All-In with Chamath, Jason, Sacks & Friedberg
GPT-4o launches, Glue demo, Ohalo breakthrough, Druck's Argentina bet, did Google kill Perplexity?
Fri, 17 May 2024
(0:00) Bestie Intros: Recapping Phil Hellmuth's birthday weekend (7:38) OpenAI launches GPT-4o: better, faster, cheaper (29:40) Sacks demos Glue: How AI unlocked his Slack killer (40:12) Friedberg walks through his major breakthrough at Ohalo (1:01:35) Stanley Druckenmiller bets on Argentina and Javier Milei: strategy, roadmap for the US (1:13:54) Jason's bet on Athena, how AI will change company building (1:22:21) Google launches AI summaries in search Follow the besties: https://twitter.com/chamath https://twitter.com/Jason https://twitter.com/DavidSacks https://twitter.com/friedberg Follow on X: https://twitter.com/theallinpod Follow on Instagram: https://www.instagram.com/theallinpod Follow on TikTok: https://www.tiktok.com/@all_in_tok Follow on LinkedIn: https://www.linkedin.com/company/allinpod Intro Music Credit: https://rb.gy/tppkzl https://twitter.com/yung_spielburg Intro Video Credit: https://twitter.com/TheZachEffect Referenced in the show: https://www.pokernews.com/news/2024/05/phil-hellmuth-60th-birthday-bash-las-vegas-45984.htm https://x.com/OpenAI/status/1790130703721521305 https://x.com/OpenAI/status/1790089513387143469 https://crfm.stanford.edu/helm/mmlu/v1.3.0/# https://www.businessinsider.com/inflection-implosion-chatgpt-stall-ai-consumer-chatbot-problem-2024-3 https://www.opencompute.org https://glue.ai https://ragie.ai https://ohalo.com https://www.cnbc.com/2024/05/07/cnbc-exclusive-cnbc-transcript-billionaire-investor-stanley-druckenmiller-speaks-with-cnbcs-squawk-box-today.html https://www.forbes.com/sites/theapothecary/2024/03/23/summers-inflation-reached-18-in-2022-using-the-governments-previous-formula https://twitter.com/DavidSacks/status/1378034932006592512 https://www.athenawow.com/jcal https://calacanis.substack.com/p/startup-productivity-in-the-age-of https://blog.google/products/search/generative-ai-google-search-may-2024 https://www.youtube.com/watch?v=XEzRZ35urlk https://finance.yahoo.com/quote/GOOG https://startuptalky.com/apple-failed-products
All right, everybody, welcome to your favorite podcast in the world's number one podcast, the all-in podcast. It's episode 1,790. Oh, wait, that's just how it feels. Welcome to episode 179. With me today, of course, is your sultan of science. I don't know if that's a movie background or it's just his favorite vegetables. What's going on there? What's the crop?
That's AI-generated.
AI generated crop. Okay.
I'm trying AI backgrounds. I'm going to try it out for a while with different crops.
Your fans are going to be crushed that you're not doing deep movie polls with us. Of course, man about town, DC, new products being launched. David Sachs, the rain man. Yeah. How you doing, buddy?
Good. Good. Yeah. Good week. Lots going on.
Yeah. Yeah, definitely a good week. Chamath Palihapitiya, chairman, dictator. He puts the chairman in dictator.
I would like to take this opportunity to wish... My child, happy birthday. I absolutely f***ing love you.
Well, now the rest of us look like s***.
Yeah, great.
I've never done that before. Sax, in your desk, in your desk is a piece of paper with your children's names and their birthdays. You want to pull it out and see? I got three birthdays a year and I've never done one. Let your winners ride.
Rain Man, David Saxon.
Love you guys. Bye.
No, no, no, but I'm saying it rarely lands on the same day. Today is the day.
Today is the day. Today is the day. Okay, good.
Today is the day. Congratulations, child. Oh, congratulations. How old, Chamath? No gender name or any other specifications, folks. We can't tip anybody off. No pronouns. We can't tip anybody off. No pronouns. Yes, absolutely. So how are they experiencing their birthday? This child has experienced a wonderful life, and this child is an incredible person.
for whom I have tremendous admiration and love and compassion and hope for the future. All right. And did you order them some chicken fingers? I cannot comment on who this person is. Chicken fingers. Are you talking, of course, about Phil Hellmuth? Your child, Phil Hellmuth. Can we please talk about last weekend's festivities in Vegas? What a disaster he is. Oh, my God.
Guys, just so you guys know.
We missed you last weekend. We missed you last weekend.
So much fun to make.
We missed you on Saturday night. Saturday night was really fun.
I had such a lovely time coming home, to be totally honest with you.
We had a cabana set up on Saturday, played blackjack by the pool. I missed you guys too. I had a FOMO when I saw the videos. It was so fun.
Well, you don't have to have too much FOMO because Phil sent the entire group chat to pokernews.com. They did an article on the group chat. Run it twice. Theflop.org.
poker dash update oh my god yeah well he kept tweeting and he leaked every single person who's there and the jets and the jet numbers he's like look here's me and elon elon came by for my dinner no no no it was worse than that no it's worth that he said i got to hang out with our guy elon for 10 minutes and 14 seconds he intercepted him at the valet
Wait, what?
How many minutes? 10 minutes and 14 seconds. He had the exact time, down to the second. Oh, my God.
Well, listen, I want to wish Phil Helmuth a happy birthday because I did miss his 60th party. Yeah, it's coming up, actually.
The good news is it wasn't actually his birthday. It was Bill Gurley's, so he just hijacked Bill Gurley's birthday.
I also got to enjoy for my first time ever the experience of Baccarat, which I've decided is the most degen game on earth. It's literally the most, you just flip a coin. More than craps.
It's flipping coins.
Well, craps is really, there's a style. Craps, you make betting decisions. All you do in Baccarat is you say bank or player, and then you freak yourself out about how you flip the cards. And the smartest people I know on earth are all sitting around this table at two or three in the morning saying, turn this corner this way. No, no, no, no, no. Turn it this way. Turn it this way. There's two dots.
And they're debating the right way to flip a card over. That's true.
No, the Baccarat sweat is the most incredible performative act ever. In the casino.
It's the weirdest thing. Yeah, you're right. Everyone's got their own little technique about how they bend the card. It's all destroyed by the end of the deck. They get thrown up.
I go lengthwise. I go like this and I try to see.
Oh, like you're curling your mustache like an evil villain? It's the evil villain. And then you call it, oh my God, no spotter. If you see it, no spotter.
We're two across. So great. And then you get to decide whether the bank turns over their cards and when they turn them over.
Then you lose a small house and then you're like, oh, let's try it again.
Yeah, you're convincing yourself that you have all this control and ways to change the outcome. You're literally flipping a card. It's high card. It's basically a high card. That's all it is.
It's a high card. It's even worse than that. You're basically sitting down at the casino's table and then they tell you whether you've won or lost. And in order to convince yourself that that's not what's going on, you have to play with the card. But really, they just tell you you either win or lose. Let me tell you a story.
And I'm watching the smartest guys we know staring at the window at the little machine that tells you whether bank or player won. And they're studying it, rubbing their chin, doing an analysis. It's been four reds. It's got to go black. Hellview's like, I'm calling it now. Bank, bank, player, player, player. And all the guys are like, let's do it.
And then everyone's like, heads, heads, tails, tails.
So Helmuth asked us to play in the high stakes poker game on PokerGo. So it was me, Helmuth, Stanley, Sammy, House, and then Jen Tilly and Nick Airball and Roble. So most of the guys from the House game plus Jen Tilly and Nick Airball. Jennifer Tilly is amazing. What a great human and actress. Well, listen to this hand. literally the second hand of the actual poker game. Gentilly
is in the big blind. No, sorry, she's under the gun. She raises. Hausenbold three bets. It comes all the way around to me on the button. I look and I have pocket kings. I ship the whole cheeseburger. Comes back to Tilly. She ships. House ships. Listen to these hands. Jen Tilly has aces. Jeff Hausenbold has kings. I have kings. Oh, my God. I've never seen a cooler hand like this in my life.
We'll all be out. In the second hand of the game. Anyways. Wow. Don't worry, guys.
I came back and I won.
Chantilly. She tripled up.
She triples up and then into lockdown mode. The first time I ever played with her, I showed up to this game.
Then I stacked her right. Anyways, I don't want to reveal the game, but it was wonderful, this one.
I show up at a mutual friend of ours game, and there's a beautiful Porsche or something in the driveway. It's a really notable car. And I notice on the license plate, it says D-Gen. but it's spelled with a J. And I'm like, oh, degenerate. What a great license plate. I wonder whose that is. I go, it's Jennifer Tilling. She is so cool.
She's very charming. Great actress. Very charming. Great actress.
What was the movie she was in? Bound. Bound. That's what it was. Yeah.
You don't have to ask me twice. Yeah, exactly.
That is a tour de force.
What a great gangster film. Yeah. With Gina Gershwin. Gina Gershon. That film.
Oh, my God.
That film. Oh, my God. Well, let's not get canceled here. It is quite a film. All right. Speaking of action. Big week. The AI industrial complex is dominating. Our docket here, apologies to Biden, Ukraine, and Nikki Haley, but we got to go AI right now. Open AI, launch chat, GPT 4.0. Monday, three days after Samwise came on all in as a programming note, and we'll go to Freiburg about this.
We probably made a bit of a strategical or tactical error in not postponing his appearance. In fairness, Freeberg, Samwise did tell us. Originally, he was coming on to talk about those things, but then it got pushed back. Anything you want to add to that as a programming note? Because people are wondering.
What happened, I've been talking with Sam for a while. A year. About coming on the show. And every time I see him, we're like, hey, you should come on the show. He's like, I want to come on the show. Okay, let's find a date. We never got a date that worked. I saw him in March and he said, hey, I want to come on the show. I said, okay, well, come on. Let me know when it works.
And a couple of weeks later, he's like, what about this date in May? And I'm like, yeah, that's fine. We can make that work. He's like, well, I've got a big announcement we're going to be doing. And I was like, perfect. Come on the show. That sounds great.
And then the night before, he asked me, he told me, he texted me, he's like, hey, we're actually not going to have this announcement happen tomorrow. It's going to be delayed. He didn't tell me how long. And I'm like, well, is it GPT-5? He's like, no, it's not GPT-5. And I was like, okay, well, you know, come on the show anyway.
Because he didn't tell me when he's doing the announcement or when it's being pushed to. So it didn't seem like that big a deal. And I thought we were just going to be able to have a good chat anyway. So it's really unfortunate, I think, the fact that the announcement happened two days after and he had to stay quiet about it during our interview. But that's the story.
I think in the future, if someone says they've got a big announcement to do, we should probably push them if they have to delay or something like that. Don't beat yourself up. But I don't think we're going to be doing a lot of these interviews anyway. I think people clearly don't love them and it's better for us to just kind of hang out and talk. Yeah.
I think if we had just gotten Sam on the day after the launch of GPT-4 Omni, as opposed to, what is it, three days before, he could have talked much more freely about it and it would have been interesting.
It was supposed to happen same day, so it's unfortunate this all worked out this way.
The other little trick is to say you can tell us under embargo. But my understanding is they were still doing the videos over the weekend. So I think those videos and stuff, they were still figuring them out. And so yeah, lesson learned. In terms of the interviews on the show, just to recap for people, we've done a dozen. Half of them have been presidential candidates.
Sometimes they break out, sometimes they don't. But we follow our interest and our passion here on the pod, it's got to be interesting for us too. So if we think this person's going to be interesting, we do it. And yeah, we understand you miss a news subject, but yeah, it is what it is.
And to your point, a lot of the people that come on and increasingly a lot of people ask to come on because they know we're not journalists. And so for all of those folks that expect us to be journalists. That's not what we are. We're for entrepreneurs. We're for business people. We're for friends. We're for technologists. We're for curious people. We're for poker players.
But we're not for journalists. And so we're going to ask whatever we feel like asking. Sometimes those things will touch a chord because it's what you wanted to have asked. And sometimes we won't go to a place whether we didn't have time to or whether we forgot or whether we chose not to. And I think it's important to have that disclaimer. Like we have day jobs and this is what we do
To coalesce a bunch of information in the way that we're thinking about the world. So we are not journalists. So please don't have that expectation.
I think what that means is that if the guest doesn't want to talk about something, we're not going to start peppering him with gotcha questions and things like that. I appeared at a conference a couple of days ago to promote glue, which we'll get to. And the first half of the conversation was like a normal conversation about what we were launching.
And then the second half was basically the reporter peppering me with fastball questions, which is fine. I knew what I was signing up for. It's a totally different style. It's a totally different style than coming on the pod and just having a normal conversation. But it's not really our job to make somebody open up if they don't want to talk.
What was the spiciest question, Sax? What was the fastball? Anything come close to your head? Yeah.
No, I mean, it's not worth really getting into. You can watch it. Yeah, I was just curious. Look, I kind of like sometimes when reporters pitch me fastballs because, yeah, you can strike out or you can hit it out of the park when they do that.
That's an important part here. I think, you know, as a former editor-in-chief journalist myself, I sometimes like to ask, I would say, a challenging question in a respectful way. I did that, for example, with Vake, you know, just clarifying his thoughts on trans and gay rights. wasn't disrespectful, was thoughtful, Would you consider it spicy or hardcore? I don't think it was hardcore.
He likes to talk about it.
No, but that's because you asked it from a position of curiosity. You weren't trying to catch the guy.
No. You see the difference? I'm actually interested in his opinion.
This is my point. That's why it comes out differently. And that's why I think people enjoy these conversations. And sometimes we don't get to the other kind of answer because I'm not interested in trying to gotcha somebody that's working hard.
I always have the same conditions when I do interviews, which is I don't clear questions and I don't let people edit it. But, you know. Everybody's got a different view on how to do interviews and feel a difference. If you like it, you like it. If you like Lex Friedman's version or Tim Ferriss's version, or you prefer, you know, Fox or CNN, go watch those interviews there.
You can have a whole range of different interviews and interview styles available to you in the media landscape. We are but one. Sam Weiss mentioned on the pod last week that the next big model might not be called GPT-5. So on Monday, they launched GPT-4.0. The O stands for Omni. It's everything you love about tech. It's faster, it's cheaper, it's better.
But from my perspective, the real show was the massive amount of progress they made on the UI UX. The O stands for Omni, as in Omnivore. It takes in audio, text, images, even your desktop and video from your camera to inform what it's doing. We can consider it like 360 degree AI.
Producer Nick will show a couple of videos while I describe them here before we go to the besties for the reaction to the announcement.
first, they made great progress in solving the CB problem we mentioned last week, that's where like when you use Siri or any of these tools, you say, you know, hey, jet GPT, what's two plus two over and you have to wait and then if you talk over each other, it breaks. They now have that working much smoother. They did an example of counting where they said speed up, slow down.
They did a translator that worked really well.
I would like you to function as a translator. I have a friend here who only speaks Italian and I only speak English. And every time you hear English, I want you to translate it to Italian. And if you hear Italian, I want you to translate it back to English. Is that good? Perfect.
Mark, I wonder if whales could talk, what would they say?
Mike, she wonders if whales could talk, what would they tell us?
They might ask, how do we solve linear equations?
Certainly, yes.
Great.
Looks like it works. I think Duolingo stock took a hit during that. Most impressive, to me at least, and also I had Sandeep Madra on my other podcast and we talked about it, was their desktop and iOS app. What this app does is fascinating. It watches your desktop on your Macintosh or on your iPad or iPhone, and it does everything. things like adaptive learning.
Here's a clip of Sal Khan from Khan Academy and his son basically using a drawing app to do some arithmetic or geometry. And it does adaptive learning and basically makes a personal coach because the app is watching his son try to solve a problem.
I'm here with my son, and I'd love you to tutor him on this math problem, but don't give him the answer. You can ask questions and nudge him in the right direction, but I really want to make sure he understands it himself. And he's here in the room, so you can talk to him directly.
Can you first identify which sides of the triangle are the opposite, adjacent, and hypotenuse relative to angle alpha?
All right, so I'm pretty sure this is the angle alpha right here, right?
Correct. Now, looking at the triangle, which side do you think is the hypotenuse? Remember the hypotenuse?
I'm not totally sure. I think it might be this one, but I really am not sure. This side AC?
You're close. Actually, side AC is called the adjacent side to the angle alpha. The hypotenuse is the longest side of a right triangle and is directly opposite the right angle. Can you find which one is the hypotenuse?
Oh, okay, I see. So I think the hypotenuse is this really long side from A to B. Would that be correct?
Exactly. Well done.
It can also participate in Zoom calls, explain charts, all that great stuff. And so it's going to be your guide on the side. It's going to be a present personality while you're using your apps. It's really impressive, I have to say. So I guess let's start, Freeberg, with your takeaways on all of these innovations that we saw.
I think it's become quite apparent that there's an evolution underway in model architecture. And I think you may remember, we talked about this briefly with Sam last week, but we're moving away from these very big, bulky models that are released every couple of months or quarters and cost a lot of money to rebuild every time they get re-released towards a system of models.
So this multimodal system basically leverages several models at once that work together or that are linked together to respond to the inputs and to provide some generative output. And that those individual models themselves can be continuously tuned and or continuously updated. So rather than have, you know, hey, there's this big new release that just happened.
This new model just got trained, cost $10 million to train it. It's been pushed. These models can be upgraded with tuning, with upgrade features, and then linked together with other new smaller models that are perhaps specialized for specific tasks like doing mathematics or rendering an image or rendering a movie.
And so I think what we're going to see is soon more of an obfuscation of the individual models and more of this general service type approach where the updates are happening in a more continuous fashion. I think this is the first step of OpenAI taking that architectural approach with GPT-4.0. And what's behind the curtains, we don't know. We don't know how many models are there.
We don't know how frequently they're being changed, whether they're being changed through actually upgrading the parameters or whether they're being fine tuned. And so this seems to be pretty obvious. If you look at this link, one of the criticisms that initially came out when they released GPT-4.0 was that there was some performance degradation.
And Stanford actually runs this massive multitask language understanding assessment. And they publish it, I think, daily or pretty frequently on how all the models perform. And you can see the scorecard here that GPT-4.0 actually outperforms GPT-4.
And so this goes counter to some of the narrative that in order to get some of the performance improvements and speed improvements they got in 4.0, that they actually made the model worse. And it seems actually the opposite is true, that the model's gotten slightly better. It still underperforms Cloud3.
Opus, which you can see here ranks top of these charts, but there's lots of different charts, all the companies published on charts, they all claim that they're better than everyone else. But I like Stanford because it's independent.
Chamath, any thoughts after seeing it and in combination with our interview, do you think ChatGPT is running away with the consumer experience? Or do you think this is like neck and neck with some of the other players?
Not to tell tales out of school, but somebody that we all know in our group chat posted something about the fact that the consumer growth had stalled. I don't know how they knew that, that maybe they got some data or maybe they're an investor. You guys know what I'm talking about. And they said that they're trying to reinvigorate growth into the consumer app at OpenAI.
Any insights as to why it might be plateauing in your perspective?
I wrote this in my annual letter, but there are these phases of growth. And when you look at social networks as a perfect example, Friendster was magical when it was first created, right? And then you had MySpace that just ran circles around them because Friendster didn't really invest the money and the quality that it took to create a moat. And then MySpace really wasn't able to compete.
So we were, you know, Facebook, we were the eighth or ninth when we showed up on the scene and we ran circles around everybody. I think what it means is that there are these phases of product development which exist in many markets. This market, I think, is going through the same thing.
And right now, we're in the first, what I would call primordial ooze phase, which is everybody's kind of like running around like a chicken with their heads cut off. There's all these core basic capabilities that are still so magical when you see them. But we all know that five and 10 years from now, these things will be table stakes, right? And what Freeberg just showed is a table that
of many companies and many trillions of market cap, all effectively running to the same destination. So I think where we are is probably within two years of where the basic building blocks are standardized. And then I think the real businesses get built. So I will maintain my perspective here, which is the quote unquote Facebook of AI has yet to be created.
Okay, and here it is, ChatGPT web visits, as you can see, have plateaued. This data is similar web. I would agree with you, Jamath. It seems like the use cases and the looky-loos who were just trying the software, because they heard about it, They've gone away, and then we have to find actual use cases. Sacks, I'm wondering.
But our friend Jason, just to kind of complete the thought, said something about the premium conversion, right? That's what he said. I don't know how he knows that.
Yes, paid, paid version. Well, yeah, so to be clear, paid versus free. And then what Sam said on the podcast last week was, it seems like whenever they come out with something new, the old stuff becomes free. In my talk with Sonny this week, he mentioned that these new models are so much more efficient that you actually can throw the old model in the garbage. Garbage. Because it's so inefficient.
And these are now becoming about 90% cheaper every year, which means every two years, these things are gonna be 99% cheaper and better. Yep, yep. And it might be that OpenAI SACS on a strategic level is going to make all this free or close to free and maybe just charge for multiplayer version. That seems to be where it's heading. You don't have to log in to use 3.5. You don't have to log in
to use Google service? No, you do have to log in still on Google services, but I think these are going to just be free. So on a product basis, what are your thoughts? And then maybe you could talk about free to pay. Do you think everybody in the world is going to pay 20, 30, 40 bucks, 500 a year, 200 a year to have one of these? Or are they just going to all be free?
Well, I think you're assuming there that the long-term business model of OpenAI is in B2C subscriptions. And I think that's probably the least attractive business model they have available to them. It's sort of the first one and the most obvious one because they put out ChatGPT and then it's pretty easy just to roll out a premium version.
But in my experience, B2C subscriptions, it's just not a very attractive business model because consumers just aren't willing to pay a lot and they have high churn rates. And there's no possibility of expansion, really. So I suspect they're going to move in more of a B2B direction over time because that's where the real money is.
And probably the way they do that is by monetizing all the apps that are built on top of it. And I think that in that sense, GPT-4.0 is a really important innovation. By the way, the O stands for Omni, which I think stands for Omnichannel. I think you may have said Omnivore, which is kind of funny too.
It's Omni, which means all the different media types currently coming in, right? That's the difference. It's not like you just give it an image or give it a video. It's absorbing all those at the same time in parallel, I believe.
That's right. So there's three big innovations with this model, right? So one is Omni-channel, which means text, audio, video, and images. Second, it's more conversational, like it understands the tone of people talking and understands sort of sentiment in a way it didn't before.
And then the third thing, which is really important, is that it's just much faster and more performant than the previous version, GPT-4 Turbo. In the speed test, they say it's twice as fast. We've played with it at Glue. We can talk about that in a minute. And it feels 10 times as fast. It is much faster.
But it's the combination of all three of these things that really makes some magical experiences possible. Because when you increase the speed of processing, you can now actually have conversations with it in a much more natural way. Before... The models were just too slow, so there'd be a long delay after every prompt.
So now, like you showed, it can do things like you point the camera at a blackboard or something with math equations on it, and it can walk you through how to solve that problem. Or two people can be talking and it does real-time translation. There's that old saying that every Star Trek technology eventually becomes true.
They've just basically invented the whole natural language, real-time universal translator. Yeah. So anyway, so those are some interesting use cases. But I just think they're going to be able to unleash a whole lot of new applications. And if they're metering the usage of the models and providing the best dev tools, I think there is a business model there.
This thing is moving so fast. They're in like Leonardo DiCaprio mode. Every two years, they throw the old model away. Okay. Let's keep... Thank you. Sacks.
Is this thing on? Did you write that ahead of time? I was in the moment.
That was good. Just one point on that is there are a whole bunch of startups out there that were creating virtual customer support agents. And they've been spending the last couple of years working on trying to make those... agents more conversational, quicker, more responsive, I think their product roadmaps just became obsolete.
Now, that's not to say there isn't more work for them to do in workflow in terms of integrating the AI with customer support tools and doing that last mile of customizing the model for the vertical specific problems of customer support. But my guess is that Hundreds of millions of dollars of R&D just went out the window.
And probably this is the best time to be creating a customer support agent company. If you're doing it two years ago, five years ago, your work has just been obsoleted.
Well, I mean, that is the thing of this pace. You used to have to throw away client-server stuff or... whatever, you had a web-based thing, you get an app out, you throw away some of the old code. But this is like every 18 months, your work has been replaced.
If you're an app developer, the key thing to understand is where does model innovation end and your innovation begin? Because if you get that wrong, you'll end up doing a bunch of stuff that the model will just obsolete in a few months.
I think you're totally right. I think that's such a really important observation. That's why I think the incentive for these folks is going to be to push this stuff into the open source. Because if you solve a problem that's operationally necessary for your business, but it isn't the core part of your business,
What incentive do you have to really keep investing in this for the next five and 10 years to improve it? You're much better off like Klarna, for example, right? We talked about the amazing improvement and savings that Klarna had by improving customer support. Release it in the open source, guys. Let the rest of the community take it over so that it's available to everybody else.
Otherwise, you're going to be stuck supporting it. And then if and when you ever wanted to switch out a model... GPT-4.0, 4 to 4.0 to Claude to Lama, it's going to be near impossible and it's going to be costly. So I also think, Sax, the incentive to just push towards open source in this market, if you will, is so much more meaningful than any other market. Yeah.
I mean, listen, you were there when I think you were there at Facebook when they did the open compute project and they just were like, sorry, guys, talk about talk about torching an entire market.
Explain what it is. So there was this moment where when you were trying to build data centers, you'd have these like one you rack mounted kind of like machines that you used. And what Facebook observed was there was only a handful of companies that provided it. And so it was unnecessarily expensive.
And so Facebook just designed their own and then released the specs online, just kind of said, here it is. And they went to these Taiwanese manufacturers and other folks and said, please make these for your cost plus a few bucks. And it was revolutionary in that market because it allowed this open platform to sort of embrace this very critical element that everybody needs.
And I think there's going to be a lot of these examples inside of AI because the costs are so extreme, so much more than just building a data center for a traditional web app, that the incentives to do it are just so meaningful.
Yeah, and I just showed it on the screen. Sax, you've actually been dancing along this line. Last night, I was using your new Slack killer or co-exister. I'm not sure. It feels like a Slack killer to me because I'm moving my company to it on... Over the weekend, we're moving to Glue. Nice. And you and I were doing some very... I may need to wet my beak on this one.
We want you to wet your beak.
It feels like a hundred bagger to me. I'm in. It is a lot killer. That's the way we're thinking about it.
It feels killer-esque. J. Cal, can you do that again in Christopher Walken voice, please?
I need to wet my beak. It feels like a hundred X sliding 500. Wow. Sacks, tell me about product decisions. Where does the AI end? And your product begins.
Yeah, well, it's a good point. I mean, I think where the AI ends, we want to use the most powerful AI models possible, and we want to focus on enterprise chat. So you could think of us as, for sure, a Slack killer or Slack competitor. It says that Slack wasn't built for the AI era. Glue is AI native. What does that mean? No channels. You know, I showed this to Chamath.
The first thing he said is, you had me at no channels, right? People are so sick of channels, you have to keep up with all these hundreds and hundreds of channels. And the real problem with channels is there's one thread in a channel that you want to see. In order to see it, you have to join the whole channel, and now you're getting all this noise. People just want the threads.
So if you look at what's the chat model inside of ChatGPT, it's just threads, right? You... create a topic-based thread in ChatGPT. The AI comes up with a name for it, puts it in the sidebar. And then if you want to talk about something else, you create a new chat. That's exactly the way that Glue works. It's just multiplayer. You just put the groups and individuals you want on the thread.
Let me just show you real quick. Here's my Glue here. And you can see that in the sidebar, I've got all the threads that I've been involved in. And like I said, you can address them to multiple people or groups. And then you've got the chat here. Now, we've also fully integrated AI.
And so Nick, who's our producer, just in this thread said, at Glue AI, what countries does SACS talk about most in episodes? Episodes is a group we created to be the repository of all of the transcripts of our episodes. And so Lou did a search and it said David Sachs frequently discusses Ukraine. What? The most. Yeah. Really?
So then Nick said, be more specific about Sachs stance on Ukraine-Russia war. Oh, boy. And it's going to overload the server. Well, it said here, David Sachs has articulated a nuanced and critical perspective on the Ukraine-Russia war across various episodes of the All In pod. Here are some key points encapsulating his stance. And it, like, nailed it.
It talked about prevention through diplomacy, opposition to NATO expansion, humanitarian concerns, skepticism of military intervention, peace deal proposal. You know, I'll copy and paste this onto Twitter X later today. But the point is, it, like, nailed it across all these different episodes. And then, this is a feature of GLOOW, It provided sources.
So it cites where it got all the information from. So imagine, you know, we're doing this for the all-in pod, but you could imagine that instead of it being transcripts of a podcast, it could be your work documents. You now have in your main chat the ability just to ask, hey, at Glue.ai,
Remind me where we left that project or tell me who the expert is on this subject matter or who's contributed the most to this project. I've actually figured out using Glue AI who's contributed the most deal flow at Kraft. It's pretty amazing. Now, let me show you like some of the bonuses. What's the answer? I'm not going to say here. I think it's Brian.
Brian's my guy.
Brian Ryan is my guy.
I want to hire Brian. If I got to get Sachs, I go Daniel and Brian. Boom, boom, boom. That's how I get Sachs on the line. Okay, you take Daniel.
I'll hire Brian. I would reconstitute Kraft.
We'll call it Kraft. All right, look, let me show you a couple.
Hold on a second. Sachs, I just looked at the chat GPT-4 Omni server, and when you did that query, it actually rolled its eyes. Keep going.
Okay. So we talked about how ChatGPT 4.0 understands sentiment in conversations, and I think this is a really good example. So Evan, who's my co-founder at Glue, was just testing this for us, and he asked, what is each bestie's personality type based on conversations and episodes? First of all, it figured out who the besties are. We didn't program that.
It just figured it out on its own, which is pretty impressive. Here's what it said about each bestie's personality. It says, Jason Calacanis, role, the moderator and entertainer, personality traits, charismatic, humorous, lively, and the glue of the group.
Say more.
Often prays for his ability to make the podcast entertaining and keep the conversation flowing. Jason's energy and ability to bring humor into discussions are key elements that define his role in the podcast. Comments he's acknowledged for his role in bringing the group together. Chamath Palihapitiya, the visionary and hustler. Personality traits, insightful, driven, persistent.
Chamath's relentless drive and ability to hustle are admired by his peers, making him a core part of the podcast and intellectual discussions. I'll skip me, but does Freeberg, the pragmatic scientist and realist. Pragmatic, methodical, and a bit reserved. Oh. Freeberg's often the voice of reason, bringing a scientific and realistic perspective to the discussion.
He focuses on analysis, logical reasoning. And then it cites where it got this from. And it says here, overall, the dynamic between the four besties creates a well-rounded and engaging podcast with each member bringing their unique strengths and personality traits to the table. I think that's pretty incredible.
How woke is this? Have you put any rails on it, or is this just pure ChatGPT 4.0 combined with the data?
Yeah, so what we're doing here is we're wrapping ChatGPT 4.0 with... glue features that we've implemented to get the most out of the conversation there's things we have to do to scope the the prompt and then we're using a retrieval augmented generation service called raggy which does rag as a service that basically slurps in our transcripts and makes them accessible to the ai
So that's basically the stack that we're using. But as the models get better and better, glue just gets better and better. Again, we just want to ride the wave.
Can I just make a comment on this? It's just so clean. JCal was the key for me in abandoning Slack. He told me two or three years ago, he called me and he said, I have, you can tell me the exact channels. I eliminated some channels that were random. There was like two or three channels that you banned that your Slack instance wasn't allowed to have. And I was like, this is genius.
And I went in and I was like, all of our companies should just eliminate these channels. And we could only get like 20% or 30% compliance. But it really started to turn me off Slack because I would get caught in these threads that were just so totally useless. And I thought, why aren't people working?
And this is really great because you cannot blather on about nonsense in Glue, which I find really useful. Well, this is what happens when Slack- We use it at 80, 90, just so you know.
So we were the- When you get into Slack too much, people start to think Slack is the job. And replying to Slacks and having conversations is the job when there's actually a job to be done.
There's a job to be done.
Yeah, and so it's important. And what I liked about this implementation facts was it's like the ability to make a feed or a data source inside of your communication platform. So the fact that you imported all of the episodes and the transcripts is great. But what I want is like our HubSpot or our CellCRM. I want our Zendesk. I want our LinkedIn jobs and our LinkedIn job applications.
I want our Notion. I want our Coda. to each have the ability. And when I was using it last night, what you do is you use the at symbol to evoke and to summon in a way. It's like summoning Beetlejuice. So you summon your AI, but then you tell it what data set you want to go after. So you say, you know, at AI, let's talk about, I don't know, How do you manage your deal flow at Kraft?
Do you use software like CRM software to manage deals? Brian, Daniel.
We just do it all in Glue. But we do it all in Glue, so it's already right there. But you're right. So the first thing that Glue AI has access to is all of your chat history, which is amazing because you get like... Then we can look at all your attachments. And we've got, I think, six integrations at launch. There'll be more. So yeah, all of your enterprise data will be there.
In the short term, you're right. You have to summon the repository by app mentioning because the AI needs a little bit of help of where to look. But in the future, it's going to figure it out on its own. So it's just going to become more and more seamless. But it'll insert itself.
So we have a discussion about sales. And then you might have a sales bot that says, hey, by the way, nobody's called this client in three months.
Well, that's where I want to go with it is I call that promptless. which is I want the AI just to chime in when it determines that it has relevant information and can help the team, even if it hasn't been summoned yet. But we need some model improvement for that, frankly. I mean, we'll be able to get there by GPT-5, but that's totally where this is headed. I'll show you just one more fun example.
i could let me just show you this so i asked it to to write a letter to lena khan to be a guest at the all-in summit and i told it mention positive things we've said about lena khan in episodes of the all-in pod and so it wrote this letter dear chair khan we hope this message finds you well on behalf of the host the all-in pod we're excited an invitation for you to speak
at the upcoming All In Summit. And then it says, in our conversations, we have frequently highlighted your impressive credentials and the impactful work you've undertaken. For example, in episode 36, we acknowledge your trailblazing role. And so the letter was able to quote episodes of the All In pod Just without anyone having to go do that research and figure out what would be the best.
Because I told it, only say positive things. Don't say anything negative. And then it said, warm regards. And it said who the four besties were. Again, we never told it who the besties are. We just said, write us a letter. So it's pretty incredible. Now, this is an example with the all-in pod. Think about any work context where the AI has access to your previous work documents.
It's pretty amazing what it can do.
Well, I mean, it is kind of in the name. Like, this is glue. Put you together. And Slack is where you Slack up. Makes total sense. The brands give you a little bit of a tip. We should have seen it coming with Slack. Totally. We have a breaking news story. It's a breaking news story.
It's an all-in exclusive today on the program.
I got breaking news coming in. Friedberg, your life's work. Sax did his product review. Now it's your turn, Friedberg. We got breaking news coming in.
I did promise you that when Ohalo decides to come out of stealth and explains what we've done and what we're doing, I would do it here on the all-in pod first before the- And all-in exclusive. All-in exclusive. So basically, by the time this pod airs, we're going to be announcing-
what Ohalo has been developing for the past five years and has had an incredible breakthrough in, which is basically a new technology in agriculture. And we call it boosted breeding. I'm gonna take a couple of minutes just to talk through what we discovered or invented at Ohalo and why it's important and the kind of significant implications for it.
But basically five years ago, we had this theory that we could change how plants reproduce. And in doing so, we would be able to allow plants to pass 100% of their genes to their offspring rather than just half their genes to their offspring.
And if we could do that, then all the genes from the mother and all the genes from the father would combine in the offspring rather than just half the genes from the mother and half the genes from the father. And this would radically transform crop yield and improve the health and the size of the plants.
which could have a huge impact on agriculture because yield, the size of the plants ultimately drives productivity per acre, revenue for farmers, cost of food, calorie production, sustainability, et cetera. So this image just shows generally how reproduction works. You've got two parents.
you get a random selection of half of the DNA from the mother and a random selection of half the DNA from the father. So you never know which half you're gonna get from the mother or which half you're gonna get from the father. That's why when people have kids, every kid looks different. And then those two halves come together and they form the offspring.
So every time a new child is born, every time a plant has offspring, you end up with different genetics. And this is the problem with plant breeding. Let's say that you have a bunch of genes in one plant that are disease resistant, a bunch of genes in the other plant that are drought resistant, and you wanna try and get them together.
Today, the way we do that in agriculture is we spend decades trying to do plant breeding, where we try and run all these different crosses, find the ones that have the good genes, find the other ones that have the good genes and try and keep combining them. And it can take forever and it may never happen that you can get all the good genes together in one plant to make it both disease resistant
and drought resistant. So what we did is we came up with this theory that we could actually change the genetics of the parent plants. We would apply some proteins to the plants and those proteins would switch off the reproductive circuits that caused the plants to split its genes. And as a result, the parent plants give 100% of their DNA to their offspring.
So the offspring have double the DNA of either parent. You get all the genes from the mother, all the genes from the father. And finally, after years of toiling away and trying to get this thing to work, and all these experiments and all these approaches, we finally got it to work. And we started collecting data on it. And the data is ridiculous.
Like the yield on some of these plants goes up by 50 to 100% or more. Just to give you a sense, like in the corn seed industry, Breeders that are breeding corn are spending $3 billion a year on breeding, and they're getting maybe 1.5% yield gain per year. With our system, we are seeing 50% to 100% jump in the size of these plants. It's pretty incredible. Here's an example.
This is a little weed that you do experiments with in agriculture. called Arabidopsis. So it's really easy to work with. And you can see that what we have on the top are those two parents, A and B. And then we applied our boosted technology to them and combined them.
And we ended up with that offspring called boosted A and B. So you can see that that plant on the right, it's much bigger, it's got bigger leaves, it's healthier looking, et cetera.
Freework, can I ask you a question? Does that mean that the boosted one has twice the number of chromosomes as A and B? Exactly right. So is that like a new species then? How does it survive with twice the number of chromosomes?
Yeah, it's called polyploidy. So we actually see this happen from time to time in nature. For example, humans have two sets of chromosomes, right? So does corn, so do many other species. Somewhere along the evolutionary history, wheat... doubled and then doubled again. And you end up actually in wheat having six sets of chromosomes. Wheat is what's called a hexaploid. Potatoes are a tetraploid.
They have four sets of chromosomes. And strawberries are an octaploid. They have eight. And some plants have as many as 24 sets of chromosomes. So certain plant species have this really weird thing that might happen from time to time in evolution where they double their DNA naturally. And so what we've effectively done is just kind of
apply to protein to make it happen and bring the correct two plants together when we make it happen.
And so- So this could only happen for a plant, right? This could never happen with an animal?
It wouldn't work in animals. It works in plants. Okay. And one way you can think about plant genetics is all the genes are sort of like tools in a toolbox. The more tools you give the plant, the more it has available to it to survive in any given second, to deal with drought or hot weather or cold weather, etc.,
And so every given second, the more tools or the more genes the plant has that are beneficial, the more likely it is to keep growing and keep growing. And that plays out over the lifetime of the plant with bigger leaves and bigger, you know, grows taller. But more importantly, if you look at the bottom, the seeds get bigger. And in most crops, what we're harvesting is the seed.
That's true in, you know, corn and many other crops. And so seeing over a 40% increase in seed in this little weed was a really big deal. But then we did it in potato. And potato is a crazy result. Potato is the third largest source of calories on earth. And so we took two potatoes that you see here in the middle, AB and CD.
We applied our boosted technology to it, to each of them and put them together. And you end up with this potato ABCD, that's the boosted potato. And as you can see, these were all planted on the same date and the boosted potatoes much bigger than all the other potatoes here, including a market variety that we show on the far right. That's what's typically grown in the field.
Now here's what's most important. When you look under the ground and you harvest the potatoes, You can see that that AB potato only had 33 grams, CD had nine grams. So each parent had 33 and nine grams potato, but the boosted offspring had 682 grams of potato. The yield gain was insane. And so you could see this being obviously hugely beneficial for humanity.
Potatoes being the third largest source of calories, Indian potato farmers are growing one acre of potato. In India, they eat potato two meals a day. In Africa, potato is a food staple. So around the world, we've had a really tough time breeding potatoes and improving the yield. With our system, we've seen incredible yield gains in potato almost overnight.
How big are those potatoes?
Those are normal-sized potatoes that you see there. Those are like, you know, table potatoes. Basically, that looks like a russet potato right there. That's like a normal-sized russet. I can tell you, you got something in it.
It started as like a little creamer potato, basically, and you blew it up into a russet potato?
Yeah, so the genetics on AB... You can see they're like little purple, tiny little purple potatoes. The genetics on CD are like these little white, you know, tiny little ball potatoes.
But when you put those two together with boosted and you combine all the DNA from AB and all the DNA from CD, you get this crazy high yielding potato ABCD, which, by the way, is higher yielding than the market variety that's usually grown in the field on the far right.
So why not just grow russet potatoes then?
We are. And so we're working on doing this with Russet. We're working on doing this with every major potato line. Sorry, the improvement you'll see is actually in yield. So it's not the size of the potato. It's the number of potatoes that are being made.
And so you'll see... Per hectare or something like that, like the Dwarf Week projects in the 60s and 70s. You know how you can tell Freeburg's onto something here? He got David Sachs to pay attention during it. Yeah. This is going to be a decacord and Saks is awake.
Saks is like, how do I wet my beaker? Saks is interrogating the potato lines. I've never... What's going on?
I think genetics is interesting. But so have you tried these potatoes? Do they taste different?
Oh, no, they're awesome. Yeah, they're potatoes. And we do a lot of analysis.
You haven't sprouted any horns yet or anything like that?
No. I mean, again, one of the other advantages of the system that we've developed Let me go back here, and I just wanna take two seconds on this. One of the other things this unlocks is creating actual seed that you can put in the ground in crops that you can't do that in today. So potatoes, the third largest source of calories.
But the way we grow potatoes, you guys remember the movie, The Martian, you chop up potatoes and you put them back in the ground. Because the seed that comes out of a potato, which grows on the top in the flower, Every one of those seed is genetically different because of what I just showed on this chart, right? You get half the DNA from the mother, half the DNA from the other.
So every seed has different genetics. So there's no potato seed industry today. And potato is like $100 billion market. With our system, not only can we make potatoes higher yielding and make them disease resistant, What we also make is perfect seed.
So farmers can now plant seed in the ground, which saves them about 20% of revenue, takes out all the disease risk, and makes things much more affordable and easier to manage for farmers. So it creates entirely new seed industries. So we're going to be applying this boosted technology that we've discovered across nearly every major crop worldwide.
It'll both increase yield, but it will also have a massive impact on The ability to actually deliver seed and help farmers and make food prices lower. Is it more expensive? No, it's actually cheaper. So higher yield, lower cost. Do you need more water? Less water, less land, less energy. Do you need more fertilizer?
Fertilizer usually scales with biomass, but these sorts of systems should be more efficient. So fertilizer use per pound produced should go down significantly. as we get to commercial trials with all this stuff. And we're doing this across many crops. So there's a lot of work to do in terms of like, how do you scale the production in the fields?
Tell us about the patents and how important patents play a role in this, because isn't it like one of Monsanto's big things, like they just go and sue everybody into the ground or whatever?
I'm going to answer you one second. I'm just going to switch my headset. It just died. Wow. We went from Saks' bots to Freeburg's crops.
I'm glad we're doing him second because all of a sudden, like, group chat doesn't seem very important.
Yeah. Wow. He just solved the whole Ukraine crisis here. We're going to be able to grow wheat in the desert and in the rainforest.
He solved the world food problem. Yeah, Saks. What did you do for the last six months? Yeah, we made our price chat a little better, but...
we added ai to enterprise chat we cleaned up your slack so yeah when you invest we've invested a ton of money this was stealth for five years we put a ton of money into this business so when you invest like um i mean north of 50 north north of 50 yeah 50 million five years and you don't have a product in market yet wow that's some we actually have some product yeah so i haven't talked about the way we've been making money in some of the business we've been doing okay
Let me just make sure this is clear. So that last photo you showed with the different types of potatoes, you had created the super huge ones, but you're saying that The yield benefit here is just you create a much bigger, hardier plant that's capable of producing more potatoes. But the size of the potatoes doesn't change?
You can control for that when you breed. So the selection of what plants you put together in the boosted system allows you to decide. Do you want small, medium, large? That's all part of the design of which plants do you want to combine.
Okay, because your goal is not to turn like a russet potato into like a watermelon or something like that.
No, the goal is to make more russet potato per acre so that we use less water, we use less land, farmers can make more money, people pay less for food. That's the goal. And so it's all about yield. It's not about changing the characteristics.
There are some crops where you want to change the characteristics, like you might want to make bigger corn kernels and bigger cobs on the corn, which is another thing that we've done. And that's actually been published in our patent. And the reason, by the way, I'm talking about all this is some of our patents started to get published last week.
And so when that came out, the word started to get out. And that's why we decided to get public with what we've done, because it's now coming out in the open.
You mentioned something briefly there about where different crops can be planted. We had these big talks about wheat and corn. They're only available in very specific parts, north of the equator, the Campion jungles, Campion, obviously polar or desert extremes. So if you're successful, what would this do for on a global basis, where these crops are made?
Because there's a whole discussion about you.
Awesome question.
Totally wheat belly of Europe, the cradle of wheat.
It's a great question. I'm so glad you asked it because that's one of the key drivers for the business is that we can now make crops adapted to all sorts of new environments that you otherwise can't grow food. Today, there's close to somewhere between 800 million and a billion people that are malnourished. That means they are living on less than 1200 calories a day for more than a year.
But on average, we're producing 3,500 calories per person worldwide in our ag systems. The problem is we just can't grow crops where we need them.
And so by being able to do this sort of system where we can take crops that are very drought resistant or can grow in sandy soil or very hot weather and adapt cooler climate crops to those regions through the system, we can actually move significantly where things are grown and improve food access in regions of-
Freeburg, when you look at a potato, how do you figure out what part of their DNA is the drought-resistant part? Yeah. And then how do you make sure that that's turned on? So even if you inherit that chromosome, is there some potential interaction with the... Generally, if we can... So these are what are called markers, genetic markers.
And so there are known markers associated with known phenotypes. A phenotype is a physical trait of a plant. And so we know lots of markers for every crop that we grow. Markers for disease resistance, drought resistance, markers for big plants, short plants, etc. And so what we do is we look at the genetics of different plants that we might want to combine into the boosted system.
And we say these ones have these markers, these ones have these markers, let's put them together. And then that that'll drive the results. One of the other interesting things we're seeing, which I didn't get too much into in the slides. It's not just about combining traits. But it turns out, when you add more genes together, Biology figures out a way to create gene networks.
These are all these genes that interact with each other in ways that are not super well understood, but it makes the organism healthier and bigger and live longer. This is why mutts are healthier and live longer than purebred dogs, because they have more genetic diversity.
So there's a lot of work now in what's called quantitative genomics, where you actually look at the statistics across all the genes. You use a model, and the model predicts which two crosses you want to make out of hundreds of thousands or millions of potential crosses that the AI predicts. Here's the two best ones to cross because you'll get this growth or this healthiness factor that'll emerge.
So how do you want to make money, Friedberg? Are you going to sell the seeds? Are you going to become the direct farmer? Are you going to become food as a service? Like, how do you... make the most money from this?
We're not going to farm. Farmers are our customers. And so there are different ways to partner with people in the industry who already have seed businesses or already have genetics and help them improve the quality of their business. And then there's other industries like in potato where we're building our own business of making potato seed, for example.
So every crop and every region is actually quite different. So it becomes a pretty complicated business to scale. We're in the earlier days. We're already revenue generating. I would like a sweeter blueberry.
No comment. No comment. Yeah. I get tilted by the quality of the Driscoll's blueberries. Let me tell you something about the Driscoll's blueberries. Also, the Driscoll, I've had only one batch of a Driscoll's strawberry that was just off the charts. And every 19,847 other batches I bought have been total sh**.
Now you want the European small ones or the Japanese ones from Hokkaido because they're rich and sweet and they're not these like monstrosity of giant flavorless strawberries. What's that about? Could you do a seedless mango?
Yes, then cut it.
So you can just cut it.
Oh my God, how great would that be?
The amount of work per bite on a mango is like the worst ratio.
Yeah. Well, somehow we made it about us. Yeah, no, no, look, I think that's, it is all about you guys. Tell us about the blueberries, sorry. Well, no, every year, Driscoll's puts out a special labeled package called Sweetest Batch. And they just had the sweetest batch of strawberry and blueberries. I don't know if they're still in the stores, but they only last for like a week or two.
And that's the best genetics only grown on a small number of acres. Really incredible.
I'm going to go in as soon as this is done.
See if they have it. So I got it a few weeks ago. It's quite delicious. Anyway, let's just say we know the berry market very well. My co-founder, CTO, Judd Ward, whose brilliant idea Boosted Breeding was many years ago, who I met because they had a New Yorker article on Judd. I cold called him and said, hey, will you come in and give us a tech talk?
We started talking and Judd came up with this idea for Boosted Breeding. And so we started the business. with Judd and Judd ran molecular breeding at Driscoll. So we have a lot of Driscoll's people that work at Ohalo. We know the market really well.
Can you go back to the patent stuff? Like, are you some seed person?
So we've spent, we've spent 50 million bucks on, you know, plus on this business today. So we have filed for IP protections that people can't just rip us off. But I would say, I think that the real advantage for the business is arises from what we call trade secrets, which is not just about taking patents and going out and suing people. That's not a great business.
The business is how do you build a moat and then how do you extend that moat? The great thing about plant breeding and genetics is that once you make an amazing variety, the next year the variety gets better and the next year the variety gets better. And so it's hard for anyone to catch up.
That's why seed companies generally get monopolies in the markets because farmers will keep buying that seed every year, provided it delivers the best genetics. And so our business model is really predicated on how do we build advantages and moats and then keep extending them rather than try to leverage IP. So I'm a big fan of like building business model advantages.
This is going to be incredible, Sax. If you think about geopolitically, what's going on in Somalia, Sudan, Yemen, Afghanistan, those places have tens of millions of people, I think hundreds of millions collectively, who are at risk for starvation. If you could actually make crops that could be farmed there, Friedberg, you would change humanity.
And then all these people buying up farmland in America... That could devalue that farmland if that wasn't as limited of a resource. Yeah, Freeberg?
No, I think... So first of all, farmland in America is mostly family-owned. It's about 60% rented, actually. So a lot of families own it, and then they rent it out because they stopped farming it. But the great thing that we've seen in agriculture historically is that the more... Calories we produce, the more food we produce, the more there seems to be a market.
It's like any other economic system. What about wheat and rice? Yeah. So those are calorie sources one and two. And there's certainly opportunity for us to apply our boosted systems there. The big breakthrough with potato is we can make potato seed using our boosted system in addition to making better potatoes. McDonald's is the largest buyer of potatoes, yeah?
So in the US, 60% of the potatoes go to French fries and potato chips. McDonald's buys most of the fries. PepsiCo under Frito-Lay buys most of the potato chip potatoes. 40% are table potatoes. In India... 95% of the potatoes are table potatoes. They're eaten at home. And the Indian potato market's three to four times as big as the U.S. potato market. In Brazil, it's 90% table potato.
So all around the world, potatoes are different. The U.S. is, you know, unusually large consumers of French fries and potato chips.
I speak on behalf of J. Cal and I said, we will gladly invest a million at a 10 cap in both of your businesses. Absolutely.
Yes.
We will work our way into J. Cal. J. Cal and I will do the deal. We'll wire the money. We'll wire the money. A little million to each of you guys at a 10 cap. Thank you. Absolutely.
You're in. It may not be a 10 cap though, but yes. You're in.
Chamath and J. Cal have secured the bank.
in the back from the besties actually doing work.
Yeah.
Well, I appreciate you guys letting me talk about it today. I'm excited to share it.
Congratulations to both of you.
I love it. It's been, yeah, building stuff is hard. There's always risk. It's a lot of work and a lot of setbacks, but man, when you get stuff working, it's great.
We're each doing the things we do best. Freeburg is solving the world's hunger problem. And I'm cleaning up your slack. Making your enterprise chat a little better.
All progress counts. All right. Stanley Druckenmiller has got a new boyfriend. Druckenmiller's got a boyfriend. And his name is Javier. And they've eloped to Argentina. Druckenmiller professed his love. Tom Cruise on Oprah's couch in a CNBC interview this week, the only free market, quote, leader in the world right now, bizarrely, is in Argentina of all places. He cut Social Security 35%.
If he came to office, they've gone from a primary deficit of like 4% or 5% to a 3% surplus. They've taken a massive hit in GDP, basically a depression for a quarter, and his approval rating has not gone down. Druckenmiller has explained how he invested in Argentina after seeing Millet's speech at Davos, which we covered. Here's a 30 second clip. Play the clip, Nick.
By the way, do you want to hear how I invest in Argentina? It's a funny story. I wasn't at Davos, but I saw the speech in Davos and it was about one o'clock in the afternoon in my office. I dialed up perplexity and I said, give me the $5,000. most liquid ADRs in Argentina. It gave me enough of a description that I followed the old Soros rule, invest and then investigate. I bought all of them.
We did some work on them. I increased my positions. So far, it's been great, but we'll see.
Yeah, that's quite interesting. Quick note, you hear Druckenmiller mention ADRs. For those of you who don't know, and I was one of them, they stand for American Depository Receipts, basically a global stock offered on a U.S. exchange to simplify things for investors and Yeah, I mean, he didn't sign a prenup here. He just went all in and he bought the stock, Chamath.
And then he's going to figure it out later. Tell us your thoughts on this love affair, this bromance.
There's a great clip of Millay. He goes on this talk show in Argentina and the talk show host, she's just so excited and greets him. And then they start making out. Have you guys seen this? What? They're just guys out of control. Full on French kissing each other. It's hilarious. Yeah, Soros has been very famous for this invest and investigate thing.
It's like a smart strategy for very, very liquid public market investors that have the curiosity that he does. I don't have much of a reaction to that. I think that the thing with Argentina that's worth taking away is when you've spent decades casting about and misallocating capital and running your economy into the ground, the formula for fixing it is exactly the same.
You cut entitlements and you reinvigorate the economy. And so the thing we need to take away is if we don't get our shit together, that's probably what we're going to have to do.
Sacks, the influence of Millais on American politics, will there be any? It seems like he has paralleled what Elon did at Twitter, Facebook, and Zuck did at Facebook. Do you think that this, you know, experiment he's doing down there of just cutting staff, cutting departments will ever make its way into American politics?
Probably not. I mean, not until we're forced to. But what Millais did, he comes in and they've got a huge budget deficit and they've got runaway inflation and they're debasing their currency. And just practically overnight, he just slashes government spending to the point where he has a government surplus.
And then as soon as he gets credibility with the markets, that allows him to reduce interest rates, inflation goes away, and people start investing in the country.
It's magic. So there is a path.
It's obvious. Listen, I mean, you can't run... deficits forever. You can't accumulate debt forever. It's just like a household. If your spending exceeds your income, eventually you got to pay it back or you go broke. And the only reason we haven't gone broke or experienced hyperinflation is because we're the world's reserve currency. So there's just a lot of room for debasement.
And there's not a ready alternative yet. I mean, everyone's trying to figure out what the alternative will be. So we've been able to accumulate more and more debt, but it's reaching a point where it's unsustainable. And what we've already seen is that the feds had to jack up interest rates from very low, practically nothing, to 5.5%. And that has a real cost on people's well-being.
Because now, your cost of getting a mortgage goes way up. I mean, mortgage rates are over, what, 7.5% now?
Yeah, 6%, 7%, depending on how much net worth and your credit rating.
Right. And so, it's much harder to get a mortgage now. It's harder to make a car payment if you need to borrow to buy a car. And if you have personal debt, the interest rate's going to be higher. The inflation rate actually doesn't take into account any of those things.
Remember, Larry Summers did that study where he said the real inflation rate would be 18% or would have peaked at 18% if you included cost of borrowing. That's why people don't feel as well off as the unemployment rate would normally suggest. So people are hit really hard when interest rates go up in terms of big purchases they need to make with debt.
And then, of course, it's really bad for the investment environment because when interest rates are really high, that creates a higher hurdle rate and people don't want to invest in risk assets. And so eventually the pace of innovation will go down. And Druckenmiller made this point in his next set of comments. He said that Treasury is still acting like we're in a depression.
It's interesting because I've studied the depression. You had a private sector crippled with debt, basically with no new ideas. So interventionist policies were called for and were effective. He said the private sector could not be more different today than it was in the Great Depression. The balance sheets are fine, they're healthy.
And have you ever seen more innovation ideas that the private sector could take advantage of, like blockchain, like AI? He says all the government needs to do is get out of the way and let them innovate. Instead, they've spent and spent and spent. And my new fear now is that spending...
And the resulting interest rates on the debt that's been created are going to crowd out some of the innovation that otherwise would have taken place. I completely endorse Drucker Miller's view of binomics. And actually, I mean, this is what I said way back in 2021.
Victory lap. Here we go. A little David Sachs victory lap. We need a little graphic for that.
Druckenmiller used the word Bidenomics and said, I give these guys an F because they're still printing money and spending money like we're in a depression, even though we're in a rip-roaring economy. And when they started doing this back in 2021, I tweeted Bidenomics equals pumping trillions of dollars of stimulus into a rip-roaring economy.
I'm not going to pretend like I know what's going to happen next, but I've never tried this before. What happened next was a lot of inflation, and that jacked up interest rates. According to even Keynesian economics, the reason why you have deficit spending is because you're in a recession or depression, and so use the government to stimulate and balance things out.
You don't do deficit spending when the economy's already doing well. So this spending, there's no reason for it.
Yeah, it's like showing up to a party that's going crazy and putting gasoline on the fire.
Yeah, I mean, more importantly, it should limit the approval or action of certain programs that you might otherwise want to do in a normal environment, but in an inflationary environment, you don't have the flexibility to do them. Student loan forgiveness is a really good example. Is now the time?
Of course not.
To do student loan forgiveness, or do we wait for inflation to temper a bit? Is now the time? So there's just a lot of these examples that actually the opposite should be true. Yeah, but none of...
All of those things get you votes. Before we move on from this, look, what we have coming out of Washington here is a contradictory and therefore self-defeating policy. You've got the Fed jacking up rates to control inflation. You move across town and you've got Capitol Hill and the White House spending like there's no tomorrow, which is inflationary. Right. Why would you do both those things?
Choose what your policy is going to be.
It's like driving with your foot on the brake and the gas at the same time. It's not a great idea for the car.
Let me just make one comment, J. Cal, before we move on about the Druckenmiller investment statement. Of course. And I just wanted to say, I think what it highlights about Druckenmiller and call it a rift in investing philosophy or skill is the difference between precision and accuracy.
What I mean by that is precision really references that you do a lot of detailed analysis to try and make sure you understand every specific thing that is going right or could go wrong.
But the problem... And so that means you, for example, might do a ton of diligence on a company and make sure you understand every dollar, every point of margin, all the specifics of the maturation of that business and where they are in their cycle. But... You could be very precise, but be very inaccurate.
For example, if you miss an entire trend, someone could invest in Macy's back when Amazon was taking off and have done a lot of precise analysis on Macy's margin structure and performance and said, this is a great business. But they missed the bigger trend, which is that e-commerce was going to sweep away Macy's. And consumers were simply, that's not possible in the analysis that they were doing.
Let's be honest, Freeberg, nobody can make that stupid of a trade to say Macy's versus Amazon over the next 10 years.
Well, yeah. And so like... Do you want to show that? No, no, no, no. Do not poke the tiger. Let's not get into it with other podcasters. The worst spread trade in history. Yeah, let me just finish the statement. But the other one is being accurate. And accurate means you get the right bet, the right sentiment, the right friend.
The problem with being accurate, you could have said in the year 2000, hey, the internet's going to take off. And you could have put a bunch of money in, but the problem was you were right. You just had to have the necessary patience. And so accuracy generally yields better returns, but it requires more patience because you can't necessarily time how long it will take for you to be right.
So a guy like Druckenmiller is making an accurate bet. He bets correctly on the trend, on where things are headed. He doesn't necessarily need to be precise, but he has the capital and his capital structure that allows him to be patient to make sure that he eventually gets the return.
And to build on your thoughts, having watched this movie a couple of times, and I overthought the Twitter investment as but one example. I had the opportunity to invest in Twitter when it was like a single-digit millions company. And I just thought, you know what? This thing is only like the headline. And I told Ev, it's the headline. It's not like the entire blog post.
It's going to be a cacophony of idiots. This thing is going to be chaos. And I was right, but I was wrong, right? Great bet. but my wrong analysis.
And so you can add precision to other aspects, like when you sell your shares or when you double down, but you have to get the trend right, which is Evan Williams, great entrepreneur, Jack, great entrepreneur, Twitter taking off like a weed, just make the bet, right?
No, the problem is you knew too much about journalism. You knew too much about the space they were trying to disrupt, and that could be a mistake. We did PayPal. None of us knew anything about payments. That was one of the reasons we were successful. All the payments experts told us it couldn't be done.
Right. Absolutely. So that happens a lot. I didn't even know what a Facebook was when I joined Facebook. It's an American college phenomenon. No, seriously. You don't have that in Canada.
But you knew Zuck and you saw some growth charts and you saw some precision in his ability to build product. And that's the way to go.
The great thing about network effect businesses is there's a trend line that sustains because it builds if it's an appropriate network effect. So you can be accurate about buying into the right network effect business. You don't need to use all of this diligence to be perfectly sound and
around the maturation of the revenue and the margin structure and all that stuff as long as the trend line is right and you're willing to be patient to hold your investment. I think Druckenmiller's point is incredible. He took a look, he very quickly made a macro assessment. From a macro perspective, what Millais is doing
is significantly different than what we're seeing in any other emerging market, let alone mature market with respect to fiscal austerity and appropriateness in this sort of inflationary, global inflationary environment. And he said, you know what? I don't see any other leader doing this. This is a no brainer bet. Let me make the bet.
And as long as he's willing to hold this thing for long enough, eventually the markets will get there and call it a spread trade against anything, he'll be proven right.
But speaking of bets, Jaykal, you told me this week that you just made your largest investment ever. Tell us about that.
Yeah. So I've gotten very lucky now because a lot of my founders from the first couple of cohorts of investing I did when I was a Sequoia scout have come back and created second and third companies. And so, you know, that happened with TK Uber and the cloud kitchens that happened with Raul from report of then superhuman.
And then it happened recently, just in the past year, my friend, Jonathan, who's the co-founder of thumbtack, He asked me to come to dinner and he said, hey, you know, you were the first investor in Thumbtack. Will you be the first investor in our next company, Athena? And I said, sure. What do you do? And he explained it to me.
And we put a seven-figure bet in, which is rare for us as a seed fund, right? Normally, our bet sizes are 100K to 50. You know, it's a $50 million fund. Why did you do it? Yeah, it's very simple. It's the fastest growing company I've ever seen. And I'm including Uber in that. It has been growing at, you know,
a rate that I'll just say is faster than Uber and Robinhood when we were investing in them, tens of millions of dollars. It's a very simple concept. When Thumbtack was building their marketplace, they used... researchers in places like Manila, et cetera, in the Philippines, knowledge workers.
And what they realized was the 0.1% of those knowledge workers were as good or better than say Americans at doing certain jobs. And so they've created this virtual EA service. You can go see it at athenowow.com. And we now have two of them inside of our company. It turns out Americans don't want to do the operations role. So it's kind of like AWS. You just give them $36,000 a year.
They give you essentially an operations or an EA. And they have ones that are kind of cheap of staff-ish. And this company is growing like a weed. So I am working with them on the product design as well. So imagine having two or three of these incredibly hardworking people who are trained with MBA class level curriculum.
They spend months training these people up, they pay them two or three times what they would make at any other company. And then they pair them with executives here. And it's kind of been an underground secret in Silicon Valley, because it's only by invitation right now, because they can only train so many people.
But if you've tried to hire an executive assistant, I don't know if anybody's tried to do that recently. You hooked me up, so I will be guinea picking this service. Yes. And I have two of them. And so it is just the greatest that you can have an operations person on board. Are these people powered by AI tools as well? Yeah.
So that's the kind of secret sauce here is they're training them and they watch you work and then they will learn how you do your job and then how quickly you can delegate and get stuff off your plate is the name of the game. So we have an investment team with researchers and analysts in it. We have a due diligence team. And then you have like executive functions in our fund.
they have now started shadowing you know you know highly paid americans in an investment firm ours and then train them up and now our due diligence our first level screening you know and our tracking of companies is being done by these assistants for what i'll say is a third to a fourth of the price i was paying previously so what that does in an organization is
We're just delegating away and then moving our investment team to doing in-person meetings and doing higher level stuff.
Yeah, you're 80-90ing. So at 80-90, we have this funny thing where we've made it a verb. Whenever you see somebody doing high quality work at a quarter to a tenth of the cost, we say, oh, you just 80-90ed it. Correct.
So you're 80-90ing the investment team. I'm 80-90ing the investment team. And you know what? It was scary as hell for them because they're like, am I going to lose my job? It's like, no. You now get to, instead of doing a check and call once a month, you can do a check and call every other week or every week.
Or instead of doing 15 first round interviews a week, you can do 25 because you have this assistant with you doing all the repetitive work.
The way that companies will work in five and 10 years, I don't think, guys, any of us are going to recognize what it's going to look like.
This is where I go. I mean, like watching Sax's demo earlier, how much progress and how seamless that product works with the features it has enabled by the underlying models. You just get to thinking how all of these vertical software applications become completely personalized and quickly rebuilt around AI. Totally. It's so obvious.
Can you imagine how long it would have taken John to write a letter to Lina Khan? Like if we said, John, invite Lina Khan, but be sure to reference all the nice things we said about her on episodes of the pod.
Oh, it'd be 10 hours of work.
It's got to go find the episodes, listen to them. Yeah, listen to them to figure out what the best quotes are.
And you got it done in five seconds. It's incredible. Totally. And then imagine building that same sort of capability into a very specific vertical application that's specific to some business function. And you can probably spend a couple minutes or an hour building that function. And then it saves you hours a day in perpetuity. Yeah.
And I think that's why these tools companies or the tools products that Google... Microsoft, Amazon, and a few others are building are actually incredible businesses because so many enterprises and so many vertical application builders are going to be able to leverage them to rewrite their entire business functions.
I got myself and my co-founders at 8090, we get this stream of emails of companies that are like, or people that are like, we have this product idea or we have this small product. One of the emails I got, this is crazy, was from a guy that's like, oh, we've 8090 Photoshop. So like we have like a much, much cheaper version of Photoshop.
And the guy was doing like a few million bucks of ARR and growing really nicely. But then it turned out that somebody saw that and then 8090'd it. So then there's an open source version of that thing. And so to your point, Friedberg, none of these big companies stand a chance. Yeah. Everything is going to get cheaper and faster.
Not because the products aren't good, but like JCal is going to go off and experiment with this. SAC's going to go off and build a product. You know, as every time that you're at a boundary condition, we're all going to explore, well, maybe we could do this with AI. Maybe we shouldn't hire a person. Not because we're trying to be mean about it, but it's because the normal, natural thing to do.
And the OPEX of companies is just going to go down, which means the size of companies are going to shrink, which means the amount of money you need is going to go down. And that's just going to create the ability for these companies to sell those products cheaper.
So it's a massive deflationary tailwind here. We had the same thing happen with compute, and now it's happening inside of organizations. I wrote a blog post about this on my sub stack called ADD. This is the framework I came up with. I told my entire team, look at what you got done every week. And I want you to ask three questions. How can I automate this? How can I deprecate this?
How can I delegate it? And the automate part is AI and what you're doing, David. The delegate part is athenowow.com. And then the deprecate is, hey, just be thoughtful. What are you doing that you don't need to do? And that's 80-90ing something. There are things inside these products that you don't actually need. What's the core functionality of the product?
You know, make it as affordable as possible. And then what's going to happen for people who think this is bad for society? You've got it completely wrong. We're going to have more people be able to create more products and solve more problems. The unemployment rate's going to stay very low. We're just going to have more companies.
So the idea, like, there was somebody who was working on very small, like, software. I want to get pitched on very niche ideas. I want to create something where people can find people to play pickleball with, right? Like a pickleball marketplace. Now, that wouldn't typically work because you would need $5 million a year to build that product. But if you can build it for $500,000 a year,
Well, now you've only got to clear that number to be profitable. So a lot more smaller businesses, a lot more independence, all these little niche ideas will be able to be built. And a VC who says, I'm not giving you $5 million to build that app will be like, but I will give you 500k.
And that's what I'm seeing on the ground in startups, the same startups that had a request of $3 million in funding five years ago are now requesting 500 to a million. It's deflationary all the way down.
Did you guys see the Google thing? Did you guys see the Google AI Gemini stuff?
ChatGPT Omni launched at the same time or perhaps strategically right before Google dropped its latest AI announcements at IO. The biggest announcement is that they are going to change search. This is the piece of the puzzle in the kingdom that they have been very concerned with and they're going for it. The new product, and they have like 20 different
products, you can see them at labs.google where they put all their different products. But this is the most important one. They call it AI overviews. Basically, it's perplexity for most users by the end of the year, they're going to have this. Here's how it works. And you can see it on your screen. If you're watching us go to YouTube. Here, they gave an example. How do you clean a fabric sofa?
This normally would have given you 10 blue links. Here, it gives you step-by-step guide with citations and links. So they're preempting the issue of people getting upset. And as I predicted, they're going to have targeted ads. Here's the things you need in order to clean your couch. You can only use this if you're using your Gmail account.
If you use a domain name on Google Docs, it won't work there. So go to labs.google.com. But they're doing citations. And I think that we're going to see a major lawsuit here. Those people who are in those boxes are going to look at the answer here and realize maybe they don't get the click through and that this answer was built on that. And now we're going to have to have a new framework.
There's going to need to be, Sachs, a new company. that clears this content so that Google can do answers like this.
The workflow stuff in Gmail also kicked ass.
The demo that they showed was you get a bunch of receipts and the person giving the demo, she said something to the effect of, well, wouldn't it be great if like, you know, the AI assistant were able to find all the receipts and then aggregated them and put them in a folder and then also actually generated an expense report or like a spreadsheet on the fly? It... It's crazy. Yeah.
I got to say, I think that it's free to change your mind. And so it's good to do that. Oh.
And I think that Chamath, in a rare moment of reflection, might do a, are we going to have a re-underwriting? Is this a re-underwriting?
I change my mind all the time.
I just, I mean, you know, because I'm- Ladies and gentlemen, breaking news, Chamath is re-underwriting his Google train.
Sorry, I know to blow your ears out. I think the Google thing is pretty special. Between last week's announcement of isomorphic labs, which let's be honest, that's just a multi-hundred billion dollar company.
So you're saying there might be many- Think about it this way, right? Multi-billion dollar opportunities sitting there dormant inside of Google that AI unlocks.
Look at a company like Royalty Pharma. So Royalty Pharma, it's a phenomenal business run by a phenomenal entrepreneur, Pablo Lagareta. But what is that business? That's buying 2% and 3% royalties of drugs at work. And you can see how much value that those guys have created, which is essentially 90% EBITDA margin business.
It's outrageous because they're in the business of analyzing and then buying small slivers. I think something like isomorphic ends up being of that magnitude of margin scale, but at an order of magnitude or two orders of magnitude higher revenue. So if you fold that back into a Google algorithm,
If you think about what they're doing now on the search side, these guys may be really kicking some ass here. So I think that the reports of their death were premature and exaggerated.
Absolutely. And the report of their death, Freeberg, was based upon people don't need to click on the ads. But as I said on this very podcast, my belief is that this is going to result in more searches and more knowledge engagement. Because once you... how to cook your steak and get the right temperature, right, for medium rare, it's going to anticipate your next three questions better.
So now to say, hey, what wine pairing would you want with that steak? Hey, do you need steak knives? And it's just going to read your mind that you need steak knives and Chamath likes to buy steak knives, but maybe you like to buy mock meats, whatever it is. it's going to drive more research and more clicks. So while the monetization per search may go down, we might see many, many more searches.
What do you think, Freeberg? You work there. And when we look at the future of the company and the stock price, Nick will pull it up, Man, if you had held your stock, I don't know, did you hold any?
I bought some. No, your original stock. Did you clear it at some point? Oh, no, I sold all my stock back when I started Climate because I was a startup entrepreneur and needed to live. So, which, you know, I recently did. I did the math on it. It was pretty, it'd be worth a lot. It would be worth billions or tens of billions? No, no.
Would it have been a billion?
No, no. Okay. You know, I was not a senior exec or anything. I think what you said is probably true. So that's accretive.
I think the other thing that's probably true is a big measure at Google on the search page in terms of search engine performance was the bounce back rate, meaning someone does a search, they go off to another site, and then they come back because they didn't get the answer they wanted.
And then the one box launched, which shows a short answer on the top, which basically keeps people from having a bad search experience because they get the result right away. So a key metric is they're going to start to discover which vertical searches, meaning like, hey, cooking recipes, that kind of stuff. There's lots and lots of these different types of searches.
that will trigger a snippet or a one box that's powered by Gemini that will provide the user a better experience than them jumping off to a third party page to get that same content. And then they'll be able to monetize that content that they otherwise were not participating in the monetization of.
So I think the real victim in all this is that long tail of content on the internet that probably gets cannibalized by the snippet one box experience within the search function. And then I do think that the revenue per search query in some of those categories actually has the potential to go up, not down.
Explain, give me an example.
You keep people on the page so you get more search volume. there, you get more searches because of the examples you gave. And then when people do stay, you now have the ability to better monetize that particular search query, because you otherwise would have lost it to the third party content page.
So for example, selling the steak knives is another is you know, it's kind of a good example, or booking the travel directly and so on. So by keeping more of the experience integrated, they can monetize the search per query higher. and they're gonna have more queries, and then they're gonna have the quality of the queries go up. So I think it's all in.
There's a case to be made, and I haven't done a spreadsheet analysis on this, but I guarantee you, going back to our earlier point about precision versus accuracy, my guess is there's a lot of hedge fund type folks doing a lot of this precision type analysis, trying to break apart search queries by vertical,
and try to figure out what the net effect will be of having better AI-driven one-box and snippets. And my guess is that's why there's a lot of buying activity happening in the stock right now. And I think they're probably all missing, to Moff's point, a lot of these call options, like isomorphic labs. I can tell you Meta and Amazon- Waymo.
Meta and Amazon do not have an isomorphics lab in Waymo sitting inside their business that suddenly pops to a couple hundred billion of market cap. And Google does have a few of those.
So other bets could actually pay off.
There may be. Look, I mean, there's Calico. No one talks about Calico. I don't know what's going on over there.
Life extension, yeah. Let me get Sachs involved in this discussion. Sachs, when we show that example, it's obvious... Google is telling you where they got these citations from and how they built their how to clean your couch, how to make your steak.
They were in a very delicate balance with content creators over the past two decades, which is, hey, we're going to use a little bit of your content, but we're going to send you traffic. This is going to take away the need to send traffic to these places. They're going to benefit from it. To me, this is the mother of all class action lawsuits because they're putting it right up there.
Hey, we're using your content to make this answer. Here's the citations. We didn't get your permission to do this, but we're doing it anyway. What do you think is the resolution here? Does all these content go away because there's no model? Does Google try to make peace with the content creators and cut them in or license their data?
What's going to happen to content creation when somebody like Google is just going to take Wirecutter or these other sources that are not behind a paywall and just give you the goddamn answer?
Well, look, this is the same conversation we've had two or three times where we're going to need the courts to figure out what fair use is. And depending on what they come up with, it may be the case that Google has to cut them in by doing licensing deals. We don't know the answer to that yet. By the way, I do know a founder who is already...
Skating to where the puck is going and creating a rights marketplace so that content owners can license their AI rights to whoever wants to use them. I think that could be very interesting.
I had a call with him yesterday and you and I will be on that cap table together once again.
So I don't want to say who it is because I'm going to let him announce his own round, but I'm only participating in the seed round. Look, stepping back here, it's interesting. If you go back to the very beginning of Google, the OG Google search bar had two buttons on it, right? Search and I feel lucky. I feel lucky was just tell me the answer. Just take me to the best result.
And no one ever did that because it kind of sucked. Then they started inching towards with OneBox, but you didn't get the OneBox very often. It's very clear now that Gemini-powered OneBox is the future of Google Search. People just want the answer. I think that this feature is going to eat the rest of Google Search. Now, it's a little bit unclear what the financial impact of that will be.
I think like you guys are saying, there'll probably be more searches because search gets more useful. There's fewer links to click on, but maybe they'll get compensated through those relevant ads. Hard to say. You're probably right that Google ultimately benefits here. But let's not pretend this was a deliberate strategy on their point.
They got dragged kicking and screaming into this by innovation and perplexity in other companies. They had no idea. They got caught completely flat-footed, and they've now, I guess, caught up by copying perplexity. And sucks for perplexity. I think they're kind of screwed now unless they get an acquisition deal. But perplexity came up with the idea of having citations
Having a comprehensive search result, yeah, which was something... Search result with citations and related questions. And they did it extremely well. And quite frankly, all Google had to do was copy them. Now they've done that. And I think it does look like a killer product.
And by the way, this was all something that I saw 15 years ago when I did Mahalo, which was my human-powered search engine. And which I had copied or been inspired by Naver and Daum in Korea. They were the first ones to do this. You know it, Shamath, because there were only three or four markets where Google couldn't displace the number one. Korea, Russia, Japan.
Russia had, what was the Russian search engine? God, I can't remember now. Japan had Yahoo Japan, which Masayoshi San had carved out. It was never part of it, and they were loyal to that. And very nationalistic Koreans and very innovative folks at Daum and Naver just made search that was so amazing. You do a search and you'd be like, here's music, here's images, here's answers, here's Q&A.
It was awesome. But, you know, it just shows you like you need to have a lot of wherewithal and timing is everything as an entrepreneur. My timing was 10 years too early and the wrong technology. I used humans, not AI because AI didn't work 15 years ago.
One thing I would say about big companies like Google or Microsoft is that the power of your monopoly determines how many mistakes you get to make. So think about Microsoft completely missed iPhone. Remember, and they like they screwed up the whole smartphone, mobile phone era. And it didn't matter. Didn't matter. Satya comes in, blows this thing up to a $3 trillion public company.
Same thing here with Google. They completely screwed up AI. They invented the transformer, completely missed LLMs. Then they had that fiasco where they had- Black George Washington. Black George Washington. Doesn't matter. They can make 10 mistakes, but their monopoly is so strong that they can finally get it right by copying the innovator.
And they're probably going to become a $5 trillion company now.
It reminds me, you know, the greatest product creation company in history. I think we all know who that was. And take a look down memory lane. Here are the 20 biggest felt Apple products of all time. The Apple Lisa, Macintosh Portable. We all remember the Newton, which was their PDA. The 20th anniversary Macintosh Super Sexy.
People don't remember they had their own video game. I was at a conference a couple of years ago that Jeff Bezos spoke at. I think he's given this talk in a couple other places. You could probably find it on the internet. But he talks about Amazon's legacy of failure and how they had the fire phone and the fire this and the fire that. And he's like, our job is to fail.
Big swings.
We have to make these blunders. But what makes us successful is that we learn from the failures and we make the right next decision.
Yeah, but I say if you're a startup and you make big failures, you usually just go out of business.
One and done. But this is how you stay competitive. If you're a big founder-led tech company, the only way you're going to have a shot at staying relevant is to take big shots that you're going to fail at. I remember this one. The iPod HiFi. You have to do things that you're going to fail at.
Right. Remember this boombox? This is one of the huge differences between startups and big companies is that big companies can afford to have a portfolio of products. They have a portfolio of bets. Some of them will work and that keeps the company going. Startup really has to go all in on their best idea. Totally. I always tell founders, just go all in on your best idea.
They're always asking me for permission to pivot. And I always tell them, go for the best idea. Don't hedge. Don't try to do five things at once. Just go all in on your best idea.
Yeah. And if it doesn't work out, you reboot and start with a new cap table. You're going to go all in. So to speak, another amazing episode is in the can. The boys are in a good mood. You got your great episode. No guests this week. Just all bestie all the time. And very important, the march to a million continues halfway there. You got us there, fans.
We hit 500,000 subbies on YouTube, which means you all earned a live Q&A with your besties. Come Coming at you in the next couple of weeks. We're going to do it live on YouTube. So if you're not one of the first 500, get in there now so you get the alert. We're going to take your questions live. It's going to be dangerous. Any questions. No questions are offered.
Who knows what could happen on a live show. And by the way, I just want to let you know that Phil Hellmuth, breaking news. Phil Hellmuth and Draymond Green just resigned from OpenAI. We didn't get into that, but the OpenAI resignations continue. Phil Hellmuth has tweeted he's no longer with OpenAI. You guys like my baby cashmere pink sweater? It's pretty great.
Are we going to get summer Chamath soon? Are the buttons coming down? Are you going to go linen? When does linen Chamath show up? The unbuttoning is about to happen in the next two or three weeks. The great unbuttoning. This is how you know. It's kind of like Groundhog Day. You know that summer's here when you lose Chamath's buttons. Almost here.
No, it's Memorial Day. Memorial Day. Memorial Day, the button can come down.
Yeah. We're going to go three buttons down. I'll still be wearing my black tee. Sacks will still be blue blazer, blue shirt, red tie. And Freeburg in fields of gold. Look at Freeburg in fields of gold. Taking us out. Staying in fields of gold. Coming at you. Two for Tuesday.
See you all in the next All In Pod for the Sultan of Science. The Rain Man, David Sachs, and Chairman Dictator, I am your Z100 Morning Zoo DJ. We'll see you next time.
Love you, boys. Bye-bye. Bye-bye.
We'll let your winners ride.
Rain Man, David Sachs.
Open source it to the fans, and they've just gone crazy with it. Love you, Wes. Nice.
We need to get merchies already.
And now the plugs, the all-in summit is taking place in Los Angeles on September 8th through the 10th. You can apply for a ticket at summit.allinpodcast.co. Scholarships will be coming soon. You can actually see the video of this podcast on YouTube, youtube.com slash at all-in or just search all-in podcast and hit the alert bell and you'll get updates when we post.
And we're going to do a party today. In Vegas, my understanding, when we hit a million subscribers, so look for that as well. You can follow us on X, x.com slash theallinpod. TikTok is all underscore in underscore talk. Instagram, theallinpod. And on LinkedIn, just search for theallinpodcast. You can follow Chamath at x.com slash chamath.
And you can sign up for a Substack at chamath.substack.com. I do. Freeberg can be followed at x.com slash freeberg. And Ohalo is hiring. Click on the careers page at ohalogenetics.com. And you can follow Sachs at x.com slash davidsachs. Sachs recently spoke at the American Moment Conference and people are going crazy for it. It's pinned to his tweet on his X Profile. I'm Jason Calacanis.
I am x.com slash Jason. And if you want to see pictures of my bulldogs and the food I'm eating, go to instagram.com slash Jason in the first name club. You can listen to my other podcasts this week in startups. Just search for it on YouTube or your favorite podcast player. We are hiring a researcher.
Apply to be a researcher doing primary research and working with me and producer Nick, working in data and science and being able to do great research, finance, etc., allinpodcast.co slash research. It's a full-time job working with us, the besties. And really excited about my investment in Athena. Go to athenawow.com and get yourself a bit of a discount from your boy, J-Cal, athenawow.com.
We'll see you all next time on the All In Podcast.