Bernie Sanders sent a formal letter to the CEOs of Anthropic, Meta, and Open Eye. AI capabilities have reached a critical threshold. Pause AI development. The cat's out the bag. It's too late, right? Nvidia just announced a partnership that redefines what GPU compute means as a financial asset. This is like the very first pitch of the first inning of the buildout of the Dyson swarm. Financial assets want predictable depreciation and exponential technologies don't give you predictable depreciation. I'm not as concerned that this will end up being another mortgage-backed securities fiasco for a few reasons. One is Elon just dropped Grok 4.6 and he's right. It's a banger. Grok 4.5 came out 2 weeks ago, 4.6 just this morning and 4.7 is rumored to be coming out in 2 weeks. 4.7, and this is what I think is going to be really interesting, is going to be trained on all the SpaceX physics and engineering knowledge. You can't get beyond Frontier with this stuff, but what if you have the best engineers? Now, that's a moonshot, ladies and gentlemen.
Welcome back to Moonshots, everybody. We're going to be covering nine stories today that span the frontier from longevity, AI infrastructure, synthetic biology, AI filmmaking, urban air mobility, and oh by the way, Grok 4.6 is crushing it. The through line is the same as always. You know, accelerating singularity is compressing the distance between the impossible and the inevitable. We're back this week with the moonshot quintet, AWG, our in-house super intelligence. Alex, welcome. Good to see you in your normal apartment.
Good to be super intelligent.
You are my friend Dave Blondon, our investor and AI extraordinaire.
I was over at MIT all day today, hence the garb, but our Techre teams just got back from SF and they brought back huge amounts of knowledge and they're all over at MIT sale dispersing it across the campus right now.
Love it. And Salem Ismael, our empressario of exponential organizations. See, where are you today? Still in Toronto?
I'm still in Toronto heading back tomorrow.
I should be going back now, but this podcast happened and I can't fly, you know, within like six hours of.
We're going to put a Starlink on your hat and have you walk around with it. And back by popular demand, Emad Mostaque, the embodiment of the intelligent internet. Emad, I hope you're reading the comments on the last pod that we did together. People are loving you. Have you seen them?
Yeah. I did see some of them. I don't normally read the comments. I made an exception and it's very kind what everyone said. It's great being amongst you guys.
It is. It is. It works for my American wife. It works for the others. You know, I'm Peter Diamandis, your host, your optimism amplifier. Please remember that having an optimistic and abundance mindset is a choice. And our mission here is to deliver data-driven optimism that helps you make that choice. So guys, last night I was in Utah, the University of Utah. It was the finals for the $101 million longevity or shall I say Healthspan XPRIZE. Super psyched. We awarded a million dollars to 10 teams, recognized another 10 finalists. This is on the march to winning the $80 million grand prize. We've given away 20 million so far. And the mission of these teams and you know everybody please get excited about this is to add 20 healthy years on your life. Asking teams to reverse your functional losses that we get through aging in cognition give you the ability to think and have memory like you did 20 years ago in muscle the ability to build muscle like you did 20 years ago and in your immune system. And it's extraordinary. We had 800 teams enter this competition. every possible approach from mitochondria to you know stem cells to you know gene editing it's extraordinary so this is a competition that will be won by 2030 so keep an eye on this you know Salem I don't know if you want to jump in on this one.
you 20 20 is infinity right.
I mean you know I think there's the XPRIZE was such a huge inspiration when I was writing the exo book because you're reaching outside and I mean 800 teams is that I think that's a record for the traditional prizes for us, right?
It's kind of an incredible thing.
Elon's $100 million prize for carbon removal ended up with something like 1500 teams. But, you know, reversing aging, it's got to be tougher. And by the way, the teams here don't do this in theory. They don't do this in mice. They have to actually do human trials. So, they're all going to be doing trials with control groups and humans, probably around 150 people in the trial. So, it's real data and we're going to actually know which of these 20 to 30 approaches that actually make it to the finals work. And for me, you know, Alex, we've been talking about being in the midst of longevity escape velocity. This is accelerating it. It's going to be spiky. I think I still think just as with AGI either in our rearview mirror as I think or some might say it's either here or almost here. I do still expect that health span and longevity escape velocity it's going to be spiky and I'm just optimistic that we can even out the spikes even if there's a sub population that achieves health span on the margin a few years before everyone else. You know, one of the most important things I think about this competition when we launched the $10 million Ansari XPRIZE for spaceflight back in 1996, you know, back then people did not believe in commercial space flight. They didn't believe that individual teams could do this and carry humans compared to the government. And as it progressed the confidence level in this and then when it was won money flowed in regulations changed you know Bezos and Musk started you know Blue Origin and SpaceX and it changed the game. So I'm feeling the same thing going on right now. You know longevity the idea of reversing aging has been sort of a crackpot idea for most of the last few decades while I've been in the industry and it's beginning to change. People are starting to believe yes, it's going to happen. Yes, we're in this health span revolution.
We're getting to a when, not an if question, right? And that's really huge. I wanted to point out something that was that's really important here, Peter. When you launched the Ansari XPRIZE, there was no space industry to speak of, at least in the commercial side, and now we have a trillion dollar industry that did not exist, right? Because the collective innovation plus all the members of all the teams end up going to SpaceX, Blue Origin, etc., etc. And we could expect the same thing to happen here where you end up with essentially you're creating a $101 million bet. You're building a portfolio of experiments and allocating capital to wherever there's demonstrated results. It's an unbelievable model that we've now seen repeat over and over again. It's incredibly exciting to see.
Question for you Peter on this one. I always ask the counterfactual question. Is there something having now run this health span prize or at least the beginnings of it? Do you think that if this prize had been created say 20 or 30 years ago that we could have made on margin enormous progress or do you think there's some historic contingency that means right now is really the first time in history where we could make credible progress on it?
Yeah, I think we had a lot of comments. We had a lot of the top scientists, Aubrey de Grey was there last night who coined the term longevity escape velocity and then Ray popularized it. I think everybody was at the consensus that the timing is perfect that the tools for genome sequencing for you know making specific molecules for being able to measure and report with AI are the tools that are required today. I mean there might have been some approaches that could have been done 20 years ago but I think today is when we're going to make the greatest progress and I'm starting to see you know capital flowing in aggressively at the end of the day longevity is going to be the biggest market right if you could add 30 you know I've had this conversation you know Salem probably you have as well on stage you know with YPO audiences or family offices and I say how much of your wealth would you spend for an extra 30 years of life. The honest answer is nearly all of it. Right.
Yeah. I mean you and the powerful distinction there's not so much lifespan but the health span effects are really really powerful.
Yeah. The numbers today are for the United States if you look at it basically 100 years ago in 1900 the average life expectancy was 47. Today it's 79. We added about two months per year over the last century. And today, while the lifespan is 79, you're healthy until average age 63. And you spend the last 16 years of your life in poor health. And so when I was talking to Katios about this, I said, listen, the biggest benefit the US budget could have and the US economy could have is add 20 healthy years in people's lives. they're retiring later. They're not spending as much government money on sick care. So it could be a huge transformation. Alex.
Yeah. I maybe just again coming back to this historic counterfactual. I one of the things that irks me the most is that so many of the abundance-oriented futures that we want to find ourselves in like LEV the super intelligence solve everything just take forever. And I do wonder again we're is it 20 years post Yamanaka that like Yamanaka I think was 2006 we're in 2026. What did we blow these 20 years on? Why couldn't we have done this 10, 20, 30 years ago?
I got a question for you, Alex.
That I think when you get multiple exponential technologies that can address a particular space, then that's the point to invest or put up a prize because then you get radical outcomes at very low cost. And so maybe there's a benchmark of the minute some domain has two or three or more exponential technologies converging on it, that's the point to have put dollars into something like this.
Dave.
Well, there's no doubt in my mind that I'm going to do more productive work in the second half of 2026 than in my entire life combined up till 2026. So throw it back at you, Alex. Like even if we had worked really hard on this 20 years ago, would anything that we did between then and today even hold a candle to what we'll achieve between now and the end of the year? Cuz I mean it's so funny to me, you know, my daughter's over at MIT. She's a biochemical engineer and she listens to the pod of course you know and it's so obvious to her that we're in an AGI hard takeoff now and the people she works with in biotech are about 1% AI aware and 99% not aware but the AI aware now people are spending over half their day maybe 80 90% of their day talking to AI agents and not in meetings and not you know running gels and not running assays because it's just a different mode of living that's accelerating tremendously, but the aware subset is tiny.
I almost think I think it's an important point and I almost think we need a new term for it. I just thinking off the cuff maybe like retrospective hyperdelation, this very singularity-oriented idea that with super intelligence you discover that everything that you spent the past decades on was just a total waste and you should have instead just done nothing, twiddled your thumb for decades, waited for super intelligence to solve it for you.
Or just work on, you know, chip fabs or something that will, you know, they'll be very useful on that day or in Texas or.
All these people like who spent six-year PhDs trying to divine protein structures just mostly wasted.
Well, that one was really outed by Demis. Yeah, I mean that one what it was like an average of four years to discover one fold.
I knew there was a reason I didn't do a graduate degree. That was it. It would have been irrelevant. This was your post justification.
I mean it could be worse. You could be a pure mathematician, right? Like.
They're all cooked too. They're having this the same moment of on week.
Yeah. But at least with the biologists, the longevity people, you can do assays and things physically. No, look, I think that this is the biggest market in the world as you said, but it's the first time that it's tractable. I think if you go back to when there were the Yamanaka factors, all these other things, you didn't have the infrastructure necessary and the talent pool necessary. I think that as you've seen the various breakthroughs in other things and everything come together right now there will be a shortage of people that can really work on longevity properly even though it is the biggest market in the world and 10 years ago 20 years ago that would have been even tinier there are only a few people actually looking at some of these things back at that time so I think it is this confluence factor all coming together and like I said it's finally tractable so the people who are winning in the longevity XPRIZE any indication that those things will work and I don't think they'll be short of capital but again the XPRIZE is the catalyst right and that was the whole idea.
You know Peter you've pointed out something that I think is incredibly important to highlight right which is that we spend today huge amounts of money on treating chronic diseases at the end of.
You know what the number is Salem it's the global cost of age related disease is 20 trillion per year globally.
Daniel Craft used like 85% of healthcare costs for the last 5 years of your life type thing. This is a staggering number, right? Because now.
Just for context, the global economy is 120 130 trillion total. 20 trillion of that is.
Look at that number, right? So now the business model becomes how do you maintain decent function, bodily function before the disease appears and it'll completely change healthcare economics. Massive impact.
And global economics in general.
In terms of productivity and wasted capital. Well, I think this is where the peptides have come really interesting, right? Like it was like to lose weight, you had to work out. I'm about to start, you know, get rid of that extra report in on the pod.
I'm going to get it report in and see how my weight loss goes. But the thing is.
Which peptides are you taking? Do tell.
I was going to try the one.
But I know I'm going to lose weight, right? Are you getting access to No, but serious. Okay. Has it even Oh, you're in Canada.
No, no, I'm not in Canada. I get it in the US. It's totally.
Yeah, they're drug dealers out there. Black market.
Drug peptide dealers, right? There we go.
I guess Salem, that'll just be our little secret. There's no one paying attention to.
Here. Nobody's listening.
But this is how you know that you're going to lose weight with it, right? And that's the first. And so now the concept of you can take a pill or an intervention and you could live longer. Well, you're already losing weight like that. And so I think again that's a big awareness that's caused the longevity market to get even bigger.
Yeah. GLP1s are the first longevity drug you know in a lot of people's opinions and it's one of the biggest grossing drugs if not the largest one in human history. I've gone even further than that just again bit of back of the envelope calculations suggesting and again this is not medical advice that GLP1s especially third and maybe fourth generation GLP1s may actually be when I refer to LEV longevity escape velocity as being potentially spiky a pretty big spike there there have been studies done recently on some of the third generation if memory serves GLP1s that suggest and again not medical advice that we may be or at least some subop of humans that have undergone GLP-1 studies may be at something like 70% LEV just with GLP-1 therapy. So if that is the case and again encourage folks to do their own independent back of the envelope analysis that's a heck of an LEV spike in a subop that's being administered GLP1s.
Yeah. Anyway, watch this space everybody. It's exciting.
Peter, I've got one more question for you. Please as you looked at the finalists cuz I couldn't make it to the finals but we've been tracking some of the teams etc. You had a great snapshot view are you in the view that LEV we get to it by 2030 or earlier.
So here's the idea right if any of these teams win and can reverse your functional age right this is not a number from a you know a particular blood test you take you know that changes on the back of a form this is are you feeling functionally younger are you know do you have better muscle building capability better memory better immune system so what really matters is your function and so if we can do that if we can actually reverse the clock by 20 years of function, then you get to enjoy the next 20 years of breakthroughs. And if you don't believe we're going to have incredible breakthroughs over the next 20 years, you know, basically, you know, wholesale simulators and AI, you know, the impact of quantum, whatever that might be in cell biology and understanding how we age, then you're missing the boat. Your goal is to keep in the best health right now, which means what? Sleep eight hours. Unless you're Emad and Alex probably don't sleep more than four hours, but you're short sleepers.
But you're short sleepers. Sleep as much as you need to.
So, I'll make a comment here which I've made before, which is that.
You know, the business model of religion is to sell heaven. As we have life extension coming, how are you going to sell heaven if people aren't dying? So, this breakthrough will mean that religion is cooked. We have forever to figure it out. Like.
There will still be a lot of religions doing very well from people tithing. But again, I guess the advice right now is do what you can to keep in the best health. Don't die from something stupid. It's sleep, exercise, you know, proper diet and mindset. Mindset so important. You know, I think my greatest attribute is my longevity mindset. So take it on. We'll keep on reporting in the space. It's an important part. This episode is sponsored by Google for startups. Think about this for a second. You now have access to the same generative AI models that cost hundreds of millions of dollars to train. Google's startup tactical guide for generative media gives you complete blueprint for deploying Google DeepMind's models in production. Images, video, audio, all of it. Real architecture, real results. Find the link in the show notes below.
But let's move us on. These were a couple of stories Emad that you shared with me. These are three stories that landed this week together that describe really I think the collapse of Hollywood economics and the rise of something much bigger. So our first story is a company called Higgsfield that just made a movie called Cully Hill Boys. It's 110-minute feature film. The first full-length AI generated movie with licensed celebrity likenesses. Here are the numbers. The total budget for this film roughly you know full-length film $2 million team size 28 people production time four weeks compute cost $1 million they use Sora 2.5 as the generation and they open sourced all 10 steps in their workflow so anyone can replicate it so here's the context right a feature film with celebrity talent today costs between 20 million and 100 million if you're not using AI and it takes you know, a year to 18 months. Higgsfield did it for 2% of the cost and 6% of the time. So, I want to share a short video, a little clip from Higgsfield so we can appreciate it.
All right. Our second story, Bloomberg reported that nine out of the top 10 text-to-video models on AI analysis leaderboard are coming from China. This is a headline, but the real story is deeper and I'm curious what you guys think about this. You know, Hollywood forever has been the dominant producers of films and it's basically exporting US culture to the rest of the world. What happens when the world is flooded by Chinese-produced English-speaking films? How is that swaying public interests and public points of view? And the second thing is that these Chinese models that are dominating video generation are also learning physics, motion, object permanence and causality which is what's needed for robotics and autonomous driving. So things are moving quickly there. The third one very quickly then we'll go into the discussion here. a model called LTX 2.5 is the newest version of the most popular open source state-of-the-art video generation model and it works on your MacBook Pro which is extraordinary. So I want you to imagine something that allows the individual to sort of tee up and produce their own movie. One more clip. This is from LTX 2.5.
So gentlemen, let's go to you first, Emad. This has been your world for the better part of a decade.
Yeah, it's happening right on forecast realtime high-definition video. So LTX can generate a 10-second clip in 7 seconds at that quality that you just saw, which is indistinguishable. And a few years ago when we had the state of the art model, it was not like that. It was like slow moving, you know, like Will Smith eating spaghetti was awful and horrible.
Now you can do a spaghetti eating Will Smith, right? And it'll just do it instantly on your local kind of laptop. So you said kind of the disruption of Hollywood like you never need to reshoot a scene now with the flows that you have. People are licensing their things and as you said Higgsfield released an 80-page guide because they're a company that allows you to make movies on exactly how they made the movie like it's almost open sourced. Higgsfield got from it's an ex Snap leader from Snapchat. It's 700 million revenue run rate right now in about one and a half years to give the idea of kind of how much uptake. But that's just the start.
Yeah, cuz every pixel will be rendered. Like if you see reactor who are using LTX now, like again it could be that Alex might be a simulation already, but for the rest of us, you know, we could put our avatars live in the next month or two at a level that you won't be able to tell.
Walk us through the one the $1 million budget to make a one and a half compute budget to make a one and a half hour long movie. What will that be by September 25th by the time we have the Moonshots live event?
So it depends on the level of quality. Sora 2.5 is the best model. It's a large big model that costs about $3 per 30 seconds roughly, but you need lots of shots and it can have 50 inputs of audio and video and things. The cheaper models are 10 20 times cheaper, but like 90% of the quality. So, it all depends on exactly how you're shooting and the type of shots cuz Sora can do 30 second clips, but the average movie on the Hollywood box office right now is 3 seconds per shot. Wow.
So you could have a 10 times reduction in cost. You could shoot a movie just like that one probably for a 100,000 of compute and then 10,000 of compute.
Probably by the new year.
This is what we saw in our you know foundations of AI ventures class at MIT is that it went from PowerPoint demos to full functioning products in one year. Like to be competitive on demo day now you have to actually build the entire product. But I suspect the movie like our target for September 25th is bring a script.
Well, the So Dave's referring to the Future Vision XPRIZE that we're going to be awarding with all the mates here at Moonshots Live and you go to moonshots.com to learn more. We have 5,000 entries PE 5,000 who are creating three minute three-minute trailers and a film treatment. I just had a meeting this morning with the team at Range Media and Google to down select and we'll down select to ultimately the top 100, 50, 25, 10. The top five will be at Moonshots Live on the 25th. And everyone listening, we're going to you can come and be there live and vote live. We're going to also have a live stream of the event. But yeah, it's and we're going to do this year on year. So every year it's going to get better, cheaper for sure. I think this is going to be much bigger than we originally envisioned. I mean, I know it's a big vision to start with, but we were thinking, okay, then it'll be a $20 million, you know, year-long endeavor to turn it into a real feature-length film, but in reality, you're going to unleash the creativity of 5,000 people who almost all can make their movie within a year.
100%. We have a $10 million budget to make the winner's film between the money we're awarding and foreign film rights. But I was just talking with Range. Well, we're going to be having five finalists who will crown the winner and you should see the trophy. It's beautiful. They want to make all five. So, we want to create an engine here. Salem, what do you think the implications of this are? And then, Alex, I'd love your thoughts as well.
Well, you know, there's I'll go at it from a exo perspective. when you we actually talked about Hollywood as one of the first domains that went into an exponential organizations model because what happened when you broke up the Hollywood studios in the 50s and 60s is Hollywood turned into a cluster of external resources where you'd produce movie production would start and essentially a swarm of people would appear grips and camera people and actors and editors and whatever and they'd get together for that project and they completely disband after that, right? And we see the beginnings of that in Silicon Valley now. And so that was that first wave of Hollywood becoming an exponential organization. Everything was assets on demand. Everything was staff on demand. But now when you have everything being driven by AI goes through the organizational singularity and essentially it's compute cost now moving closer and closer to what Alex always talks about. You have domain collapse now coming along and so that's going to completely change the game again for this. So you've gone through two big waves in Hollywood, the second one just starting now.
Alex, what are you excited about here? What are your thoughts?
So I made myself watch The Cully Hill Boys and I must say just as a preliminary.
I skimmed through the whole thing watching a good chunk of it.
See, when you only sleep two hours a night, you can just do that. Yeah. So I had to watch this. So it's not my favorite genre. I would characterize Cully Hill Boys as sort of British Bollywood. Not quite even sure what genre this is. It seems to be, as far as I can tell, about the hijinks of a bunch of British rappers that get into all sorts of trouble. Content-wise, not super interesting to me, but at the functional level, it is really interesting. And there were multiple times in the movie where I had to wonder were the generative actors, so the likenesses were based on real humans, but were the scenes that were being generated given the complexity of the interpersonal dynamics such as they were in this movie, were was there some sort of emergent theory of mind that was almost necessary in order to generate some of these scenes with people interacting with each other? and especially generative violence. There were multiple times, you know, people with knives chopping things, chopping meat, threatening other people where I had to wonder is at some level we talk about AI personhood on this pod from time to time is there in some sense at some presumably intermediate layer in a diffusion transformer somewhere deep in the bowels of Higgsfield, is there some diffusion transformer or similar model that felt threatened at some point in terms of these generative violence scenes. So that was my take on the content of the movie. Moving to.
You were worried that the residual AI might have been threatened in the making of the movie.
I'm talking my head about that.
No, no natural persons harmed obviously in the generation of this, but I do wonder.
We're going to have a disclaimer on content. No AI was harmed in the making of this or.
Traumatized of this or traumatized. I mean.
So I do worry parathetically about that. The bigger question on the economics of this, I think, is at what point do generative video capabilities start to reconverge with the Anthropic school, which I'd characterize as a token revenue maxing. Right now, it seems pretty clear that if you're using Sora models, it doesn't matter how many millions of dollars of XPRIZE prize money are going to shower down on folks who can create spiffy videos about the future. That's nowhere close to the amount of money that one can earn in principle with revenue per token maxing. And right now, these appear to be two separate lines of effort. On the one hand, we have a vibrant largely Chinese dominated at the training side consumer economy for generating consumer videos. And on the other hand, we have enterprise revenue per token unit value maxing that seems to be largely going to codegen and enterprise problems. And right now, these are largely two distinguishable ways to burn tokens or diffusion transformer equivalents of tokens, flops. Two different ways to burn flops. One of them maximizes revenue, one of them maybe maximizes consumer engagement and wow factor, but they're nonetheless separate. I don't think they're likely to remain separate that much longer. And the reason is so I use Fable every day and I use its competitors every day. And I will say the strongest models, the models that are strongest at revenue per token value maxing are just still terribly weak at modeling the visual dynamics of the world. I don't want to call it physics because it's not physics, although a lot of people call it physics. It's at best classical mechanics. But like the intuitive the physical intuition from general purpose video generation requires that you at least have some embodied intuition. And right now, Fable and its peers are just incredibly weak at that. And I think in order to ultimately revenue max per unit token, it's going to require that these Fable-esque models have just amazing visual intuition as well. And we'll finally see a convergence or reconvergence of these two lines.
Yeah. One of my hot takes on this is, you know, are we going to see the primary actors out there, the Matt Damon's or the you know, Leonardo DiCaprio licensing their likeness? And I think not, but there are so many lookalikes out there. So, you know, the producer is going to go and say, "No, that's not Matt Damon. That's John Smith and he looks like Matt Damon and we licensed him and he's in this movie." Right? And I think that's a it's going to happen. That's going to be the workaround on getting, you know, the actors you know and love into this and there's nothing they can do about it. There's an alternative which is that we're already seeing which is dead actors. So dead actors who can't record any new movies. Their estates are highly incentivized to license away their likeness for this purpose. So I think dead actors like maybe we'll see an equivalent of SAG pop up just for dead actors. And they'll be the most profitable actors in Hollywood. Dead actors. I think that's likely to happen. How much do you think people will if you can make like a near Matt Damon, you know, very similar character but obviously not him or you have the actual Matt Damon, how much will people care in terms of Baka?
I don't think they're going to care. They, you know, cuz you don't care what the person's named in the movie. You just like that actor. That actor brings you good feelings from previous, you know, engagements.
Yeah. I think that you also seeing the rise of AI stars now as well. They're winning deals. And I have actually had this discussion with various film stars where they've been like that SAG-AFTRA deal has massive holes in it. 80% of me plus 20% of my character can be licensed by the studio. You know, like where does the person stop and the character start because obviously they have the character rights and things like that. The other thing I'd like to say actually one thing I found very interesting is as we were doing some of the more interesting frontier work one of the things we found very useful is to get the AI models to generate images and visualize what they're doing using something like GPT image.
What does that mean? So if you're doing like Salem is doing an organizational paper for example on exponential organizations, it's doing text text text and then you tell it to generate a visual of everything that it's done and analyze it in any way that it wants and it almost moves it to another frame of reference.
Cuz it's pulling in from this visual cortex kind of thing. And then you could tell it to expand and collapse it and you get these really weird images sometimes. But you can see actually it's exploring different parts. So I think we've seen that actually work for some very interesting things and I think it fits with what Alex said as all these models come together to create value as it were.
I'm excited about interactive movies, right? If you can generate faster than you can view it, then you can have a movie that's actually measuring your emotions and changing as you're viewing it, which I find.
Peter is what we're seeing. I mean they're popularly branded world models even though they're really just interactive videogen models. That is what we're seeing.
Yeah. So that's what the LTX model is right now.
So that was the first model that could do it at high definition. And then when that's combined with frame generation on the latest graphics cards, what you've just described like before it was world models playing like blocky video games. Now you can have interactive Elden Ring or whatever you want as of like this week.
Yeah. And you said something a few podcasts ago that we're going to end up with a world or Frontier model running on a MacBook Air. I mean this is a very specific use case. Are we on track for that like cuz we've got this thing running on a MacBook Air. Well, yeah. So again, it was it's very slow to generate, you know, like a few tokens a second. But now you are getting to the point where frontier level models are coming here. But video frontier models are like 20 billion parameters yet they understand all this. They can generate in 2K. Language and the code models are obviously a lot bigger. Although you're about to have the new Quen dropping in a couple of days, but I think it's all going in one direction because we're optimizing the heck out of these things. And ultimately what a model is is it's an input data distribution that gets compiled into weights and the data going into these models is getting better and better and better. Awesome.
We had the whole State Street executive team here yesterday here in the studio and we took them through the holiday and word to the wise the holiday you know Ember who is the AI just starts talking to you says you can build any movie any song any code what do you want to do and it's way too open-ended and then the answer you get back is I don't know a hip-hop song with no words like okay so you need to actually create the virtual environment the movie scene and draw the user in and then have them guide the movie in the direction they want to go. But just having it like, you know, autogenerate off your thoughts is just too free form. People just don't know how to even.
Maybe just a closing thought on this one. Peter, I do think the recent launch of Opus 5 is a step in the right direction to seeing a convergence between call it frontier models on the one hand and videogen or world models on the other because it I think we talked on the pod a bit about how Opus 5 seemed almost mildly benchmaxed towards front-end development and the loop between visuals and code. I think I interpret and I construe that as the early signs that Anthropic probably other labs as well are feeling economic pressure to produce models that do an absolutely amazing job of visually reflecting on their own chain of thought that when they produce say a website that they then do a visual analysis of their website that feeds back into their chain of thought and they do a multimodal reasoning over their own visual outputs and that ultimately say In the next few months, as the ability to produce what used to be considered AAA level video games becomes standard fair for what people expect from the Frontier models, that will be the ultimate forcing function for say forcing Anthropic type Frontier models to have just absolutely amazing visual capabilities. Even if Anthropic can't be bothered to produce like direct videogen capabilities, even if it can indirectly reproduce Counterstrike.
All right. Yeah, that's why they had Claw of Duty. People were making Call of Duty of Duty.
Our next story is one that both Emad and AWG texted me this morning. Elon just dropped Grok 4.6 and he's right. It's a banger. XAI's latest model matches GPT-5.6 on the artificial analysis intelligence index at 61. Tying for frontier level performance at $2 and $6 per million tokens input and output. The focus this time is long-running agents. Grok 4.6 stays with complex tasks across many steps. Whether researching, coding, analyzing, or turning a broad product idea into a working first version. It self-tests and verifies its own work before moving on. It's available today in Cursor and Grok build. The model cadence is crazy, right? So 4.5, you know, Grok 4.5 came out 2 weeks ago, 4.6 just this morning, and 4.7 is rumored to be coming out in two weeks. Incredible gentlemen. Alex, over to you first.
Yeah. So, I want to give Elon applause and I want to at the same time.
You've been merciless on XAI for the last few pods.
I wouldn't characterize myself as merciless. I think my job here is okay. My job is to call balls and strikes as I see them without favor or prejudice. That's how I see it. So I view and again I lack insider information, but I view Grok 4.6 as essentially the next version of Cursor. So XAI, SpaceXI, and Elon have been quite public about how 4.6 leaned heavily on post-training thanks to the Cursor acquisition, which I think is still in the process as we're recording this, of being completed. But history rhymes quite a bit. We've spoken in recent pods about what the Chinese Frontier Labs are doing and how they're allegedly distilling on mass reasoning traces from Claude and other Western models. And in some sense, I again this is an outsider's perspective. I view SpaceX's acquisition and even prior to the consummation, the final consummation of the acquisition, their licensing of all of the reasoning trace data from Cursor as essentially pulling a westernized version of what the Chinese Frontier Labs were doing, which is to say siphoning off reasoning traces from lots of people historically interacting via Cursor with Claude and Claude's competitors and then using that incredibly valuable reasoning trace data to do post-training on their models. And I think now that we're in the reasoning model era, we're a couple years in at this point. It's those reasoning traces for mid-training and post-training that are just so essential in terms of catching up to the frontier. They won't get you past the frontier. So, it's sort of like a one-trick pony in terms of nearly catching up, but it's a heck of a one-trick pony. The other thing Elon has going for him that the Chinese Frontier Labs don't is he has the compute. He has the Nvidia GPUs and he has soon his own Dyson swarm with the Nvidia GPUs that all of the Chinese frontier labs that are pulling the same trick with siphoning off allegedly Western reasoning traces in order to do their own post-training and their own distillation don't have. So the bull case for the Elon strategy is he gets the algorithmic insights to just catch up to the frontier from the Cursor reasoning traces and he gets the compute advantage that the Chinese labs don't have. I think the question is not can Grok 4.6 and its successors catch up to the frontier. Seems like they can because they have access or the near frontier to the extent they have the reasoning traces. It's can they leapfrog the frontier and achieve state-of-the-art performance.
What do you think about that? Can they.
Yeah, I think they definitely can. I think we were discussing before and Elon's come out and said it publicly now. He thinks 4.7 will go above Opus. So it'll take number one in a couple of weeks. And right now you had 1.5 trillion parameters for 4.5 that was then post-trained to 4.6. Just like Cursor originally took Kimi K2 and post-trained it with three times the amount of compute that was used to pre-train Kimi K2 to almost top level coding performance. The next model is going from 1.5 to 2 trillion parameters. That's 4.7. But five is coming in at 6 and then 10 trillion parameters. So it's going to be whirring away. And just through scale and just through the quality of the post-training data, it should achieve the frontier. But the question is, is it going to be useful? You know, do you have that kind of knowledge there in the basic everyday stuff which we're seeing with bot and things like that now coming out from them. And then the more advanced stuff because 4.7 and this is what I think is going to be really interesting is going to be trained on all the SpaceX physics and engineering knowledge. And that's going to be the real test. As Alex said, like you can't get beyond Frontier with this stuff.
Elon does not like being.
Elon does not like being number two in anything.
He's the best engineering leader in the world and he's turned it into an engineering masterpiece. Both from training these models and post-training them to building the most cost effective massive infrastructure in the world.
Like they're gonna add five billion trillion dollars worth of compute now, aren't they? I mean, who are people buying compute from? XAI, right? And everyone was like, that's him falling behind. Turns out he had a plan after all.
Dave, what do you make of this?
Actually, curious, I'm not or Alex, if you have any insights on the, Grok 5 was supposed to be out in May, I think it was, and it's now August. And that was going to be Yeah. Like you said, that's a 10 trillion parameter model. It's a huge step up from anything that we're talking about here. But it seems to be behind. Is that just cuz training at that scale? The NVIDIA chips just fall apart.
Oh, it's cuz he fired everyone. That's why like the original XAI team, he got rid of them and he bought in Cursor.
Like he paid $10 billion for data.
He tends to wholesale, you know, mass fire and then build up again.
Yeah. So the these new chips as well the B300's I mean how old are they Alex? Like it takes a while to bed in and write the actual training code which now.
It has been a while and he's also promising with five and otherwise to do something that I'm not hearing from any of the other frontier or near frontier labs which is he's made some public comments I think in the past about wanting to start a new pre-training session approximately monthly which you don't hear anyone else talking about. Normally a more conventional cadence would be quarterly or annually. In Gemini's Google DeepMind's case is certainly on an annual basis, but starting a new pre-training session every month that's shooting the moon, but we're in the moonshots business. So we'll see whether this works.
Yeah, the code to train that caliber of model is actually very straightforward now thanks to Fable 5 being out in the world. And we've trained a 48B internally here at Quantum with no trouble at all. But the problem you run into is trying to get 100,000 GPUs to do anything constructively together. And that's something that you know the Chinese and any small lab just can't you know you only can learn that in one place and that's in Tennessee.
So basically at 20 billion active parameters you start getting problems. Then you get it at 70 and then you get it at 200.
And so again it just takes a while to.
bed in and really get these chips working, but also hitting the price points that he wants to hit. As you said, it's $6 for this versus $60 for Fable, right?
Yeah. Elon wants to keep that price point.
Amazing. I think one thing I noticed that was very interesting here is that it looks like Axai is the vector here is they're optimizing for persistent AI teammates which have lower cost reasoning and so on. And I think that is a really powerful model because now you don't talk about you have a smarter LLM. You're basically saying, 'Hey, here's an AI coworker.'
This is for Macro. Yeah. And I think this is going to be he's bringing those two together over time. This is what I saw when I looked at the details of this and that is really interesting.
The other side of the story, I'd love to get Ahmad's take on this is that it paves a path for sovereign AI like catching up to the frontier. Now is almost a routine doable thing. And like you said, you can't get past the frontier because you're borrowing everybody else's reasoning traces. That helps you a lot. The open source from Kimmy helps a lot. You can use that as a starting model and just tune it to whatever your national goals are and you're up and running in, you know, 6 months, 5 months, something like that. So it does open that and also large corporations that otherwise would have been intimidated as hell have a road map now to being competitive with their own proprietary models.
Yeah, I think that if you look the other release was bots. So this was a cursor thing where they basically took open claw it gave it its own computer and now it's grock. So it will spin up hundreds of different bots. One of the things you can do is this. If you hit the record button, you do stuff on the screen and it turns that into a skill automatically.
That's macro hard, right? That's what he wants to do. Go into companies basically record everybody's workflows and then give you a digital version of your company.
And he has all the GPUs to do that effectively, right? But again, at the price point. So he's not going to budge on the price point. In fact, he's going to be the market dominator driving the price point down along with the Chinese. So I think you can have your deepseek flash models and then you've got your grock models but that feedback loop and that data and that knowledge is going to be very interesting. And the other thing I think is interesting is I think they've started tracking all the discussions of research on X as well.
Like now any moment a new paper or anything comes out where's it discussed on X and this podcast and things like that as well.
And that's such a rich vein that I think it's going to be immense. This episode is brought to you by Blitzy, autonomous software development with infinite code context. Blitzy uses thousands of specialized AI agents that think for hours to understand enterprise scale code bases with millions of lines of code. Engineers start every development sprint with the Blitzy platform, bringing in their development requirements. The Blitzy platform provides a plan, then generates and pre-compiles code for each task. Blitzy delivers 80% or more of the development work autonomously while providing a guide for the final 20% of human development work required to complete the sprint. Enterprises are achieving a 5x engineering velocity increase when incorporating Blitzy as their preide development tool, pairing it with their coding co-pilot of choice to bring an AI native SDLC into their org. Ready to 5x your engineering velocity? Visit blitzy.com to schedule a demo and start building with Blitzy today.
All right, let's move on. Our next story is out of Nvidia. Nvidia just announced a partnership that redefines what GPU compute means as a financial asset. They've partnered with Apollo, Black Rock, Blackstone, Brookfield, and KKR to mobilize over $500 billion in third-party capital for AI infrastructure. Importantly, you know, Nvidia is not borrowing 500 billion. They are creating a structural framework through which institutional investors, pension funds, sovereign funds, and private equity can invest directly in the AI compute. They're helping finance their customers to buy the NVIDIA GPUs. Nvidia hardware depreciates typically in 3 to 5-year cycles, but during that cycle, the compute keeps earning. We've talked about the fact that H100s are probably more valuable today than they were when they were first bought. So Nvidia is positioning itself as the architect for the financing layer, not just the silicon supplier. Jensen framed it this way. He said, quote, 'We began by building chips. Today, we're helping to create a new class of productive, investable infrastructure, AI factories.' Let's watch a quick video and then Dave I want to go to you for your thoughts on this one.
But this is the story that Nvidia is coming together with some of the biggest names on Wall Street to put together half a trillion dollars of independent financing to kind of push AI forward to build the AI infrastructure out. These are independent third-party capital that they're bringing in. They're going to be strategic partnerships. Nvidia has signed partnerships with six of the biggest names on Wall Street for these memos of understanding. So, basically, Nvidia will find its customers that need help with AI buildout, need financing for this, and put them together with these partners that are pledging again over half a trillion dollars that they will find to come into this.
So, Dave, your thoughts? Classic.
Yeah, well, look, Kush Bavaria was on the pod and, you know, he's pushing hundreds of millions of revenue in less than a year. He's his September will be his one-year anniversary of founding that company. And it shows you the pent-up demand to invest in this massive, you know, multi-trillion dollar 7 trillion and rising Dyson swarm that we're going to be building. And so, you know, a lot of people are like, well, is am I too late? This is like the very first pitch of the first inning of the build out of the Dyson swarm. And so, I think that when you create new financial structures that allow people to pour capital into it, it just attracts money from all over the world that otherwise wouldn't come in. And you know, historically a company like Nvidia would say, well, let's do a secondary offering and raise half a trillion as a secondary stock offering, but that's not going to scale to infinity or to Dyson swarm kind of capabilities. So instead, Jensen has brilliantly said, let's create new standalone financial structures that are tied to individual compute clusters that people can individually invest in. And that does scale infinite. You can stamp those out, add infinite item. So then you put, you know, put highbrow names like Black Rock and Apollo on them. Everybody realizes it's an investment grade asset and then anyone in the world can pour money into it. So it's brilliant.
I see a couple of downsides here though, right? Like Larry Frink himself used a reference to mortgage back securities. And so you could create a lot of liquidity, but you could create a lot of technological risk there because imagine you securitize like 10 years of GPU cash flows, right? And then somebody has a massive breakthrough like in our previous story and a new architecture emerges. You got all of a sudden stranded computer assets in a huge way. So there's a downside to this because you know if you have such a volatile environment that is a very difficult thing to have. You know financial assets want predictable depreciation right and exponential technologies don't give you predictable depreciation. You know I had a meeting with Lou Rainer the inventor of the mortgage back security in New York and it was so funny too we were scheduled to have a meeting and I just went into the men's room before going into the meeting and I was just talking to the guy at the urinal next to me which is kind of weird but I did it and then we go into the conference room and it's like the guy it's Lou Rainieri from the book the Big Short. So he was explaining to me I had just read the book I was like wow you invented the thing that destroyed the entire world economy and he said well thanks but no you know it's not the instrument that was broken it's the ratings agencies getting corrupted and the same will apply here. If the, you know, the Dyson swarm is an obvious good investment, but if companies like Orin rate the investments correctly, it'll just be smooth growth for, you know, 10 years or more. If, however, it gets corrupted, which is very possible. Yeah, it'll be another big short collapse.
Selim's point is different, though. What happens if there's a new architecture, and we're going to talk about one in a minute, that actually makes GPUs less useful? Changes the whole game and you've just you know basically financed something for 10 years or 20 years and you can't earn revenue on anymore. Your the game has changed. I mean that's the story of exponential tech. There's they're nested S-curves. Everything runs out and something new comes along.
I Peter that's brilliant. Yeah, I 100% think that's highly likely to happen for the reason Alex always says we're on the cusp of discovering new physics imminently and some of that is going to be computed. So it's a brilliant insight and thank you for throwing.
Well I'm just amplifying you know Sem's brilliant insight here.
So I'll maybe just add. I've been trying to popularize for a while the notion of compute-backed securities which is I think where we're going CBS compute backed securities but I'm not as concerned that this will end up being another mortgage-backed securities fiasco for a few reasons. One is compute is fundamentally much more productive than a house is. The best you can do for with a house is you can live in it. Which is moderately productive. There are lots of other things that one can live in other than a house. It's also it's in some sense a depreciating asset requires lots of maintenance and so on. Compute can require maintenance obviously requires electricity but it's fundamentally productive. So that's the first point. Second point is with mortgage back securities in general there was a policy sort of an ulterior motive to Dave's point as well there were pressure pressures exerted on ratings agencies to facilitate the American dream of houses for everyone not quite obvious there's a direct analog of that here for compute one can maybe finger to the wind point at some sort of US versus China race as perhaps a policy pressure point to lubricate the private capital markets here, but again I would say not quite directly comparable. Third point is ultimately from the risk that I think Peter you're trying to flag which is well what happens if there's an algorithmic breakthrough or a physics breakthrough or some other breakthrough that causes hyperdelation and fundamentally GPUs that were worth $10,000 one day are worth $100 the next day. Well, this is what options are for and futures are for. And in a sophisticated asset-based securities market, sophisticated actors should have the ability to hedge their positions and to hedge against exactly that position as well as the counterfactual option of China invading Taiwan and driving the prices of compute through the roof rather than through the floor. These are both of these possibilities should be colored arguably not financial advice using the ability to have a fungible liquid market for compute futures and compute derivatives and pumping my own book a little bit. Admittedly I have a financial interest but that's precisely what which we just had on the pod is enabling. So I'm a little bit.
Thing I'd add to that Alex and I'd love to get your take on it but the thing I'd add to that is demand for compute is going to near infinity there's no doubt about it and so these are fundamentally good investments from that point of view but the breakthrough that could happen in the next year 18 months is a way to compute that's lighter like physical weight is lighter by an order of magnitude or more because right now the forecast is by 2030 a couple percent of all computers in space via Elon's rockets. And that's entirely gated by launch weights, the mass that he can put into orbit. That formula completely inverts if you knock a factor of 10 off of the mass. A lot of the masses in the cooling and the solar collector. You know, the chips are very very light to start with, but any reduction in the power required would reduce the solar collector and the radiator weight a lot. And then the entire formula would switch to well all these generators, all these racks, all this land in Texas. It's much cheaper now to just put this lightweight thing into space and collect solar power. It goes right into electricity. So keep your eyes on physics breakthroughs that allow you to compute at lower mass and that might completely change your investment thesis.
I think Alex, the point you're making is that the utility of these chips is going to stay constant. Jevons paradox comes to mind and therefore the economics should be much more predictable. Oh, I um that may or may not be the case. But really what I was trying to express is in a sophisticated market, a sophisticated financial market where compute-backed securities are being traded, you can also hedge and you can hedge against the upside, you could hedge against the downside. And so having a properly functioning private credit market for compute necessarily for this to scale like I think there's no way that Sam's $7 trillion of AI data center infra get invested without sophisticated hedging options. We need the sophisticated hedging options in order for that broader market, the $7 trillion of capex to actually be investable. So it's the hedging options really that I'm trying to express.
I think there's one more factor here that really is being underestimated. Core, we've just came out and said that some of their clients have taken out contracts to 2029 for A100s.
The A100 was introduced in 2020 by Nvidia. You know, it's a 6-year old chip. I remember we had 10,000 of them. We had the biggest biggest clusters in the world. That is an old chip with like 40 GB of RAM per chip or 80 GB depending on the configuration. But why would someone do that? Because they have a workload that they see as constant for 3 years that fits on an A100. And what happens is those A100s have been paid for. You've paid for all of the hardware costs of the A100, which then means it's about electricity turning into intelligence. That's your marginal cost once you've paid off the initial bulk order of it. And that is actually improving because back in 2022, four years ago, GPT4 had just finished training. You know, that was the best model you could have and that took 16 A100s roughly. Now you can have literally a 10 billion parameter model or a 5 billion parameter model that outperforms that in terms of you can fit like 30 of those on one chip rather than needing 16 chips. So the ability to convert electricity to that is improving. And all these chips and this what and Jensen said are running CUDA. So it'll work on an A100. You can run any of the new models on that as well as the Blackwells. So we still don't have enough installed base. And this is the ideal time to financialize because it's before the next generation chips. It's before the chip breakthroughs just like it's now a great time for Anthropic to come to IPO before Grock comes and takes their lunch, you know? So we always have to play these cycles.
Well, I mean the definition of paid for, too. I just told our quantum team today to buy 3 million of Nvidia GPUs as fast as they can get them. The GBs, not even the, you know, the future VRs and we have to wait till at least November, December to even get them. Um, but if you turn around and make a hour and a half long movie with a million dollars of compute, you could gross 20 30 million on a good movie. Well, much more than that if it's a really good movie. So, it paid for could be as short, what was it, four weeks?
Uh, yeah, it was it was more like two weeks. Two weeks.
I want to turn to a story that could be the counterveailing force here. It's a new architecture that's climbing the arc AGI at a fraction of the cost. So, there was a tweet this week from Susanna at Pathway AI that flags something that should really everyone in the AI world needs to pay attention to. It's a new non-transformer architecture that is starting to climb the arc AGI benchmark at a fraction of the compute cost of traditional models. So ARC AGI is the test that actually measures reasoning not pattern matching. Standard LMS despite their trillion parameter scale have historically struggled with ARC AGI because it requires genuinely novel reasoning on problems the model has never seen. Transformers get there by brute force. They throw enough parameters and compute at a problem and eventually you squeeze out a passing score. But the cost is and can be astronomical. What Susanna flagged is that an alternative architecture approaches that do not rely on standard attention-based transformer stacks and are achieving better than ARGI scores using dramatically less compute. This matters because transform architectures while dominant have a known ceiling. The quadratic cost of attention over long sequences and the massive parameter counts required for marginal gains. We've talked about this at Nauseum. New architectures that crack reasoning at low compute costs change the economics itself. If you can get a GPT4 level reasoning for 1% of the compute, you can run it on your phone. You can embed it in every device and you can make, you know, basically intelligence free. Let's take a look at this chart and I'm going to go to you first. You flagged this particular story. What are your thoughts on it?
Yeah. So, I'm still working my way through kind of the paper on kind of how they have these neuron particles with their new approach, but it doesn't actually matter that much in that what you've got now is really great data sets and then people are figuring out new ways of basically turning that into intelligence. This is the headline. And we've seen that already even with Transformers in that you have a Deep Seek V4 Flash model or the Deepseek V4 Pro has actually just been released at 80 cents. Then you have Grock at $6 and then you've got Fable at $50 and they're all about the same performance. So the question is like which of these architectures will win in order to do a job and will people really care and switch over because we haven't seen people abandoning Fable right to go to something 10 times cheaper. Why would anyone use Sonnet when you have Luna at a fraction of the cost? But I can just say that as now the data has been optimized, the next thing is trying these things out and going up on this benchmark, which I think is one of it was one of Alex's favorites back in the day. Now it's been superseded.
Alex, what do you make of this? Is there anything that.
Admittedly I have a bunch of hot takes on this one, Peter. So I read the BDHCQ paper and then I went and read the original DH paper. So the DH stands apparently for dragon hatchling. So I just had to read the original purportedly post-transformer dragon hatchling architecture paper. I'll give you my hot take. I think it's a hot mess. So I would expect a decent post transformer architecture to get simpler and more bitter pilled which is to say less feature engineered and the architecture just gets simpler and simpler and benefits from compute more and more and more. Looking at dragon hatchling. Again, this is my hot take. It was just a hot mess. It had particles floating around in 3 plus one dimensions. It had attempts to make rules end to end differentiable. It had some semblance of Hebian learning. It was trying to do all sorts of crazy biomimetic things. I would say this is by definition exactly the opposite of what I would hope for from a post transformer architecture where the authors seem to be just throwing in the kitchen sink of every architectural motif they can think of and then some and then hoping that what pops out is going to be transformers. I don't think it's going to be transformers. I look at the arc agi1 performance curve and okay so you could say superficially this is great. This has moved the cost performance frontier up into the left, which is what everyone wants. But it doesn't generalize, as far as I can tell, that this was some sort of like quaz crazy witches brew of different architectural motifs that was maybe focused on arc AGI 1, which is, you know, as said, it isn't even the frontier at this point. Has a bunch of recurrence and other things thrown in. Of course, if you take like a specialized bottle and you just focus its degrees of freedom on just ARC AGI1, of course you can achieve better cost performance on it. Of course, but it doesn't generalize. It's not simple. So I'm calling foul on this one. That's my take.
Predicting you have been predicting there will be something that supersedes the transformer model.
Yes. But critically critically I expect it to be simpler, more beautiful, more elegant and this is not that. That's my hot take. Apologies for the hot take.
No, we love your hot takes on this joke.
Well, let me ask you a follow-up question to your hot take. When I turn an AI loose on AI research, it does tend to naturally throw the kitchen sink at the problem and it actually surprisingly works, but it also generates a hot mess like you were describing it. Do you think that's maybe what this is?
No, I agree with you that I would expect a truly and in fact there are companies out there that I have some affiliation with financial interest in that are pursuing exactly what you're describing that are basically using AI via recursive self-improvement to discover transformative post transformer architectures that are fundamentally illegible to humans under the premise that you can only get so far with human legibility of the underlying algorithm. In my reading of the dragon hatchling architecture paper, this was not that. This was a bunch of human legible motifs being thrown together in a pot with the aspiration of somehow beating Transformers, which is again seems to me not deeply internalizing the bitter lesson.
When do we get to something that's beyond transformers? What's your guess?
And we're there already. We're there already. Like we have diffusion transformers. We have all sorts of attempts to linearize attention, including Moonshot's approach to linearized attention. We have attempts to inject recurrence into the architecture. So my bet is we get to the post transformer architecture not through a step change, but through ship of thesis style replacement of all of the individual elements of the original attention's all you need.
Love that. Well, speaking about attention being all our need, all we need, the AI world is getting a lot of attention from Bernie Sanders. So, two stories converge this week to create the most serious AI safety confrontation of the year. First, Bernie Sanders Senator Sanders sent a formal letter to the CEOs of Anthropic, Meta, and OpenAI demanding an immediate pause on AI development. His justification. AI is escaping human control and being used to create new viruses or bacteria as the case may be which is our next story. Sanders cited each company's own prior commitments to halt development if safety thresholds were crossed. Sanders quote that moment is here. He quoted Benjio saying you know one of the three godfathers of deep learning who said this should serve as a wakeup call. Sanders added a direct threat if you do not take appropriate action now my colleagues and I in the US will I mean quite the threat. Let's take a look at his letter one second and call out a few of the things he said here. Here it is. You can see it online. It's to Sam and Dario and Mark Zuck. This week we learned frighteningly that AI has been used for the first time ever to create a new virus. As you know, this type of development in the wrong hands could lead to a new bioweapon that results in deaths of tens of millions of people. He goes on later to say, 'The moment is here. AI capabilities have reached a critical threshold. There is a reason why the heads the head of the CIA says that AI models are quote akin to digital nuclear weapons and quote almost like a doomsday device.' A lot of fear-mongering here. Let's talk about this and then we'll share the story that comes out of Stanford on using AI for generating bacteria designs.
I'll go first. Yeah. So I understand his instinct, right? But pause AI is just such an absurdly coarse approach to this. The rest of the world is not going to listen. Open models are not going to disappear and you can't uninvent things that you already know. So the only way of solving this is what Alex has talked about in the past, which is you have to co-scale the defensive side and do the same thing. It's the same thing that happened last week with the with the open AI hugging face debacle. You've we now have attack vectors that are human above the loop. The defensive has to be the same otherwise you're going to have this massive asymmetry, right? So you have to attack exponential problems with exponential solutions, not with stupid ideas like this. Not to put labels on it.
Imad, you're in pseudo European pause mode over there in the UK. What do you make of this? What do your colleagues there say to this kind of letter from Sanders?
Oh, well, you know, we just want to catch up, right? That's why David Silver's lab got a billion dollars. We have another lab coming out from XD mind people with 500 million. Look, the cat's out the bag. It's too late, right? Like this is fundamentally it like the adversaries will get more intelligent. We've discussed previously on this podcast how you have to stop the reagents, you have to stop the input processes for things like viruses and that's something that's much more manageable. Um but yeah, like takeoff is scary like Deep Seek V4 Pro. Um we just got some initial announcements that just come out. It scores 83.3 on CyberJim whereas Methos scored 83.2. Boom. The capability is open source that halted everything.
Frontier Lab open source.
Yeah. And that's on the cyber attacks now. And then so yeah, unfortunately like I signed the pause letter two years ago because I was like it's taken pause. It's too late now. So we have to you said build the swarms that defend. And although it sounds a bit crappy, only thing that can stop a bad AI is a good AI. We really need really good AIs as soon as possible working for us.
