated by Nano Banana or I think

Pierson Marks (09:10)
It was like,

there were some workflows that people did. There was like nano banana, but then it was, then you take those images and then you move the camera up. So looks like you have like kind of like a square.

Block, I don't know. They looked so cool. I'll probably put them in this show. I'll try to find some put them in the show notes, but like they were such cool little like snippets you took New York City and you imagine you condensed New York City into like a little square and it was kind of like clay, not claymation, but it like that soft sort of subtly vibrant

color scheme and you had like either the space needle or you had the Empire State Building or the Golden Gate Bridge and it was like this really cool thing that was going around that people were creating and I liked them.

Bilal Tahir (09:47)
⁓

yeah, I mean those kind of worlds are so cool and that may be I think...

I think I did see the post about how you can use that to kind of explore cities and stuff. It's kind of like Google Maps, but in a fun way. You can just see a block. Actually, one of the most famous, I think, 3D models, if you try to go play around with 3JS, which is the most popular ⁓ 3D JavaScript library, and you look at their main page, one of the top five, I think, video models is this really cool mini Japanese town kind of setting. There's a train that's

Pierson Marks (09:55)
⁓

Right.

Bilal Tahir (10:19)
goes around the square and then there's a shop and then and it's just so cool because it's a very small town but you know not but the idea that you can just you can like imagine people are living here and this is the train that goes from east to west or whatever right it's just so cool seeing that and it opens up I think the reason I think it's popular is because it opens up possibilities for people to go wow I can create this world for myself you know

Pierson Marks (10:19)
Mmm. How well?

All right.

Totally. No, it is really cool. mean, it opens up so many possibilities. It's sick. Okay, so we talked about Runway. So we talked about, yeah, okay, so let's just dive into Runway a little bit more. Cause Runway was one of the first image or video models out there in like 2023. I remember using Runway and like seeing Runway videos back like two plus years ago. And...

It was like you are on crazy drugs. Like you look at some of those videos and it looks like these worlds are like morphing. It looks like you're on like, I don't know. It was wild. But now it's just like so nice. so I think, so they don't have an API. And so that's your criticism of them. I think they've decided to potentially to do that because it probably looks like they're working with the studios in Hollywood. And so.

Bilal Tahir (11:26)
I think they're just kind

of like going the mid-journey angle. They're like, you we do this well. We have very premium product and you we want us, you you want to use this through our UX. And I actually think they might have an AP, but it's one of those where you have to contact them and partner and right, or something like that. So, but yeah, I mean, hey, I mean, it's just, you know, respect, you know, they're doing their thing. They're doing, you know, amazing stuff.

Pierson Marks (11:41)
and pay a million dollars or something. Yeah.

Right.

Bilal Tahir (11:51)
But I think to your point, they're basically the choice for filmmakers and studios. It's like, you know, kind of like the mid journey angle where, you know, people who are actually into seriously making, you know, editing videos and stuff, you know, like you want to make a ⁓ 30,000, you have a $30,000 budget for like a five minute commercial or whatever, you know, yeah, you're going to, whatever, you'll spend the five, 10 grand creating the shorts or whatever, you know.

Pierson Marks (11:51)
Right.

Totally

totally. Yeah, so physical actually Cinematic realism all those things I think does really well and talk about this there was did you see this absurd company they were on TP a TBPN the other day like it's a YC company that was doing Did a call she had not PJs called she had but like a different call she had But they were like saying that they were charging Studios like $30,000 per video. It's like all AI generated and stuff. So it's like costing them nothing

but they have like the taste on how to use these models and string them together. And so they've kind of created like a services consulting company. It's called absurd. And the whole thing is like, we make AI videos. And did you watch that one? I think you've had to have seen this one. It was like a founders fund-esque, like America, like.

Bilal Tahir (13:01)
the

one with the moon landing and the... I think that is the culture, right? I mean, it was American dynamism and all that, right? Basically, yeah.

Pierson Marks (13:04)
Yeah, I think they created that one.

Right, alright, alright. Yeah.

This was like kind of cool. Like if anyone's watching this that's like wants to see...

Bilal Tahir (13:16)
I didn't know that was AI.

I just thought that was just, I guess, yeah, there's part of it that was AI generated.

Pierson Marks (13:21)
Right, I think

it was inspired by the old Founders Fund thing and then they made a new one. But yeah, check out absurd.com. have like, I'm on their website right now, has nine videos. They have one that's like Mondami and Cuomo in the Knicks. ⁓

Bilal Tahir (13:29)
It's a signal range I have, but yeah. ⁓

I mean, but

to your point, mean, taste is like, it's one of those last modes remaining, you know, some people just have that, you know, ability to know what resonates with an audience and, you know, they, they can use these tools. So can I envy that? Having that.

Pierson Marks (13:48)
totally

So

on the other side, mean, okay, we, you were just talking to me about like this artwork show that you went to. I want you to talk about it. Okay.

Bilal Tahir (13:56)
Yes, but before jumping into that, I also want to

touch on one. ⁓ I think one thing to close the loop on there is the editability. Because we talked about how editability wasn't. I mean, there were other models. Like there's Ray 3.

Pierson Marks (14:03)
⁓ yeah.

Bilal Tahir (14:09)
just cool kind of expensive you know that would let you edit a video but this is just taking it the next step I think link is traditionally I think I just think it's up there with Google in terms of how good their video is and so the fact that they really focused on this model to be editable you know making editability a first-class you know feature means like I think it's gonna open up a lot of interesting remixes you know people have used Sora for remixing and stuff like that I think we're just gonna see a lot more of that where you're gonna take

your favorite video, whatever, maybe you switch Tom Cruise for Denzel Washington or whatever, right? You know, like.

and all this like taking old, not old, but like classic content we grew up with and stuff and just remixing it, it just opens up a whole Cambrian explosion of content. I think studios, some studios will fight it, but a lot of studios will just embrace it because it actually, if you take Jurassic Park and somebody makes a studio give you Jurassic Park, people are gonna watch that and then suddenly Jurassic Park is in the side case again and then they're gonna go back and watch the original. So it actually helps the original content in a lot of

Pierson Marks (14:49)
Totally.

All right.

Bilal Tahir (15:11)
ways

so I think we'll just see ⁓ probably this being buttoned up in terms of royalties and stuff like that but also studios actually making it easier at least the forward-facing studios you know who embraces technology they'll probably be like yeah use Mickey Mouse whatever and then some people may probably like the studio give these other who are very focused on hand-drawn stuff they'll probably fight it and probably lose in my opinion

Pierson Marks (15:13)
Totally.

Right.

Well,

you know what's really interesting too? So this was a few months ago. There was a video game company. I forget the name, but the original Mickey Mouse, the Steamboat Willie animation entered into the public domain after 75 years, right? And so it was like that classic black and white Mickey Mouse style, very different than how Mickey Mouse is portrayed today. Similar, but different. But they, the moment it went to the public domain, they created a video game with Mickey Mouse and that.

like Steamboat Willie style with like a first person shooter game.

And it was really cool because like you can't do anything about it. It's public domain now and whatever. as like that's like the Steamboat Willie was like the first, you know, IP pretty much of like the modern era in media. But let's say like 10 years from now, even 20 years from now, when some of this IP that gets starting to get more released, like you could see this kind of exploding where like all of this old stuff that from the, I don't know, 50s, 60s starts to become available.

Bilal Tahir (16:36)
Yeah.

Pierson Marks (16:38)
publicly for anybody to use.

Bilal Tahir (16:38)
Yes. No, 100%. And it's funny you mentioned that because I think I just saw the article on.

news the other day I think in November is when the next year's copyright the public domain happens so I think all the artwork until 1926 and there's some rules so there's there's a 75 year rule and there's one if when the like in domain artist dies plus 50 years or whatever so depending on which it's I but majority less just for keeping simplicity anything before 1926 now is gonna be in the by entering the public domain so it's always an opportunity whenever this happens over deck I definitely recommend

our users check out what's entering in the public domain because usually there's some new stuff, the interesting stuff, works being entered into public domain that you can jump on and take and do your spin on it. with AI, it just makes it even cooler. It enables you to do even more stuff, so check it out.

Pierson Marks (17:29)
Totally,

it's gonna be super cool. And just like, at the end of the day, if you could get out every, like, Netflix spends so much money to create shows that come out week after week. You Severance, there's a new show on Apple TV. There's like, you know, yeah, yeah, play-along, right. But like, everyone's talking about it, I know. But you know, if you could tell a studio like, you could snap your fingers and just have more high quality content out.

Bilal Tahir (17:43)
It's pretty cool, interesting though. Yeah, everyone's talking about it.

Pierson Marks (17:55)
No studio is going to that down. I mean, like there becomes an area where it's like there's too much out there, but we're not nowhere near. And if you could just create higher quality content that comes out more frequently, going from like a release every Friday for the whole year versus like, you know, have a series, you have season one and then season two comes out the next year. Like if you could just keep that series going, if it makes sense from your story perspective, but like every single studio would want to have, you know, more high

quality content that they can publish. know, like nobody would ever want to turn that down. So like if you told me there was a new Star Wars movie that came out once a month, I would watch a Star Wars movie once a month. Like as long as it's good, you know? And so just like...

And we're entering to that future where it's like, it's not gonna be once a year, you're have a new blockbuster movie. It could be once a month, it could be once a week, like The Mandalorian or like all these things that they're putting a massive budget to. So it's gonna sick.

Bilal Tahir (18:51)
Yeah,

it's interesting. also think I also get the other side of it, you know, because there's something about value and scarcity. It's like, know, and so when something when you get a Starbucks like a Star Wars movie once, Star Wars movie once a month, you after you might be excited for the first six months, but then you might just get it might be too much. And then you'd be like, ⁓ I mean, I'm kind of getting sick of Star Wars. So it'll be interesting to see how that balance plays out to.

Pierson Marks (18:58)
Totally.

Right, for sure. Right.

Totally.

absolutely. Right.

And that's a good problem to have, you know, if you can just because it's no longer limited by like the actual like limiting factor of production, it's more becomes a business strategy decision like when do we release this stuff? How frequent? We still want to have that scarcity. And so it just eliminates one variable in that equation and makes it much more easy to plan out, you know, your media strategy.

Bilal Tahir (19:39)
Yes, yeah.

It's a classic binge versus time release schedule. I think Arun put out this post and he was complaining about why are these old studios, they still do release ⁓ episodes once a week. This is old fashioned, we should just binge it. And my reply was like, I'm a binge watch, I love bingeing, but.

I will say this, if you want to keep a show in the zeitgeist, if you're doing Game of Thrones episodes one week at a time, you put out an episode, then people get a chance to discuss it and then talk about, what's going to happen next? It's annoying, but it also builds the show up. So in a way, you need that. It's not just legacy stuff. So in a way, you want that artificial, let's wait and see. I'm going to give you just little bit tidbits of stuff. So we'll see.

Pierson Marks (20:07)
Right.

Yeah.

Totally.

100%. Yeah. I like once

a week shows. I'm like, okay, cool. I could only spend one hour a week watching this thing. I'm not going to spend like 10 hours binging this thing over the weekend. So, but cool.

Bilal Tahir (20:33)
Right.

Yeah, yeah, but yeah, that'll be true.

Maybe in the future, they put out a show and then everyone remixes it and then they try to predict like which version like, you know, is the one, the true one. So sometimes.

Pierson Marks (20:47)
Well, that's what

Disney's doing. know, that's what they announced. Bob Iger announced on their earnings call is like now you're able to remix Disney characters. So from your favorite shows, you're going to see this really quickly. Like it's going to be on your phone. It's going to be on your TV or something. I don't know how they're going to do it. like imagine taking your Zootopia characters and just like, hey, wait, what if I forget the name of the rabbit, the bunny, but like what if

she'd made this decision at this scene instead, like how would the DMV interaction work or like, know, yeah, Judy, yeah, Judy, ⁓ So.

Bilal Tahir (21:15)
Yeah, Judy, think. Yeah. But yeah, and

who knows? Maybe if you make a sick idea, Disney calls you up and they're, can we buy your remix? Like, we actually think this is sick and we want to introduce, you know, you get promoted to cannon or something like that. That'd cool.

Pierson Marks (21:32)
Yeah, they're not giving you money. They're like this is our IP you can make whatever you want, but we're gonna take your idea. Yeah 100 % well, okay, so cool. So there's a few one like So one thing that we talked about in the past that I think we just want to touch on I mean So open AI if you're following in the the world

Bilal Tahir (21:35)
That's fair. But hey, at least you get bragging rights. You're like, I made the best bunny or whatever. Nice. So yeah.

Pierson Marks (21:56)
Sam Altman issued, guess, some code red. It's like, ⁓ Google's catching up. I mean, no brainer. Like, it's everything that we've been talking about for the past few months. I saw your tweet the other day. like, all you bandwagon folks now saying Google's going to win, OpenAI's going to lose, bearish on OpenAI.

Bilal Tahir (22:03)
Yeah.

Yeah, I mean, I love how they come out and say, contrarian

bet. think Google might. I'm like, yeah, now it's contrarian for you. It was contrarian six months ago and everyone was just like, you know, still on the opening of Bandwagon. You know, people, it was pretty clear. And who knows, maybe it changes now. Whenever everyone gets on board, then I get nervous. Now I'm like, I mean, I'm a Google bull and now like, maybe, maybe I got to, I actually took some chips off the table. I was like, I don't know. But, who knows?

Pierson Marks (22:21)
All right.

Hmm, it's just saying right. It's trading around like

300 something I think it's like 3.8 trillion dollars. It's kind of crazy But I know we sometimes devolve into stock not financial advice, but like it's crazy four trillion dollar company. Yeah

Bilal Tahir (22:42)
Yeah. Yeah.

Hey, everything is financialized. ⁓

The Colshe founders just became, they both became billionaires, and Luana Lopez, think, started last time. She's one of the Colshe founders. She became the youngest self-made billionaire. I mean, she beat Taylor Swift by four years. She's 29. Just like, god damn it. Which is, mean, prediction markets. I think, I generally, I mean, it's kind of like,

Pierson Marks (23:02)
Mm.

Wow.

Bilal Tahir (23:12)
late stage Roman Empire vibes but I do think in a post-AGI world we just maybe just gamble and stuff you know like that that's just kind of what you end up doing.

Pierson Marks (23:18)
Gamble away

your your universal basic income Yeah, ⁓

Bilal Tahir (23:23)
Yeah, I know, because you're just bored. I mean, whether it's

good or not, I mean, can debate about that. it's coming. I think basically everything becomes indexed on the exchange and something you can predict. It kind of just makes sense in a world that's just hyper-connected and overfinancialized. So yeah.

Pierson Marks (23:45)
I mean you tokenize a lot of things. It's interesting but Yeah, so open AI Code red cool talking about Google

Bilal Tahir (23:51)
Yeah, so

sorry, just to kind of like what happened. Apparently there was a down hole, Sam opened it and he basically Google had released the Gemini 3. that actually, the reason that he said it's a code red is because if you look at the charts and stuff, a couple of things. Chad GPD's mouse have, they've been stuck at 800 million for a while. They were projected to get to 1 billion, haven't.

Pierson Marks (24:15)
I wish we

were stuck at 800 million.

Bilal Tahir (24:17)
Yeah, I

know, know. Same Jelly Pop, you know, same, we're having the same problem too. 800 million is just a hard number to break, you know. No, but they've been stuck there, they've plateaued there for a while. And the second thing is their Gemini has gained like 10 % share or something like from Gemini 3. Sorry. Yeah, just from that model release. And I actually don't think it's Gemini 3. I think it's Nano Banana. Because you know, image models, we saw it with OpenAI with the...

Pierson Marks (24:21)
I know it is.

Bilal Tahir (24:41)
their GPT image model as well. That's something about just creating cool imagery and editing it. It gets in the zeitgeist very fast. I think that Nano Banana is what propelled Gemini forward. Gemini 3 obviously helps as well. So they need to respond to this. I've said before, OpenAI's mode, if you think about it, is consumer, which is a very fickle mode. At the same time, they do have the, they're the Google of the.

the AI space, know? People say, apparently young people, say, ⁓ you know, they call chat, say, use chat. You know, so it's like Google it, the equivalent of Google it. So they just say, I use chat. So it's a sticky product, but at the same time, I think the problem with OpenAI isn't that they don't have a good product and it's not sticky. It's the fact that they've just committed.

Pierson Marks (25:19)
Interesting.

Bilal Tahir (25:29)
to getting to this insane revenue and they've signed all these partnerships, their good partnership deals and they could very well end up in a hole, even though they're making a shitload of money, but just because they basically have lost 270 million, billion dollars this year and they're projecting they need to basically turn that around and make a profit by 2029. So.

Pierson Marks (25:55)
Right.

Totally.

Bilal Tahir (25:55)
That's kind

of the VC bet. It's like a great essential VC bet, you know, that they can't just be like, we're profitable. They're make a shit ton of money to make their investors happy.

Pierson Marks (26:02)
Right, right,

And the other dimension too, so like, know, OpenAI, Johnny Ive, their consumer hardware products, whatever that is and when it comes out. You know, I respect Johnny. think Johnny, you know, the iPhone and everything like amazing and the work that he's done at Apple. It'll be difficult for me. I don't know. I try to see. I try to, you know, run the gamut of like what these potential features could look like. And I mean,

I don't see ⁓ them coming in and creating a device that blows out like Apple or Google. Like Google has Pixel. They have the Android systems. They have Waymo's. I think that like if somebody's going to win consumer, it's going to be Google.

Google's not an enterprise play, like Microsoft Azure maybe, like Google has Google Workspace, and we'll talk about that in a second, but like, if you, one, Waymo's are expanding like crazy, and like, people don't like think about Waymo and like AI together, they kind of do now a little bit more, because obviously it's self-driving car, but we saw this massive expansion of Waymo this past week, like all of California, so.

Bilal Tahir (27:04)
all the way to 10,000.

Pierson Marks (27:07)
All the Bay Area got approval from the DMV in NorCal essentially and then in SoCal all of it from LA to San Diego got DMV approval. Full self-driving in Houston. A lot of the parts in the Pacific or in the like Southeast like Florida like all these places are starting to get Waymo's and Baltimore and like it's going to be really really cool to see more people experiencing like this. That puts Google like people think about Google they think about I'm just going to Google on the internet. No Google is going to come into

how they get around town, how they interact with the world. And so if you're in a Waymo and you have that screen there, like, like, welcome, like you start talking to Gemini, like, ⁓ like, you know, talk to Gemini, you have a screen, you could like do stuff like it. I think that is a different consumer dimension that if you're getting around places, like you're stuck in a car, like you're going to be forced to interact with like Gemini or whatever, it's going to be there. So it's interesting.

Bilal Tahir (27:59)
agree.

And this is just a consumer angle, is why I think the B2B angle is even, as we know, mean, B2B is just better than B2C. You just get better margins, better customers. And that's why Anthropic is in a way better position, because they just focused on their API platform and their mode. And they're fundamentally a B2B business. Google, we're going to touch on.

⁓ Touch on I guess we won't speak too much about it by Google release the gems Google workspace launch which basically bakes in Gemini's intelligence under the hood So now you can and this is how is you're gonna use AI going forward You're just gonna go to Gmail and be like hey, you know if this then that but in a smart way like if I get a an email that sounds like spam You know spam, but if it's from this one company forward it ping me Otherwise, don't bother trash it like you're gonna be able to create these rules. And so they release this Google

Pierson Marks (28:26)
Alright.

Bilal Tahir (28:48)
workspace update which makes in a lot of these features and Microsoft will probably do the same you know and I think the ecosystem will become a lot more important to open-S point they've tried to do it with Sora I mean it's just hard to tap into that distribution right if you're not a Facebook or a Google right it's hard so it's an uphill battle and on the end they basically are not

B2B business because these other competitors have taken that space. And it's funny actually, I don't know if I wanted to mention, but it is really funny. There's an interview that Dario Amadai, the CEO of Anthropic, just did yesterday. It came out on the New York Times Deal Summit with Andrew Ross Sorkin. And it's just fascinating.

how much he hates Sam Altman. I had no idea. I know they were competitors, he was just like, well, there are some companies who prefer YOLO, but I like to be calm and collected. There are other companies. And Sorkin's just like, what other company? We all know what company you're talking about. He's like, I can't say the name, but whatever. It's just very in your face. He hates Sam Altman. know. Yeah, I agree. I agree.

Pierson Marks (29:51)
They're both sketchy people. Honestly, think Dario's sketchy, Sam's sketchy.

Honestly, Dario gives me like a little bit more of the heebie-jeebies. Like the other day when they like released the whole stuff about the cyber...

Like we re like they said they released a report about there was a state sponsored like thing and it was kind of like debunked kind of like Like they're playing the good guys with like the regulatory capture Sam Altman's playing the other side of like Like it's crazy. I mean Google's kind of just out there like you guys go like do this and we're here to eat your lunch like we're the big giant in the room with a wave coming you guys got to keep running keep swimming because this wave is coming and I don't know

Bilal Tahir (30:08)
Right.

Yeah.

Pierson Marks (30:31)
No.

Bilal Tahir (30:32)
Yeah, I agree.

mean, definitely said it before, Anthropic is just, gives you this vibe of people taking themselves too seriously. I mean, just, this comes down to like, you know, it's like they're the savior of humanity. I'm like, geez, get over yourself. But still, great models, Opus 4.5. I don't know if you get a chance, amazing model. So they do, they have figured out something about the RL training is figured out coding and they're really leaning into that.

Pierson Marks (30:42)
Yeah, totally, totally. And one of the... For sure.

Try Opus 4.5.

Hmm.

You said that the design was good. I mean, I got it to produce some purple lucid icon design slop and I was just like, whoo, the sick shit.

Bilal Tahir (31:04)
Wait, wait. Did you use

Cloud Code though? Because Cloud Code still uses Sonic 4.5. You gotta use just the Cloud regular.

Pierson Marks (31:13)
I was using a cursor.

Bilal Tahir (31:15)
Or cursor, I guess.

Yeah, yeah. Super expensive. I blew through my budget in like a day. yeah, it's good. Codex is good, too, would say. OpenAI did respond, but it just takes so long. It's very fascinating that what OpenAI just, it's like that meme, know, look what they have to do to mimic a fraction of our power. You know, like the, what do call it, the show.

Pierson Marks (31:23)
totally.

Bilal Tahir (31:38)
the SuperAV show, forgot to say, ah, so good. it's because, invincible, that's the show I saw. But it's because OpenAI will take so long to, because they lean on the thinking, and they'll spend so many tokens and just thinking, and then it'll do the writing, but then Opus 4.5 just does it, you know, without thinking too much, so, yeah.

Pierson Marks (31:49)
Mm-hmm.

Right.

Bilal Tahir (31:59)
You're like, wow, what

Pierson Marks (32:00)
Totally.

Bilal Tahir (32:00)
if Opus 4.5 had the thinking of OpenAid? It's like with your powers combined.

Pierson Marks (32:03)
for

sure. And the last thing I know I want to just touch on right now is just because we're still on Google and I saw this thing this feature today about in Chrome. It was the split tab view.

Bilal Tahir (32:16)
What is that?

Pierson Marks (32:16)
and it

was really interesting. So you know how like if you have, if you want to open two windows, you want to see two browser windows on the screen, you take, you open up a Chrome tab, you open up a Chrome instance window and you have another window and you open up two windows, you put them side by side on your desktop, right? That's what you do. I just saw today they released this thing was split Chrome tabs. So you can take one tab, click on it, right click on the tab and say, add tab to split view. So if you do this,

I don't know if it's like rolled out fully, but maybe if you right click on the tab and then you click

create split view, it'll actually keep one window, like one Chrome window and split that one tab into two tabs. So you have like a split tab view. And it was really cool because I'm like, interesting. Now you're going to have one tab that's active in one Chrome window instance where you can see something on the left and see something on the right. And why is this important? Because you could really clearly see them this laying the foundation to integrating Gemini within your tabs. And it's going to have like, it's not two windows where you have two Chrome

instances windows that don't have like connection to each other it's literally you have one window open and the tab is split into two and you can see on the right hand side probably you could just search internet like normal but like that will be where Gemini lives if you want Gemini to live there and you have your browser and you have your website on the left and like they're just laying the foundation for this to come out very soon so I think the agent browser is coming more that quicker than you know

Bilal Tahir (33:36)
Yeah, that's interesting.

So just to clear it, you don't have to toggle the tabs. You can see both tabs at the same time, but it's like half and half.

Pierson Marks (33:44)
It's literally

that you have one tab open, but that one tab is split in half where you have like two things.

Bilal Tahir (33:50)
And the initial site is the same site, but then you can do stuff here and that doesn't change the other view. Wow. Yeah.

Pierson Marks (33:57)
Right, right, right. And I know

from Chromium, like, I don't know what the security that they did here. Like, I hope they're isolating both of these tabs, but I know in the past what happens is like one tab is essentially one container where like a website and one tab can access the website and the other tab. And so hopefully the split tab is doing the same thing, but I could see it maybe they're figuring out a way where one tab can interact with the other tab through like some APIs or something where...

Bilal Tahir (34:12)
Mm-hmm.

Pierson Marks (34:25)
Gemini can interact with the website or just the tab itself because you can see maybe Gemini can like use vision and interact with the DOM and it doesn't share cookies, it share scripts. Because that's the danger, it's like when you start sharing scripts and you have cross site request forgery and you have like cross site scripting and all these things. But it'll be, it's really cool. I was stoked to see this today. was like okay, Gemini's coming to browser very quick, very soon.

Bilal Tahir (34:47)
I think you're going to see AI baked in basically in every function. And the browser is the most obvious entry point where most things happen, which is why I've always been not. been an ARC or browser. I mean, I've never really tried it in fairness. I know Chrome is just a feature. Chrome is going to have, actually.

Pierson Marks (34:55)
For sure.

Alright.

Yeah, so that'd be sick. I think that note ended on Google, talked about runway, talked about a foul or.

Bilal Tahir (35:15)
Yeah.

Yeah, jam-packed

episode, like, you know, lots of updates. We only touched on the video stuff and the spicy gossip, but other, you know, lots of other new models came out, image models we didn't talk about, but a quick, a Sieve Dream 4.5 came out, very good. There's a model where, you know, I'll just mention Zimage, which is very fast and very cheap. I think it's like the cheapest model out there. And Flux 2 came out as well, you know, which is updated in Flux Dev and Flux Pro.

Fluxtnl, which is the cheapest model anymore, but the Flux2 dev is pretty cheap enough. I don't think they need Snl. And you can actually, there's a new one called Flux2Flex, which actually lets you edit the number of inference steps, like even more granularly. So you can actually kind of create,

your own level of granularity in terms of quality you want. very cool stuff, lots of cool models out there. Check it out, know, unfollow or replicate.

Pierson Marks (36:11)
Totally. Well, cool. A lot of stuff. on that note, we'll see you all next week. Cool.

Bilal Tahir (36:18)
Yeah. All right. Take care, guys.

Bye.