Runway CEO: Gen 4 video model solves consistency, enabling solo filmmakers to produce short films in days

joining. I I have a million questions, but let's just start with can you introduce yourself, the company, and uh what you launched today? Yeah, I mean, you did kind of a pretty great job. OG, you've been using Runway since since a couple of years. Yeah, that's great.

Uh yeah, so we just introduced Runway just a couple hours ago. Uh sorry, Gen 4 just a couple of hours ago. Um it's our latest like video model. It's basically a new frontier, I would say, of not only video generation, but of storytelling.

It's uh it solves one of the most pressing things of uh AI video these days, which is consistency. So you can create really consistent worlds. And the way we wanted to like emphasize this idea of consistency is that uh besides just releasing a model and I think it's kind of the model part is like the first stage.

Uh we put together a collection of like different shorts and films that if you watch them, if you guys haven't watched them, like definitely try watching them. It's just like a few minutes. They're short films. It's crazy.

It's such a trip because you're looking at it and you know it's AI and it's does not We're passing the uncanny valley for sure. and faster than the VFX stuff did. I mean, we that that took 20 years and it feels like AI video generation, it's been like a few years.

There was an uncanny valley for a while, but we're getting the other side really fast. Yeah. I mean, we weren't going to run for almost seven years. So, it's it's definitely been like more than like a couple like two or three years, but uh but yeah, I would say the last two or three years have been like really fast.

And uh I don't think we're even close to like the final stage of her media will come. Yeah, I think there's still a long way to go. But but the point is really to point that at uh once you start like seeing how good the stuff that you can do runway is and you layer in a really good creative mind.

You can make stuff that you just like forget how that you're watching something that's completely generated. Yeah, we need a new eval. We had uh the Will Smith eating spaghetti. We blew past that. That's photoreal. The next one I think should be uh born identity style shaky cam.

You know that because I think AI could not do that today. I think the next evil should be just like how good the movie is, how good the shirt is, like you know, like how many awards, how many people are watching it. That's the that's evil.

Like I think that's the that's our internal goal is basically to make sure that you judge by the quality of the story more than anything else. Uh I'm sure last Wednesday whenever the Giblly meme started ripping, I'm sure some of your investors, big and small, reached out, hey, did you see this?

You know, I'm curious what uh if you had to answer that question, like, you know, what what was your immediate reaction to it? Uh it's amazing that you guys so quickly followed it up with something that is, you know, uh Yeah.

So, you come on the show, we ask you the stupid questions and then you don't have to answer the stupid questions from your investors. Yeah. Yeah. You just one time, you share the clip. So, so I I have to be honest. Uh no one reached out to be honest. Like I don't think like Yeah. It wasn't like a like a thing.

I think it was it I think there's a separation between like memes and like more of the casual like fun stuff that I think it's great. You can just kind of like have fun.

It's like a trend, but I think if you're making like good professional content, if you make a living out of making content, like making a meme or like one image, it's not going to get you far. It's like you need a suit of different tools. Yeah.

Um and so I don't think we've I I saw the trend and I saw the images and I think they're great, but I don't think they're No one really thought about us to to be honest. Yeah.

Uh the thing that the thing that I think was what what part of what made that so viral is that it didn't require sophisticated prompt engineering. You image oneline studio Giblly. You can spell like a text uh replacement on my iPhone a shortcut.

Now I just upload a photo SJ SGS and it automatically expands to Studio Giblly style and then it just does it. Yeah. But but so so more importantly as it as it comes to using runway and and video uh how important do you think prompt engineering is in the context of runway long term?

Is I'm assuming it's just not going to be a thing. You just sort of describe exactly what you're kind of feeling off the cuff and it creates something beautiful. Yeah. Exactly. You know, I I so prompting is a way of like sampling something out of a model, like basically getting what you need.

But there's many other ways of getting what you need with much more control because language doesn't really like have all the nuances of like describing a composition or the style, the mood or the feel of something.

So the idea, for example, with Gen 4 is that I saw like the demo, but you can combine images of real things. You can take a photo of something you've actually like have.

You can take a color grade and a kind of like a moody scene and you can take a location and you can there's no prompting just like put them in and like create like the composition that you like.

Um and and you you can get a much more like specific detail thing rather than just like trying to use the words correctly to like get to where you need to go. So it really depends on like what you need. But I think prompting I think prompting is overrated.

Like just language as a unique interface uh will not get you too far because like storytelling and films particular in video is just so hard to just describe in words, you know. Yeah.

How how uh how many years away do you think we are from being able to oneshot a a movie that could, you know, win uh you know an award-winning movie, right? It feels like so many movies in Hollywoods are sort of derivative. hero's journey, its set of sort of characters, but it's somewhat predictable.

You could imagine that like from a storytelling standpoint, a lot of models could already deliver something that's like pretty good from a story standpoint. And so to catch up, you need uh the the I mean, from what we're hearing, like the GPU cost alone would be like hundreds of thousands of dollars right now. Sure.

Sure. But but you know, if you're a movie theater today, I mean, we we're so we're here in LA. We've been actually touring studio spaces and almost every single studio space is completely empty because like it's just too expensive to film in LA.

So I I I can see a world in the future where a big studio is just like run does these expensive, you know, sort of prompts that then they get maybe they remix it or whatever. But, you know, I'm curious what your take is.

Yeah, I mean, if you look at the the lonely little flame, the short I was just mentioning before, that's production value before Gen 4, we're talking about hundreds of thousands of dollars. That's a very expensive thing to make.

It took us like two to three days to make uh like the first rough sketch of that and then like another few days to just like get the final comp. So, you can and by version of one, the whole thing was done by one person. Wow. And so, um I think you you are there.

you can technically just take that same process and it's something we're doing already and do it like for longer future films.

It's more of like you're going to sustain the attention of your viewer not by how fancy or how expensive the thing is like by just how good the story is and just iterating on the shots takes you a lot of time.

And so I think the issue with the way we've been traditionally making like very high-end production things is that um the budgets have like skyrocket to the point it's unsustainable.

Like I mean to your to your point like you're spending hundreds of millions of dollars trying to make something and then if you need to make something you're going to bet for the thing that the safest the safest which is like superheroes and and and franchises that you know will return which I think from a from an like uh film making perspective has kind of like prevented like new independent stories from actually like being made because it's more much more risky actually.

Um, and I think part of uh part of this new wave of models is that it will kind of like flip that a little bit and allow this like independent set of people actually don't don't use those studios just like use runway and like make the films to yourself. There's no excuses like you just do it yourself, you know. Yeah.

Do you think uh the big Hollywood studios are reacting quickly enough to the advancements of runway and other you know are they do you feel like they're sort of properly planning around the sort of the biggest change to film making in our lifetimes?

Yeah, I mean I I think so many are I mean Bob said it was the most uh important technology of her company uh just a couple days ago. I think many more that we work with are pretty much on that same vein. I think some are more advanced than others.

I think for a long time it wasn't clear to many what this meant and and to be clear I think that's fair because if you look at like video generation four years ago no one took it seriously like no one cared it was so I mean maybe you were like using runway for like rotoscoping and green screen but like that was a small amount of people right and I think now I can show you a film that makes you feel something and it was done like in two to three days all generated and that's when you wake up and I think some have woken up before but some are still like dormant.

Uh but I think yeah everyone's really paying attention to it. Can you talk a little bit about uh just like style transfer animatics and the importance of like generating the first image in something like images and chatg.

Uh, I could imagine a world where it's a lot easier and maybe even more efficient to block out a story in Minecraft or in Roblox and really direct it and be very very solid on I kind of program the camera move. I know the shots.

I know the layout, but then I just want to take it to photore and that style transfer is what you guys are great at versus I'm going to be purely prompting and really hands off. Uh, how do these things evolve? What's most important? Is there an important like step function where we work through one to get to the next?

No, I think you're right. There there's many people who've been using runway to do exactly what you're saying. You block the scenes by using either like 3D software or just like props. Uh, we have a a way of rendering content using existing video.

So, I don't know if you guys have seen it, but you you use a performance of yourself. It's like advanced motion tracking and then that drives a performance of the generation that you're creating. Um, and that's great because it allows you basically to like you become your storyboard.

You are defining the shots in the way you want it. And I think the the essence also is that um once you create like a mood there there's a few for example another mood another short film that we released called the herd. It's like the story of like two guys like hiding from each other in like this dark moody scene.

uh we start with one image and then you can use Gen 4 to kind of position the camera in different parts of the scene and you kind of like it's interesting because the the the model becomes like a cinematographer.

It's kind of helping you understand or provoking provoking you with ideas of how you want to do it which is different from how you do film making traditionally because you can't afford to do that in traditional film making.

You have a kind of like a set list of shots that you need to make and it's very expensive to like not do those. here you can just like wander around and kind of like find your angle. And I think that's also interesting. This changes the nature of how you make things in the first place.

It's a much more experimental like you're kind of finding ways and you're getting to stories that you never thought you could.

That that to me is the most exciting thing because historically if you're a director and you're like I'm going to bring you know these you know superstars out to this like desert and we have one day to get this shot and then we got to go this other location and do this other shot. storyboarding.

You're basically like you could have a better idea midway through the shoot and not actually be able to do it necessarily without increing the producer and being like, "Hey, I need another 20 million bucks because we miss the same. It's like one person sort of deciding in the moment.

set and if you think about how different startups have been historically where startups are this sort of like daily iteration and you can adapt the strategy and you can adapt the product versus film making which is like well you write the script and yeah you can change it along the way and you can change it in post but then every decision is just so much more expensive.

Exactly. Yeah. I mean I've been thinking about it as um like you're pay you're you're playing in a in a jazz band and you're improvising and so you're you get a shot you see something and then you can go anywhere you want. It's like whatever you want to fill and then like take it there. Um it's very fun.

It's it's something I haven't filled for in totally in really a while. H how do you think about the the the kind of shift in or the ratio or breakdown of creative video content as it relates to fiction versus non-fiction? Like YouTube and the iPhone has exploded the amount of creativity.

Uh and we've gotten hours and hours of like non-fiction content. Everything from like Mr. beast videos to video essays, histories, podcasts, all this stuff that's non-fiction. Fiction stuff hasn't really taken off on YouTube in the same way. That's still kind of the domain of Hollywood.

Of course, Tangerine won uh Academy Awards. It was shot on an iPhone. It took a number of years to get there from, you know, the iPhone being introduced, but it but it did eventually democratize film making to the point where an iPhone filmmaker could win awards.

Um, do you think runway is more of a of a benefit to people that want to tell fiction stories or do you think we'll see kind of the same breakdown where there'll just be a ton of non-fiction content that goes through Runway and you're seeing like illustrated video essays or I can just pull up history of, you know, Yura Mate and all of a sudden I just get this beautiful story that's visually intriguing but it's still non-fiction.

I think Renui is made for stories. That's what the ethos of the company is. And I think fictional or not fictional like stories are just stories. And I think we're humans are like obsessed uh with stories and will consume anything that tells a story. And so I think I think you're right.

YouTube hasn't been like yet good at perhaps like AAA content or like professional like uh fictional content and it's better to do like more informal like blog type of things because it's just cheaper. Mhm. But I think this is particularly interesting because like what gen models like Gen 4 do is break that barrier.

Now you anyone basically anyone out there who has a YouTube channel right now can create a type of content that looks like someone made it in a studio with $50 million. Um and I think that's the that's like the last unlock I would say media wise. The first one was like everyone can do like this.

Everyone can just like tell like create their own like channels and tell their own stories. But then professional grade content has been for the for the last couple of years, the last kind of like stage and it's only been accessible to this very sort of companies and now it should be everyone.

It should be basically anyone who has great stories. It's like creative metor whoever has the best stories will win. What happens to movie stars, right? You can imagine uh early on with some of these AI generated films, somebody's planning to generate a film using entirely AI.

They don't plan to have any onsite shoots or anything like that, but they want to go they go to Brad Pitt and they say, "We're going to give you 10 million bucks to star in this film because Brad Pitt, if they just completely use his likeness and his voice and whatever.

" Uh, but then over time, people start realizing like, hey, it's actually much more valuable if I kind of create my own potentially like IP or own superstars from the ground up because I can instead of paying Brad 10 million for one movie, I could invest a lot of money in this sort of Yeah.

I can make my own gig of Chad. Yeah. And uh I don't have to pay any of the fees and maybe there's some R&Dish to sort of generate that character. But what happens to our our our stars? Yeah. Uh well, I don't think that changes that much to be honest. I think we we had that capacity of doing that for a long time.

I mean, animation and like CGI and anything that's non like live action has uh we've been able to do that for quite a long time. It's more the cost of it is really expensive. I think you can still have a space in a world where you can have both. You're going to have characters and actors performing the way they do.

And you're going to have this new medium. It's a new format. It's like a video game that allows you to create infinite worlds. Maybe the two might overlap. You may bring some elements of the real world into this. You may create some elements generated into and it will swap. I I I don't see them as a replacement.

I don't see them as like, oh, we're going to suddenly like stop hiring actors. Actually, for most of our shots, you actually require actors to do like act one or performances that you can transpose into the way you want.

And so, um, I think it's more of like a change more than like, oh, we're going to suddenly like stop doing everything we're doing and and and do this thing now. That makes sense. Uh, do you have any type of sort of like long-term cost target for a 90inute movie in 4K? Is it something that in in a decade it's a dollar?

Uh, or is, you know, how how long just comput wise or like people time? Yeah. Yeah. Ignore ignore the sort of like uh the prompting. Yeah. OPEX of like like you used to have to pay by the foot for film and it was very expensive to get it developed.

Now anyone with a phone can record 200 hours of footage for you know a couple hundred used phone. So so I think maybe the way they need to think about cost is to think also to the point so to understand how like the models and the optimization of these models is like progressing.

you're going to get really soon to a point where you can run these models in real time. And so by the time you run them in real time, you're streaming the pixels as you're watching them, interacting with them. And so if you do so, then like the story first of all never ends. There's no 60, 90 minutes.

It can be days that you can just like start generating. And it's a cost of inference. It could just be as cheap as however uh Nvidia wants to optimize their like GPUs to make their faster for everyone and cheaper for everyone.

So, um I think it's just a couple dollars if you want to make uh it's like a L file that you just like import and boom you got it and you're just like sort of wandering around. It's amazing. Well, I mean what does the business of runway look like at that point?

And how how are you thinking about the importance of like value acrruel in the application layer? Uh are you thinking you go more B2B and try and get big contracts? Do you want to go uh B to C and just get every consumer uh on you know some you know $10 $100 a month plan?

Uh how are you thinking about pricing and customer base? So far for now our focus is being professionals mostly making sure that like those who make a living out of telling stories can use Runway to accelerate like their work.

Um I think there's still a lot of like creators and individuals that will like start becoming professionals because of this. But I mean I've I've I now could do things that will require like an army of BFX artists. Um and there are many other people who can feel the same right now.

Um our plan is mostly to make sure that those who want to tell something can actually like tell a story. Um and that's a wide spectrum of folks.

So we work with like Lionsgate, major production studios to like independent studios, indie studios, like filmmakers themselves, just groups of like two, three people making amazing stuff. And then we have an iPhone app so consumers can also make stuff.

Um I think like the mission of a company is like look this is the most important technology for like art like I I've started working on this because I had the idea that neural networks applied to art would like eventually become a new form of art and I want to be part of that and I think the consequences of that go beyond just B2B and B2C.

It's like everyone out there should be making great stuff. Are you optimistic about any of the hardware developments kind of speeding up runway rendering or generation?

Uh obviously there's like new Nvidia chips coming, but then there's also startups that are focused on transformer models baked into silicon like etched, there's Cabbrus, there's there's Gro, not the AI XAI thing, but the the the the chip company.

Uh and then there's like the deepseek approach, which was like it seemed like they mostly just worked on algorithmic improvements to speed up inference. Um, where are you most optimistic about getting the next 10x improvement in your models and your systems? I think those are all like interesting developments.

We've been watching some of them really closely.

I think right now um cost is mostly for us at the scale of what the business is and the like the research that we do the main factor and so if you give me like X performance but it's unproven or like I will need to spend like six months trying to like debug into like much performance and whatever or I already know um that makes it really hard to switch um and so I think it's it's it's cost mostly what cost in the sense of like how cheap can I get just like my computer and for how long can you like guarantee me that that price and maybe a price better after.

Um, and then performance improvements if they're like significant will matter a lot, but I don't think we've seen something so significantly yet that will make us like either switch or move to other like hardware.

I do think it's important like I do feel like all of the chips that are being made to to compete with Nvidia are relevant, but for us it's still like very hard to to see where we can switch to be honest. Makes sense. Last question or This has been great. This has been awesome. We'd love to have you back.

This is really informative. Thanks. so much. I' I've I've I've been a fan for so long. I wanted to talk to you. I'm really glad I got time for you. Seen we make we make some pretty cinematic videos at some point. We're And I want I I I really want to do the style transfer thing on some of our older content.

You can now you can now you can just now it's animated. So yeah, we'll definitely be uh racking up those runway bills when we get to hitting render it all. I want every piece of B-roll we've ever filmed in studio Giblly style. I don't care what it costs. I want a full of it all. It's worth it. Yeah, it's it's awesome.

It's it's such a fun time and there's so much that you can do with this stuff. So, uh thanks for powering it and thanks for stopping by the show. Yeah, congratulations to Congratulations on the new model. Perfectly timed. Dr. All right.

Uh we can do some timeline and then OpenAI drops some news about a new openweight model with reasoning in the coming months and uh I'm going to use the rest. Okay.

Well, uh, while Jordy is doing that, I want to take a look at this video from Isar Aerospace, which is a company I hadn't heard of before, but they had the most cinematic rocket crash I've ever seen. Uh, from NASA Spaceflight. com, uh, drone and pad footage from ISAR Spar aerospace spectrum launch.

You can see it avoided the pad when it came down. Uh, Andrew McCall, friend of the show, says, "Insane footage quality. Sorry about the loss, but serious credit to your media team. Top tier work. I couldn't agree more. It we'll have to pull it up because it's it's the most cinematic just footage ever from this drone.

It looks like they color graded it. It looks amazing. And uh and Ashley Vance uh also chimed in and said uh and said, "This site, this launch site slaps because it was great. Uh and it's just a beautiful, beautiful uh website. " But speaking of beautiful things, go to bezel do getbbezzle. com.

Shop over 23,500 luxury watches. Fully authenticated in-house by Bezel's team of experts. Uh, get a luxury watch, folks. Um, anyway, how are we doing on the footage of Isar Aerospace Spectrum launch? I'd love to have the founder on the show. We're we're we did defense week or defense day.

Uh, we're thinking about splitting that up because we kind of did manufacturing in there and then also weapons systems. I think we need to break it up further. do space launch day specifically. Maybe we'll do space satellite day specifically.

Um, but aerospace would be would be a fantastic company to include in the uh space day if we do it because I want to talk to these folks and I want to see what cameras are they using to film this stuff because it looks great. And we have the video here. So check it out folks. Isar space.

I think you got to pan down more, Ben. Oh, here we Oh, we're good now. So, look at this. This is so beautiful. I can't get enough of this. Look at this. The rocket takes off. Very sad that it didn't make it to orbit, but very cool video. And look at this. The drone just going higher and higher.

This is why you got to watch the video, folks. Sorry for everyone on the RSS feed. Um, but uh the rocket's going up up up and then starts flipping over and winds up crashing. I actually don't know where this was. It looks like it's in Antarctica or something, but I think it might be in Europe. I'm not sure.

Um, but beautiful glaciers. They got to get Harry Stebings over to the launch site. Have him figure out what's going on. But, uh, this this post uh came. Look at that. Wow. Big like like look look at the teal and orange. It's just the classic Hollywood color grade. Looks like a Transformer movie. Uh, fantastic.

And so if you're doing something cool, like the the takeaway is just like spend the money on the nice camera because you're going to get a lot of attention and attention is all you need. Uh rad backwards. This one it it was unrelated but it felt related. He says uh caring is cool.

You should care a lot and work really hard. That's the coolest stuff ever. And I saw those folks blowing up that rocket and I was like you know what it didn't go well but they care. They care and I love that. Try again. And uh should we go to Gokul Roger? He says AI shrinks the skill premium.

A year ago, a portfolio co would ask me every week on tips to hire senior engineers. They found it really hard to attract season builders. For the past couple months, they have stayed conspicuously silent on this topic. Today, I asked them why. Had they finally found their dream senior engineer hire.

They said, "We don't need senior engineers. Our junior and mid-level engineers are able to use AI to solve every logic and system design problem. " This is what we were talking to uh Patty about. Um chat GPT is his colonel and uh cursor is his army or something like that. I like that model.

But really, it's like, yeah, uh, you know, you have these tools and systems that plug in that help speed up the autocomplete and the implementation of your ideas, but how do you how do you reality check and and, uh, it's the rubber duck test.

Can you explain what you're building, what your architecture is, and the these reasoning models are fantastic at that. So, goes on and says, "This will be one of the pervasive effect uh pervasive effects of AI. AI will massively erode the premium charged by senior professionals in any white collar role.

You might still need the 0. 01% of talent to build worldscale systems, but the top 1 to 10% are going to see massive competition from the middle 50% as AI uplevels the latter skill group. I don't know if you had to take Jordy, but I thought it was interesting. Yeah, I mean uh it's interesting.

Last night I had a question that I wanted to ask a friend of mine who is a a uh very much an expert in the subject and I felt like and and I I reached out to him. I said, "Hey, can you chat about this thing? " But I also asked the same question to a language model. Yeah.

And like hours before my buddy responded, I got a like satisfactory or potentially even better answer than my friend would have given me. Yep. And I think that that scenario is basically similar to what what goal is is describing here. Yeah. Um which is good.

It's sort of I mean again it's sort of leveling the playing field in many ways. Uh the internet as a technology leveled the playing field globally where like you didn't have to be in the United States to compete with US workers for jobs.

You could be a web developer in UK or India or whatever competing for work in the US. So yeah, I want to go to this post by Dylan Patel. Uh he says, "Data Center capex is going to zero. Grock struggling deploying anthropic extremely low limits for paid plus cursor. Google not offering 2. 5 on API.

Open AI not serving 03 struggling to serve 40 image. Azure still not serving many models including 01 03 deep research 4. 5. " Uh and this is so funny because he's the most respected analyst in uh in semiconductors. Uh Dylan Patel obviously the the founder of semi analysis.

Um but the first line is a joke and if you don't have a good irony detector uh you would just read this completely wrong.

And uh and fortunately Brad Gerson chimes in and says, "Bizarre how the GPU doomers all somehow understand true demand better than the companies who all have massive computing down their GPUs and scrambling to find supply.

" And so I mean, yeah, if you look at the uh how long it takes to inference some of these models, it's all very slow. It does seem like demand is very high.

Just anecdotally, I think we saw that everyone is using all of these models all the time and they are intense from a GPU demand perspective and you can just Gary Gary Basin actually has a good point here saying they're looking at two different time scales.

The short-term squeeze is priced in but like that capacity for the short-term squeeze is in many ways being coming online or being built out. Yeah. There's also this this interesting dynamic where some of the founder mode companies do have an incentive to overstate the the the level to which they are overstrained.

You know, it's like the classic steed stage founders like I can't sleep. I my phone's ringing constantly. I'm the most popular startup. We're growing too fast. Like you kind of know that if you signal that to the market, you're going to get more term sheets and stuff.

Um, but I don't believe Google is in that mindset with like, oh yeah, like if we don't offer 2. 5 on API, like the market's going to see that and price us higher. It would be very odd if all of these dynamics were uh fake.

It would be it would be reasonable if it like just one was, oh, that person's kind of overstating or they're a challenger and so they're saying, oh, we're so popular.

Um but when they're all struggling clearly there's there would be an economic incentive for one of them to break rank and say actually we have we have plenty of GPUs so come over to us we would love to have your business when they're all struggling it's definitely a sign that data center capex is not going to zero in fact uh and speaking of uh 40 uh images and chatgbt uh there was an interesting post uh by technium saying 40 image generation can do calculations during its image generation somehow This was community noted because uh the first LLM prompt so it it's a picture of a calculator and the prompt is uh make an image of a calculator app for the calculation 53* 88 and the result of that calculation is 4,664.

Uh and this would be shocking if the LLM had just been able to or the the the the image model had just been able to do this at image generation time.

But in fact it appears that in the prompt reasoning step before the image is generated the uh there is an internal uh prompt interpol interpretation that runs that that basically hydrates the prompt and adds extra context.

And so um someone else shared that and said that when you if you give the exact same prompt in the screenshot, you can see that they pre-process the prompt and that that prompt actually uh does the math which is still impressive that it can get it right. Uh let's not be wrong about that.

Um but it is it is interesting and this is a this is obviously just like a huge unlock and and obviously pairing LLMs with image generation.

This is the future and this is I think why um this new image generation the studio giblificication of chatpt it didn't happen in dolli 3 or dolly 4 like a separate app it just happened in chatbt and it takes advantage of all the other chatbt features and eventually you're going to see consolidation so you can upload a PDF have it search the internet do some deep research generate an image with images in chatp it's all going to happen in the same app and same window, which is great.

Um, but but interesting like kind of technical deep dive here. Um, is there anything else you want to run through on the timeline because we are past 2 p. m. We've been streaming for three hours.

I wanted to take an opportunity to talk about another new partner which some people have called out in the ticker which is Poly Market. That's right. And uh we're very uh excited to be partnering with Shane and Matt overarket. We've been very happy to have them as a partner. Thank you.

Um, and the reason that we were so excited to partner with them is that the Poly Market platform is news, right? Like it is a way to understand the world and understand news. And um, so it's a natural fit to integrate into the ticker.

It's like everything that's happening in tech, like here's how to get an understanding of it. Uh we're going to be sort of updating the ticker, making it um yeah, we're going to be improving it basically like improving it uh day by day like we do. Uh and we're also going to be creating some of our own models. Yep.

Um and they already just have like uh one of my favorites that's going right now is AI wins IMO gold medal in 2025 at 61%. That's really high. Uh which is really high. And what that would mean is that everybody basically gets the average ramp employee. Yep. or a average cognition employee on their team.

Um, and so yeah, there there's a bunch of other stuff. Some of this stuff feels like um I probably can't say this, so this is not an ad, but it's just commentary. Some of the stuff feels like uh free money uh just because it's like, you know, you look at the probabilities on some of this stuff.

Um but um and I think it'll be cool because Poly Market's been known for political prediction markets, but we never talk about politics. And genuinely, this is not a political show. Uh, and so I'm just excited to build markets around tech predictions. I love, you know, will the next Starship make orbit?

How many Starship launches will there be in 2025? Like, these are the markets that I've always cared more about and I'm excited to highlight on this show and kind of there are plenty of other shows that are partnered with uh Poly Market that can highlight the important political markets.

Uh, I want to drive the tech markets forward for sure. So that's what I'm excited about. Yeah, there's so much other stuff. I mean, which company has the best AI model by end of April? Uh, Google's sitting at 51%, which again is not what I think the average person on X would even tell you. Dangerous.

It's forbidden knowledge. It's forbidden knowledge. But look at this. I mean, Sundar Pachai posts 2. 5 Pro ship today to everyone. Find it in the drop down here. Gemini. google. com. Still he puts up 5k likes, a million views. built different. Built different. He does it. He's so confident. He throws the link in there.

He's got the hilarious uh image. I'm I'm grow. Sundar's growing on me, man. I I like the I like that he's going direct. I own links. Yeah, I am the internet. I'll post this. Whatever. It's also hilarious that it's in this drop down on a subdomain.

Like, you're taking it seriously, Sundar, but like how seriously are you taking it? Uh should this just be baked into the Google search box? Yeah. Right. Um but hopefully soon and I'm excited. I mean it seems like they're they are hearing the message we need better productization.

Google legendary team in terms of AI algorithm development. Uh attention is all you need. That's a Google paper. They created the transformer. Time to capitalize guys. You got the TPUs. You got the team.

You got the Poly market has a poly market for Google's whiz acquisition being blocked before July sitting at a very six uh only six% chance. But and that's an interesting one because you normally if when a big acquisition is going on you can actually determine you can extract that market from the stock price movement.

So like you can see how Activision is trading and understand okay the market expects that Activision will be acquired. This is what happened with the Twitter acquisition.

Like it went right to 46 billion or 44 billion or whatever Elon was going to pay and then it would it would drop by a little bit based on the probability that it would not go through.

But Whiz is a private company so you don't have that you don't have that real time data on how the share price is trading which you can use to address whether or not the deal is going to go through. Poly market comes in there. Very cool. We got it. We're also sitting at a 25% chance that X relaunches Vine before July.

Do you think they need to do that? I feel like they have the video feed. They have their they have their secondary apps are rough, you know, to stuff it all in one app. And they're they're very good at getting you to open up one video, it autoplays, then it scrolls you to the next one.

Pretty soon it's just all slop, but you're watching it anyway. Those unreged user seconds are dropping to zero. Uh but uh we'll see. I mean, I I' I'd love a Vine comeback. uh the Vine team, one of the just legendary crew of uh entrepreneurs. Yeah.

The other thing that we're going to be paying attention to as April starts is just what's happening with Tik Tok. Oh yeah. Poly market says today there's a 30% chance that the Tik Tok acquisition is announced before May. Wow. That's pretty high. Yeah.

I mean it's high, but that's also what everybody's been orienting around this early April sort of deadline to get a deal done. So we'll see. Um and there we go. Oracle Walmart. Ben Thompson says says Walmart should buy them. People think there's only a 1% chance that Mr. Beast is the winning bidder.

So, no faith in the beast man. No faith in Jimmy. Oh well. Uh well, that's a great show for you guys today, folks. Thanks for watching. Leave us five stars. Apple podcast, Spotify. Stay tuned. Repost all of our clips. Like and fave everything. Do it. Do everything. Uh thanks a lot. Uh last shout out.

Yeah, Chris Amdon, we saw you got an aid sleep. We're competing with you now. Yep. I want to see you put out those numbers. I want to see those scores, Chris. Let's see it. That's it. Thank you, folks. We will see you tomorrow.