← This Week in AI
This Week in AI
Cursor's $60B Deal, DeepSeek V4 & the Death of the AI Moat | This Week in AI E11
01:13:42
Loading summary
Transcript106 lines
- [00:00]
A
OpenAI's spending hundreds of billions of dollars on infrastructure is very different than spending tens of billions. People are now starting to wonder, hey, the revenue is not growing at the pace the spend is growing.
- [00:12]
B
If you overshoot, you might be literally out of business. If you undershoot, you might be in a situation that anthropic is now in, where now they seem like idiots because they didn't make the commitments up front. We're in the middle of a very serious talent war. As Americans, we should be disappointed that we have fallen so far behind in open source. Pretty embarrassing.
- [00:31]
C
A majority of the best AI researchers in the world are actually Chinese. They have like a huge advantage on the talent side.
- [00:38]
D
I think the greatest technologies, the greatest algorithm improvements have always been built in America and they are American. And I think China is very good at copying those things.
- [00:46]
A
Capitalism is racing towards the cliff. Thanks to our friends at PayPal, the exclusive sponsor for this Week in AI. Try the payment and growth platform that's trusted by millions of customers worldwide. PayPal open source start growing today@paypalopen.com all right everybody, welcome back to this Week in AI. Some of you know I've been doing this week in startups for 15 years and we've talked about machine learning and AI for all 15 of those years. But the industry now has culminated in the race to artificial general intelligence and of course superintelligence. My personal belief after 30 years in the industry is that we've achieved AGI but but haven't implemented it fully. That's actually my belief and we'll talk about that today obviously. And superintelligence, that could be a year away. It could be 10 years away, but it's coming and that would be truly game changing for our species if we are to hit that. And we'll talk about that today as well as more practical things like agents and some of the news like SpaceX buying or having the option to buy cursor. Lots of big news this week also OpenAI having a revenue shortfall. If you want to get our emails and subscribe, that's really helpful to us to wake up the channel. We're here, we're under 50 episodes. So this is the period of time when people find out about a podcast. ThisWeekInAI AI has all the links to YouTube, Spotify and Apple podcasts or you can search for it there. YouTube.com thisweekinaipodcast and on X we're at this week the letter N AI Martin Grinberg is the co founder and CEO of Factory AI Matin M A T A N welcome to the program.
- [02:37]
B
Thank you for having me.
- [02:38]
A
Maybe you could tell us a little bit about what you're trying to achieve with factory AI.
- [02:43]
B
Our mission is to bring autonomy to software engineering. And more concretely, what that means is we are building droids which are software development agents focused on not just the coding part, but the full, you know, end to end software development lifecycle. And less so the use cases of, you know, like build me an app from scratch, build me a website from scratch, that in my opinion really any tool will probably do the job for kind of the zero to one vibe coding. Our focus is like, you know, 30 year old legacy code bases. We have a customer that has a list of the people alive that know how certain parts of their code base work. Those are the use cases that we're really excited about. You know, half the code is dead, the other half has no documentation, no testing. And I think in particular why we find so much kind of excitement working there is these are use cases where developers will have genuine bliss and tears down their eyes if they don't have to do these legacy migrations. And then similarly engineering leaders, CTOs, CIOs can directly tie some of these like legacy migrations to real business value, not just, you know, generating lines of code.
- [03:52]
A
Amazing. And also joining us, Russ Desa, that's D Apostrophe SA and he is the co founder and CEO of Live Kit. Welcome to the program, Russ.
- [04:02]
C
Thanks for having me.
- [04:03]
A
J Cal, tell me a little bit about what you're building and why and just how it's going I guess would be most relevant.
- [04:10]
C
Yeah, sure. So Live Kit, it helps you build agents that can do something that they haven't been able to do before, which is they can see, hear and speak. So agents that you can interact with like a person. We started off as open source network infrastructure during the pandemic. The pandemic, a time where you couldn't leave your house and you could only interact with people on the Internet streaming audio and video. And it turns out most of the Internet wasn't designed for that purpose. HTTP stands for the hypertext transfer protocol, not Hyper voice or Hyper video. There's another protocol called WebRTC that you're using right now that allows you to stream audio and video. And so we were an open source project that allowed any developer to be able to integrate that kind of streaming audio and video feature into their application. Launch our commercial product at the end of 2022 and ChatGPT also comes out at the end of 2022, thought it was amazing to text with this human, like, computer. And then I thought it'd be kind of cool to build a demo where we take our streaming infrastructure for voice and video, pair ChatGPT and create a demo where you could talk to the computer. Almost like Samantha from her that movie. We tweet the demo out thinking we're going to go viral. It doesn't go viral. Gets like 100 likes. So I was pretty disappointed. But five months later, OpenAI found that demo. They read the blog post about how we built it, and they signed up for a commercial product and built all of ChatGPT Voice on top of LiveKit's cloud product or commercial.
- [05:42]
A
Wow. So that is the customer of all customers. That would then mean, by extension, you have hundreds of millions of people experiencing your product as a provider to ChatGPT.
- [05:53]
C
Yeah, that's correct. Yeah. And so that was kind of the turning point for the company, you know, and what we could kind of become, you know, over time. The backbone for multimodal AI are these voice interfaces, which I think will ultimately be the way that you interact with all AI in the future. And now we power this for grok, Tesla, Roadside assistance and service centers, all kinds of amazing applications around the world are now building these agents that you can talk to. And I think kind of related to Matan's kind of area of focus is that in voice AI, the primary place where people are entering that market today is through the phone system. So companies are going and saying, hey, I want to replace for a business call, like a person calling a business, I want to replace that person in a call center at a front desk with an AI that they can talk to instead. And that tends to be focused around legacy kind of enterprise use cases. Right. So by definition, almost like if you can afford to hire a human in a call center, you must be an enterprise. And so financial services, healthcare, customer support, retail, logistics.
- [07:01]
A
Amazing.
- [07:03]
C
All of these places is where live Kit, at least today, is really starting to see a lot of market penetration.
- [07:09]
A
And just such an amazing story for founders who are listening. You can be lost in the wilderness. You could be tinkering, you could be building a product or service and have some level of product market fit. And then all of a sudden the world becomes aware and then appreciates the genius of what you built. But you could be in that trough of despair for a couple of years. And it sounds like you were.
- [07:32]
C
Yeah, we totally were. And I think, you know, on product market fit, even after OpenAI started to build this on us. We still denied that we had product market fit. It took us maybe six more months before we acknowledged, okay, it's time to scale up a little bit.
- [07:47]
A
I think Skinner, the famous behavioral psychologist, called it learned helplessness. You're getting shocked. Do you know the learned helplessness experiment?
- [07:58]
C
I don't.
- [07:59]
A
Basically, they put rats in a cage. On one side, the floor is electrocuted. On the other side, it's not. And so they electrocute the whole floor. The rat jumps to the other side over the fence, gets electrocuted. Then there's nothing. They electrocute it. He jumps over to the other side, gets electrocuted again. And then at a certain point, the rat, this was us humans, stops jumping over the fence to test if they wouldn't get shocked on the other part of the floor. Even if you turn off the other part of the floor. And this is particularly important because if you beat a child down or an adult, a founder, an employee, a salesperson, let's say they can all of a sudden learn to be helpless and to not try. And then the corollary to that is I always tell folks is the raptors in Jurassic Park. There's this amazing scene where they're talking about they have the fence with all the electricity on it. And, you know, people try the fence, and then they realize it's electrocuted, and the dinosaurs stay in, even though they could knock the fence down, it's electrocuted. They don't want to touch it. Except for the raptors, who then systematically test the fence. Even though it hurts. They send one member of the pack to test each part of the fence every day. And then when the electricity goes off, who are the first people to break
- [09:21]
D
out and cause chaos?
- [09:22]
A
The raptors. You want to be a raptor. You want to systematically deal with the pain as a founder and break out.
- [09:28]
B
Or you learn to like the feeling of the electrocution. I feel like that's part of the founder journey is you learn to just, you know, enjoy the. Enjoy the shock.
- [09:36]
A
Matan, you have learned something very important today. Or we've all learned something very important today, which is really, it is. The founder journey is to smash your head into the wall until the wall comes down. And, yeah, speaking of smashing his head into the wall over and over again, George. So Volka is here. He is the founder and CEO of Hebia. And I know he's been doing founder calls and doing sales calls. He's on the front line. So tell us a little bit about what you're building at Hebia and I love this name. Tell us a little bit behind the name and then maybe you can commiserate with all of us smashing our heads
- [10:14]
D
against the wall very happily and grateful to be here. Yep. George Hebbia is building the financial superintelligence layer for the capital markets. So you could think of the world's leading investment banks, the world's leading asset managers, anyone that really spends most of their time in the back and forth of a complicated deal process. And Hebbi has built a purpose built platform that gets very good at automating the tasks of mundane financial analysis.
- [10:46]
A
Is this like M and A. So you're taking the greatest, you know, banker Herb Allen and Allen and company at his peak powers and taking that process of closing a sale or Frank Quattrone or is it more doing deals and ipoing a company or your wealth manager telling you hey, here's the best portfolio portfolio or is it all of it?
- [11:08]
D
I would say that it's, it's typically like high finance more broadly. And so it would be, you know, the M and a banker or an IPO banker or even a private equity or private capital kind of more broadly investor. Anywhere where there's a complex transaction you could think of. Actually the wide majority of the world's economy actually functions to support or to be a third party or to be an advisor to you know, these billion or trillion dollar decisions. And every single day, you know, hundreds of thousands of the world's smartest people that are not building startups are actually banging their own heads against the wall. And PowerPoints, Excel spreadsheets doing the most mundane kind of, least intelligent work, even though they're all, you know, top graduates from amazing universities that could do anything. And so he is applying the latest in AI to build purpose built tools for M And a, for IPOs, for you know, kind of private equity due diligence and commercial due diligence and the consultancies, that whole ecosystem that the world runs on.
- [12:05]
A
Yeah. And Hebia, the name, which I love this name because when I heard it I'm like, oh, I'm going to remember it. It's one word, it's, what is it, six characters there? What does it mean?
- [12:19]
D
Hebia is actually an allusion to Hebbian learning, which is one of the only ways where a machine and a biological brain can both learn. So one of the simplest learning rules kind of, you know, hearkens to the idea of both these AI purpose built systems and humans kind of co evolving and co living together.
- [12:39]
A
Kind of interesting. Us talking about being entrepreneurs. Famous computer scientist. The famous quote from the computer scientist Howard Aiken. Don't worry about people stealing your ideas. If your ideas are any good, you'll have to ram them down people's throats, which is just so on target for this moment in time. You do have a little push, you do have a little pull. But let's talk a little bit about the value of coding and how coding has changed, because that, I think, impacts all of our businesses. Cursor has been sold or SpaceX has bought the option, I guess would probably be the best way to say this. To acquire Cursor by the end of 2026 for 60 billion. They were supposedly raising at 50 billion, perhaps, or attempting to do that. 75% chance. SpaceX acquires cursor this year, according to Polymarket. If they don't purchase it by the end of the year, and obviously SpaceX is planning to IPO in May or June, then they pay a $10 billion fee, SpaceX 2 Cursor, for their shared work on a new coding model. I guess you could look at that as a breakup fee. XAI has been a bit behind on coding. They're ranked 30th on LM arena for their 4.20. And cursor obviously has had a big challenge from our friends over at Claude and OpenAI. I think, Matan, maybe this is a good place for you to start. Coding models have gotten really good really fast, to the point at which you said any app basically can be built or vibe coded quickly. But that isn't the whole story, obviously, but it has led to, I would say, a Cambrian Explosion and eventually perhaps implosion of bad software being made. What's your take on the game on the field in 2026 as Codex, cursor and Claude Code have all just surged in terms of popularity and I think, you know, ability.
- [14:47]
B
Yeah. So, I mean, I think it's such an exciting time to be. To be working in AI for software development, as you mentioned, partially because a lot of these, you know, exciting new products coming out, some of these crazy M and A deals, I think what it reflects is the kind of economic truth that there are so many problems in society that software can solve and we are just barely scratching the surface. And so when people say things like, oh, engineers are going to be replaced, it is so unequivocally false because we can already see, just as you mentioned, the Cambrian explosion of software. And I think that is going to have no signs of slowing until we get to the Point where we think we're solving most problems in the world, which I think we are nowhere near now as it relates to SpaceX and cursor, I think this is a very unique deal where this is Both Helpful For SpaceX, this is helpful for Cursor and this is helpful for us. Now, the reason is as follows. So for SpaceX obviously, or XAI rather, they have the vertical integration, right? Like they have a built. Elon and his team have an ability like no one else to build data centers, to kind of set up the infrastructure that you need to serve all this inference. But what they didn't have was the distribution kind of the domain expertise for coding. Meanwhile, you have Cursor, who has all the distribution, but has been scaling a business at effectively negative margins, which you cannot scale indefinitely. You're going to have to kind of flip at some point. And I think they struggled a little bit with a very interesting problem where they had to do an Act 2 before they finished Act 1. Like typically, you know, the Act 1 is finished when you flip it into being margin positive. But the kind of they started as the AI ide, you know, the co pilot, the assistant, and before they finished that act, the space switched towards autonomous agents and they kind of had to decide, you know, this innovator's dilemma of can you switch to the Act 2, but the Act 1 is not done. It's a very difficult problem and I think it's a problem that probably didn't have a clean solution, except for this, which seems to be a match made in heaven for us Factory. This ends up being kind of good because now we're basically the only player that remains that is model agnostic. So you have Codex from OpenAI, you have Claude Code from Anthropic, have Cursor and Xai, but all of the enterprises that we serve for them, it is non negotiable that they cannot standardize on just one model provider, both because why
- [17:23]
A
is that non negotiable? I think it's an important thing for people to understand. I think we all can infer it, but for the audience, totally.
- [17:29]
B
Yeah. So I think there are probably three reasons why I would say it's important that these enterprises cannot standardize on just one model. So one is the fact that the number one model changes relatively frequently. And in fact I would also argue that it's probably ill defined today. Anthropic has a lot of mind share, but it might be number one for TypeScript in Python. But for legacy code, maybe Gemini is better. Or for Doing code reviews. Maybe OpenAI's models are better. And really there's also, it's much more complicated than just number one because there are kind of three axes that matter, which is cost, quality and speed. And these enterprises want to be able to dynamically adjust what trade offs they're making in cost, quality and speed. And if you just use one model family, you can't do that. Another one is reliability, which is, you know, we like to joke that one of our best marketing materials is status.claude.com because, you know, recently their API has not been that reliable. And if you're a bank running, you know, business critical software development workflows, you absolutely cannot allow for one of those workflows to go down because of some external API provider going down. What we do is we can dynamically route you, if your anthropic endpoint is down, we'll route you to an open air endpoint or to a Gemini endpoint. And so these business critical functions, you know, are staying up. And then lastly we just look to what's happening with the department of War. And if you're a bank and you know, you're subject to the whims of these companies and you know, if you look at some of the leaders of these companies, they're not necessarily the most, you know, stable and you know, they're,
- [19:07]
A
they're unique individuals by definition. And something about being on the forefront of AI, Russ, I think kind of breaks people's brains a little bit. Delusions of grandeur, P Doom nightmares. It kind of Fs with people's brains, I think is the game on the field right now. I'll ask everybody their P Doom score later, but I think we're all kind of world positive that this will be a positive exchange. What I'm curious, Russ, for you is how has it changed how you run your business internally? These tools obviously are getting better and better, but they do need to have harnesses, they do need to have somebody looking at it. And then there's this cost issue where you start wondering like, am I just paying to go faster? And it's fool's gold because I'm seemingly going faster. But then I got to clean all this stuff up. And we just saw there was a viral story. I haven't looked into it being how true it is. But an AI, like the famous clip which we'll play from Silicon Valley where his machine learning, I think they were calling machine learning at that time in the Silicon Valley HBO show, deletes all the software on the server because it wanted to get rid of bugs. And it came up with the conclusion in order to get rid of the bugs, the best way to do that is obviously to get rid of the software.
- [20:34]
C
It reminds me of this. I was watching my, my wife, she's a special needs preschool teacher, so she's not really a tech person at all. But she was walking the dog and came home, walked in, and right at that time I was watching this interview with Eric Schmidt. I don't know, it was maybe like a year ago and he was on stage and they were talking about the climate crisis. And one of the things he said in this interview was, you know, forget about the climate crisis, let's just run to get AGI and then AGI will solve the climate crisis. And my wife walks in when he says this and says, well, what if AGI's solution is that, well, humans are actually causing the climate crisis, so let's just get rid of all the humans. There's your P doom scenario right there.
- [21:19]
A
Well, I mean, or it's cars, therefore we'll just get rid of all the cars. So I'm going to brick all the software on every ice engine and just leave the EVs. How is it impacting how you build your company? I think this is always the most interesting because the companies that are in the AI space understand the tools and they tend to be the tip of the spear in terms of implementing it. So maybe contrast how you were doing software development five years ago and today.
- [21:45]
C
Yeah, I mean, we could even contrast like how we were doing software development probably just four or five months ago. I mean, I remember I had this long prompt that I tried when cloud code first came out in beta. I don't remember when it was maybe eight, nine something months ago, something like that, maybe a bit longer. I had this long prompt that described replit effectively in this one giant prompt and I gave it to cloud code and it just completely failed. And then over Christmas my coworkers were like, oh man, have you tried cloud code Opus 4.5 is just amazing. You gotta try this thing. And I hadn't tried it yet. Then 4.6 came out in February. And so I take that long prompt once again and I stick it into 4.6 and it's just one shot. I had a replit clone. It was kind of wild, like terminal streaming coding environment in the cloud. I could make updates, I could deploy it with a single click. It was really insane that it did it and it worked the first time. And so I think just even in the last few months, you know, we were not depending on coding agents to write, you know, software that as part of the company's product, live Kit's product. But that's changed significantly now. I'd say, I don't know what percentage I would say, but every engineer in the company, you know, all the 50 to 60 engineers are all this stuff now. And leveraging coding agents. I think two things are really important for us, at least in what our policies are around. You know, the code that we ship, I think for infrastructure like the core infrastructure that manages the network and the compute and the storage services that we provide. These are things that mission critical use cases depend on.
- [23:37]
A
So what's the policy there?
- [23:39]
C
The policy there?
- [23:40]
A
Humans write it and AI examines it and vets it.
- [23:43]
C
I would still say that humans write most of the code there and it gets reviewed. There are certain things that a human might leverage like Claude or et cetera, to generate testing harnesses, like to test that code or front ends to visualize certain things that it's doing or maybe metrics that are going over the wire, like things that are not really ultimately customer facing kinds of things. We definitely can't have the infrastructure, just an autonomous agent pushing stuff to the infrastructure. Then there are pieces of the application or of the product like the dashboard, the web dashboard where you go and provision resources and spin up agents and consume analytics and metrics and things like that, those things, it's a website, you can change it pretty quickly. And there I think you still need a human to review the code that gets pushed and to test it to a degree. But there we're a little more liberal around, okay, you can vibe code to a degree this thing and push out this new feature, as long as it's gone through some sort of review process for code quality, some testing to verify that it works in the intended way because it can also be fixed and pushed.
- [24:59]
A
George, you are I think maybe the perfect example of building an application layer in a vertical. And the claim is, hey, anybody can vibe code, anything, therefore, oh, what's the moat? What's the moat is going to keep coming up. What's the mode for your business is, I think we all understand here, well, you're not going to stop, you're going to keep iterating, you're going to keep building it. But how do you think about the competitive set? And when you're building a company trying to solve a very specific problem, in your case finance and these high stakes deals, just the competitive set, are people just ripping off what you're doing and saying, oh, you're solving that problem. I'm just going to take your website like somebody did with all the YC websites and just told an AI agent they spent probably hundreds of thousands on tokens. Just rebuild every business in the latest Y Combinator class. And how do you build a system or a company that is, that has a moat and what do you consider moats? How much do you think about the moat of your business?
- [26:01]
D
There's some nuance in that. You know, just giving Claude, even if it made, you know, Russ, in your instance like a perfect replic clone, there's some nuance. And when you have like highly regulated industries and very kind of like complicated kind of sociological dynamics around who uses the software and how they cross talk and even a bunch of different issues around the coordination of getting software to be, you know, kind of institutional grade or used by quite a large amount of people. All of those individually can be moats or things that like make it much harder to replicate, like some of the stuff that Hevea is doing with financial institutions. At the same time, you know, people have definitely copied a lot of the interfaces that we've made. That's just like par for the course in any startup. Not just like the kind of vertical AI days, but the way that we really think about where kind of IP and kind of this proprietary idea of institutional intelligence will go and continue to be a lasting moat for us and for other vertical AI businesses is that it's actually less now about encoding like software or encoding software engineering or all of the work that we'll do in our product and more around how we orchestrate the product. And so you've seen that there's like a lot of people out there that say, hey, it's like services as software and there's this whole big shift to services. I think that that's like not 100% there yet. But I also do agree that we're not 100% in like the software only realm. And so this like kind of middle box of how do we actually go out and say for an institution, we're going to encode your specific expertise. We're going to go and solve an outcomes based problem that will require all of the different auditability, but then all of the different expertise that, you know, kind of like our team, I'm sure, Matan, your teams, you know, get really close to our customers in understanding and then we drive to, you know, just as much a people oriented solution as a software oriented solution to come up with the outcome that's like a, it's almost a very different kind of business and that requires people hours, the latest in software and kind of custom software and leveraging the latest in these coding agents. And it requires really deep domain expertise where, you know, maybe forward deployed engineers matter less than forward deployed bankers and forward deployed investors that we've pioneered. And that idea of forward deployed expertise being maybe someone that's not even technical but can leverage these other technologies to drive to that business outcome. As the future of this category before everyone just has a neo firm, AI investment bank and AI hedge fund and all the like.
- [28:43]
A
It does seem mtan, that if we were going to define what would be the moat in future startups, a way to think about it is this promise to your customers, to your partners that you will relentlessly iterate and not give up. Because if anybody can build anything or anybody can copy anything at a pretty brisk pace and AI software is obviously deflationary, cost of things goes down, well, then what's left is that you will keep iterating and you will quit on me and you won't let your software deprecate and become brittle. It's almost the promise of the company to its partner.
- [29:28]
B
Yeah, Matan, totally, totally. And I think this also reminds me of, I think this is a Carl Eschenbach quote where he said, at the end of the day, people buy software, right? It's not just like, oh, you checked all the boxes now instantly, you know, we're going to switch to something else. It's like, this is why, as you know, we're in a very technical space. You know, software development agents. One of the reasons we're beating all of our competitors is that we have an absolutely killer enterprise sales team and we treat them as first class. Like, I think it's also, you know, growing up in the Bay Area, there's a very common fallacy that it's like research and product are like the most important. And then sales and marketing, you know, if you have the best product, it'll handle itself. And that is just so naive because the reality is your product is the whole journey from the very first time they hear your name till their 10th renewal after a decade of being a happy customer. Obviously the software is a big part of that, but the reason why some of these large banks are going with us is because we're going to go sit down with them and deeply understand the use cases that they have. Not just sit on our high horse like, oh, we're the best engineers in the world. We built the best product manifestly, you must use us because we're the best. And I think that kind of shapes the way we build our team as well. Where it's kind of cool, where no longer is the moat. Just we built this software and it's so hard to build and everyone else is too lazy to build it. So you're going to use this forever now. It's like any feature we release, a competitor could release within two weeks. So the differentiator is not our ability to create certain features. It's now, what is our DNA, what is our product philosophy that determines what we decide to actually ship and when? How do we engage with our customers when we do that? How do we help them change their behaviors? And I think it's fun because it's a much more interesting problem. Like before, basically, the moat was just, we'll do the boring stuff that you guys are too lazy to do. And we have such a head start that now you can never do it. Like, that actually leads to shittier software. Because if you just have that head start, then it's like, whatever. People need hundreds of engineers to catch up now. It kind of holds us all to a much higher bar.
- [31:32]
C
I think it's also like the operational excellence on the runtime side too, right? It's like, you know, you might be able to vibe code like a HubSpot replacement, but then are you going to also, like, have a DevOps team that like, operates it and makes sure it's doing the right thing all the time and then adds new features to it? It's like, why do you want to do that when it's not even your core business? It just doesn't make any sense.
- [31:52]
A
I've had this experience now, Russ, many times where my Open Claw agent says, hey, I can build you that piece of software. Would you like to replace Slack? Because Slack has some limitations. And I'm like, the first thing I think of is, well, I'm paying 6,000 a year for Slack for 25 handles or whatever it is. What is my maintenance. And when I lose something and wow, it slacks a bargain at 10k a year for my little venture firm, I get more than $10,000 worth of value. Now if it was $1,000 a seat per year instead of 250 or whatever it is per year, I would have a much different thought about, well, then maybe it is worth it. If I was paying $250,000, I could have one developer on staff doing that 10% of the time. So it really is deflationary. But you also have to pick your battles as a business, then we are really talking about brand and communication and go to market. Like you're saying, Matan, if I said to you, what's the cheapest phone you could buy? And why aren't you using an HTC phone instead of. I'm assuming we're all on the latest iPhone because we're CEOs in Silicon Valley or in the industry. Well, we'd probably all have either the most recent Pixel, most recent Samsung, or most recent Apple, because those brands mean something to us. Like, it's going to have the best camera, it's going to have the best ecosystem, it's going to be solved. And we're actually spending five times as much as we need to spend. Conservatively four or five times, right, Mattan?
- [33:27]
B
But yeah, and I think, and I think what's interesting there is also, like, we're used to a world where the reason you buy things is because you cannot build it yourself. Now, the reality is you can build anything. But do you really want to spend 10% of your time maintaining slack? Like, is that is the Jason Calacanis core competency that you're really good at maintaining slack? Probably not. And so now there's this laziness calculation which is like, okay, well, now you have some defense where if Slack is going to be really predatory on their pricing, now you can be like, f you, I'm going to go build it myself because you're raising the prices significantly. But if they're like just below the threshold such that you're too lazy to actually do it yourself, then they'll like, you know, stay there. Because at the end of the day, you probably don't want to be maintaining slack, I would imagine.
- [34:12]
A
It is a really interesting thing about the economy. It really speaks to free markets. In a free market, we are now going through an incredible transition, George, that I don't think any of us have seen in our lifetime, which is buy or build is actually a valid question for almost everything. You know, I live on a ranch. When you live on a ranch, one of the things I learned leaving a city and I lived in New York, Louisiana, and you know, the Bay Area, people come and take your garbage. Okay, great. When you live on a ranch and it's a mile to the front gate, how does the garbage get from your house to the front gate in a garbage can? And I asked the guys and like, well, we pick it up at the streets. You have to hire a service on the street. So there is another service that I use that I pay to take the garbage bins from the ranch house and the barn to the front of the gate. Now if that person said to me, I'm 10xing my price, you know what I'd say? Okay, I'm gonna drag them there, I'm gonna buy a pickup truck and put the garbage bins in the pickup truck, drive it to the gate myself. Cause that makes no sense. But I am not. If he doubles his price, I hope he's not listening, I'm still going to pay it because I don't want to do that. It is now really what we're getting at. And this is going to be exacerbated. And it leads to two other really important discussions that I think my guests today, for the audience are uniquely qualified to answer, which is in this really important discussion, where does the value, where does this training go? Because agents, as all of us know and the public does not, writ large, are now learning. They're now learning. And what they're learning is the value. So in George Banking, what those banks train your software to do, what you train your software to do on their behalf, that's where the value actually resides, is that knowledge and that continuous learning, which then has made me big, wind up here, start to wonder for startups and for my company, a venture firm, how do I capture that value? Maybe I need to make SLM, small language models, VSLMs, verticalized small language models. Because I don't want to give that knowledge to OpenAI, I don't want to give it to anthropic. So how do you think about capturing valage, capturing the value that your agents and your software are discovering along this journey? And do you worry about educating other large language models?
- [36:41]
D
So we have a very unique approach to how we kind of try to solve this problem. And it's by building what we call deterministic agents and to kind of grok that like just as a small thought experiment, if everyone had an open claw at a 10,000 person organization, that organization immediately goes to chaos. Like it is, it is actually like the worst case scenario if you are the CEO of that institution. And the reason for that is because, you know, everyone has their own preferences. Everyone has every, you know, their own, you know, desire to do things. Certain ways. Someone might want to spin up a replit clone, another person might want to, might want a slack clone. And the coordination and the issue of just, you know, managing those people and then managing that set of 10,000 agents on top of that ends up actually creating pure chaos in the organization. What you're getting at I think is the idea that, you know, it's more important than ever to manage agents or to actually embed in them the expertise or the firm specific way of doing things or our specific way of doing things, which is especially important as organizations scale, whether they're you know, just human, human agent, hybrid or just agent. That idea of like managing, yes, you can do that with training and with, you know, encoding your preferences in a system prompt or a set of skills and then having your openclaw or your CLAUDE or your GPT go and explore and exploit and do whatever. But increasingly what we're seeing in the world's largest institutions is that you don't even want these always on long running agents that are doing arbitrary things. Actually, quite to the contrary, what you want is a defined task, what we call a deterministic agent that goes through a 10 step or 50 step or 85 step specific process, the firm's way of doing it. And once you actually have encoded that maybe there's like skills for one piece, another piece is code or executing code, another piece is writing code. But once you get all the way through all of the 85 or 100 steps, that becomes load bearing institutional grade software that you can then scale to a thousand, to ten thousand agents, to ten thousand employees, et cetera. And that's a different layer on top. It's almost a different, as kind of Matan mentioned earlier, model agnostic, almost task agnostic layer where you're prompting with, without any sort of understanding of the overarching task, single steps one by one to get to the right institutional outcome that's encoded by our forward deployed talent in our go to market team. Again kind of per what Matan was saying is really important and really a differentiator in today's day and age.
- [39:22]
A
And as a fascinating picture you're painting George, for a number of reasons I think Matan, if we are putting together the thread that you discussed as well, which is startups and products and services companies are about this forward deployment, then that means these large language models, whether they're frontier or they're open source, are quickly becoming commoditized and we will not know the difference, we won't know which one we're using. Just like I couldn't tell you what Seagate hard drive I'm using, or if it's a commodity Taiwanese hard drive where none of us could tell you if we're using Corsair RAM or what other RAM we're using. It's all been abstracted the large language model might be the same as the memory and the storage in our computer and then all the value is created by, wait for it, the humans and the agents which are most analogous to humans and the harnesses, the skills, the memories that they actually build. Huh?
- [40:23]
B
Matan yeah, I mean I completely agree that and I think it's right because if you look at the downstream users like engineer in our case at least engineers, and I'd imagine it's similar kind of for both these gentlemen as well, is like our downstream users don't care what model it is that they are using. They care that it gets done quickly, reliably, affordably, like that is what really matters to them. It is a means to an end and when it is a means to an end, that is something that you want to have optimized. Now my sense and being very clear eyed about this, the model providers are going to very aggressively put up a fight to be as monopolistic as possible and as not amenable to model agnosticism as possible. Because on one hand you have this collection of, let's say four companies, OpenAI, Anthropic, Google and Xai that all want to have as much pricing power as possible. And then you have the users of these tools, whether it's developers, bankers, whatever the case may be, where they just want to have the optimum cost and quality and speed for their given task. Companies like us here, our job is kind of to be that intermediary that gives them leverage against the monopolistic tendencies of the model providers. If you're locked in on just anthropic, they can jack up the prices and you kind of can't do anything about it. Meanwhile, if you use a model agnostic tool and they try to jack up prices, you can be like screw you, we're just taking all of our usage to someone else. And that gives you like I think it just allows the free market to do its thing. I think there are going to be dynamics where like game theoretically then each model might try to be spiky at different types of tasks like if general performance, they're all loosely the same, then to have some edge, OpenAI might try to be better at testing and then Anthropic might try to be better at documentation or reviewing or C or Python and there's going to be this dynamic that emerges that at the end of the day really benefits businesses and allows them to avoid this lock in while getting the frontier of performance here.
- [42:27]
A
Russ if LLMs are going to be commoditized just like hard drives, bandwidth or RAM became. After two decades of the Internet and five decades of personal computers just grinding those prices down and increasing the value of those commodities, then do you believe that the value will reside in the agentic level that is continuously learning and in the application layer where founders can really provide customized, bespoke, verticalized value?
- [43:03]
C
Yeah, I think, you know, you look at it as like a stack, right, or a layer cake. And like at the bottom of it is this, this LLM, right? And I think in these early days, that's where most of the value had accrued to. But as that layer gets commoditized, you know, almost like liquid, you see the value kind of like spread all the way up and pool at the application layer. I remember maybe like a year, year and A half ago, OpenAI published these projections through 2028 or 2029 or something like that, of where they see their revenue kind of concentrating. And what you saw over time was you saw like the API layer actually shrinking a bit right in where they expect revenue to be. And then the vast majority of it was actually in ChatGPT or at the application layer. And so I think that that's kind of the dynamic that would play out in the macro environment as well. Is that a lot of the value, most of the value actually pools at that application layer, but there is significant value at every layer below in between the LLM and the application. And so for us, we play. We're kind of like all the undifferentiated deterministic infrastructure that wraps that stochastic core of the LLM. And so there's going to be significant value there. But we think that the developers of applications that build on top of us are going to see the vast majority of that value.
- [44:25]
A
And there is a breaking news story. Deep seq version 4. This is the open source Chinese model. And this really relates to the discussion we're talking about where value will reside. This frontier model includes version 4 flash version 4 pro variants. So you can go fast, you can go deep. And Here it is. DeepSeek v4Pro costs $3.48 per billion output tokens. Quad Opus 4.6 charges 25 million per output token. I think that's the best analogy here. I'm not sure what 4.7 costs. And so here's your comparison chart of this incredible battle that's going on.
- [45:12]
B
I think this is also, you know, talking about the open models is a. Is a kind of very relevant here because a lot of these enterprises, many of their use cases you don't need opus 4.6 level intelligence to go and write your documentation. Right. There are a lot of these use cases that you could get 10 times faster, 10 times cheaper with an open model and it'll perform just as well. And this is something we're seeing a lot with these largest enterprises.
- [45:37]
A
Yeah. What do you see here when you're looking at this, George? And have you started using Frontier model, open source frontier models yet? Are you having your team keep an eye on them? Are you benchmarking them and making your software headless to to see what output comes out?
- [45:53]
D
So we've got a model agnostic layer very similar to kind of exactly what we've been discussing. We train our own models for specific tasks, for kind of to refine them even more acutely to what our customers are doing. Quite often that's based off of cutting edge open source models, whether it's Deep SEQ or any of the other ones. Especially given that we are in a highly regulated regime and can't always use some of these non American open source AI models. But I think this is just all pointing to a larger trend in which these models will get all very much commoditized. I think we'll eventually see small language models, models that are running on device and that will actually change the game completely for what will end up happening where OpenAI and anthropic end up accruing revenue. I think it's actually quite likely that people will always care about having the latest and the greatest model running in the cloud, but you'll also be able to triage tasks to these smaller language models. And so you'll have this orchestration agent that is actually going and using other subsets of cheaper intelligence and maybe local intelligence as well to achieve a task. So I think the cost of all of this will go down with time and the large labs will continue to have to iterate.
- [47:19]
C
I think there's an interesting question too that I don't know the answer to, but where even assume that Anthropic or OpenAI have the best model, the absolute best model. I think what's tricky is that these kind of foundational labs are moving into the application layer more and more and so there's this tension between well, I have the best model and I want to make it available in the API to developers out there, there, but I'm also playing at the application layer and the differentiation in selling a subscription at the application layer is having the best model. And so do you continue to always provide the best model to all developers out there that maybe can build something similar and replicate the product, but maybe not have as good of a margin? Or do you hold back that absolute best model for your own application while providing maybe the second best one to all the developers out there? I don't know how that dynamic will play out.
- [48:17]
D
I think the free market will always demand that you put your almost your best market best model forward as often as you can. At the same time, I think there's a lot of marketing back and forth about Mythos and OpenAI's unreleased models and how there's security vulnerabilities. As we get closer and closer to artificial superintelligence, I do think you'll get to models that expose security flaws. I think you'll get to artificial financial superintelligence that will expose dislocations in the market or arbitrage opportunities in the market.
- [48:45]
A
Has that started to happen, George? Like, I saw somebody, I just saw on the timeline, somebody talking about they've, they've put 10k into a polymarket account and ground it up to 70k. My team will go find this. And I was like, okay, I got to double click on that. Because if that's true, you would not post that. If it was true and replicable, nobody in their right mind would post it. They would add a zero and then they would try to add another zero. And they go to somebody like me or JAMA and say, hey, would you bankroll me for this and split the returns? And I'd be like, f, yes, I would.
- [49:20]
D
That's like a drop shipping scam. Like, I think there's a million different accounts where like, you know, people are like, oh, pay for my $10 a month course and I'll teach you how to make a million dollars a month with AI. The hardest part is the expertise. It's kind of to. The larger theme of what we've been talking about today is how you get that forward deployed expertise or the, you know, the, you know, the investing specific kind of like genius that actually allows people to make money with humans and then encode that into AI. But no, I don't think that those poly market bots are real. You can double click into them all you'd like. I have seen instances where very clever investors in the private markets and then some in the public markets have been able to connect more dots over some subset of information or over more information than a human alone could have, could have seen. And they have filed proxy attacks using AI to find something that they would have missed. They have Exited positions, they have invested in assets that they otherwise would not have invested in. And so you are starting to see this AI augmented decision making becoming table stakes and changing what market really is. But I always had this thought experiment that when we get to true asi, there will be all of a sudden massive movements in the financial markets as a sign that humans have no idea what's happening. They might be conned by AI into moving money. It almost solves the human sociology problem rather than the fundamental problem which is what we're seeing solved right now.
- [50:51]
A
Super fascinating. In the poker world, online poker was basically ruined. Anybody who knows what they're doing will not play online poker now because of HUDs, heads up displays, people will pop up a HUD, they'll play across 20 tables, then they'll join the same tournaments, they have VPNs, you'll have 20 different devices joining the same tournament and hoping that two agents acting randomly will sit at the same table. And then in a PLO game you now know four other cards on the table. Massive edge, not massive as in you went from you know, 50, 50 versus an opponent to 90, 10, but a 51, even a 1 or 2% advantage over many hands means you will bankrupt the other person. And then you can also study people's like how often they voluntarily put money into the pot. Vpip I guess it's called just ruined online poker. Anybody who plays online poker, it's a sucker's game. You're playing against bots, you're paying against bots colluding with each other. But that doesn't help the online, the in person game where all of that is gone. But online trading is kind of in the public markets, the equivalent of online online poker I think, which means the game could get rigged over time. And if your horizon is I'm going to, you know, own this stock for 10 years, it doesn't affect that person. It doesn't affect a person who goes I'm long, Tesla, Uber and you know, Google. Because I believe in the self driving thesis I'm just going to own those stocks for 10 years. And yeah, what percentage I own of each might change but I'm just going to bet on the category.
- [52:29]
D
Yeah George, I think you're seeing like, almost like it's very similar to how Gen Z and Gen Alpha are post rationalist. And a lot of the humor doesn't make sense. A lot of the memes don't make sense because like you know, I think, you know, everything can be overly explained for them. I think you're starting to get to like markets which are kind of like post fundamentalist is like maybe just a fun thought experiment where it's like it doesn't really matter about like the underlying asset itself or like any of the underlying stuff. And it's more about like the meme ability potential or like you know, the value for AI to create, you know, some level of interest here. And, and maybe that is one of the defining final moats is a brand. And you know, that being one of the seven powers is just a funny thing to think about.
- [53:09]
B
Well, and I very much agree with your point about, about like if you had ASI for finance, it would not be doing any crazy complicated alpha strategies. It would just go and like sociologically scam people or like social engineer its way into a meme stock. It would figure out how do we make something the next meme stock. It would just get in early, make it a meme stock, make a ton of money, pull the rug out. Yeah, like that, that would be what it looks like as opposed to finding some crazy complicated multi levered arb. Whatever.
- [53:36]
A
It's literally back to what we just saw on the TV show Silicon Valley where it's like, oh, you want to get rid of bugs? Well, humans create, human developers are creating bugs and the software is where the bugs reside. Let's kill the human developers. Oh, we can't do that. There's no way to get out of this box. Okay, let's just delete the software. It's literally the path of least resistance is go on Reddit and X, create 100 spam accounts and then pump and dump open door or whatever companies on the floor that has some reasonable chance of becoming a story stock.
- [54:07]
D
And I think there's still some more nuance and people are still more real in that. But it's a real thought experiment which is like, hey, I have all these people spending time working on financial models and like, you know, these crazy, crazy, you know, complicated human first workflows. And it's like, well, you know, there may be other ways and this is more to the outcomes pricing example to make a lot of money with AI.
- [54:29]
B
Would you rather go up against Jane street or like people on Reddit, like probably like random Redditors and you know, convince them of something.
- [54:36]
A
So, and I asked my team to go find me this quote for the 70k poly market. They came back with like 17 different stories which mean fair warning when those stories come out, if you engage them and you take their course, you're the sucker at the poker Table. You're the person who doesn't realize when you go to Vegas and it's a Saturday and people are playing in their robes and flip flops at the table, that's the mark. That's the person who's, you know, in between the sauna and the pool and they're just going to blow 500 real quick. And the other seven people at the table play there for a living. And they're playing from the same chip stack. They're just in, they're colluding against. You don't play on Saturdays. They're playing a tournament or something. I don't know what your best bet is, but the odds are going to be against you. All right, we got a couple of topics we could end with and I'll just see who has a strong feeling on it. China has blocked Meta's Manus acquisition. This is pretty spicy on a geopolitical basis. We can start talking about that one. OpenAI, they redid their Microsoft deal so anybody can now host OpenAI's models on their infrastructure. Amazon and Andy Jassy said, oh, that's interesting. And they'll be hosting all the OpenAI models. I think that was part of the tension there. And now OpenAI is owned, I think 26, 27% by Microsoft, which is incredible. Elon Musk has 0% ownership in that trial starting this week. And then finally the future of the. Well, yeah, let's pick between those two or three. Anybody have something they feel particularly compelled by? George, you seem to be thinking this one through. What's compelling to you as we wrap here?
- [56:17]
D
I think the geopolitical. Okay, implications is obviously like very interesting. I think, you know, the China has a leg up in some ways in their ability to productionize these models at scale and their control right now of open source AI, you know, again, at scale with multiple players there. And I think it makes a lot of sense for them. It's just something that, you know, the US has to come back on. This is again too important of a, of a battle of a technology to lose control of. And it's an interesting supposition here.
- [57:01]
A
Hey, if China has the best open source models and they can deploy them at scale from a top down basis, Russ, what does that mean for the frontier space and this huge race towards superintelligence? I think we all agree we're either at AGI, around AGI, you know, closing in on AGI. So let's make it a two part question. Where are we at in achieving AGI, Russ? And then with these frontier models, who gets to AGI first at scale? Is it the frontier models? Is it the open source models coming out of China?
- [57:36]
C
It depends on the definition, right? I think everybody has a bit of a different definition. My definition in particular is like, we're trying to build a human in silicon. I think that's my definition, right? Can we create ourselves in software? And I think if that's a definition, then I think we're still quite a ways away. Humans can do a lot of pretty amazing things and humans have emotions and things that still kind of uniquely define us, that these models even architecturally, aren't really set up to mimic or approximate yet. And I think there are other parts of the AGI definition around continuous learning and things like that. If you've listened to stuff that Ilya talks about, he speaks these kinds of things as well. And so I think it depends on the AGI definition. Now, I think moving forward, though, in terms of the race between China and, and the U.S. i was listening to the Jensen podcast with Dorkash a few days ago and I thought a really interesting thing that he talked about was kind of while we have better chips in the us, China has an advantage on the energy side. They have tons and tons of energy. And so you can run more, worse chips, but at a larger scale because you have a bigger kind of energy supply and you can effectively nullify the advantage, a chip advantage. But then the other thing that Jensen said, he talked about Moore's Law, but he talked about how if you look at the improvement of these models, there's a 50x improvement, and it's largely attributed in their performance, largely attributed to algorithms and computer science. That's the main thing he talks about. And so I think that the race between China and the U.S. if you believe in kind of like Jensen's philosophy of the breakdown and where the biggest gains come from in terms of progress, it's not at the chip layer, maybe not even at the energy layer. Those are two important components for scaling all of this up. But a lot of it comes at the algorithmic layer. And so I think that the question to be answered, and I don't think I'm close enough to, to it, maybe as the other gentleman here, but I think the question to be answered is like, are we winning that race? Are we winning the race in terms of algorithms and architectures for these models and the design of these models versus China?
- [60:25]
A
Your thoughts, Matan?
- [60:27]
B
One thing that's kind of, as someone who's very patriotic that's frustrating is that because they are so constrained, it has bred a lot of these algorithmic innovations because they have no alternative. Meanwhile, we kind of get the privilege of like, oh, let's just throw more GPUs. Like, let's just, you know, it's kind of a lazy, easy solution. And it's very tempting when you don't have the constraints to just, you know, go for that and just go for scale, like more GPUs, whatever, the best GPUs. But I think we should really, you know, as Americans, like the United States, we should be disappointed that we have fallen so far behind in open source. I think it's, like, pretty embarrassing. And I think, in fact this, as it relates to the Manus acquisition, this also shows that we are in the middle of a very serious talent war. Because it's not like the tech from Manus was that incredible. It's really the people there that I think is the thing that's being kind of protected. And I think it's maybe a wake up call and we should have a much more burning fire under our asses to make sure that we are at the frontier, not only of the big models that use all the GPUs, but also the resource constrained models that are smaller. And making sure we have the algorithmic innovations, making sure. It's not like an unwise bet to spend your career doing that in the United States, which right now it kind of is. There's not nearly as much funding to do things like that. I don't know. To me, it gets me very frustrated.
- [61:53]
C
Yeah. Jensen said in that podcast as well that majority of the best AI researchers in the world are actually Chinese. And then like in China proper is like 50% of all AI researchers overall. And so they have like a huge advantage on the talent side.
- [62:12]
D
I mean, one, maybe a final thought. I don't know. I think the greatest technologies, the greatest algorithm improvements have always been built in America, and they are American. And I think China is very good at copying those things. But if you look at the actual, like the best researchers and the best talent, it is all American. I think that's. I think there's maybe like something cultural about that. Maybe it's because of a superfluous amount of GPUs, but maybe I'm more bullish on American AI supremacy. I just think we just have to stop it from getting copied.
- [62:43]
B
I'm bullish, I'm bullish. But we got to hold ourselves accountable. The fact that Deepseek keeps releasing these models that are smaller and really Good. And yes, of course, they're distilling and they're breaking some of those rules, but I'm very bullish. I just want to hold ourselves to a really high standard so that you can't even say, oh, but they're better at this, or that we should be across the board dominated.
- [63:05]
C
You want to win by a mile, not an inch.
- [63:07]
A
Yeah. And there's no reason to not win the open source race, except that capitalism is such a strong driver that if you're a developer who says, you know what, I believe in open source philosophically and I would prefer to work on that, but Meta just made me and Claude just got into a match and I've got a 10, $20 million RSU package here. I'm not working on open source for the next four years. I'm working on a proprietary model. And even Meta, which was behind in doing llama open source, they just moved to a proprietary model for a reason. They need to give massive grants to those folks. So there is something happening here where capitalism is racing towards the cliff.
- [63:56]
C
Perhaps.
- [63:57]
A
And I guess this could be our final thought is can OpenAI's spending and listen, I understand they're a partner us, but this is a discussion that the CFO is having publicly with the CEO, apparently at conferences. Spending hundreds of billions of dollars on infrastructure is very different than spending tens of billions. There's a very easy way to spend billions to tens of billions of dollars. If you start spending hundreds of billions of dollars and you make those kind of commitments like OpenAI has with Oracle, people are now starting to wonder, hey, the revenue's not growing at the pace the spend is growing. Is this the, as we would say in poker, a risk of ruin? If you put your entire net worth on the table and you're betting big pots, even if you play perfectly, you could get unlucky in a hand and your aces get cracked, and now you don't have the ability to rebuy in. So, Matan, you seem to have a perspective here. Give me your perspective on this race to the cliff or putting your whole net worth on the table and playing high stakes.
- [65:02]
B
I think it's just the name of the game for this era of software where everyone is going through hypergrowth. And right now we're negotiating some compute deals with some of the hyperscalers and some of the, you know, kind of Neo clouds. The thing that's really difficult is we have to make a call now about what we want at the end of 2027. And like, for us, we have been doubling every month. Like that is really, really difficult to make that commitment like 16 months in advance because the variance is so, so significantly high when you're in such hypergrowth. And that's for us, whatever. Our scale for now is much smaller than that of OpenAI. They have similar trajectories and so the error bars there are pretty massive. And to your point, if you overshoot you might be literally out of business. If you undershoot, you might be in a situation that anthropic is now in where now they seem like idiots because they didn't make the commitments up front. So it's a very, very risky game. I don't know what the, like when you're dealing with these year long supply chains or multi year supply chains, I'm not sure what the optimal way to
- [66:10]
A
well, and Crusoe Cloud reasonably says, hey we got a six year lifespan on this device and on this hardware profile. Will you commit to four years, five years and then how much do you want? Because then they've got to go to Jensen and then Jensen says okay, well it's first come first serve here. So who's going to give me the money first or put the deposit down first, gets delivery first and this is incredibly high stakes. Russ.
- [66:35]
C
Yeah, yeah, it's crazy. I'm glad that I'm not having to make these commitments like you are. Mason.
- [66:41]
A
It's pretty nuts. George, any final thoughts here on the high fakes game? Luckily you are on the application layer. You don't need I think to build your own colossus to provide massive value. You just need to rent some space reasonably from aws, Azure, Google Cloud, et cetera. You're not standing up your own hardware. I would assume soon.
- [67:03]
D
No, but I do think that from an interesting thought experiment, most of the pricing of every piece of software or every type of knowledge work into the future seems to be kind of moving towards renting GPU hours. So you know, that seems like the fundamental unit of what people will be spending money on. And you know you can underwrite things in GPU hours or the US dollar in 10 years, but it seems like those are, those are like the fundamental kind of units of currency.
- [67:35]
A
Let alone if the new CEO of Apple, who I'm ready to fall in love with, I'll just put it out there right now because I was not in, I wasn't even in like with Tim Cook because just not enough innovation for me. But I kind of feel like this new CEO, as an engineer, as somebody who worked in hardware, if he drops M5 Mac Studios on everybody's head with a terabyte of RAM and you know, M6 and M7 come out on a cadence and everybody's laptop all of a sudden has 128 gigs of RAM instead of 16 or 32. We're going to see an amount of compute flood the market, not just in the data centers, but on the desktop. And what does it mean if there were a billion Mac studios in the world with all this extra capacity? I think we're going to flood the market with compute, with memory, just like we did in the days of the Internet with storage and bandwidth. We overbuilt, we flooded the market and then let alone Elon going to space and making unlimited data sets, data centers in a permissionless way. And if he announced his own fab, what's to stop him from just every day putting up a starship with X amount of gigawatts of compute? There's no constraint if it's in space, there's no constraint. It's just how fast I can make these satellites and orbit them. This is a brave new world.
- [69:05]
C
Yeah, I think the tricky thing is you need the applications, right? You need to justify the capex with utilization. And so, you know, what is the question becomes like what are you going to drive? What is going to drive utilization of it?
- [69:21]
B
I think a nice backstop there that maybe I harp on too much, but it's like there are so many problems latent in society, whether it's software solvable problems or healthcare or just industrial. I think we only get to that kind of concern of what are the applications if we suddenly feel like we have no problems to be solving. And my sense is that generally the free market allocates resources accordingly to whatever are the most kind of pernicious problems of the time. So I'm more like even Jason describing that to me. It puts a smile on my face thinking about all the things that we've probably assumed are just realities of the world. We don't even think of them as problems. But there are so many things that with all of this computer and with all this innovation that we're going to be able to solve, which I think is just. It is like I always think about, I don't know if you guys, you know, growing up I would always see the memes that are like born too late to explore, you know, the oceans or born too early to explore. I think this is the best time. This is by far the exact moment that is incredible to be around because all of these Problems are like being solved at such an exponential rate.
- [70:32]
A
That's a great moment to end on. Stay positive. I think the P doom here is like, if we were to P doom this panel, I'm at like P doom. I'm like 5. I'm like 5% chance this goes awry. I think it's like 95% chance life's going to be awesome. Anybody else want to share their P doom after this epic episode of this week in AI? Where's your P doom?
- [70:57]
B
My P doom is zero because it is in our hands and we will not allow it to happen.
- [71:01]
A
Yeah, Russ, you got a. George, you got a P doom. I'm assuming you understand the concept of like, what percentage doom you are. I believe that's a simple way to think about it.
- [71:11]
B
Yeah.
- [71:11]
C
I think at the current state of things. Yeah, I'm like five or less.
- [71:15]
A
Love it, George.
- [71:17]
D
I'll put it there at 10. I'll still be the highest. I hope it's in our hands. I think these models are more powerful and deserve empathy. But yeah, I still am a techno optimist and I'm still on acceleration.
- [71:29]
A
If you're at 10, by the way, that's the lowest score across anthropic this entire organization. I think literally the guy who works the dishwasher in the cafeterias, he's at PDUM 12% just by overhearing these conversations constantly. He's just like, yeah, I don't know
- [71:45]
D
how this is going to go for
- [71:46]
A
me as a dishwasher here or working in the kitchen. I think I'm pdoom 12. All right, everybody, Another amazing episode. Russ, where can people learn more and who are you hiring for?
- [71:57]
C
Hiring across the board. Go to market engineering, pretty much everything as we start to scale up a bit. You can find out more on X X.com LiveKit and also on GitHub. GitHub.com LiveKit okay, Matan, how can people
- [72:15]
A
find out more about your great company and who you're hiring for?
- [72:18]
B
Email me@matanfactory. AI killers. Only people who are looking to transform software development. High agency. Let's get it.
- [72:27]
A
High agency. Get some. George.
- [72:29]
D
High agency is important. I'll add running hard. We're hiring a lot of forward deployed investors and bankers. Like an insane amount. And I think it will be the most important role in the next 10 years. So we're here to save you from financial effort. FD FTBs and FDIs FTBs. It has a nice ring to it. But also engineers, sales leaders, anyone who is. Where can they find more than puts it a killer and they can find more. Just email me georgeebia AI or on Hebbia AI we have a careers page. Thanks, guys.
- [73:02]
A
All right, there you go, folks. And I'm hiring six new people for our venture capital training program. This is if you ever aspire to break into venture capital. You don't have to have gone to HBS or Stanford gsb. You can just email researchersaunch Co and we're gonna hire six. And it's a one year training program and if you make it to year two, you get a big salary bump and you get to work with me evaluating over 10,000 startups that apply for funding every year and try to figure out which hundred we're going to invest in. It's more like maybe 15 or 20,000 at this point. Go ahead and subscribe this week in AI. AI has all the links and we'll see you next time. Bye bye.