0:18
Hey, [music] hey, hey. Hey, hey, hey. Heat. Heat. Heat. Heat.
Search this video
1:11
[music] [music] [music] Hey, hey, hey. >> [music] >> Heat. Heat. Heat. Heat.
3:15
You to this world [music] to reach the stars.
3:25
We came to this world to see a future.
3:35
[music] We came to this world to reach the stars.
3:45
>> [music] >> to shape our future.
3:50
[music] We came to feel the new >> [music] >> I see a large IPO on the horizon. >> Triple glaze. >> UAV [music] online.
4:39
>> [music] [music] >> You're watching TVPN.
5:00
Today is Wednesday, November 19th, 2025.
5:02
We are live from the TVP Ultradome, the temple of technology, the fortress of finance, the capital of capital. Ramp. com. Time is money. Save both.
5:11
Easy to use corporate cards, bill payments, accounting, and a whole lot more all in one place.
5:15
Uh, thank you to um the good folks over in Australia.
5:21
>> Ben, >> Ben Sans from Strong Compute sent a whole crate of Violet Crumble.
5:25
If you know, this is my favorite piece of candy in the world. It comes from Australia.
5:31
It's their greatest export.
5:31
It's why we need to defend them at all costs.
5:33
It's why they belong in AUS.
5:35
It's it's the backbone of uh geopolitical uh protection in the Pacific.
5:41
Uh so strong compute for TBPN visualize every data center announcement interactive in real time for GPU cluster users.
5:48
See and control all GPUs in all clouds.
5:51
Ben Sand sends this from there says visualize any cluster.
5:56
Thank you to the team for sending >> very thoughtful >> enough violet crumble for a lifetime. >> What a crazy.
6:03
>> What do we got today, John? What's your take?
6:06
>> My take is does do you want iMessage in Gemini 3?
6:11
Do you want iMessage in your AI assistant in your personal super intelligence?
6:15
After Meta Connect, we left saying, "Wow, the virtual reality, the Call of Duty heads-up display is here. It's arrived.
6:23
The Meta Rayban display."
6:24
And and the technology was really cool, like like the glasses didn't look that crazy.
6:29
Uh and the heads up display, like the actual HUD was really high quality.
6:35
Like you could actually read what was going on there.
6:38
Um but where we left it was, wow, if it doesn't work with iMessage, I can't imagine wearing that because my whole life is iMessage.
6:45
And and I and I was just kind of reflecting on this idea that like iMessage has kind of emerged as my personal ERP system.
6:51
remember when VCs used to be like, "Oh, we need a personal CRM."
6:57
And it was like, >> you're you've just turned every one of your personal relationships into a business relationship and now you should be using an actual CRM.
7:06
Uh, and many VCs do use actual CRM.
7:09
Even if it's like catching up with coffee uh with a buddy from their MBA program or whatever, like people will track that because it makes sense.
7:17
These are professional relationships, so they should be professionally managed.
7:19
Um, maybe in a CRM like Adio, adio.
7:21
com, [applause] >> the AI native CRM. >> The AI native CRM. Where is Adio?
7:26
Here >> I have a new uh I have a new list.
7:32
>> I'm getting I'm getting the blood flowing this morning.
7:35
>> I'm glad I'm enjoying some movement.
7:37
>> But um personal CRM never took off.
7:37
And I noticed that like iMessage has kind of become like my personal data lake, my personal ERP system.
7:46
Like it's my single pane of glass.
7:48
Like if it's it's the it's the source of truth. Yeah.
7:50
It's like the it's like the system of record for my personal life and also we use it for business and stuff.
7:55
I don't know how unique I am.
7:57
I feel like a lot of people are are are stumbling into this world, sleepwalking into this world where uh they they bought the iPhone.
8:03
They were like, "Yeah, it's cool. It's got all these apps.
8:05
Like I I could switch to a different phone."
8:08
And like truly you can't if your whole life is in iMessage because there's so many different chats.
8:13
There's so many different like you know the images and like iMessage has really really grown to the point where it's not just like one-on-one text messages.
8:18
It's all these group chats.
8:21
It's sharing of locations and and documents files files that were shared, you know, PDF that was [clears throat] shared over a year ago. >> Totally. Totally.
8:29
Uh and so um and so my question is like it seems like like iMessage is important for the heads up displays for the for the smart glasses.
8:40
uh will it be important for Gemini?
8:40
And we were debating this like right now uh iMessage when you go in there like the the only AI experience you see is like those Apple intelligence summaries which are uh sometimes very funny.
8:52
I I was laughing about uh it it's summarizing one is it declared over because you someone if someone says it's so over it will just like rewrite these.
9:01
It doesn't get the jokes.
9:02
Uh, other times it'll just say PNG image shared and like sometimes those funny, sometimes it's a little bit useful.
9:08
But, um, in general, I think that all the Apple intelligence features will get better with Gemini 3.
9:14
We saw on the benchmarks, we demoed the product.
9:16
Uh, Gemini 3 is definitely a great model, the best model potentially right now.
9:20
Um, Apple will be able to implement that all over the place and they just won't have to worry about like, do we have a good foundation model to build on?
9:28
Um, so they'll be able to stuff it everywhere.
9:29
But what does the actual flowback look like?
9:31
because uh Google and Apple are famously like walled gardens.
9:35
Like you can't really just interface with them.
9:39
>> Some of the best walled gardens of all time.
9:41
>> Some of the best walled gardens of of all time.
9:42
And I was wondering about if you if I'm if I'm using So the average consumer will just see Apple Intelligence and they'll really just see Siri and they'll be like, "When I ask Siri the history of the Roman Empire, it does a great job giving me the history of the Roman Empire.
9:59
doesn't necessarily uh get confused and hallucinate because it's using Gemini 3 under the hood.
10:03
But the the consumers I don't think will will expect if they wander over to Gemini 3 hosted on Google Cloud Platform or Google AI Studio.
10:14
Go to AI do uh go to what? Gemini 3 Pro.
10:19
Google's most intelligent model with state-of-the-art reasoning, next level vibe coding, and deep multimodal reasoning. AI. studio/build. That's the URL.
10:29
Um the uh um I I I think I think people won't necessarily expect that that if they're interfacing with Gemini over in Gemini world in the Gemini app or in Gmail, they won't expect it to connect to their iMessage, even though it's the same model that's powering both of those.
10:47
And Apple will say that that's for privacy reasons and consumers won't know to ask.
10:52
But I'm kind of curious about that because that would be an interesting feature and I don't know if you would even want that.
11:02
Like would you want to be able to go to the Gemini app and have it be able to pull a, you know, a file that was shared with you in an iMessage group chat and then do something with that in the Gemini app. Is that a feature?
11:16
The only thing that I can think is I I feel like uh my entire life runs on iMessage >> and it doesn't feel like Apple is super motivated like actually building for power users.
11:30
And so if there was a way to get more value having that data within Gemini, right?
11:34
like, "Hey, draft me like text message responses to >> uh people that I've texted, you know, more than more than one day, uh that I haven't responded to in the last two weeks and have draft a bunch of messages that I can then just go through and >> uh at least like look over and respond to." >> Yeah.
11:54
>> Um but I I don't know.
11:54
I'm I'm I I have zero faith that there will be portability, >> any jumping of the of the wall.
12:01
And the reason for that is Apple's paying Google >> to white label >> to effectively yeah white label the model leverage Gemini in the next version of uh uh of Apple intelligence. >> Yep.
12:13
>> And uh they're just going to be focused on integrating it within >> their ecosystem deeply.
12:17
And I think if if they weren't paying for it, Google would have been able to negotiate for quite a lot more and potentially more interoperability between between the products. >> Yeah.
12:28
I I I feel like they're there might be some magic that comes out of uh you know like a deeper integration between these two things.
12:36
It does feel very different than Google search because the models are actually intelligent and could uh it it's I I think that the the obvious like you know draft a summary like the example that you gave um draft a response to a text message.
12:51
I don't know if anyone would even want that and and I do think that Apple intelligence will be able just to do that out of the box.
12:56
Um, I I I'm imagining more of like of like when I when I go to an LLM to prompt it for a gift guide, if it has access passively to iMessage, it can understand, oh, like people have been sharing these links with you to things that could be gift.
13:15
Here's the context around the context.
13:17
Maybe they shared that link with you being like, lol, I would never buy this someone for someone for Christmas.
13:21
or they could have been from a family member saying, you know, this has been like I'm I I would write to Santa for this and they're like alluding to the to you actually wanting to buy them for that.
13:32
Uh so Tyler, what do you think?
13:34
>> Um I I think like when when I think of like AI in like communications generally, I think it's more like >> the vision is like um let's say I'm trying to set up a meeting with Jordy.
13:44
It's like I have an agent, my agent talks to Jord's agent. Yes.
13:46
They sort everything out if we should meet, when we should meet, where we should meet, and then it's kind of like done completely separately from like iMessage even. Yeah.
13:54
>> So, I I think that's more of like my kind of ideal vision of like what LLMs and messaging like look like where it's basically like I'm not even doing the actual messaging.
14:04
>> I'm not sure how important it actually is that it interfaces with iMessage.
14:06
I mean, obviously, it's like good to search through your messages.
14:09
That's like useful, but >> yeah.
14:11
I just wonder like like the reality of everyone's life is that they use multiple messaging systems.
14:15
They use email and WhatsApp and Signal and then iMessage and Twitter DMs and there's never been a successful unification of these.
14:27
Um, but I was laughing to myself thinking about like a humanoid robot because like a humanoid robot you could literally just like be like here's the phone, here's the passcode, go respond to every message on my phone and like it could do that and it would be impossible to like there's no like data wall that you can put up at that point really. Yeah.
14:45
I mean, maybe if you're like world coin scanning constantly to, you know, like eyeball scanning to get into the actual uh the actual app or something.
14:53
But um it reminded me of like George H.
14:56
Hot was saying that like at a certain point the the the full self-driving like it's like uh you don't need to worry about car compatibility because it's just a humanoid that gets in the driver's seat. You want a driver.
15:07
I thought that was such a funny take.
15:08
Uh because it's like yeah like uh right now Toyota I believe it's Toyota but a few of the car makers are basically saying like no third-party self-driving kits like we are encrypting our OBD2 ports like the actual port where you control the car.
15:21
We're not going to let anyone build on top of us because we want to own the self-driving stack on top of our vehicles. Yeah.
15:29
>> Uh so no third party u kits and uh it's just very funny to imagine like well how are you going to stop a robot from just sitting in the driver's seat and shifting the gears and uh and and and pushing the pedals.
15:41
Um anyway Reream one live stream 30 plus destinations.
15:44
If you want to multiream go to reream. com.
15:46
Uh, Lisa Lissan Algib has more Gemini context.
15:53
Said, uh, Gemini 3 Pro is the first LLM to beat professional human players at Geogesser. Wow.
15:58
We got to watch uh who's that who's that the the amazing Geogesser guy?
16:04
Does he just go by Geogesser?
16:06
What's his >> Oh, I know who you're >> You know what I'm talking about.
16:08
Um, the greatest game of Geogesser. This is the guy. He's in the thumb. >> Rainbolt. >> Rainbolt. Yeah, Geio Rainbolt.
16:15
Uh I I I I want to see his uh his reaction to that and see how how he's doing.
16:20
Um >> he's just crying crying on stream.
16:26
>> He's done a he's done a few like uh >> this is this is one of those things that I think is actually still going to be wildly entertaining >> even when even when the like chess, right?
16:35
Like watching him >> figure out where something is down to a single street >> is still going to be impressive and and probably entertaining.
16:44
>> It's a pretty cool benchmark.
16:44
I'm I'm surprised by this, but uh what is this?
16:48
Oh, so it got a higher score, but lower country percentage than a professional player. That's fascinating.
16:53
I wonder I wonder what that says.
16:56
So So it outperformed on score, but it underperformed on guessing the country.
17:02
And I wonder if that's something like uh it's using different heruristics uh that are like less intelligible because a lot of the heruristics that you'll watch the uh the the geogers use the really good professionals is that they will be able to identify like this color of of signpost is only used in this country.
17:23
So even though it looks like it's a tropical like that helps me understand it's this country and not that country.
17:29
And that might be something that uh Gemini 3 Pro is not picking up on, but it's doing it's still doing a better job of understanding just the the references.
17:36
Um also, I mean this this feels like it has to be like overfit on geoging because like didn't Google create all the geogesser like data source, right?
17:46
[laughter] >> Yeah, it's all just Google Maps.
17:49
>> It's Google Maps and like it has to be in the training data like perfectly.
17:50
So, uh, even if it's like not intelligently thinking, like the beauty of watching someone play Geogesser is that they're they're not just doing memorization.
18:00
They're not just like, "Oh, I know that street.
18:02
I know every street because I've memorized every street."
18:03
They're they're actually applying a whole bunch of huristics and patterns and matching.
18:09
>> Yeah, that's probably true.
18:09
But also, I remember with the I think it was the um GP5 release, people would would like submit just a picture they took like on their phone of like themselves. It's like where am I?
18:19
So that's not like actual I mean that's not from Google and it would still do like incredibly well. >> Okay. Yeah. Yeah.
18:24
Also, this is >> Yeah, they nerfed that pretty quickly because there was so much there was it was could easily be abused.
18:29
I remember I uploaded a picture of of outside my house and I could I could tell I could tell by its response that it knew exactly where it was even though there's no street view >> um cuz it's a private neighborhood and like I it was basically like saying where it was.
18:47
Like I knew it knew exactly where we were.
18:49
>> I knew it knew, >> but it was it was just wasn't giving like specifics, but it so much like it it was within like at least like a mile. >> Yeah. 2.
18:58
6M says we should play a round of Geogesser on stream.
19:01
Uh we should we should get we should figure out how to actually wire up uh like games.
19:04
We've done it once before uh and it was pretty fun.
19:09
>> I'm also curious where um the deep think model uh ends up on this because this is still just this is just three pro.
19:14
>> Deepthink must be do even better, right? >> Yeah.
19:16
Yeah, I mean you would imagine. Yeah. >> So, yeah.
19:18
How how would you benchmark the uh the 3 Pro versus GPT5?
19:20
Because it seems like 3 Pro is not equivalent to to 5 Pro.
19:30
5 Pro is more like deep think. >> Uh yeah.
19:33
If you're looking at like price and like the >> how long it takes how long it takes to generate up. >> Got it. >> Yeah.
19:39
So 3 Pro is like five instant or is it like five thinking?
19:45
>> Uh it's five thinking.
19:45
And then three flash if that comes out or three light. Yeah. Like like 2.
19:49
5 light or flash or there's flash light also. >> Okay.
19:53
>> Uh that's more the instant model.
19:55
>> So it feel Yeah, it feels like most of the labs are coming out with like three variations on speed right now.
19:59
Maybe something along those lines.
20:03
>> And then maybe a deep research product adds like a fourth to the end.
20:04
Uh but that's like more more of a specific uh >> yeah like anthropic has uh sonnet, haiku and opus.
20:12
Those are like the three.
20:12
And then there's like thinking on all of those, but >> it's kind of a similar breakdown. >> Yeah. Um, fun.
20:18
I wonder I wonder if Gemini will do a model switcher at some point.
20:24
Like right now, I mean, I guess like AI mode has some of that, but uh maybe they just are they they just don't have to worry about the actual uh GPU cost at this point.
20:35
So, they're not authority needs it.
20:37
He he couldn't figure out how to find the the thinking model. >> Oh, yeah.
20:41
You need the you need the switcher. you need the switcher.
20:42
Uh it is it is funny that um >> to to I asked the model >> what model are you and then it said that it didn't have access to Gemini 3.
20:54
>> Yeah, it is that is something that they should like hardcode in because it is very frustrating.
20:59
It's happened a number of times where >> it just makes it feel not intelligent. >> Yeah.
21:04
Where where where said like okay like like like I want to use the latest and greatest.
21:07
How do you actually do this?
21:08
They should uh they should definitely like make that URL or that explanation like available in the prompt so that it it can answer questions like you need to sort of like bake in an FAQ since you imagine that people will be interacting with the chat directly. >> Yeah.
21:23
Well, it seems like there's some difference between the naming conventions, right?
21:25
Where like uh the the like lab like uh DeepMind wants to come out with it's like a new model, right?
21:31
So it's three it has a number but then on the product side you see it's like numbers are kind of confusing. faster thinking.
21:38
But then for people who like the the the the name scheme is is very funny right now.
21:48
There's I mean everyone has like different different models fast and thinking but then there's also like uh deep research which is deep and then there's deep think deep thinking and deep research and that's very hard to communicate the difference between there unless you're following the stuff very closely.
22:05
and then the create videos with VO, but then instead of create images with nano banana, it's the nano, it's the banana emoji and then just create images.
22:14
And so there's like not a lot of uh like symmetry in the U in the way the UI is laid out because um I think everyone's moving so fast in this category that it's like just get it out, ship the code word, oh the code word leaked, we got to go with it.
22:27
Like there are still people who know strawberry in the context of open AI which is like a wild thing uh to to be at the level where like no one knows like the code word for the next iteration of the diet coke can or whatever like I'm sure that internally there was some project for this but like there aren't like d people following the industry that closely maybe there are but certainly not on the consumer side.
22:50
So uh yesterday Google announced uh Google anti-gravity their new uh agentic development platform.
22:58
Marvin vonhagen >> uh one of the most powerful names intact said which IDE did they use to build anti-gravity windsurf or cursor uh and Silus over at Cognition said so Google just forked the Windsorf codebase and they even forgot to remove the Cascade branding in some places.
23:15
Cascade uh isn't is a is a part of uh of Windsurf's product which is obviously now in by cognition.
23:22
This is funny uh that that that that they kind of missed this and I think it's fair for the cognition team to dunk on it.
23:30
That being said uh they of course Google did buy you spend uh however many billions on on acquiring the the Windsurf IP.
23:40
So >> not super surprising.
23:42
But yeah, I mean you'd think like like step one is >> find and replace, you know, just find and replace and just like anywhere in the codebase remove the old branding and >> put in the new branding. Do you have anyone?
23:56
I think it's actually kind of hard to do this.
23:58
I remember like >> I mean it should be really easy, but I I remember like months after the um Twitter XT takeover. Yeah.
24:05
>> You would still find on on docs Twitter branding.
24:07
I mean that was still like >> months ago.
24:09
I I would I would see that is true. >> Yeah.
24:12
less imperative to actually make those changes in my opinion, right?
24:16
It's like >> also that's a living that's a living breathing service and and like that might be a little bit difficult if it's like you know twitter.
24:23
com is baked into some DNS and if you switch it live like you're going to have a bunch of downtime or something like that.
24:29
Uh like like this is a new product like you can you could just like the codebase is just dead.
24:36
It's just sitting there like waiting to run and then you're just about to ship it.
24:40
You think you do control?
24:42
>> Also, I think this was in their launch AI.
24:44
If you got the best AI, you think you'd say, "Hey, go and fix this. Go make this change."
24:50
>> I also think this was part of >> Wasn't this a part of Google's launch?
24:54
Wasn't it in the launch video?
24:54
I'm pretty sure the screenshot is from the launch video. >> Oh, really?
24:59
>> Which makes it >> No, no, no, no. I don't think so.
25:00
I I I I don't think this is in the uh the launch video.
25:04
The launch video is like very minimal.
25:05
Uh and and and this is like clearly has like a streamer in the corner like looking at it.
25:10
Um >> but anyway, whether you are uh excited and bullish or bearish on Google because of this, head over to public.
25:18
com investing for those that take it seriously.
25:21
They got multiasset investing.
25:22
They're trusted by millions.
25:24
Uh Kyle Chan says, "This is the big story here.
25:27
Google train Gemini 3 Pro on Google's own TPUs.
25:30
No mention of Nvidia chips." This is pretty crazy.
25:33
I mean, they've been doing this for a while, but uh Nvidia's announcing earnings today and um >> it's pretty crazy.
25:42
The biggest story in AI is not is not really that relevant to the biggest company in AI.
25:48
>> Best model ever created. Yeah.
25:48
From a benchmark standpoint. >> Yeah.
25:53
>> Didn't use didn't use Nvidia chips, which are supposed to be a monopoly, >> right? >> Yeah.
26:02
>> And uh and so yeah, I don't know.
26:02
I this doesn't feel fully priced in yet >> to either company. Yeah.
26:09
>> But then again, right, it's so hard to predict demand over the next 5 10 years that maybe maybe it doesn't even matter. >> Yeah.
26:16
I wonder I wonder how much because if if TPUs are not for sale, Nvidia does have an monopoly.
26:23
Like you can if there's, you know, a monopoly on if if Nvidia truly is the only seller in the market because Google is not a seller.
26:33
Uh then yes, they still extract they still extract monopoly power from every other buyer because every other buyer says, "Yeah, I'd love to buy TPUs, but I can't."
26:41
So, you're the only game in town still.
26:43
Um but it's a very weird dynamic where you do have two very clearly performant products that are not uh that are not actually driving down cost.
26:51
Uh it must be very frustrating if you're somebody else.
26:54
But that's why every other all the other labs are working so hard to uh to develop their own chips or you know bring AMD online.
27:00
And there's a whole bunch of different uh uh efforts in this.
27:05
>> Do you know the background here on on uh from this post?
27:08
It's extremely Google that a flagship consumer product is named as a reference to inner org drama that happened three years ago.
27:15
>> Well, there's lots of people saying that they require context.
27:17
Let's see if anyone >> anti-gravity. >> No. Oh. Oh.
27:21
Is anti-gravity the reference? Zodiac.
27:24
The zodiac Gemini refers to twins.
27:27
Google's Gemini is a reference to two formerly distinct labs, Google Brain and DeepMind, that were merged into one lab, Google DeepMind. >> I think that's it. >> Um, yeah.
27:40
And I guess the interorg drama that happened three years ago was just this idea of of you know, Deep Mind was acquired in, but Google Brain was still running.
27:48
This is Isn't this a reference to Gemini as in the constellation of the Gemini twins referring to the consolidation of twin organizations? I like that.
27:55
That's actually a pretty good name.
27:57
And uh I mean it it this original post makes it sound much more dramatic like interorg drama. Yeah.
28:03
Um but in fact it's it's it's sort of a way to keep uh keep the lore going basically.
28:13
>> Um anyway, let me tell you about adquick. com.
28:15
Out ofome advertising made easy and measurable.
28:17
Say goodbye to your headaches. Out of home advertising.
28:19
Alex he has a Q&A with Dennis over at DeepMind.
28:25
He says world models Alex is on a tear. >> He's on a tear.
28:28
Um he's averaging like the challenge when writers go and when when journalists go independent >> Yeah.
28:35
>> is actually figuring out a way to get enough scoops to justify a subscription business model as a standalone company.
28:43
>> Alex has been it's basically been at least a scoop a week or or like very interesting content. >> Big stuff too. Yeah.
28:50
Um, and uh I don't know just like it's interesting seeing like uh there's a bunch of there's a bunch of interesting things.
29:00
I mean uh he did that interview with Mark Zuckerberg that was uh like seated hour-long, you know, in-depth interview.
29:08
There were some big scoops that came out of that.
29:10
Some funny takes about the uh about the the the bubble basically.
29:12
I think Mark was saying like, "Yeah, we might overspend."
29:16
Uh and that was sort of a viral moment.
29:18
and then um doing some Q&As's.
29:20
Also, uh it feels like maybe great timing too.
29:23
Uh just you look at the stats on this um 57,000 views, 477 likes link in the core image. I mean, I get it.
29:34
It's like it's a it's an important scoop. It's an important story.
29:37
Uh but um it feels like a year ago this post would have been buried by the X [laughter] algorithm.
29:44
uh and so people will show it three people like like really really great timing on that as well.
29:50
So uh just just catching you know different uh opportunities and capitalizing constantly.
29:56
Uh so very very exciting.
29:58
But the actual uh quote that Alex Heath is sharing from his piece and sources.
30:03
Which you can go subscribe to is uh he says world models are the thing I'm spending most of my research time on. I'd love more TPUs.
30:09
Uh, you look at seed rounds with just nothing being tens of billions of dollars is not quite logical to me.
30:17
[laughter] Taking shots, shots fired. Do we have the gun? Do we have the gun? >> No, we removed that. >> Oh, we removed it.
30:24
Okay, >> might have to add it back. >> I like that.
30:26
We need a taking shots one. >> Jumping in. Just a note on TPUs.
30:28
Alex says, "When you talk about the constraints, Google has more computing access with TPUs than most companies.
30:34
I would think that Google could just go all in on your team's work.
30:37
But Google also gives TPU access to other startups and even rival AI labs.
30:41
Do you ever just go give me all the TPUs?
30:46
>> And Dema says, "I'd love more, but there are business requirements to balance.
30:49
There's short-term and long-term revenue, and all of these things need to be balanced and smoothed out." It's a huge advantage.
30:54
We have TPUs in our own stack, and we co-design the TPUs with the TPU team based on where we know we're going software wise.
30:59
But yeah, there isn't enough compute in the world, as we all know, for everything that we want to do.
31:04
There are always competing things and then there's the question of what is the return on that amount of compute.
31:10
It can be a research return, a new product investigation return or direct revenue.
31:13
Genie is still in the exploratory phase in terms of what we may eventually do with it.
31:18
So >> anyways, >> well if you're looking to manage a bunch of TPUs, get linear meet the system for modern software development.
31:26
Linear is a purpose-built tool for planning and building products.
31:29
Greg Brockman uh looks like he's hanging out in DC.
31:33
This has to be Washington DC.
31:35
>> Yeah, this was last night. >> This was last night.
31:36
Uh says he says the future is bright and he's pictured with David Saxs, his wife, of course, uh Elon Musk, Jensen Wong.
31:42
Um what a fantastic photo.
31:46
>> Tyler did a green line analysis on this.
31:49
>> I think Tyler was getting a little wild with this one.
31:50
It's barely barely a >> not beating the >> I don't know.
31:56
I I think he's standing up pretty straight.
31:58
He's a little leaned over.
31:59
Not fully, but we got to pull it.
32:01
Elon is the is standing up extremely straight with some wild shoes.
32:04
People are saying, "What are those shoes for?"
32:08
>> It's Nvidia earnings day.
32:08
You got to look for any signal they can possibly get. >> Yes. >> Pull it up.
32:14
It's at the bottom of the timeline. Squad. >> Uh here we go. Here we go.
32:17
So, everyone is is is is pretty accurate line.
32:23
>> You did him dirty with that line.
32:23
>> I drew the line perfectly. >> I'm I'm with Tyler.
32:26
He's It's all about center of gravity, right? center of gravity.
32:32
>> Anyways, I I'm I I I think uh Jensen will put on a show later uh right after the show ends, so we will uh look forward to finding out more. >> Okay.
32:45
Well, uh Elon was was pictured wearing some very very crazy shoes.
32:48
Uh these are his SpaceX shoes.
32:51
I don't know who made these. Look at these.
32:53
Jordan, would you Brock these?
32:55
Could you pull these off? the team likes them.
32:59
I'm not I'm not I don't think >> Are these are these like uh were they made in collaboration with some another brand or these just >> I don't know.
33:09
>> Is he vertically integrating drip? >> I don't know. Yeah.
33:11
Did he make his own shoes? I have no idea.
33:12
Um >> these would uh these would go for I I have a feeling they'd go for quite a lot.
33:21
>> Yeah, they seem pretty cool.
33:21
Uh, let me tell you about fall to build and deploy AI video and image models trusted by millions to power generative media at scale.
33:32
>> Bobby says they look like Yeezys.
33:35
>> They do look like Yeezys. That's right.
33:37
Uh, Quarter the Quarter app has dropped the Oh, I got to follow that.
33:41
Why am I not following them?
33:43
Um, Quarter App has dropped a uh a an announcement that the Nvidia earnings call will be tonight at 5:00 p. p. m. Eastern time.
33:51
As soon as we log off this stream, you can head over to Quarter and start streaming the Nvidia earnings call.
33:58
Uh, Jensen is there pictured. Uh, all eyes on Wong.
34:04
And, uh, they've done a fantastic job developing this image style.
34:06
I feel like uh it's been 2025 the the meta on X has been exploding in terms of like image macros.
34:17
We've had a ton of fun with the trading cards.
34:19
Um these have done extremely well.
34:22
Uh anything where you can bring design and just tell a little bit more of a story, give a little bit more context, texture, something breaks through.
34:33
And every time they post one of these 3,000 likes, people love them.
34:35
Um, >> scroll down if you can cuz somebody ran this graphic through midjourney and it's pretty >> crazy.
34:42
[laughter] So bad by comparison.
34:45
>> I mean it's it still goes pretty hard. >> Yeah.
34:48
So I mean >> the arm there is looking a little >> it's also an interesting testament to uh like I I know that the quarter designers use midjourney, they use AI, but they are really really deep in the SEFS.
34:59
They obviously have a whole bunch of different stylized prompts and then it seems like they're also doing a ton of work in post-processing, layering text on top of it.
35:10
Uh they've they've created like a visual style that's distinct.
35:12
I'm sure people will copy it, but um it's definitely created um its own its own sort of style and broken through.
35:21
And at the same time, like it it doesn't feel like yes, there's AI involved, but it doesn't feel like um if you just threw, you know, this this prompt at a random person, okay, hey, go make one for, you know, uh Coca-Cola next week.
35:36
I don't know that they could necessarily pull it off even if they had a midjourney subscription.
35:41
Like there's still a lot of like inspiration to understand like what is the texture?
35:45
What is the what is the style?
35:48
>> Right now on Poly Market, will Nvidia beat quarterly earnings?
35:50
is sitting at an 87% chance.
35:51
The real question is how will it trade >> after the fact? >> Yes. Yes.
35:58
It it feels like we're in this weird market where you can beat on earnings and then sell off because nothing is ever good enough for the street right now.
36:06
Um but uh but we we will see the the the Nvidia Nvidia is such a big story now that uh just the fact that they are going to have earnings is uh essentially front page news at least of the business and finance section.
36:20
Uh Nvidia and jobs data coming reports will provide key signals for investors after a market pullback.
36:28
The fog masking the direction of the American economy and future of the artificial intelligence boom is starting to lift after mounting scrutiny of stratospheric tech investments as well as a blackout of federal data during the longest government shutdown in US history.
36:44
Wall Street awaits two reports that stand to reshape its outlook for the months ahead.
36:48
AI poster child Nvidia is due to report earnings after the closing bell Wednesday, offering a snapshot of demand for chips that are in that are a lynchpin in the tech mania that has lifted markets and helped buoy the economy.
37:02
Also, with the Nvidia news, it's like how much can you actually read into AI demand based on Nvidia earnings?
37:10
Because I feel like we're we're projecting out like these deals five years in in advance.
37:13
We buy the chips, then we install them.
37:15
like are we really seeing like if like the that whole rumored uh you know uh decline in or deceleration in chat growth like if that is real and that's happening and and Chachi PT usage is starting to plateau from 800 million weekly to hey next year it's going to be like 900 mill a billion like it's not going to be five billion next year.
37:42
next year. uh if that's happening uh are we expecting that to show up in the NVIDIA data this quarter like probably not right because like OpenAI has projected out five years of demand for GPUs so I don't know it seems hard to actually read into Nvidia's earnings as
37:58
a as a as a real uh snapshot of demand I mean I guess demand for chips certainly yeah a sell-off in Nvidia has dragged down indexes with Peter Thiel's macro hedge fund and others dumping shares sort kind of crazy that that's that that's that's in the journal. >> Yeah. Especially especially when when >> Yeah.
38:15
Especially especially when when it's the equivalent of, you know, the average person in tech selling like a a $10,000 position in the company, you know, >> it's like not not like super notable >> with no statement either.
38:26
It's not like it's not, oh yeah, he was also on Rogue and Trash. >> It's like nothing.
38:30
Uh the tremors extended beyond other AI names into crypto, gold, and more.
38:35
Uh, even Warren Buffett's latest big bet on Alphabet hasn't staunched the bleeding.
38:40
America's richly valued stock market has retreated in similar fashion several times during its yearslong runups.
38:48
Uh, in every instance, bargain hunters snapped up stocks, tech companies out profits, uh, and the economy kept on motoring ahead.
39:00
>> Um, the fact that there's Yeah, we can move on.
39:03
Yeah, I mean the reason there's fixation Nvidia's currently we it's like 8% of the S&P 500. >> That's crazy.
39:11
So like it it just it matters more than any other.
39:16
>> This this feels like the most important earnings call of the year given given the sell-off in Neoclouds, given the uh the just like pressure and debate around open AAI, given >> uh given Google's investment and progress with the TPU.
39:32
I mean there's so many different factors.
39:33
Uh in related news, uh it got announced this morning.
39:39
Musk's XAI and Nvidia to develop a data center in Saudi Arabia.
39:43
>> Um, it's a 500 megawatt data center in Saudi.
39:47
Uh, XAI is working with Nvidia and a Saudi Arabian partner to develop a data center in the kingdom.
39:52
Musk said Wednesday at an event with the Crown Prince.
39:56
Uh, they're teaming up with Saudi Arabia's AI company, Humane.
39:59
uh that's going to be 500 megawatts or enough electricity to power several hundred,000 homes for a year.
40:06
The announcement came at the US Saudi investment forum.
40:11
Of course, uh the crown prince announced a trillion dollars of investment in the US uh yesterday.
40:19
>> President Trump touted Saudi's investment in the US and the partnership between the two companies uh countries.
40:25
My question is like what like there's no information here on on how this data center is going to be used.
40:33
Is this do do we expect XAI to be operating and and competing as like a AI cloud or is this going to be something that they're they they want to have a local version of Grock, right?
40:46
And I would and and and to me it seems much more likely that they're just going to be in the >> like they just want to be in the in like the data center business. >> Yeah.
40:55
Oh, that Yeah, that's a very interesting >> and to me that's always made sense because Elon is clearly fantastic that >> pretty much best in the world.
41:01
I mean he was he was mogging Microsoft for bragging about how many million work hours 15 million work hours more than more than uh and so clearly very good at at like large scale physical infrastructure buildouts getting getting access to energy doing things on a ridiculous time horizon and so >> uh in order to support XAI's valuation I could see them trying to get into uh get into that game. >> Yeah. Yeah.
41:29
Yeah, I mean there's also the uh the possibility that uh if there is strong US uh inference demand um but latency is not an issue like it might be valuable to actually just colllocate the uh the data center next to the oil.
41:42
Um so because maybe the energy is cheaper uh midjourneys I believe been doing that for a very long time doing in inference internationally because uh the data center demand uh during peak hours in the United States is more expensive than yeah across the world.
41:58
Um let's pull up let's pull up this video and while we do let me tell you about graphite.
42:03
dev code review for the age of AI graphite helps teams on GitHub ship higher quality software faster.
42:08
Let's go to Elon Musk saying AI and humanoids.
42:12
actually eliminate poverty.
42:16
>> Eliminate poverty >> and Tesla won't be the only one that makes them.
42:18
I think Tesla will pioneer this, but companies that make humanoid robots, >> but but AI and humanoid robots will actually eliminate poverty.
42:26
And Tesla won't be the only one that makes them.
42:28
I think Tesla will pioneer this, but there will be many other companies that make humanoid robots.
42:33
But there there is only basically one way to uh make everyone wealthy, and that is AI and robotics.
42:41
And we can't talk about robotics without AI. >> What do you think?
42:47
>> All all problems in the world solved by one product. >> I love it.
42:54
>> I mean, it's it's not the craziest take uh over over a long period of time.
42:57
You know, you give everyone uh the ability to sort of marshall anything.
43:03
Uh it does it it you know it I wonder if we'll redefine poverty at that point. Yeah.
43:11
>> Uh poverty will be not having a beachfront property, a beachfront mansion or something.
43:14
Something that's truly scarce that uh even even an army is my land thesis.
43:19
>> Even an army of humanoids can't necessarily uh give this is this is you know we joke about land a lot.
43:26
We joke about it being the most undervalued asset by the current uh generation of investors.
43:33
But land is the one thing that even with an army of humanoids like you can't as easily like copy and paste, right?
43:40
Like it's just it's just it truly is scarce.
43:42
It's not like land it's not like land on the blockchain where people were like no like >> you can buy this plot of land on the blockchain and that's yours forever and somebody's like what if I just make another blockchain. >> Exactly.
43:52
This is the most ridiculous.
43:54
>> And I can also get from this piece of land I can get from this piece of land on this blockchain to this other piece of land on this blockchain in a second. >> Yeah.
44:01
[laughter] It's ridiculous.
44:01
If you have enough humanoid robots though, then land is actually not that hard to to get. >> Why?
44:06
Uh because >> you're saying like you would just put enough dirt in the ocean and like it's like oh go you had this ocean or you have a robot >> robot army.
44:15
>> Oh and then you just steal >> then you just take but I think I think what Elon's saying is like if you assume universal basic humanoid army everyone gets 10 humanoids and so the humanoids can cook for you, they can give you shelter. They can clothe you.
44:28
They can give you health care.
44:30
So you get everything that you know would typically be bucketed in poverty, but there's still scarce resources.
44:37
There's only going to be one Mona Lisa and so you got to fight over that.
44:43
>> Uh Orange is in the chat says the humanoid form factor silly make it an R2-D2.
44:46
I'm actually >> surprised. Yeah.
44:49
So one Maddic I got I got to give a shout out to Maddic.
44:51
My Madic has been running in my house every personally been running it in my house daily >> for months now. That's fantastic.
45:02
>> And it has worked flawlessly. >> Yes.
45:04
>> Uh I I I had like, you know, I feel like everybody's been disappointed by like a vacuum robot over the years.
45:12
>> Uh and so I didn't have the highest expectations even though setting it up was like fast and and it got to work quickly.
45:18
But I've been shocked at at how uh just well it's worked and haven't had to think about it.
45:23
You replace the little bag every once in a while and it's great.
45:27
So, uh, but R2-D2 form factors, would love to see more of that.
45:32
>> What do you want out of an R2-D2 though?
45:35
Because the Roomba form factor, the Madic form factor where it goes and cleans, like that's pretty useful.
45:41
Uh, but if you like is it an R2-D2 that it can fold your laundry, do your dishes, like you you need to sort of define a few different because clearly we're we're going to be in the age of like spiky intelligence and also like spiky humanoid usage like they're going to be good at some stuff and >> I think R2-D2 form factor you could reduce the number of motors that you need.
46:03
It could carry more weight, right? >> Carry.
46:06
So it's going to carry you.
46:06
Are you going to ride it? >> No. explain what it's doing.
46:12
>> Because in the classic Star Wars, R2-D2 is like basically just like a hard drive that like carries like a video.
46:20
>> Like that's all he does the whole movie.
46:22
>> Yeah, but I can you can imagine it has it has a number of different >> What does it have? What does R2-D2 have?
46:28
Have you seen the movies?
46:29
>> Doesn't it have like a screwdriver?
46:30
>> It has a [laughter] screwdriver. has a screwdriver.
46:32
It's basically a USB cable that comes out and like plugs in when it's like you could just use wireless to hack the network or whatever. >> That's true.
46:40
No, the the other the other challenge with that form factor is is uh a screwdriver >> you have.
46:45
So, so it's great if you have like a one-story home.
46:46
If you have a second floor, >> R2-D2 is kind of cooked.
46:49
RTD2 starts asking like >> totally cooked.
46:52
[clears throat] >> He's like, "Hey, we're going to need some more capex. I need some capex. Elevator, please. Elevator." RD2.
47:01
I think the op, if you want to talk about like Star Wars form factors, I think the optimal is uh >> General Grievous or whatever the one that has a bunch of arms.
47:07
Basically, he can walk around like a normal human except it has more arms.
47:11
>> No, no, I completely agree with that.
47:13
Bunch of arms form factor is >> Yeah, everybody wants to make a humanoid.
47:15
Nobody's trying to make the the General Grievous.
47:17
[laughter] >> General Grievous >> way better way better lightsabers.
47:24
>> You you distilling R2-D2 as it has a screwdriver.
47:28
[laughter] It's like the most just completely mogged.
47:29
Uh, but I mean truly other than just being like cute, RGD2 can't even speak English. Think about that.
47:37
[laughter] >> RGD2, it just goes beep boop boop.
47:41
Again, >> I mean the the implication was like the general form factor of something that can roll around and has the a lot of different capabilities built in.
47:50
[laughter] >> It doesn't have any capabilities. >> It has no capability.
47:53
>> Can we pull up Can we pull up General Grievous on the screen? [laughter] Okay.
47:58
RTD2 hacked the Death Star and saved Luke from the garbage disposal. Yes.
48:02
Two things that could have been done with wireless networking.
48:04
It didn't need to plug in for that. Why? It is a hard drive.
48:09
I'm getting into battle with the with the chat right now.
48:12
Um we uh uh C can we tell the story of us risking our lives yesterday? >> We really should.
48:20
>> Yeah, this was this was truly incredible stuff.
48:23
So, we we're looking uh we're we're in the Ultradome here for at least another year, but uh we're starting to think about our our second uh the next Ultra Dome.
48:33
We want to get slightly more space.
48:37
There's a number of different things that we want.
48:38
>> There's General Grievous.
48:38
That is the ideal humanoid form factor.
48:40
And if you're not building that, it's a zero. What are you thinking?
48:46
I think I I think Optimus would look way better with six arms. Not scary at all. It is crazy.
48:53
They they they do the quadripeds, but no one's no one's really working on like the six-legged, six armed, like the really crazy creepy stuff.
48:59
There's been a couple humanoid robots that look really scary where they were like, "Wow, let's put it up on the meat hooks." Remember that one?
49:05
That [laughter] was crazy. That was a wild one.
49:09
>> Was that the video where it started going >> No, no, no. That's different one.
49:11
But this was the company that was like, "Here's our here's our presentation.
49:15
Like, we're ready to release our humanoid."
49:17
and they were like hanging it up on meat hooks and the and it looks so spooky spooky because it was using muscle fibers basically.
49:24
>> Nathaniel Smith is uh very bearish on R2-D2.
49:26
A cell phone can do most of what R2-D2 >> completely agree. I R2-D2 cute for sure.
49:32
Like definitely like fun to have around.
49:34
Uh but more of just like a toy companion and and you know we we looked at that lamp and that lamp we were kind of like what is that lamp?
49:41
And I think there's just like it's just delightful.
49:43
Like it's just nice to have around.
49:44
it's just nice to have around. It's like this this turbo puffer here thing search every bite you know serless vector and full text search built from first principles on object storage fast 10x cheaper and extremely scalable like the the turbo puffer I mean obviously we're
49:57
sponsored by them but but this is something you might have in your house just because it's cute people people people like having cute things have asked and and having an R2-D2 in your house would be cute until it runs into a stair and goes tumbling down and smashes into a million pieces. Uh, anyways, so
50:10
Uh, anyways, so we're looking so we we found a space that we love.
50:14
It's it's dome like we're looking for a space in LA that is fit for the Ultradome.
50:20
There's not a lot of things that qualify.
50:23
>> And so we had looked at the space a couple times.
50:25
Uh I had seen it with Ben.
50:28
Uh John and I drove by it and then we went back to look do another walk through and we're getting like I'm like extremely excited space on here's where this thing goes. Here's where this goes.
50:40
You made us on the way to the show in the morning.
50:41
You make me poke p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p poke my uh head through the window.
50:46
>> Then we go back at the keys, we go in.
50:49
>> Really selling John on it.
50:49
Uh it's a beautiful beautiful space.
50:51
It's like a few minutes from where we are now. Made a lot of sense. >> Sorry.
50:55
Ethan says R2-D2 was the original digital guy.
50:57
[laughter] [gasps] >> True.
51:02
>> Yeah, digital guy is incredible for sure. Sorry.
51:04
Anyway, >> uh so so anyways, we go for the third time to this space and I'm just selling John on every every inch of the space.
51:13
I'm like, "This is what we're going to do here.
51:14
This is what we're going to do here.
51:16
Here's where the truss is going to go.
51:18
Here's where the production team is going to go."
51:20
>> And we're just walking around kind of get getting uh getting a feel for it.
51:24
And we're basically wrapped up.
51:24
Like we're super excited about it.
51:26
Not necessarily ready to make an offer on it, but uh but certainly like we're like, "Okay, this is by far the best option that we found.
51:33
We've looked at a bunch of >> He checks a bunch of the boxes.
51:36
>> Checks a lot of boxes.
51:36
And uh right as we're about to leave, John like looks over and there's like a closet door with a key in it.
51:43
And you just like walk over.
51:45
I just walk watch you walk over and like open it up and you start looking looking around.
51:48
And first I make the joke.
51:51
I'm like, "Oh, this is like the the intern closet cuz it's like this really long narrow like hallway thing that's just like a it's it's like the worst room you can imagine."
51:58
And and so the idea of putting Tyler in it was was uh was at least entertaining.
52:01
Uh, and then we're like, "Wait, what's that humming sound?"
52:05
And there's like this box that's like covered up.
52:06
And it's just like this like not super loud, but just like constant humming sound.
52:10
And we asked the the broker, we said, >> "It's super weird because it was drywall."
52:15
Like you walk into this to this big room. It's a big room.
52:17
And then within that big room is a massive drywalled box.
52:23
>> And so with no entrance, >> no entrance to the box, but it's drywalled.
52:26
Like you don't usually see drywall inside of a room that's not doesn't go all the way to the ceiling. And so it's very clear.
52:31
So, we walk into this room.
52:34
>> Well, they were hiding something basically.
52:35
>> And there's no there's no purpose to the room. >> Yeah.
52:38
>> Other than it just stores the box.
52:38
It has no entrance and it's humming. >> Yes.
52:43
>> And and we look around and John's like, "What's in the box?"
52:44
And the broker says, >> the broker says, >> "Oh, that's just the machine that cleans the soil."
52:51
[laughter] >> And we were >> No, no, no.
52:53
She said she said, "Uh, that's just the machine.
52:59
That's just the machine."
52:59
And we're like, "Oh, like >> what kind of machine?
53:03
>> What kind of machine is in there?"
53:04
>> And she's like, "Don't worry about it."
53:05
>> She's like, "Don't worry about it. It's not a big deal." >> Yeah.
53:07
Just like, you know, buildings have machines sometimes.
53:09
There's a machine in there. It's a machine.
53:11
>> It's always on, but you don't It's It We took that out of the square footage, so don't worry. >> Oh, yeah. That was a wild one. >> We're not billing.
53:16
We wouldn't bill you for it.
53:18
>> And so, what type of machine is it?
53:20
>> And she And then she goes, "It's a machine that cleans the soil."
53:23
>> And we're like, "Is this on like some sort of haunted burial ground or something?
53:27
Like, what are we doing down there?
53:28
this is a hazardous waste site.
53:30
And she goes, "Again, really not a big deal.
53:32
I would worry about it if you were going to buy the place, but since you're just planning to lease, don't worry about it."
53:38
And then we were like, "Okay, like the more you tell me not to worry about it, like I I kind of want to know more.
53:45
So, what's it cleaning up?"
53:45
And she's like, "Oh, I mean there there there's it's it's 85% of the way clean."
53:52
We're like, "What's what's getting >> when did this process start? How long will that go?
53:56
Oh, has it been going for a hundred years?
53:58
Is the box 15 years start an hour ago and it's just going to be 15 more minutes?
54:03
Like, you gave us no context to actually project out what 85% of the way means.
54:09
Uh, and and finally she's like, there's there was a laundromat here ago.
54:13
And we start piecing it together and we kind of like don't want to press her on it too much.
54:19
So, we leave and start doing some googling.
54:20
figure out that it's not a super fun site, but apparently there was a uh there was a uh laundromat there that was using toxic chemicals that >> No, it's a machine shop. >> Oh, a machine shop.
54:31
Oh, that's what we figured out.
54:33
So, they said laundromat and apparently laundromats uh can give off toxic chemicals that if they get in the ground can be very cancerous for a very long time.
54:41
This was apparently a machine shop like almost 100 years ago or something and they're working on cleaning the soil.
54:46
Uh, but I still don't even understand how you clean all of the soil under an imassive building without causing a collapse.
54:53
Is it like a whole bunch of tunnels that are digging around?
54:57
>> A bunch of R2-D2 robots.
54:58
>> Maybe it's a bunch of R2-D2s, honestly.
55:00
>> Anyway, so so she she's still saying, "Yeah, I really wouldn't worry about it.
55:04
It's just like not that big of a deal. It's just a machine. It just runs.
55:06
You won't even know that it's running.
55:08
We'll keep the door closed."
55:10
And the and granted, the machine would be like 10 feet from the set.
55:13
So we'd be sitting here doing the show and and you just have the the death machine running right there always.
55:20
So anyway, >> it was very very bizarre.
55:21
It was one of the funniest like just like jump scares ever. Uh very very good.
55:28
>> It was just it was such a good bit too because I'm I'm far more health consscious I think than you.
55:32
And even you were were thinking there's no way we're going to lease an Ultra Dome that has a death machine that always need that needs to run.
55:42
I I just I I found it so fascinating that it could sit there and clean the soil for years with a massive machine the size of a giant room. Uh I want to learn more.
55:54
I want to know what that machine is.
55:56
I want to know what invest in that company. Exactly.
55:59
That's what we got to figure out how to make.
56:01
We got to have the CEO of whoever makes that machine on the show.
56:05
I want to get to the bottom of it.
56:07
Uh we need to do a deep research report.
56:09
Tyler, can you fire off Gemini 3 Pro deep thinking max 247 mode where it works for ages? It works for eons.
56:18
>> So, so I found two groups.
56:18
CDE group, soil washing equipment. >> Okay.
56:23
>> Our wet processing equipment extracts maximum value from hazardous soil and [laughter] >> but so is it just that corner that has the hazardous soil?
56:32
I what I want to know is is it going under the building and then over so that underneath us over here the machine's here.
56:41
Is it digging a tunnel that goes underneath the building and then washes over here too?
56:46
Are are there is there a network of tunnels under that building? I have to know.
56:53
[laughter] We have to go back.
56:53
We have to lease this thing.
56:54
We have to buy the building just to get to the bottom >> to get to the bottom of it. >> I have to know.
56:58
I I I I'm I'm ravenous for information. >> Yeah.
57:02
The cool thing is they use physical and chemical meth methods to separate heavy metals. >> That's super cool. That's super cool.
57:09
>> That's exactly what we want.
57:10
>> That's exactly what we love.
57:10
Um well, in in other news related to >> I found I think I found the machine. We got to pull it up.
57:18
I [laughter] can't we can't leave people hanging.
57:21
>> Yeah, I dropped I dropped one of the makers of these machines.
57:24
>> It sounds like fracking. Yes.
57:24
If we could if we could frack directly some natural gas out of the soil and then use it to power a natural gas turbine that we use to uh you know run the show and power us. I'm I'm down for that.
57:37
Um while you're looking that up, let me tell you about Finn.
57:41
AI, the number one AI agent for customer service.
57:43
If you want your AI to handle customer support, go to Finn. ai.
57:49
>> Um so in the water news, okay, you want to pull up that and then we can go into the water news.
57:53
I got to talk about Andy Massley at some point this show.
57:55
I >> I just I want to see your reaction when you start to see >> the scale of this >> the scale of this contraption and how it pretty much perfectly fits into >> the box. >> Yes. Yes. Yes.
58:08
Fracking with extra steps. Language, please. Was someone swearing? I don't know.
58:13
Um anyway, uh let's pull that up.
58:17
Let me also tell you about profound.
58:19
Get your brand mentioned in chat. GBT.
58:21
We reach millions of consumers who are using AI to discover new products and brands.
58:24
Um, let's see about uh this water story.
58:28
Andy Masley is going back and forth with uh what's what's her name? Karen.
58:35
Um the uh the AI and the environment somewhat related to our own environmental story that we could kind of go through.
58:45
Um if >> How we doing, boys? >> But >> there we go. Look at this, John. Okay. Okay.
58:52
>> This is from GN Separation Core Equipment for Contaminated soil washing.
58:57
And you just look at this machine.
58:57
This is pretty much exactly what would have been in the >> in the in the room. >> Soil washing.
59:06
You have to wash all the soil.
59:09
>> And there's a graphic if you scroll down a little bit.
59:12
>> I want to know how much like this.
59:14
[laughter] >> It's a simple process. It's 85% done. It's >> It's 85% done.
59:21
>> We just need to get the hazardous waste into the decanter centrifuge and then get it into the non-acceptable solid second wash.
59:27
Then take the acceptable solid up to the coarse screen into the washing fine screen and then take the washing chemical and bring it up into the washing reaction tent tank.
59:36
Put it back in the centrifuge.
59:38
Push it down into the soil filter press dewatering screw press. Okay.
59:45
>> And then move it back up through the hazardous waste.
59:47
John, and you're you're good. You're good, >> Doug.
59:49
So So Doug is asking uh if it's behind drywall, like is that because it generates fumes? We have no idea.
59:54
Maybe it does generate fumes.
59:58
>> We don't know how they access it without knocking down all the drywall.
1:00:00
>> Yeah, we don't know how they access it. Is the drywall just up?
1:00:01
And also, I really want to know like was there another entrance that someone could go into like like what if the machine breaks while we're there?
1:00:08
Does someone come by and change out something?
1:00:10
Does this machine need to be turned off at night?
1:00:14
Does it require is it fully automated?
1:00:16
Does it just run for years?
1:00:19
Would we have never seen a technician come by? What if it gets jammed?
1:00:21
Like, is it just the most flawlessly built machine in the world that never breaks? That seems unfathomable. All machines break.
1:00:29
All machines need some level of of attention from time to time.
1:00:31
But maybe it's the most perfect machine possible.
1:00:36
>> And the machine is of course made by Hey GN Solids Control Co.
1:00:39
Uh, which is a China based company. >> Wow.
1:00:43
Well, we don't know that this is the actual machine, but who knows?
1:00:44
Um, anyway, let's go over into the environmental impact of artificial intelligence.
1:00:51
Uh, there was a very funny post from Henry Thunberg who says, "Whoa, I had no idea that AI uses 5,329, 584 water per year.
1:01:04
[laughter] That's insane."
1:01:06
Like, it uses just one water.
1:01:11
Uh, yeah, people are all over the place with the water thing.
1:01:13
It's so interesting because uh no one is debating that it uses a lot of energy.
1:01:19
Like you could just have all the same discussions about energy.
1:01:21
Like like we're actively burning natural gas for a lot of this AI stuff.
1:01:26
Like like all of the old school don't don't uh cause global warming by burning fossil fuels.
1:01:33
Like all of those all of those like claims apply to AI today.
1:01:36
Like you could just make those claims.
1:01:39
But instead, everyone seems to have been like caught up in this water.
1:01:42
Oh, the water usage is so bad.
1:01:45
And it's like you had right here, which was like like we're burning fossil fuels and that's bad.
1:01:50
>> Is it because water feels more scarce to people than electricity? >> Maybe. >> Energy in general.
1:01:56
>> It's like it's like if I can't drink water, I die.
1:01:58
But if I can't access natural gas, like I can still live. Maybe. >> Yeah.
1:02:02
Or or the sun beams energy on the earth daily. >> Yeah.
1:02:06
And maybe it's easier to spin move out of that being like, well, we're doing nuclear and solar tomorrow.
1:02:08
Next year we're doing we're doing nuclear and solar.
1:02:12
So, so like you don't it's not a gotcha that I'm using natural gas today because tomorrow I'm going to be using nuclear and solar.
1:02:18
Um maybe maybe whereas the water issue might be like more like it's not as concise to to wrap up in a bow.
1:02:25
Um but anyway, >> [clears throat] >> um >> we covered this story yesterday a little bit.
1:02:30
Uh and I and I wasn't able to pull up the original post.
1:02:32
Andy Masley called me out.
1:02:34
He put me in the truth zone.
1:02:34
He said, "John, you follow me.
1:02:36
How do you not know where the story broke? I broke the story."
1:02:40
Uh, and yes, Andy Masley, you did break the story.
1:02:42
And so, we wanted to run through a little bit of this post.
1:02:46
Um, to actually understand the claim about what he's saying went wrong here.
1:02:51
Uh, and basically the the high level is that uh he says, "This is the single most massive factual error in a major book I've ever personally noticed on my own.
1:03:02
And I think I'm the first person to notice it.
1:03:04
Empire of AI asserts that a data center is using 1,000 times as much water as a city.
1:03:09
In reality, it's 22% of the city's water."
1:03:15
Uh, and so the chapter turns to Chile.
1:03:18
We talked about this a little bit.
1:03:18
It's a unique combination.
1:03:20
Um, look at this line again.
1:03:22
So the line says, in other words, the data center could use more than 1,000 times the amount of water consumed by the entire population of Cerillos, that Chilean city, roughly 80,000 residents over the course of a year.
1:03:34
How justifies this number in the notes saying, in other words, the data the goo the Google environmental impact report to SEA stated that the data center could use 169 lers of potable water a second or 5 million Oh, it's right there. That's the same number.
1:03:49
five mill uh five billion liters a year.
1:03:51
Um according to the water service authority in Cerillos, the municipality consumed 5 million liters in all of 2019.
1:03:58
The Google uh the year Google sought to come in.
1:04:02
5 billion liters a year divided by 5 million lers equals 1,000.
1:04:09
Um something isn't adding up here.
1:04:09
It doesn't make sense that you could use 1,000 times the amount of water used by that city.
1:04:15
And so Andy Masley has successfully put uh uh these this book Empire of AI in the truth zone and we thank him for his service.
1:04:27
>> Uh let's go back to the timeline, but first let me tell you about numeral. com.
1:04:30
Let numel worry about sales tax and VAT. Um numeral.
1:04:34
com uh new product from Travis Kalanick. That's exciting. >> Big trypicnick. com requestpicnick.
1:04:41
Uh Travis says, "I'll come out of Twitter retirement for this one. Picnic at work."
1:04:46
Uh LFG, great job with >> Picnic is delivering lunch directly to your office floor with no fees and no tips every day from 50 plus restaurants.
1:04:58
Sign up your office for free.
1:05:00
>> Okay, there's only one benchmark for this stuff.
1:05:03
We got to look at the benchmarks.
1:05:04
What's the max amount of protein? >> Is it over 200?
1:05:08
>> Are they protein maxing? >> Is it over 200?
1:05:09
because we saw the a major major jump in in uh in the amount of protein in a bowl yesterday with sweet greens. Sweet greens at 108 now.
1:05:19
This is the most important benchmark in the bowl economy, [laughter] which I'm a huge fan of.
1:05:24
Uh but uh are we seeing acceleration?
1:05:27
Are we seeing a fast takeoff in the amount of protein?
1:05:29
I want to be seeing 200 grams of protein, then a thousand, then 10,000, then a hundred,000.
1:05:34
It should be 10xing every year. Just 10x that. Yes. Exactly.
1:05:37
So, uh, >> everyone's always talking about fast takeout, but we need to be talking about a fast takeoff. >> Casual takeoff.
1:05:46
>> No, just a fast takeoff in protein per serving. >> Yes.
1:05:50
Uh, anyways, I I think this is this this has to be built on top of cloud kitchens. >> Cloud Kitchen.
1:05:57
>> I wonder if it's a separate company or it's just a subsidiary kind of a front end for cloud kitchens.
1:06:02
And uh but either way, I think people just don't like paying delivery fees and uh and tipping too is still uh you know debated. >> Yeah.
1:06:17
So you get >> and part of it part of it is like I feel like a lot of a lot of these things if you just build it into the cost of the food, people feel better about it.
1:06:23
But when you're [cough and clears throat] when people are forced to make the decision around tipping for something they want to do every single day and it's like well uh you know maybe maybe it's great sometimes maybe it's not but you're setting these things oftentimes before. >> Yeah.
1:06:39
>> Uh so >> yeah a lot of the tipping stuff it just it needs to be uh like injected in the UI at the right time and a lot of the apps don't necessarily uh like like prompt for the tip at the right time.
1:06:53
Like if you ask if you ask for the tip before the service is rendered, it's hard to use the tip as a as a quantitative feedback mechanism. >> Exactly. Exactly.
1:07:01
So, so I will when I order I order delivery uh from a grocery store. Yeah.
1:07:07
>> And I >> tip like up front. >> Do they see the tip?
1:07:11
That's always >> That's the other thing. I don't know.
1:07:13
I don't In theory, I'm like I'm going to tip because I want you to not throw the drinks a bunch. >> Exactly. [laughter] Exactly. I do that.
1:07:22
Um but but then yeah, >> getting the the fact that we've just normalized getting getting an exploded bag of drinks in a in a in a bag is just is funny.
1:07:32
>> Uh back to the press release economy.
1:07:34
Today's press release is out.
1:07:34
Brookfield today announced the launch of a 100 billion dollar global AI infrastructure program in partnership with Nvidia and the Kuwait Investment Authority.
1:07:44
There are tons of press releases going out every single day.
1:07:49
Um, uh, Daniel Tenrio says, "Running a business is all about partnerships.
1:07:54
It's all about announcing partners."
1:07:56
>> It's not even about you don't even necessarily do need to do the partnership.
1:07:58
You just got to announce the >> I mean, the partnership economy is going crazy right now.
1:08:02
Like the prediction markets are obviously the the the the most heinous offenders with a new a new partnership like every single day.
1:08:08
Uh, it's hard to keep up with.
1:08:10
We obviously are partnered with uh Poly Market and we want to celebrate them when they do great things, but uh there's a lot of these things going on.
1:08:18
[laughter] Uh and so we tend to give you a little bit of a higher level review >> on the prediction market front.
1:08:24
Somebody just leaked a bunch of screenshots from Coinbase's upcoming prediction market.
1:08:31
Uh everybody's getting into this game.
1:08:33
There's different approaches.
1:08:33
Some are some are like partnering with existing prediction markets.
1:08:36
Others are building it, you know, entirely themselves.
1:08:38
Yeah, >> I'm interest I'm more interested to see if when Coinbase does their uh prediction markets product, how are they actually running it under the hood?
1:08:48
Are they taking on the responsibility of actually like managing the markets being a market maker?
1:08:53
What is that actually going to look like?
1:08:55
>> They should just hire a guy on the other side of every trade [laughter] where you just go to Coinbase and you say, "Look, I want I want 50 bucks on the eagle."
1:09:02
And the other guy says like, "Yeah, I'll take that.
1:09:04
I think they're going to lose." And that's how it goes.
1:09:06
It should be a guy that you call >> almost like a bookie.
1:09:08
They should acquire my bookie. ag. >> Maybe my bookie.
1:09:11
ag should pivot to not having any digital experience and just lean into label. No, no, no.
1:09:17
Lean into just the guy me. >> Oh, the guy. Yes. Become [laughter] a guy. Become a guy company.
1:09:25
>> Well, uh, Cloudflare uh had unfortunately had an outage yesterday. We were not affected.
1:09:30
Although Tyler, do you want to take us through uh how we we seem to be dodging exposure to uh internet outages lately? What's going on?
1:09:40
>> Well, so I mean to be clear, all my systems were fine, but I I recently moved uh some of the kind of backend processes we use >> uh onto like a local machine. Yeah.
1:09:49
>> And then I used onrem. >> Yeah. On prem.
1:09:50
And then I used Cloudflare tunnels uh to like help do like API stuff. >> Okay.
1:09:55
So I was worried for a second that I the stuff that I moved off AWS onto on-prem was actually going to go down because of a cloud.
1:10:02
>> Yeah, >> because AWS was down what two weeks ago, three weeks ago or something.
1:10:06
>> It I think it was longer. >> Maybe longer. That one was rough.
1:10:07
I I feel like I feel like that day we actually did cancel a bunch of guests.
1:10:13
There was there was a lot of stuff going on.
1:10:14
We've had a few we've had a few rough outages, but let let's read a little bit of the uh the postmortem from the journal uh because it did make front page news.
1:10:23
Obviously sending our best to Matthew Prince over at Cloudflare and the team uh hoping that uh for a swift recovery because we love the internet and we love them.
1:10:31
An outage that knocked swaths of the internet offline was resolved Tuesday after drowning social media sites, disrupting retail sales and stalling transportation networks.
1:10:40
Users visiting sites including X, Chat, Door Dash, IKEA, Metropolitan Transport Authority in New York City were met with error messages related to Cloudflare, a cloud provider used by major companies for security tools that protect from cyber attacks and traffic surges.
1:10:57
A spokesperson spokeswoman from Cloudflare said an unusual rise in traffic to one of its services at around 6:20 a. m.
1:11:08
Eastern time caused traffic passing through the company's network to experience airs.
1:11:12
The bug was fully resolved by 9:30.
1:11:14
Uh she said in an update for several hours Tuesday, users were unable to access sites and services from retail and social media to financial services.
1:11:23
The outage echoes problems with AWS.
1:11:24
Cloudflare and AWS services were effectively invisible to users, but their tools underpin many people.
1:11:31
So, I don't know if there's a full um full breakdown here.
1:11:33
Last year, a bug in a in a tool used by uh cyber security company CrowdStrike upended computer systems across the world.
1:11:41
Um there's just a lot of these going on, but um we I don't think we have like a full postmortem.
1:11:47
I would love to know exactly what happened. It's always interesting. We failed.
1:11:50
I'm I'm interested to know what what happens to the business when they have these outages because on one hand it's a great way to tell the world that the entire world runs on Cloudflare or at least like a large amount of >> sort of a Super Bowl ad. >> Yeah.
1:12:03
It's like super very very much so. >> Yeah.
1:12:06
You got just like give everyone >> and then and then you talk about the stress from the Cloudflare team where anybody that's you know built a a software product has experienced the product going down and the stress around that.
1:12:16
But it's it's like when your product goes down and then and then you know many of the services that people use and love across the country and the world also go down it's even more stressful.
1:12:25
But it also probably brings like a ton of you know ton of traffic to the site uh and people might might start evaluating some features and say hey maybe this is a good solution.
1:12:34
I'm going to watch I'm going to sign up and see how they kind of react to this.
1:12:37
>> Well I mean the the Cloudflare team like reacted very well.
1:12:39
Uh they got a lot of uh praise for uh their response.
1:12:41
uh praise for uh their response. Dne connect here says or nect uh he's the CTO of cloudflare uh he says I won't mincech words earlier today we failed our customers and the broader internet when a problem in cloudflare's network impacted large amounts of traffic that rely on us the sites businesses and organizations that rely on cloudflare
1:13:02
depend on us being available and I apologize for the impact we caused transparency about what happened and we plan to share a breakdown with more details in a few hours in short a latent bug in a service underpinning our bot mitigation capability started to crash after a routine configuration change we made that cascade into a broad degradation to our network and other services. This was not an attack that
1:13:24
This was not an attack that issue impact impact it caused and time to resolution is unacceptable.
1:13:28
Work is already underway to make sure it doesn't happen again.
1:13:32
But I know I caused real pain today.
1:13:34
I know I caused I know it caused real pain today.
1:13:37
Uh the trust our customers place in us is what we value most.
1:13:42
just taking full responsibility here.
1:13:44
Lulu says, um, "Well done response."
1:13:48
And the comments reflect that.
1:13:50
People in the comments are very happy.
1:13:52
Um, Mert, of course, always having fun. Okay, thanks Dane.
1:13:57
But have you considered that blockchain's handling 0.
1:14:00
001% of your load did not go down? >> Very [snorts] funny.
1:14:05
Um there was another company has a picture here from uh from Shogun I think uh says Cloudflare's comm's playbook I ask permission to commit sepoku [laughter] because they're just like fully throwing themselves down just being like yeah we're 100% responsible we won't mince words pretty sweet >> uh we should uh get into Mr. for Hobart's piece. >> We should.
1:14:30
He's joining the show in just a few minutes. Tyler, what?
1:14:32
>> Uh, before I There breaking news, OpenAI new model. >> New model. >> Uh, yeah. GPT 5. 1 Codeex Max. >> Okay. >> So, it's firing back. >> They're firing back. We were debating. Did they have the juice?
1:14:43
>> Well, what what's interesting is that uh Gemini 3, the one benchmark that it didn't outdo open uh Anthropic on, it was better in a lot of benchmarks, but it wasn't better at Sweetbench. Correct. >> Correct.
1:14:55
And so uh and so that was of course a testament to like anthropic being really really great at doing just something special in code.
1:15:02
Obviously that's aligned with their mission of reaching super intelligence through self-replicating code essentially.
1:15:06
Um but a fascinating like you know durability of their business that even with this Gemini 3 thing that's so good at all these different things anthropic still on top in SweetBench but do we know how open AI is fairing in this bench? Is there any reaction?
1:15:21
What can you tell us about the latest model from OpenAI?
1:15:26
Cuz we got to get to the bottom of it.
1:15:27
And while you look it up, I'm going to tell everyone about Vanta automate compliance and security with the leading AI trust management platform.
1:15:35
Also, Sunno raised $250 million to build the future of music.
1:15:40
I'm going to hit the while Tyler pulls up the reaction.
1:15:44
[music] >> Great hit to open up the day.
1:15:47
While Tyler gets into that, uh, there is some, uh, breaking news.
1:15:54
Glue has hit the public markets.
1:15:57
Christian Tech Group tests investors faith in AI deals on Wall Street debut.
1:16:03
Shares in a company backed by formal former Intel chief Pat Gellzinger waiver after scaleback IPO.
1:16:11
Uh, I didn't realize that Glue was uh, uh, IPOing. >> No, I didn't know.
1:16:15
Pat Pat's company uh shares in a company developing AI software to connect Christian organizations across the US wavered in its Wall Street debut. >> Oh.
1:16:24
>> Following a scale back initial public offering. >> That's very cool.
1:16:28
>> Uh Glue counts Pat Gellzinger as exact chair and video rental store Blockbusters former chief operating officer Scott Beck as chief executive rose as much as 5% after it began trading on NASDAQ on Wednesday morning, having raised 73 million from investors.
1:16:45
Uh the average uh share price pop for a US IPO that has raised 25 million or more this year is nearly 25%.
1:16:49
According to Renaissance Capital Management, founded in 2013, Glue hopes to pull the Christian faith into the digital age by using values align generative AI to distribute content and sell marketing services to ministries and community outreach groups.
1:17:04
There's an imperative to shape technology for good.
1:17:06
On its own, it isn't good or bad.
1:17:08
The question is what it's used for, Beck told the Financial Times.
1:17:13
Uh, and uh, anyways, so wasn't tracking this one.
1:17:17
Uh, but really enjoyed having Pat on the show a while back. >> Yeah. No, he's he was amazing.
1:17:23
Uh, I'm very I'm very excited for for him.
1:17:26
He's just like I don't know, just fun to fun to talk to.
1:17:30
Um, let's run through Burn Hobart's economist piece.
1:17:33
Um, but first, Tyler, did you give us any >> uh Yes.
1:17:38
So uh previously the codeex bench verified was 73.
1:17:41
7 >> and now with like the highest reasoning it's at 77. 9. >> Okay.
1:17:48
>> Uh and then sonnet 45 is it's always kind of hard to tell like what exactly it is because people measure it differently. >> Yeah.
1:17:55
Do they take some of the questions out sometimes? >> Yeah.
1:17:57
Sometimes they do that or uh so so sonnet 45 is like officially 77.
1:17:59
2 so that's lower but then >> uh with parallel test time compute it's at 82%.
1:18:07
Okay, so it's kind of unclear what that like really means, but it it's definitely better.
1:18:11
Um, >> so this is a big improvement.
1:18:14
>> Isn't parallel test time compute just a a real guy who's just kind of sitting there being like, "Oh, don't actually don't do it like that. Do it like this."
1:18:22
I [laughter] >> I think that the main headline, um, is that they said, uh, >> yeah, tool use.
1:18:26
You kind of find you find a dude who's a bit of a tool and you tell him, "Hey, I can't solve this. You got to do it, human.
1:18:32
You I got to kick this one out to you.
1:18:33
Do this arc puzzle for me."
1:18:35
Uh, a AJ, our our incredible broker in the chat.
1:18:37
He's talking about the the office [laughter] debacle.
1:18:42
He said, "LMO, I still can't believe that happened.
1:18:44
Maybe something the landlord broker should disclose before tours." Lol.
1:18:47
[laughter] >> He's in the chat watching us talk.
1:18:51
>> AJ's been incredible uh finding us every every possible uh every possible space in uh the greater Los Angeles area for the next Ultra Dome.
1:19:00
highly recommend if you're in the LA uh office market. >> Yes.
1:19:06
>> Um >> I also recommend Figma.
1:19:06
Think bigger, build faster.
1:19:08
Figma helps design and development teams build great products together.
1:19:10
So [applause] >> uh we have our update.
1:19:12
We will keep monitoring the the GPT 5. 1 CEX Pro Max.
1:19:18
>> One more thing, there's an interesting headline.
1:19:20
They said um there were some tasks they found that uh the model worked for more than 24 hours.
1:19:23
Ooh, >> which is like that's, you know, if if you're if you're um following that that one meter chart >> where it's that uh time horizon.
1:19:30
This is >> super interesting.
1:19:34
>> Definitely a good sign. >> Okay.
1:19:35
>> Have you ever worked for 24 hours straight? >> Buckle up, buddy.
1:19:40
It depends on how you take it for a spin.
1:19:42
See what that brings >> uh I I would love to know what it actually is doing for 24 hours.
1:19:48
I want to know the prompt and I want to know the output.
1:19:52
>> Yeah, that's what people are asking.
1:19:52
Of course, >> they didn't say that in the press release >> because I mean it's like just just sit there and like >> it was working on the easiest problem just trying to debug it because it's so >> Yeah.
1:20:01
>> Yeah. Or I mean I mean there is a world where it's like hey the prompt is like just go take a crack at every single open GitHub issue on every repo for as long as you can and work on it and then you're basically just wrapping another
1:20:16
for loop around it and it's like is that one like there's obviously a lot of >> even if it is that like having a task that you can continuously work on is having a like kind of plan that you can maintain you don't get lost is still like a big improvement. >> Yeah. I know. Uh I mean in general I I >> Yeah. I know.
1:20:30
Uh I mean in general I I would imagine that there there's maybe some SAS productization, but there's also just some a ton of value to having agents that sort of roam through your organization continuously and clean up data or or look for different errors or just do opportunistic tasks. That seems very bad.
1:20:49
>> Trey Trey says, "Count for 24 hours."
1:20:51
Yeah, Tyler count for 24 hours. >> Mr.
1:20:53
Beast has literally done that.
1:20:53
Count There's a YouTube video up.
1:20:56
It's It's I think it's 24 What about what about uh doing one rep of 135 every 5 minutes for 24 hours?
1:21:05
That would probably get >> that would probably really >> get absolutely brutal by I don't know.
1:21:09
I don't I don't know if that would >> I think I think you would I think most people it sounds like doesn't >> it Yeah, it sounds very easy but I imagine it'd be very difficult.
1:21:20
>> Anyways, we have to talk about >> Burn Hobart.
1:21:23
>> Burn Hobart the legend talk. The king of Bub Talk. >> Yeah.
1:21:27
He wrote the book on bubbles.
1:21:29
>> He wrote the book on bubbles.
1:21:31
>> Uh yes, >> he's in the economist.
1:21:33
>> He says, "How I learned to love financial bubbles by the author of a book on bubbles."
1:21:38
>> So says, "Tech stocks have sold off this week over fears of frothiness and artificial intelligence. AI." I love that.
1:21:45
[laughter] >> Economist adding that in.
1:21:48
>> Some investors were no doubt surprised by this, but for the others, >> some some some reader out there is just like, >> "Thank you.
1:21:54
I ne I never put that together that artificial intelligence was the thing that people were talking about.
1:22:00
>> They have a very broad audience of The Economist, but I I absolutely love The Economist.
1:22:04
I've been a subscriber for probably over a decade.
1:22:05
Uh the signs of an AI bubble have been there for some time.
1:22:10
Clue, whose original product was a tool for using AI to cheat during Zoom job interviews, raised $15 million and then dropped its cheat on everything tagline and pivoted to being a more benign AI meeting assistant.
1:22:22
More serious AI labs have been able to raise 10 figure sums and 11 figure valuations, not just pre-revenue, but be pre-product.
1:22:31
Individual researchers have reportedly been offered nine figure signing bonuses.
1:22:36
And in the past year, the spending commitments made by a single company, OpenAI, total about 1.
1:22:40
4 trillion, a sum equal to 1.
1:22:43
2% of global economic output.
1:22:47
A frenzy like that is enough to make you long for the relatively sane and responsible days of the Pets.
1:22:54
com sock puppet or the synthetic CDO squared.
1:22:57
You want to continue reading? >> Yeah.
1:23:01
Uh, but bubbles are tricky things.
1:23:03
The default school of thought is that they're driven by irresponsible speculators who aren't trying to invest in great companies, but to buy something they can flip to someone more gullible.
1:23:12
A more benign theory is that they're a wealth transfer from rich investors to everyday consumers.
1:23:16
People who bought telecom firms, junk bonds in the late 1990s, lost their shirts, but the rest of us were blessed with bandwidth cheap enough to support the likes of YouTube and Netflix.
1:23:25
This is one of my favorite takes of his is that is that in the bubbles uh like when bubbles pop uh rich people actually get hurt more than main street.
1:23:34
Uh which I think is not how it's framed most of the time.
1:23:37
It's because yes there are some retail traders that go crazy and put you know they they have a five figure net worth and they put it all in the most risky NFT and they do lose it.
1:23:49
There are some anecdotes like that, but in general, most people have a pretty diversified uh you know uh you know asset base, whether it's their house or their you know their stocks in a retirement fund and those fluctuate a lot less than someone who's in the most riskone positions.
1:24:10
>> Y >> um >> we're Doggler, the first AI native dating vibe coding platform for dogs.
1:24:16
We've raised $150 million as part of our seed.
1:24:18
[laughter] >> As part as part of our >> part of our seed dog is a good thing.
1:24:24
[laughter] >> That's hilarious.
1:24:26
>> Um there's some truth to this.
1:24:26
To this day, America and Britain benefit greatly from rail networks whose construction turned out very badly for the original investors.
1:24:34
But there's another way to look at bubbles.
1:24:36
The participants in the AI race are all building products that are economic compliments to one another.
1:24:41
You need the turbines that power the grids that power the chips that run the models that power the products.
1:24:45
And you need firms to build their growth and hiring plans around the expectation that ever more of their work will be done by AI, but that every company and every employee will be automating different sets of tasks.
1:24:57
If TSMC builds hugely expensive chip factories, but the big AI labs all decide they've spent as much as they need to, those factories are a stranded asset.
1:25:06
But when asset prices are loudly signaling that the technology is real and the economics will be compelling, it encourages those complimentary investments that actually make it happen.
1:25:15
There are countless historical examples of this.
1:25:18
The car industry's growth implicitly subsidized oil production and vice versa.
1:25:22
Electric electrification followed a similar path.
1:25:27
Appliance manufacturers had to operate on the assumption that utilities would wire up more households and those utilities had to bet that once power was available, GE, RCA, and the like would give people something to plug in.
1:25:38
During the heyday of Moore's law, chip companies raced to build ever more powerful chips, and software companies rushed to ship products that would use them.
1:25:46
It's hard for any of this to happen without entrepreneurs getting excited about a business based on its hypothetical future rather than its present profits.
1:25:53
And it's impossible for this process to keep going unless investors too get excited.
1:25:57
Naturally, one side or the other will overshoot.
1:26:01
This hasn't been a technological revolution in history.
1:26:03
There hasn't been a technological revolution in history that didn't at some point get overhyped. There's always there.
1:26:09
That's always obvious in retrospect, but less so when we are in the cycle.
1:26:12
An investment researcher once circulated an essay called a home without equity is just a rental with debt.
1:26:19
warning that house price appreciation was driven by loosening underwriting standards and would inevitably lead to collapse.
1:26:24
But it was dated June 2001. >> Wow, that's crazy.
1:26:31
>> Even at the >> so early because it's true, but it's seven years too early or six years too early.
1:26:39
>> Even at the post crisis low a decade later, the case Schiller index of American house prices was still 18% above its level when that piece was published. >> Wow. >> Wow.
1:26:48
>> I mean, this is like the Bitcoin bubble stuff. >> Yeah.
1:26:49
always like, "Yeah, it's going to crash."
1:26:51
And it's like, "Yeah, it crashed from 100 down to 90 or whatever." Yeah.
1:26:54
>> It's like, >> similarly, media coverage of.
1:26:55
com's described trading as nutty and quoted an investor saying, "I don't really know anything about the company, but that article, the Wall Street Journal on the Netscape IPO was published in the summer of 1995." Yep. >> At its post.
1:27:08
com low in 2002, the NASDAQ 100 was still 40% higher than it had been then.
1:27:16
Signs of a bubble aren't necessarily signs that it's time to sell because they precede the peak of the mania by an unpredictable amount.
1:27:22
Anyone who read the quite cogent arguments against buying a house in 2001 or buying tech stocks in 1995 would have benefited financially from completely ignoring them. >> Love it.
1:27:33
The famous dictim uh >> Apocryphully attributed to John Keynes is that markets can remain irrational uh longer than you can remain >> sol didn't say that. That's funny.
1:27:45
I always thought that >> really >> Yeah.
1:27:49
I mean that's that's >> that's like his >> Well, no it's not apocryphily attributed to him.
1:27:54
We'll have to get to the bottom of that.
1:27:56
>> But this presupposes that everyone has the same information and that irrational traders are simply ignoring it.
1:28:00
It's more in the spirit of canes to argue that the economic growth is partly a matter of believing that it will happen.
1:28:07
Recessions and when people and companies start to spend as if they're over >> animal spirits >> and booms persist when some participants are building the infrastructure that others need to make that boom happen.
1:28:16
When OpenAI announces a splashy new scale up or Meta declares that it has found yet another opportunity to raise its planned capital expenditures, they're signaling to AI users, coders, lawyers, writers, whoever, that they'd better be prepared for smarter models.
1:28:30
The more people and organizations gear their behavior towards a world in which AI is even more powerful and ubiquitous, the more they're locking in the demand that justifies all of those eyepopping expenditures.
1:28:39
In the end, a bubble functions like an industry cluster that exists in time rather than space.
1:28:43
If you want to be a movie star, you move to Los Angeles.
1:28:48
If you want to start a hedge fund, you move to New York.
1:28:50
And if you want to be if you want a part of being first to something in AI, first to build, first to use, first to profit from, asset, prices are insisting that now is the time to act. I love it. >> Fantastic.
1:29:04
Oh, >> someone in the chat was saying earlier uh they were expecting uh Nvidia to beat and then trade down 5 to 10%. >> Yeah.
1:29:13
which I feel like is the consensus view now, which means that I'm I think I think we might see uh something else >> something else happen. Who knows?
1:29:22
It's all been very unpredictable.
1:29:24
Um well, I'm very excited for Burn to join the join the join the show.
1:29:28
I'm also excited to tell you about Julius.
1:29:30
ai, the the AI data analyst that works for you.
1:29:33
Join millions who use Julius to connect their data, ask questions, and get insights in seconds.
1:29:39
>> It really is like a UAV for your business. >> It is.
1:29:42
Although that's night vision.
1:29:45
That's not the UAV sound. Play the UAV. >> UAV online.
1:29:49
>> That's the UAV for your business, baby. Um, crazy. >> Ben got me this.
1:29:55
What What do we What do we got here, Ben?
1:30:00
>> That's bub talk, baby.
1:30:00
Wow, that's actually a lot of bubbles. I like that. >> Whoa.
1:30:03
All right, we will be ready to go when uh when Burn joins the show.
1:30:09
Hopefully those don't get on the camera.
1:30:11
There is a uh there there's some pretty crazy news.
1:30:14
The founder of an ADHD startup is found guilty of conspiracy in an aderall case. What a crazy story.
1:30:20
Uh Ruthia he >> got to give credit to Will for like predicting this >> like years ago at this point.
1:30:28
He was just saying like all this stuff is yeah like seems deeply >> so back in 2021 Will Manitis uh said uh in on October 3rd 2021 so four years ago he said tele medicine psychiatry startups have driven an unprecedented wave of amphetamine abuse.
1:30:47
So he was worried he was sounding the alarm bells four years ago about ADHD medications being overly prescribed too easy to prescribe.
1:30:58
He said, uh, after tweeting this, an executive at helloahed.
1:31:01
com DM'd me from an anonymous account details of my care history with them, asking that I delete the tweet or caveat that they are not bad.
1:31:11
This is an unimaginable violation of patient privacy and an odd threat.
1:31:17
>> Just the worst person to >> It's also insane because he didn't specify.
1:31:22
He didn't call anyone in particular out, but then he got a threatening message from one person in particular. So, that was very rough.
1:31:26
Um, >> just say you're responsible. >> Yeah.
1:31:30
[laughter] Uh, and so Will has followed it up and said, "Worth remembering that in 2021, 2022, many major healthcare venture investors funded a cabal of internet pill mills that operated with mafia tactics to silence regulators and drive an unprecedented wave of amphetamine dependence in the United States."
1:31:45
Well, today there has been some justice I done, I suppose, for uh these these ADHD startups.
1:31:53
And so a jury found Ruthia he guilty of conspiring to distribute controlled substances after her startup Dun Global became a ready source of aderall prescriptions for more than 100,000 patients.
1:32:05
The jury found he and Dun's former top doctor, David Brody, guilty on two conspiracy counts and four counts of distributing controlled substances.
1:32:14
Uh the former CEO was found guilty of conspiring to obstruct justice.
1:32:18
Uh, so the company was the subject of a series of articles in the Wall Street Journal from 2022 to 2024.
1:32:24
Maybe they got maybe they were reading >> interesting tweets.
1:32:26
Tech and Venture basically decided like doctors were a bug, not a feature.
1:32:31
It's like, yeah, why waste time talking to a doctor >> just to get the medication that you want and that you know you need.
1:32:38
and that you know you need. It's like, oh, actually like having having somebody that is like pro, you know, even if it's slower, like having somebody that's there and actually understanding the patient and >> uh having like some personal connection
1:32:51
with the patient feels very much more and more like a feature >> and and also having the economic incentive of the doctor being uh like they get paid a lot of money and live a great life just to give great advice and follow the hypocratic oath and be like a pinnacle. like a member of their society
1:33:08
like a member of their society to increase conversion >> rate. Exactly.
1:33:12
As opposed to like peace meal, how many scripts did you write?
1:33:15
That's the pill mill model. Yeah.
1:33:15
Uh and so uh defense lawyers argued that he uh so during [snorts] a seven-week trial, prosecutors argued that he sought to enrich herself by making it easy to get aderall and other stimulants, while the government classifies uh this as a controlled substance with high potential for abuse.
1:33:33
The startup collected more than hund00 million in revenue.
1:33:35
Um and uh and the defense lawyers argued that they just wanted to make it easier to get the drugs when there was a shortage of providers.
1:33:46
Quote, "I think the goal we want to optimize is to help patients manage their ADHD in a convenient way."
1:33:53
And there's some good reasons for that. Not everything.
1:33:55
Sometimes you actually can't get to a doctor.
1:33:57
There were good arguments on both sides, but um in this case, it does seem like they uh they uh pushed it way too far.
1:34:04
Uh there's some crazy crazy quotes in here.
1:34:06
Uh whoever is the first person to get arrested, I'll buy you a Tesla.
1:34:10
Ruthie told concerned staffers uh saying like, "Don't worry. Uh you know, bend laws.
1:34:17
>> Wait, don't if you get arrested, >> a former company executive testified that the CEO encouraged staff to quote unquote bend laws."
1:34:25
>> Okay, I'm encouraging everyone here who works for TVN, never bend the law at all.
1:34:30
>> Operate within the law.
1:34:31
>> Operate within the law. Tyler, >> 100%.
1:34:33
looking at you, >> Tyler.
1:34:35
>> Operate within the laws of physics. >> Yeah.
1:34:37
Don't bend the laws of physics.
1:34:39
>> You're studying physics.
1:34:39
Make sure to operate within the laws of physics always. >> Yeah.
1:34:43
Whoever is the first person to get arrested, I'll buy you a Tesla.
1:34:45
That's a crazy thing to say.
1:34:49
>> Was that in writing or was that just like a quote from from one of the employees?
1:34:53
>> No, that was the former custom former executive testified on the stand that this was said by the CEO.
1:34:58
Uh, which is uh pretty pretty crazy.
1:35:01
Anyway, um fortunately the bubble in ADHD medication is uh winding down as the justice is being served.
1:35:10
Uh we have Bern Hobart in the reream waiting room.
1:35:13
Let's bring him in to talk about other bubbles. More positive bubbles. More beneficial bubbles.
1:35:19
Welcome to have you here. >> Great to have you on. >> Brought the bubbles. >> Awesome. We brought bubbles. Great. >> Bubbles. The bubble king.
1:35:29
>> It really goes everywhere.
1:35:29
Uh for those who don't know you, please uh can you kick us off with a little bit of an introduction on yourself and and thank you so much for taking the time to be here. >> Yeah, absolutely. So, hey everyone. Um I'm Burn.
1:35:39
I am probably best known for writing the newsletter, The Diff, which you can check out at the diff.
1:35:43
co, covering topics in tech, finance, everything adjacent to them, everything in between.
1:35:48
Um also a partner at Anomaly, an early stage frontier tech venture capital firm.
1:35:54
Um, also co-authored with uh Anomaly Partner, co-authored the book Boom: Bubbles in the End of Stagnation, published late last year by Stripe Press.
1:36:02
So, uh, yeah, I'm the bubble boy.
1:36:05
>> The bubble boy on a roll. >> How can we talk?
1:36:08
How should we set the table?
1:36:10
Do you want to talk about uh just it feels like you've been uh sort of defined as like probubble.
1:36:16
So, I feel like asking you the question, are we in a bubble is a little bit irrelevant.
1:36:20
Uh, but would you agree that we're in the bubble? in a bubble.
1:36:24
Maybe at the start of a bubble, maybe at the end of the bubble, but we are.
1:36:26
It feels like we're in a bubble and it's safe to say it now. >> Yeah, totally. It's great.
1:36:30
Um, and yeah, that that is like [laughter] that that is that is the curse of uh co-authoring a book that is trying to rehabilitate the image of bubbles is that every time the NASDAQ hits a new high, people start calling you and asking you if that's good.
1:36:43
And um yeah, my my obligation here and and the the model advanced in the book is to say that yeah, it is pretty good.
1:36:51
Not to say that stocks will always go up forever.
1:36:54
Not to say that everybody's um options or their weird quasi equity participation units um will all be valued at uh at their present prices.
1:37:05
But yeah, so the general argument we advance in the book, I guess you can rewind a little bit and say you can you can go through these different ways of talking about bubbles.
1:37:12
One is just say they're stupid.
1:37:13
It's when people are they just get over excited about some new technology or in the case of like housing, they get excited about some very old technology that is uh newly easy to finance and they just lose their heads and by the end everybody knows they're overpaying.
1:37:27
Everybody assumes someone else will overpay more in the future and just when you run out of stupid people, prices collapse.
1:37:31
So that's like the bubbles are really stupid vein of thought.
1:37:35
And then there's one which is a little bit more nuanced which is hey yeah they're stupid but they're actually a wealth transfer from hedge fund people venture capitalists etc to everyday consumers.
1:37:44
So I lived in San Francisco in 2015.
1:37:46
I remember I I I long for the incredibly cheap you know universal basic Uber where every you know you could get anywhere for like $8. It was amazing.
1:37:55
And um that was yeah it was great.
1:37:58
It was like you know wonderful wealth transfer from the u Saudi Arabian sovereign wealth fund to me and I really appreciate it. Thanks guys.
1:38:06
Um but but that >> well it's also not notable that they they they've you know done fairly well even though there was like it it was it was a it was a bubble.
1:38:15
It wasn't necessarily sustainable but it created a enduring business or at least one out of the out of the out of the two. >> Yeah.
1:38:24
And that's that's often what we get.
1:38:25
Um now Uber is a little bit of an abstract case.
1:38:28
What we often get from bubbles is we build too much infrastructure, but that means we have the infrastructure and we have enough infrastructure to build the next thing.
1:38:35
And that was the case in the 19th century with railroads.
1:38:36
That was the case in the late '9s with telco infrastructure.
1:38:39
And um you know that that could end up being the case today with GPUs.
1:38:43
But there is the probubble argument which is that what bubbles really do is they coordinate different market participants um founders, employees, investors, regulators, customers, suppliers.
1:38:55
They convince everyone that this is happening.
1:38:57
It's happening right now.
1:38:59
And that if you build something, if you overbuild for today's demand, you will still have underbuilt for future demand.
1:39:04
And if everybody's doing that at every layer in the supply chain, then you actually do build enough to satisfy future demand.
1:39:12
And so like making it more concrete, if if TSMC does not buy into the idea that AI is a really big deal, they're not going to build enough fabs.
1:39:18
Nvidia will not be able to ship enough chips.
1:39:20
The next models will not be quite as good or we'll have to do more of a trade-off between training and inference.
1:39:24
and the whole thing slows down.
1:39:26
But if everyone is wildly optimistic, then they do build all of that infrastructure, you know, all the way from the power generation to the end use cases.
1:39:34
They're building all of that and it all it's kind of like just in time manufacturing of the future.
1:39:39
And the prices like the crazy prices, they are this signal that this is the time like it's happening now.
1:39:44
If you build something on the assumption that OpenAI is going to keep shipping better models and that they will need a lot more compute and that will need a lot more power to make those models work.
1:39:54
If you operate on that assumption, you're making the right call.
1:39:57
>> How how was how have you been processing uh some of Ben Thompson's maybe jitters around the idea that the infrastructure that gets left behind in this particular bubble might not be something that's with us for a hundred years.
1:40:10
He's been advocating for, hey, let's do energy.
1:40:15
Let's do nuclear, solar, let's build out a lot of energy.
1:40:17
But if we're just like, "Yes, we way overbuilt like a bunch of H100s and then, you know, years later we're like those actually aren't that valuable.
1:40:26
They don't they maybe depreciate over a few years."
1:40:28
Um, that's sort of how he's articulated some of the fear of the overbuild not being as durable.
1:40:35
Do does that resonate with you or how have you been processing that?
1:40:38
>> Yeah, I think I think you can look at these different lags and you also look at how generalizable the use cases are for these products.
1:40:44
And so it is true that GPUs depreciate, but depreciation it is this economic concept that's tied or it's an accounting concept that's tied to an economic concept.
1:40:51
And it actually ties in a couple different things.
1:40:54
One of which is just if you use a machine there is friction, it can break, it can overheat, whatever.
1:40:59
And so eventually it's trash.
1:41:01
And then the other piece more on the economic side and much more relevant to GPUs is if the GPU is producing fewer tokens per watt and that just relative to newer ones, it can be economically worthless even though it can still actually do something useful.
1:41:17
And so if the GPU buildout slows down, that actually decelerates the depreciation for all of the world's existing GPU fleet.
1:41:23
And power is harder to slow down.
1:41:25
The just the lags are a lot longer.
1:41:28
takes a long long time to build a new power plant and to build all the equipment that goes into it.
1:41:33
So you could have this case where bubble pops, open AAI has to do some weird recap at a much lower valuation.
1:41:40
NASDAQ's down by half or probably not down by half, but down by a lot.
1:41:44
[laughter] And um you know, a lot of a lot of people who felt very smart as of like a month ago um look pretty stupid, myself included.
1:41:50
stupid, myself included. And um you know we could have that but there will still be this increase in power generation capacity because that stuff is locked in and the like the gas turbine companies they they can really lock their
1:42:03
customers in because there are just not that many places you can go to buy one and so if you're going to buy one and they say okay you have to actually guarantee that you'll pay for it even if you really regret this that's pretty much what you'll have to do. So you
1:42:11
much what you'll have to do. So you could have this case where what actually happens in the aftermath is power costs decline and so these GPUs actually become higher margin when they're doing inference and so you get really really cheap abundant intelligence at today's model capabilities and that that is still
1:42:29
>> yeah the GPU is fully depreciated and power costs have come down which would make them you know the the the concern around depreciation is you just get a chip that is so much better that it's just non-economical to to run one of these old GPUs, but their the scenario that you're pointing out uh you could they could potentially be valuable for for much much longer. >> Yeah. So, if you look at a company like >> Yeah.
1:42:52
So, if you look at a company like Coreweave where their business model is we stack a bunch of GPUs in a data center, we lease them out to various people in varying terms.
1:42:59
That is actually kind of a bet on this narrow slice where AI is not a complete flop.
1:43:06
we don't find out that it was actually just Sam Alman typing those answers really fast all along but it's also you know doesn't completely revolutionize things because that like if there's another generation or two of GPUs or if TPUs take more share of inference then maybe those GPUs end up being economically stranded but if there's a
1:43:24
world where we're not building many more GPUs but we are using we do find all the use cases for the ones we have that might be a world where actually core is you know reporting pretty nice gap profits and um and their investors are happy And when you look at Corwe, one of the interesting things about them is on their cap table. So they one of their
1:43:38
So they one of their big backers is this hedge fund magnetar which does incredible has done incredible things with structuring various bets.
1:43:45
Like if you control F their um their perspectus, Magnetar is actually mentioned more often than Nvidia.
1:43:51
>> And Magnetar likes to make interesting bets on like the relative volatility of different things or the time relative timing of things.
1:43:57
And so you could see this as them making this kind of esoteric bet that AI is actually both a really big deal and somewhat overhyped.
1:44:06
Now it's always always hard to you know read into like what you're reading into is Magnatar is involved in this.
1:44:09
They have done a bunch of really interesting deals throughout the core cap structure but it is it is still striking that they are very sophisticated about this exact kind of trade. >> That's interesting.
1:44:21
>> Uh what is like what is the right way to view a bubble?
1:44:25
Is it this monolithic structure in which there or or or are you viewing it as like because in in my view it's like we we just read your article or essay in the economist right before you congratulations by the way.
1:44:38
>> Uh it's great great kind of summary of everything but uh >> high high honor to be in >> but you're you're saying like there's pieces of a bubble that are feeding into each other and making the possibility for durable value creation to be higher and higher because you have all these parts combining.
1:44:51
parts combining. I feel like another view maybe is is similar but slightly different is like you have these rolling bubbles that are all kind of like building up and right right now there's like we have a private credit bubble maybe there's a a neocloud bubble maybe
1:45:06
there's an LLM bubble right like do we it's unclear if we need uh you know the the 50th uh uh closed source LLM right I I don't know maybe we do right it's it's hard to predict but what what is like the right kind of like way to even just visualize the bubble uh the AI bubble broadly. >> So like a lot of other bubbles, it
1:45:25
>> So like a lot of other bubbles, it starts out as this really differentiated unique thing where most people do not know, you know, like 5 years ago, most people did not know or care that much about AI.
1:45:36
about AI. It was kind of this thing where you would listen to the quarterly call from Google or from the company then known as Facebook and they would talk about how they're an AI company and you'd think okay like I'm glad you have
1:45:47
your science project nerds but I really care about more ad clicks and more dollars per ad click so good luck with whatever whatever robot experiments you're doing and then when it starts growing what it starts doing is actually connecting with the rest of the economy. Now like the marginal dollar of AI capex
1:46:01
Now like the marginal dollar of AI capex is increasingly going into general purpose power generation infrastructure.
1:46:08
So and and meanwhile AI is getting much more broadly distributed like initial initial use cases were one it was a really good autocomplete for coding and two if you needed to create original content in order to spam people or if you were replacing like the lowest value bloggers you could do it and it was cheaper.
1:46:26
But then it became this thing where it's like a lot of there are just a lot of things where you wish you could apply a little intelligence to it.
1:46:32
It's really not worth your time, but if you can get the right answer easily, then you should do it.
1:46:37
And just like a lot of a lot of cases where you'd want like I use it a lot when I'm writing as a research tool where it's like I want examples of this phenomenon or I want >> a research tool, not a writing tool. >> Yeah. Right.
1:46:50
Are you even using it as a first draft or is it more like you have a bunch of facts here and then you are actually typing out the sentences that you want?
1:46:58
>> We're like blood hounds for AI content and we did not and no alarm bells went off when we were reading >> Oh yeah, the economist article was like >> I've just been I've been surprised the most clearly human written.
1:47:10
>> It's so it's it's extremely notable that like using AI for writing has become the most low status thing that you can do on the internet.
1:47:17
>> It's like it just feels like disrespectful.
1:47:18
It is lower status than making just like sloppy memes.
1:47:21
Like >> there are status things you could do. >> Yeah. Okay.
1:47:27
Like [laughter] adult content >> or like use my coupon code to sign up for prize picks, [laughter] you know, here's my parlay. Who's writing?
1:47:36
>> No, but but still it just communicates it.
1:47:39
It it it it you know, people feel disrespected by because it's like, hey, you put this out.
1:47:43
you wanted me to read this and if it's completely obvious that that a computer generated it, it's like well was this even worth my time, right?
1:47:53
Like if you couldn't have said it in your own words.
1:47:56
>> So there there is this dynamic where there just sometimes when there is increasing efficiency with something we find out that some of the effort was loadbearing and that doesn't mean the technology is bad.
1:48:04
It means we do have to adapt.
1:48:06
And so in the case of writing, one of the things that used to be the social norm was if you can produce a grammatically correct lengthy document about some topic, that is an indication that you probably know what you're talking about.
1:48:17
And to get into a position where you can do that, you have to read a lot.
1:48:20
So you get you acquire knowledge.
1:48:22
And if you want to write something persuasive, you probably have to talk to a lot of people and find out what's persuasive to them, what's persuasive to you, etc.
1:48:27
And if you can just just ask a model to admit that, then you can basically write at a level that is much higher quality than your ability to think.
1:48:36
You can write well beyond your wisdom.
1:48:38
It's kind of like when people use um some peptides and steroids, they end up getting weird injuries because they're just like mechanically their body is not actually suited to lift the weights of their muscles can move.
1:48:48
So they do get serious injuries unless they train pretty rigorously.
1:48:52
So, um, I have a a nine-year-old who has in the past used Chat TBT to write emails to me explaining her side of a fight that she had with one of her siblings.
1:49:00
And the email is very clear, very articulate, lots of M dashes, lots of it's not X, it's Y.
1:49:07
And it's, you know, that I think if someone that age sat down and write this coherent letter explaining their side of an argument, that would actually be impressive.
1:49:16
Like, you'd say, okay, this person's actually thinking seriously about what happened.
1:49:19
But in this case, it's like she can write two sentences in chat GPT and you know answer some follow-up questions from it and then produce this nice coherent looking document.
1:49:28
So, um, >> do you how much do you worry how much do you worry about a a new we have, uh, uh, kids younger than that, but how much how much do you worry about potentially a generation of young people never like, you know, maybe in a classroom setting, teachers can be like, put your phones in this box and you guys are all going to write a paper on this.
1:49:48
And it's possible that writing will become like highly supervised because the only way to prevent somebody from just generating the written word, >> it literally already is in many schools. Yeah. Yeah. Yeah.
1:49:57
But but but even even then it's like when I think about uh growing up and being forced to think deeply about topics, >> often times it was because I was assigned to write an essay on something and I didn't have the world's best autocomplete tool and I just had to sit there and kind of wrestle with an idea and actually learn about it and I had to read a book or read a bunch of essays and really put it together.
1:50:19
And I think it's possible that just a lot of time spent like deeply thinking is just fully lost forever.
1:50:28
>> Yeah, I think it comes back to that loadbearing effort question.
1:50:30
So I do tell my kids that there is just a qualitative difference and also that when they get an assignment at school, it's not because the teacher has this burning desire to read an essay about, you know, whatever about Charlotte's Web or something like it's not like the teacher is absolutely, you know, they have been pining for this.
1:50:49
It's like the the point is the effort and the point is the way that you think about things and that writing is actually just a very useful way to think something through.
1:50:57
I I don't really understand why that is.
1:51:01
Like I don't know why it is that if you just try to talk to yourself for 20 minutes straight about a topic, you won't get to the same level of clarity that you do if you type it out.
1:51:09
Even though the typing it out process is just really similar, it is one word after another and then a little bit of editing sometimes.
1:51:15
Uh so I think some some of what the education system has to do which different schools do to different degrees is um is actually explain to kids what the purpose of what they're doing is so they understand what that purpose is.
1:51:29
And then we also do have to make this adjustment of sometimes there are things that it is it used to be necessary for basically every adult to be able to do no longer as necessary and fewer people will be able to do it.
1:51:40
and maybe the ones who do it will still take a lot of pride in their craft, but they they won't strictly have to.
1:51:46
So, think of it as like I don't know, things like manual labor and I don't know, wilderness survival skills, things like that.
1:51:54
There there was a time when being physically strong and knowing how to like being able to navigate in space and figure out which way is north if you're lost was actually a pretty important skill that a lot of people had to have.
1:52:06
And there's actually you'd be mocked if you could. >> Right. Right.
1:52:09
And so then you go to this generation where there is a lot of mocking, there is a lot of bullying, but the nerds are actually probably right that this thing is not so important.
1:52:16
And then the next generation, it's only the hobbyists who who do this.
1:52:20
Um Thomas had this this argument about um I think it's in um his book on knowledge and decisions where he's he's talking about how if you live in a really if you live in a subsistence level tribe somewhere, you actually have to have this incredible breath of knowledge.
1:52:33
Like you've got to know all the landmarks, how to get from one place to another, all the signs of danger, everything you can eat, everything you shouldn't eat, and you know which which local tribes are friendly, which ones aren't.
1:52:43
And you just don't need nearly that level of knowledge to survive in a modern city today.
1:52:50
There are all kinds of things about where your food comes from and what is safe to do and not to do that you simply don't have to know because you're not exposed to any of the risk.
1:52:57
And so we um we actually have just a much lower knowledge requirement in in more advanced societies.
1:53:03
On the other hand, we have much higher returns from having that having unique kinds of knowledge because now that whatever value you can create can be advertised over a much larger number of people and there's just more stuff to go around.
1:53:14
So the the rewards from being really really smart are a lot higher.
1:53:18
And you know I hope that when I talk to my kids about this stuff and I basically say like there's going to be a cognitive overclass and a cognitive underclass.
1:53:27
You can opt into one of them and it's super easy if you tell your kids that it's amazing.
1:53:32
>> You must escape the cognitive underut that way.
1:53:34
[laughter] >> It is true like it is so easy to go through life without thinking and it will only get easier and so you you have to decide knowing that the thinking part is increasingly optional in a larger and larger number of domains.
1:53:46
Do you want to be the kind of person who thinks because you like thinking and you like creating and discovering new things or do you want to be the kind of person who has just a much easier more relaxing time because they don't?
1:53:58
uh very there's that we could continue uh on this conversation for a long time but I wanted to ask you about what scares you about this current bubble like things that are not necessarily like uh bad today but could get bad to me like te tech you know indulging in in uh in leverage for the first time uh may maybe as an industry or as like a lot of the leadership has not act they weren't in the they weren't participating in the telecom bubble.
1:54:30
They didn't get blown up.
1:54:32
Maybe they've never gotten blown up by leverage and and maybe that's uh a concern, but I'm curious how you think about it.
1:54:39
>> Yeah, I'm I'm less concerned about that.
1:54:42
I I think the current generation of tech leaders, there's a lot more tech history that they can know about, and they just seem more interested in tech history.
1:54:48
You can actually go back and see that the people who were more obsessed with tech history tended to do better.
1:54:52
Like Steve Jobs was obsessed with the story of Polaroid.
1:54:55
It's this beautiful consumer device.
1:54:57
Changes everybody's behavior. Really simple tool. You look at it.
1:55:01
You know exactly how to use it.
1:55:02
You know what it does and it does what it's what it looks like it's supposed to do.
1:55:05
Um Jeff Bezos gave a TED talk when that was a much cooler thing to do.
1:55:08
I think right after the com bubble had rolled over where he's talking about the early days of electrification and how the internet is like that partly in the sense that we we did not know how to use it.
1:55:21
We didn't know all the applications.
1:55:23
And he I think he he said I think that's where I I heard that the original appliances like if you bought an iron originally it would actually plug into a light socket.
1:55:31
Like you'd unscrew a light bulb, screw in the iron, and then iron your clothes in the dark and then screw in the light bulb again.
1:55:38
Or I guess you'd iron your clothes during the day.
1:55:40
But anyway, [laughter] like we it was very janky.
1:55:43
And so you could have looked at it at that time and said like this is just a clown show.
1:55:46
like, okay, sure, electric lighting, I get it, but what are you doing with all these other weird gadgets? And who needs that?
1:55:53
Like, we already had irons. They were fine.
1:55:54
Um, so I I I think that a lot of tech people are actually pretty keenly aware of history.
1:56:01
And a lot of them are just they're they're way more obsessed than you would think with the prospect of their company becoming irrelevant in six months and a total failure in two years.
1:56:08
So, I think we're, you know, it is riskier to borrow than not to borrow, but we're probably safe on that front.
1:56:11
I think one thing that could go wrong is some combination of um corporate behavioral norms and regulatory norms and investor assumptions where we decide that this stuff is really dangerous. We should not touch it.
1:56:25
It will blow a giant hole in somebody's balance sheet because we know it happened and it'll happen again and it just becomes untouchable for a while which did kind of happen in the dot space.
1:56:33
And I think people underestimate that when they look at things like Mark Zuckerberg starting a a social network in 2004 is that that was you could have looked at that as really like now it looks really forward thinking.
1:56:46
At the time it kind of looked dated.
1:56:48
It kind of like the example I use is like if you if in 1999 you moved to Seattle to start a grunge band like you missed it.
1:56:56
You were you were way out of date and that's what it looks like.
1:56:58
Um so it was still a kind of contrarian thing to do and it was still a company that was started in the aftermath of this dotcom bust when people were were cautious.
1:57:08
So um but it with AI the capital requirements are so high that it is actually a really big deal if investors decide that the space is uninvestable progress actually stops.
1:57:18
Whereas you just don't need a lot of capital if you're in your dorm room on your laptop just slaying PHP. >> Yeah.
1:57:26
Or or you can at least monetize much much earlier.
1:57:27
And you see that with like the Google uh earnings like preipo is like a massively profitable business undeniable didn't need any permission.
1:57:35
Uh what do you think about sovereign AI international uh how bubbles spread internationally?
1:57:39
internationally? I was listening to Tyler Cowan uh talk about uh one of the weird side effects of tariffs is that other countries might copy America's tariffs just for sort of mimemetic reasons and America might be in such a powerful position that tariffs might not
1:57:58
actually wind up hurting America because of its position in the global economy but if another country says oh let's copy that they might be hurt more uh I'm wondering about how bubbles propagate uh at the same time a lot the telecom uh magnates in foreign countries that just kind of copied our telecom build out. Well, they're the richest people in
1:58:15
Well, they're the richest people in those company in those countries now.
1:58:18
So, how are you thinking about like the bubble spreading internationally?
1:58:23
>> Like I think it is a it is a really cool toy for pro states and some of them have actually done some really impressive work.
1:58:28
So, um you know I I don't really begrudge that.
1:58:30
I'm not sure how many how many general purpose models the world needs.
1:58:34
I suspect what the world needs is lots and lots of special purpose models.
1:58:39
And that can be the level of okay, this model just knows Rust, but it is insanely good at Rust and it has not polluted its mind with any bad habits from C or C++ or anything else. It's just pure Rust.
1:58:49
And then you could also have even more narrowly scoped models where it's like this model is this one person and it will give you the best approximation it can of what this one person does.
1:58:58
And if you have a lot of different models and you and people who interact with models interact through a router where the first thing the router does is figure out which submodel to send things to and it can do many iterations of that and eventually might be sending some things to you know maybe delegating some things to an agent that ends up talking to an agent at some third party service.
1:59:18
Um so like I'm thinking of things like if you are planning I mean everyone says if you're planning a trip let's say you're planning a really complicated tax-sensitive global M&A transaction.
1:59:25
So maybe you need like the French tax law bot to interact with the US tax law bot and they both need to make sure that the economics of your weird tax thing also make sense in that world.
1:59:38
You could actually have this great diversity of models with a great diversity of model use cases.
1:59:44
But for the general purpose stuff, like I don't I don't think there is I I think that there is enough room for customization at the user level that we probably don't need 50 different models that are close to the frontier. >> Yeah.
1:59:58
>> What are your labor displacement timelines?
2:00:02
Because every CEO over the last year has has used AI as the reason behind layoffs and I think everyone has been calling BS on a lot of that.
2:00:10
It's just like they need a good reason to do a round of layoffs for other more real reasons.
2:00:19
Um, and everyone I think has seen the chart by now of of of job openings versus uh uh uh you know when when HGBT was released and at the same time you know if you've used these tools uh you know you're not a lot of people >> it doesn't feel like a drop in replacement. >> Yeah.
2:00:36
Meanwhile, you have engineers like, you know, LMS are incredible at coding and you have engineer if you're a talented engineer or even a high agency engineer, you probably have more opportunity than ever.
2:00:46
And but I'm so I'm curious about uh how you're thinking about timelines.
2:00:52
Yeah, like this stuff takes a surprisingly long time to deploy because one of the loadbearing inefficiencies is that if something required intelligence, there's a single there's at least one human being whose judgment is implicitly tied to the output of that product.
2:01:05
And it's really hard to go from there is some specific person to blame like if a mistake was made, some someone made it to if you scale up your work by, you know, 100x and now 95% of the time you do just fine and 5% of the time you mess up. Is that your fault? Is that Claude's fault?
2:01:23
We don't want to blame Claude. Claude's so nice to us.
2:01:24
Um, [laughter] we don't know.
2:01:26
So like we actually have to rethink how people get judged.
2:01:30
There's this sense in which everyone becomes a kind of engineering manager who like everyone in software becomes this engineering manager who is describing what needs to get done and vetting what has been done but is writing less code themselves.
2:01:40
On the other hand, LMS are actually pretty good at doing the opposite where you are the junior coder.
2:01:45
You are doing the grunt work and what it's doing is looking at your overall architecture and telling you what things you missed and what design mistakes you have made that are just a lot easier to fix up front.
2:01:55
But like a lot of organizations, they they don't want they they don't want the risk of their workers are massively more productive, but they're also producing some mistaken things and that's actually going to be a big hit to the company's reputation.
2:02:10
So you'll you'll probably see what I think you'll see is that a lot of AI deployment is that there will be a legacy version of something.
2:02:17
There will be an AI native version of that thing.
2:02:20
The AI native version will sell to smaller customers.
2:02:22
Those customers will grow faster than legacy companies.
2:02:25
and then the AI native product gets sold to all the legacy companies.
2:02:28
So this is kind of the strike model where they started out doing payments for early stage companies that had pretty simple requirements and had some tolerance for error and then they as long as they stay good enough to maintain whatever their biggest customer is they are necessarily building out the feature set for other companies the size of that biggest customer.
2:02:46
So you get some deployment that way, but it has this it it actually takes a while because the big companies, they just they want to be somewhat cautious on this.
2:02:56
And you sometimes have this case where there's a top- down mandate at a big company saying everybody's got to use AI.
2:03:00
And there's also this bottom up insurgency of I can use AI and it makes my job more effective.
2:03:06
And >> also there's going to be a dynamic there's a dynamic too where we will see scenarios where employees say, well, I don't want to adopt this AI.
2:03:16
this one's a little too good.
2:03:16
I'd be worried about losing my job, right?
2:03:18
And so I think we're going to see like more friction between even even with tools that actually can replace labor like truly not just being like a co-pilot and the friction to adoption because the people that would be adopting them and and that's probably you know years out.
2:03:34
Certain investors have been underwriting early stage private market bets to uh uh they're saying like labor is the TAM.
2:03:43
Like how do you view that framework?
2:03:43
is is that like it it feels overly simplistic to just say like any dollar that is spent, you know, that goes out through any type of payroll system today is up for grabs.
2:03:53
Um but there seems to be some some element of truth to it.
2:03:59
>> Yeah, there there's a little truth to that, but I think it's in the same way that Netflix says that time is the TAM and their biggest competitor is sleep.
2:04:06
Like it is it is broadly true marketing that Yeah, like it is marketing, but it's also a way to frame the scope of the opportunity.
2:04:12
So what I would say is that when the labor in question is mostly delivering value by producing a sequence of tokens whether that is writing a document or building an Excel model or writing some code that that is the addressable market for an AI tool.
2:04:27
But the real world just has enough complexity that models have to develop a really good world model.
2:04:32
And one way you can think of it is in software they have a really good world model because that is their world.
2:04:38
like their world is this abstract world that is defined by whoever wrote the compiler.
2:04:41
And to a lesser extent, that world has some complexities if you're actually working with real world physical systems where someone can trip over a cord and unplug one of the servers in your distributed system.
2:04:52
And that is just not contemplated in the purely software world model.
2:04:56
But as soon as you move out of pure software that is running on one machine for one user, you start to get some real world complexity.
2:05:04
And then when you're trying to automate something like building up a financial model, you need pretty tight feedback between what assumption works in the real world, what actually maps to economic realities, and then what assumption is the the most probable token in this cell that needs a token.
2:05:20
And I think that we'll like as AI gets deployed in messier parts of the world, what you'll actually see is that more of the world will get structured in an AI friendly way and that more more of GDP will be in that world that is already pre-structured for AI.
2:05:34
But then you still have the rest of the real world where it's just really hard to get eagle onboarded.
2:05:40
And you can actually see that with things like when when the company that noticed Facebook was growing internationally, one of the obstacles they ran into was in many places almost nobody has a computer.
2:05:49
So that's one of the reasons that they went into mobile early, but they also realized they could market themselves through internet cafes and that the apparently for a while in the developing world if you went to any internet cafe half of the unused computers would have the Facebook logout screen and that was actually a huge source of user acquisition for them in develop in developing markets.
2:06:10
And then once smartphones came out those people migrated onto smartphones and then Meta was able to keep them and continue to sell them.
2:06:17
So somebody would be on Facebook, they would use a computer in an internet cafe.
2:06:21
They would get up and leave and then somebody would sit down and they'd be like, "What is what is Facebook?"
2:06:24
And then they would just create >> Exactly. >> Yeah. >> Yes.
2:06:27
There's there's a great story in Chaos Monkeys about this and about how they wanted to have an ad on the logout page because they're like, "This is otherwise just wasted real estate."
2:06:34
And it turned out the logout page is this missionritical thing in all the in many of the non- US markets.
2:06:39
So there's a big internal fight on that. >> Interesting.
2:06:42
And they did end up doing ads on the logout page only in developed markets where growth had slowed down enough and they were already a dominant market share.
2:06:49
But like they they needed the outside infrastructure to catch up with the product.
2:06:54
And once it did, the product was already there kind of waiting for that infrastructure and saturated it really quickly.
2:06:58
But this is this is another thing that happens with bubbles in general, technology bubbles in general.
2:07:03
It's like you you don't consider the podcast an electricity company.
2:07:07
Like you don't think of yourselves as that business, but the business doesn't really function if you can't plug something into an outlet or use a battery and actually get power from it.
2:07:15
>> Nothing can stop us from podcasting. Let's be clear.
2:07:17
We will >> do it without microphones, without cameras.
2:07:21
>> We'll just find a crowd of people and scream at them.
2:07:23
>> Megaphone on the rooftops.
2:07:26
>> But like so in one sense, the 1920s bulls who were like, I'm all in on electricity. This is the future.
2:07:31
They were absolutely right.
2:07:33
But if you transport a trader, a stock trader from 1925 to 2025 and you're like, okay, go buy all the electricity stocks, it's like, well, that's everything.
2:07:41
Like every company uses this, so [laughter] there is no real way to make a direct bet on it.
2:07:46
And that to the extent that there is, the direct bet is now a totally different bet.
2:07:49
Now, actually, that that particular time traveler, if if he arrives in um in 2023 instead of 2025, actually his his 1925 thesis of just buy levered power generating companies and put all your money to that. It's actually brilliant.
2:08:01
So these things, you know, the cycle does repeat itself a little bit, but they as it as it disperses, you've got a little bit of AI and everything.
2:08:09
And it's, you know, internet is the same way.
2:08:10
Like you don't consider Target an internet retailer. They do a lot of ecom.
2:08:13
All the physical ret basically all the physical retailers do a lot of um online sales.
2:08:16
The the fast food restaurants do a lot of their sales through apps and through kiosks.
2:08:24
So there is just this convergence where by the time the bet is such a big scary bet that you're like the whole economy is dependent on this.
2:08:30
You're also like well it's just mixed in with the whole econom like you can't actually take the AI part out of the US economy and the US growth story without completely breaking things and at that point it kind of converges.
2:08:42
It settles >> yeah makes a lot of sense.
2:08:44
>> I have one more question that is probably worthy of like a 10-minute answer but we'll see.
2:08:50
We only have a few minutes.
2:08:52
Uh, how how is it the CEO's job to disconnect the stock price from reality?
2:09:02
>> Well, it's partly the market's job to tell employees where their equity where they should go if they want to max out the value of their equity comp.
2:09:09
And this is something that I I used to not really believe.
2:09:12
And what happened with Meta in 2022 kind of converted me this view where Mark Zuckerberg did not actually have to care that his stock was under $100 a share.
2:09:20
He it's not like the board is going to vote him out even if even if he didn't have voting control.
2:09:25
They're just not going to kick him out.
2:09:28
But it did mean that it was harder to recruit people.
2:09:31
And so if your dream is we're all going to live in the metaverse, we're going to have this legless utopia.
2:09:35
You could only hire the people who make that possible if they think your stock is going to go up.
2:09:38
Otherwise, you have to pay them entirely in cash.
2:09:39
And then your stock goes down even more.
2:09:41
And suddenly you're in this position of making really hard decisions that you don't want to make. So sometimes you Yeah.
2:09:47
Yeah. you take a foot off the gas pedal in terms of massive capex for something investors are skeptical of and as long as you're still in the lead and this is what like investors would send like hedge fun people there was a great hedge fun letter that was completely plain was
2:10:00
like Mark even if you cut your metaverse spending in half you'd still be spending the majority of the world's metaverse money like you you know you're still a winner you're still you still get the trophy but please just give us some free
2:10:11
cash flow buy back some stock it's cheap >> yeah it feels like such like disconnecting your stock price from reality uh uh at least to the positive uh can be like a massive advantage and and you can see different like like um uh uh Palanteer is like a good example
2:10:28
of this or or Tesla is a good example of this and if you can keep it going it's like tremendously effective because investors want to be in companies where the stock price is not necessarily always going to be tied to fundamentals and even employees can benefit. Uh and
2:10:39
Uh and maybe like the the opposite side of that is like is like Dylan Field with Figma.
2:10:46
like I feel like he just wants he wants to be valued like like fairly and like accurately and just wants to make the business better and better and better every day.
2:10:53
Um but uh of course it's a double-edged sword and it's great when it's uh disconnected uh uh to to to the higher end.
2:11:00
But anyways, this was super fun. Thank you.
2:11:02
Thank you so much for joining.
2:11:03
Would love to have you back on again soon.
2:11:05
>> Let's do this again soon. This is fun. >> Absolutely. >> Great time.
2:11:08
Have a great rest of your day. We'll talk to you soon. >> We'll do. You too.
2:11:12
>> Uh before we hop on with our next guest, let me tell you about Privy.
2:11:13
Privy makes it easy to build on crypto rails, securely spin up white label wallets, sign transactions, and integrate onchain infrastructure all through one simple API.
2:11:21
Our next guest is Glenn Hutchkins.
2:11:25
He is the co-founder of Silverlake Partners and uh the chairman of North Island, North Island Ventures.
2:11:29
I believe he's in the Restream waiting room.
2:11:30
We will bring him into the TVPN Ultra Dome.
2:11:35
We were keeping him waiting just a few minutes.
2:11:37
We'll keep the bub [music] talk going if he is on the line. Glenn, good to see you.
2:11:44
Sorry for keeping you waiting. Welcome to the show. How are you doing?
2:11:50
>> We don't have audio audio.
2:11:50
Can we check how we doing the mute button?
2:11:53
Check on that and see if we are getting audio through the um the call.
2:11:57
I will give some more context.
2:12:01
Uh he is the uh chairman of North Island North Island Ventures, the co-founder of Silverlake, the vice chairman, lead independent director of Sander. There we go.
2:12:12
>> He's also the lead independent director and he's here on the show. Welcome to the show.
2:12:16
Thank you so much for taking the time.
2:12:18
>> I just want to say I'm a big fan of your show. >> That's amazing.
2:12:21
>> So, it's so much fun to be here.
2:12:21
Really a real pleasure to meet to meet you guys in uh almost in person. >> Yeah.
2:12:27
Well, next time you're in Los Angeles, please uh feel free to stop by the uh TV Ultra Dome here in Los Angeles.
2:12:33
>> We are so excited to have you on.
2:12:33
So much so much to talk about. >> Yeah.
2:12:37
Why don't we uh start with uh just a little bit of your career arc?
2:12:39
I know we're going to want to talk about the the dot bubble, the dot boom, your experience there, but walk me through uh your career up to co-founding Silverlake in I believe it was 1999, right? >> That's right.
2:12:53
Um just really briefly, >> please.
2:12:57
>> Basically, when I got by Silverlake was my third and now I'm on my fourth essentially startup. >> Yeah.
2:13:03
>> In and around investing largely called private equity.
2:13:05
First one, >> I was a junior partner to a guy by the name of Tom Lee, founding what people look back on now say is one of the first private equity firms.
2:13:13
>> Uh then I um uh took some time off, worked in the White House for Bill Clinton.
2:13:17
Uh and then was recruited to come to a young little firm called Blackstone. >> There we go.
2:13:25
>> That was getting into the private equity business and wanted to build a private equity platform.
2:13:28
Um about 5 years later, the Blackstone guys helped me start uh Silverlake, which was the they invested in it, uh which was the first large scale organization to combine private equity type of investing with technology.
2:13:44
>> Uh and now I'm on my fourth, which is my platform called North Island, uh which um has one very difficult limited partner, which is me.
2:13:53
Uh [laughter] I um we're doing and I'm doing so it's my fourth startup in investing. >> Yeah.
2:14:01
>> Uh and you know the maybe we can get into this a little bit later but you know >> uh the origins of private equity might be something worth talking about at some point if you'd like but we come back to that.
2:14:10
What's your next question?
2:14:12
>> No let's start let's start there.
2:14:12
I' I'd love I'd love your view on it.
2:14:14
Uh and and it really is uh quite quite funny to think that you couldn't have maybe picked better stepping stones uh across the whole across the whole journey.
2:14:24
Must have been you must had some good intuition. >> Yeah.
2:14:28
>> You know it's better to be lucky than smart.
2:14:30
Um but um so you know the one thing I would say is can I speaking can I get a little geeky for you guys with you guys for a moment?
2:14:37
you guys for a moment? um we can come back to the more personal dimension but there were four or five in real advances in largely quantitative approach to finance that enabled the creation of kind of what I've done over the years um especially in the early ' 80s when we started thinking about private equity
2:14:55
and the first was the capital asset pricing model which allowed us to really do very good in-depth uh valuation of equities which had not been done before the second was to were was the Black Shoes option uh pricing model which allowed us to value options and really understand what embedded options how to value embedded options inside of equity securities. Often times when you bought
2:15:16
Often times when you bought a private equity company you paid for the company and then you identified something inside the company had a real upside and how to value that and how to pay for that was the question.
2:15:26
>> Second or maybe third was understanding fixed income.
2:15:29
a fellow by the name of Marty Leez came up with something called inside the yield curve which let us really value um fixed income and then Mike Milin really unders did really good work on understand the riskreward associated with high yield securities which became the tool that we were able to use >> to to build these companies.
2:15:45
Michael Porter at Harvard Business School did a bunch of research using uh standard economic analysis.
2:15:52
Um uh about the five forces that you could use to extract value from companies which wasn't being done in a very systematic way in those days.
2:16:01
And then finally modern portfolio theory with sharp ratios and efficient frontiers were adopted by places like Harvard and Yale.
2:16:09
Uh and a key part of that was having um an allocation of private equity.
2:16:14
And as that model was uh rolled out across first pension funds and then sovereign wealth funds, a huge amount of money flowed behind us.
2:16:22
>> That makes a ton of sense.
2:16:23
>> We figured out how to value companies.
2:16:24
We figured out how to use debt.
2:16:24
We figured out how to extract value from the companies >> and then we had a big flow of money coming in to back us doing it. >> Fascinating.
2:16:32
>> And so that was that's a kind of one way to think about what happened over the last >> four years.
2:16:36
How how quickly did those ideas and methods actually disperse?
2:16:40
And tying that to the present, it feels like, you know, stay ahead of what people do uh who copy me.
2:17:18
One of the things I say is that um my uh very my my best ideas are the ones that people dislike and my very best ideas are ones they hate intensely cuz I know if someone really hates something I'm I'm thinking about it and I know it's like could be really good.
2:17:35
Um and then by because by the time it turns out to be generally accepted that's when I sell what I bought before to them.
2:17:40
[laughter] >> Uh if you know what I mean. >> Yeah.
2:17:43
So you basically like if they if they if they think an idea is dumb or silly that gives you a window of opportunity to signal to to could be good. >> Yeah.
2:17:51
Well and it gives you a window to like you know get as much value out of that idea before it becomes common knowledge or an accepted approach.
2:17:59
>> Occupy that territory before they get there.
2:18:02
When they come there then you sell to them the beachfront property that you've already purchased. >> That's >> right.
2:18:07
>> right. Uh uh and then but to go back to it the halflife of innovation on Wall Street I AI is a little bit different but the halflife of innovation on Wall Street is a time it takes someone to read a perspectus >> and then copy what you did and so like you know someone does a spack and
2:18:22
everybody does spaxs y >> someone does a digital asset treasury thing and everybody wants to do a dat >> um it's like on uh in on Park Avenue in New York City as soon as it starts raining >> the guys with the umbrellas come out it's almost like they knew it was raining and then blocks on either side of Park Avenue. Everybody's with
2:18:38
Everybody's with umbrellas and you got to figure out something else to sell because the umbrella's already there.
2:18:41
But the um >> so you know when we first started doing and private equity was a way of exploiting value that was latent inside companies because you didn't have the financing to be able to purchase these companies >> and you didn't have the um toolkit to extract value from them.
2:18:58
That's those are the issues that we resolved with what I just what I talked about earlier, especially when Mike Milin untapped this high yield market that we could borrow from to to finance these companies.
2:19:08
Um then then people rushed in uh and in part the Blackstone idea you'd have to talk to Steve Forsman was to build a platform that you could take to scale where you could raise an amount of money that people who would come into the space couldn't match you and so you could like mine a different vertical layer of companies that were immune yet to private equity disciplines by getting to scale >> in the enterprise. You see what I mean?
2:19:38
very controversial in Silicon Valley to call that out nowadays, [laughter] >> right?
2:19:42
>> There's a big discussion over platform funds and funds that might be doing exactly that. >> Yeah. Yeah. Yeah. Okay. We'll come back to that.
2:19:47
And then and then what I and I decided that another path that the technology industry had reached a point where there were scale companies where you could use debt and more private equity style skill sets to buy the companies.
2:20:04
First big one we did was something called Seagate. Yes. Yes.
2:20:06
>> Where we borrowed a bunch of money to buy a big tech company. Yes.
2:20:09
>> Um but if you look at companies like so in when I was coming up people looked at companies like Microsoft and they said, "Oh, these are very risky company."
2:20:17
Steve Balmer and Bill Gates were college classmates of mine by the way. >> Yeah.
2:20:21
>> Um we're all you're lucky you're a lucky guy.
2:20:24
>> Class of 77 at Harvard.
2:20:24
Steve and I graduated build, but he did better.
2:20:29
>> [laughter] >> um >> on on on the on the financial innovation, it feels like a lot of what you identified, black shores, modern portfolio theory, sharp ratios, all all of that. Uh that's all pre999.
2:20:43
Uh I'm interested in understanding what what what was the key unlock to bringing private equity to technology specifically?
2:20:49
Were you thinking about metaf's law, network effect, zero marginal cost?
2:20:53
Were you looking at businesses that fundamentally differed from the traditional widgets business or industrials business and had different structures or what what else was going on there? >> Really good question.
2:21:05
So at that point technology was in a part point of the transition.
2:21:09
This is like the future is here.
2:21:11
It's just not evenly distributed.
2:21:12
>> Yeah, >> the it was thought to be an area two things.
2:21:15
One, it was thought to be an area of expertise where you and it was true.
2:21:20
Yeah, >> you really had to have specialized expertise to understand the companies to invest in them successfully.
2:21:23
You couldn't just wander off of out of Wall Street with your pinstripe suit and sort of think you could figure go to a couple conferences and think you can figure out how to buy, you know, a tech technology company because the the the process of evolution was so rapid.
2:21:38
>> Um and then secondly, to that point, people did not understand how technology companies had evolved.
2:21:43
That point technology companies were big um they consumed huge amounts of cash. >> Mhm.
2:21:49
in investing in R&D to build the products. >> Yes.
2:21:53
uh and they had very volatile earning streams um as a consequence of um being pioneers in a space that came and went very quickly and so people looked at that from it was a venture capital gig but look at that from a private equity perspective and say you can't do it but at that point Microsoft for instance that's why I was talking about Microsoft
2:22:12
>> got to a level of scale where it was one of the greatest economic enterprises in world history the um where you make this piece of software that comes comes out of someone's brain has almost no capital expenditures and associated with it no kind of fixed cost and sell it a billion times >> right I mean that's and just this massive flywheel of cash comes into that
2:22:36
company and we I you know and >> when was that when was that like light bulb for you no but for you personally >> for me it was I I observed it in the 90s >> um I had the benefit then of living in Boston and the venture capital business was pretty sc um uh vibrant there then it moved later primarily to Silicon Valley but there was a big footprint in Boston in those days uh because you
2:23:00
remember um data general and digital equipment were kind of there the micro companies the sort of the mini computer companies um and so you could watch it happen and you say you know that's a better way to make money than just trying to extract value from rationalizing legacy industrial companies that have been poorly managed. >> Yeah. widgets business, >> Yeah.
2:23:19
widgets business, >> right?
2:23:20
And then the other thing is remember is that people will in thinking about exiting businesses uh the market will pay you more for companies that have good this was an insight in those days. Not now.
2:23:30
By the way, in those days to borrow money, you had to have assets to back it with, you know, like um you know, inventory, working capital.
2:23:40
>> You couldn't use you couldn't use the company that and and and then yeah, you couldn't use the cash flows in this in the in a software business.
2:23:45
Didn't you know, they maybe had a a lease for an office, but not a lot of like servers.
2:23:52
>> But so what we had to do was teach the markets to lend against cash flow.
2:23:57
>> Oh, >> actually lend against assets.
2:23:57
So cash flow lending became this kind of new thing that we had to teach people how to do.
2:24:03
And when then once you got that >> then you realize that if you had a rapidly growing company like a Microsoft that had an extraordinary cash flow engine huge barriers to entry at that point people when I came in the investment business people said tobacco is the best business to invest in because I'm serious because it was very stable.
2:24:23
It had stable cash flow stable pricing and it didn't vary with recessions.
2:24:27
I said, "Let me get this straight.
2:24:28
A businesses that addicts and sickens its customers is better >> than Microsoft." No, I'm sorry. I don't agree with that. >> Right. >> Right.
2:24:38
You got to look at the modern world and understand that these cash flows are sustainable and these businesses are extraordinary because it doesn't because the the the product comes from someone's head.
2:24:47
>> They don't have to [clears throat] build a factory >> to build the thing. >> Fascinating.
2:24:51
>> And so we just built this built this business that got that that had that set of insights.
2:24:55
And as a consequence, we were able to build a so the other idea there was to build a strategic competitive advantage, a commanding heights that you could occupy that made it very hard for anybody to compete with you.
2:25:07
>> So um yeah, so help us bridge to uh some of the debt financing that's going on today.
2:25:13
I think that there are a lot of folks in the tech community that are very used to uh bunch of 20% dilution equity rounds, maybe a growth equity round and uh the idea of bringing on a partner like Blue Owl for some massive deal.
2:25:28
deal. it just doesn't map to the traditional like tech startup uh like path and yet folks who are trying to understand where AI is going and where the big hyperscalers are working start have to grappling with a with with debt and how debt is coming into this
2:25:45
generation of this >> Sam Alton has has said like we maybe need new he he said we need new ways in we need financial innovation not just technological innovation a lot of people have >> uh you know, kind of shunned him for suggesting that. But I think based on
2:26:00
But I think based on what you've been describing of what enabled this wave of like value creation Yeah.
2:26:07
>> and unlocking the value of these private companies and the value of their cash flows, >> it can be done responsibly.
2:26:12
>> It can be done responsibly. Yeah. >> Yeah. >> Wow.
2:26:15
That's a really really good question.
2:26:16
Um and I know maybe we'll have to do a second show just on that because because this is that's a complicated topic, right?
2:26:24
But um it is it technology is very I like technology, you know, gotten into it full-time for now, you know, 25 years ago.
2:26:33
Um I still feel like I'm new to like I'm still new to golf even though I've been playing it about the same period of time.
2:26:40
>> Um the uh but one of the things great about is it's constantly changing and you have to constantly adapt your thinking and develop new modes of sort of how of investing.
2:26:49
And so this AI thing is come brings us back to the future.
2:26:56
>> Um that which is it's techn it's a techn softwaredriven LLMs technology enterprise that requires a scale of capital investment that we've never seen before.
2:27:09
>> Uh and that's a really unique kind of challenge.
2:27:11
It's one of the things that's drawn me into the kind of investments that I've made there.
2:27:13
It reminds me a bit historically it reminds me a bit of when um the fab the semiconductor companies went fabulous about 25 years ago. Yeah.
2:27:24
>> Um and the industry split into companies that design semis and and basically TSMC, right?
2:27:29
Uh and TSMC succeeded largely because the the country of Taiwan was willing to essentially lend them the credit rating. Mhm.
2:27:42
>> The scale of capital necessary to build a fab that could design these that could manufacture these wafers with a nanometer scale that they had at prices that were cost competitive that could continue to drive adoption of um technologies based upon semiconductors was only approachable by a national credit rating.
2:28:02
>> TSMC had a backs stop >> basically had the the backing of the government of Taiwan to go get this done.
2:28:08
There's a reason why it's in Taiwan.
2:28:10
You couldn't do that in those days that that the capital wasn't available to do something like that in those days at that scale for that kind of enterprise. >> Mhm.
2:28:19
>> So, very similar today, which is the scale of financing that's that's required to do to in to [clears throat] build all these um fabs, not fabs, I'm sorry, factories, data centers.
2:28:31
I call them factories because they're factories manufacturing data now.
2:28:34
And what I say to my people, what I say to my friends is America's now the leader in the world in advanced manufacturing because we're building these data centers to manufacture data. >> Y >> right.
2:28:44
Uh and that's kind of what it is.
2:28:46
It's a massive factory manufacturing LLM and applications for both training and inference.
2:28:51
Um >> yeah, >> that's kind of one.
2:28:53
The second point um would be you um that people compare this to so the question is are we in a bubble? >> Sure.
2:29:02
That's kind of underlying the your thing you raised, right?
2:29:04
What kind of bubble is it?
2:29:05
And the question you have to make a decision whether or not this is more like subprimes in08 or more like the internet in 1999 2000. >> Mhm.
2:29:17
>> You know, whether the subprime is just kind of something that's not real.
2:29:18
It's going to collapse and when you're left, you're just left with a bunch of debt and no value there because the home values all went down.
2:29:25
>> Um I am more in the internet camp. >> Yeah.
2:29:29
uh which means that um of course there will be uh companies that will be formed that won't be successful.
2:29:39
>> Of course there will be investors who put capital in bad places and lose money.
2:29:42
Of course there will be some number of scoundrels and shysters who come in because money gets moved around and they get attracted to this. Right.
2:29:49
Um but there are um uh one major you mentioned blue alowl um the one major difference today between the buildout so this what was happening simultaneous with the internet was being when the dcon companies were being built could say the the LLM equivalent today the fiber optic networks were getting constructed all around the country the selex and those all went to zero and people lost their money on it.
2:30:15
people lost their money on it. The major difference between that and people use that as analogy today and and maybe the railroads is another analogy but the major difference between that and today is every one of these data centers almost all of them has a counterparty a
2:30:31
solvent counterparty that is contracted to take all the output they're built to suit >> y >> not if you build it they will come >> okay uh Microsoft has I think the world's best credit rating if you sign a deal with Microsoft to take the offput for your data center. >> Satcha is good for it.
2:30:48
>> Satcha is good for it. >> He's good for it. Yeah.
2:30:51
>> And by the way, Microsoft's gonna survive if that has a collapse at some point before it comes back again. >> That's a good point.
2:30:56
>> So, it's a very different kind of financing structure.
2:30:58
And the the last point I would make and just finish this is that each of these deals so far as I understand it is done in a way that essentially generates in in the four to 5 year period of the deal generates about a two times multiple of money on the cost of buying the GPUs and standing up the data centers.
2:31:17
>> Oh interesting >> right so the cont and they're about four to five year contracts. Yeah.
2:31:21
to five year contracts. Yeah. uh and the output is has a now and and and then and talk about okay embedded options and how you value those right and the and then the owner of the of the the GPUs in the data center has an embedded option on the value of the used GPUs which will be
2:31:37
worth something I mean your 5-year-old iPhone is still worth something >> of course >> even though people are buying the new ones right y >> uh and so the each of the model each of the business each of the contracts and builds right now has an has a has a commercial proposition in Mhm. >> And when done well, these companies that
2:31:53
>> And when done well, these companies that are doing this, like Cororeweave, are putting one of building a wall with one of those bricks on top of the other. >> Yeah.
2:32:01
>> Do you see what I mean?
2:32:01
So, it's not it's not analogous at all to the Selex where they put a bunch of money in the ground and then went to get the customers and the customers weren't there. >> Sure.
2:32:11
>> That's a very different thing.
2:32:12
>> That's a really good point.
2:32:12
I haven't I hadn't considered that.
2:32:14
That makes a ton of sense. That's great.
2:32:15
Uh yeah, I feel like a lot of people in uh in tech are just struggling to, you know, there's been this narrative for a while that chatbt is the new Google and then you look at how capital consumptive OpenAI will be before profit comes or cash flow comes versus what happened with Google where they were throwing off millions of dollars in cash like well before IPO and the prospectus just looked so clean.
2:32:41
in this like super high margin business very fresh out of the gate.
2:32:46
Uh and it's just a very different world that we're in where we're delivering as something similar.
2:32:52
It feels just like a >> partly because OpenAI has to compete with Google.
2:32:57
>> Yeah, [laughter] maybe. Maybe.
2:32:58
But it's just a different it's just a it's a capital consumption changing a little bit. >> Yeah.
2:33:04
>> Each wave of technological innovation companies are created that don't obsolete the company that went before them.
2:33:11
They do something completely different. Yes. >> Right.
2:33:14
And they're sometimes very different like you know so you've got you know the the the soft the Microsoft software was unlike that was the operating system the applications was unlike anything we had before because we didn't have the PC. >> Yeah.
2:33:28
>> And then Amazon was not anywhere near like Microsoft.
2:33:31
It was a whole different kind of innovation that was based upon the internet that was built. >> Yeah.
2:33:36
>> Uh and then Google and was different was something new and Facebook was something entirely new.
2:33:41
So these aren't companies that say I your thing away from you, >> right?
2:33:47
Each one is a very different kind of unique unicorn type of business that occupies a niche itself and eventually obsoletes the other businesses because they stop growing. >> Yeah. >> Right.
2:33:59
Like Facebook might stop growing if consumers go to Open AI, but it's not because they're going to open AI because it's a new social network. >> Yeah.
2:34:09
It's because as a different use case is valuable to them today.
2:34:13
>> What's been the biggest learning surprise sort of update to your mental model from working with coreweave?
2:34:26
>> That's a really good question.
2:34:26
Um the the the pace of change the scale we talked about.
2:34:35
I mean the thing that just amazes me is the scale at which this thing is growing >> uh and the um uh this the rapidity >> uh that you have to uh have in order to act to act at to to be successful at this kind of scale with this kind of growth.
2:34:59
It's unlike anything I've seen before.
2:35:00
before. you look you saw the adoption curves of um you've seen the adoption curves of open AI versus Google versus other things right >> uh and it's just like this asic thing going to 700 million customers right overnight all the infrastructure to
2:35:14
support that is like unlike anything we've ever seen before >> yeah and it's it's still under discussed how much bigger and faster the outcomes can be when you have the internet as a distribution as a distribution engine so like during when like you know you founded uh Silverlake in 1999. I'm sure
2:35:32
I'm sure you've looked at a bunch of companies that had a lot of potential that if there was already billions of people using the internet they would have done very well.
2:35:41
And the challenge at that point is there maybe wasn't enough internet users uh to support even ideas that were struct like structurally good ideas just missing enough enough of a of a user base.
2:35:53
How how much time do you spend uh finding and and meeting and backing new managers?
2:35:58
I feel like every new technology cycle, you know, the the hottest hedge fund of the year is situational awareness or at least at least on X and that feels like um I I imagine there will be more of those.
2:36:13
And so I'm curious uh how how many how much like new fund formation you're seeing and and what you're most excited about on on the GP side. Yeah.
2:36:21
So, I've got so I have investments.
2:36:23
I don't do venture capital investing per se.
2:36:28
>> So, I've got investments in some of the major venture capital funds, you know, in my investment platform and I know all the people and watch what they do.
2:36:34
But my business model outside of that is to find a small number of companies where I can put a fair amount of capital and be engaged helping to create the outcome. >> Mhm.
2:36:50
>> See, that's kind of where I spend my time.
2:36:51
So, I'm not doing the whole build the massive portfolio thing. I'm picking my spots.
2:36:55
And you mentioned, uh, the European Bank that I'm the lead independent director of. >> Yeah.
2:37:00
>> Um, you know, we've got that stock up 3.
2:37:02
5x in the last three years since I invested. >> Hit the gong. >> We have a gong here. We'd love to hit. >> Right.
2:37:08
So, you know, there you go. Oh, that's great. Thank you. >> Thank you.
2:37:15
>> I don't But I got to tell you, I didn't get founder mode, guys. I don't know.
2:37:16
I [laughter] don't know what's going on here. >> Founder. Here we go. >> There you go.
2:37:21
I didn't get founder mode. Come on. Wait for that.
2:37:22
[laughter] >> You got to do founder mode for you. >> I love the love show.
2:37:28
>> Is also there for you. [laughter] >> I love it.
2:37:32
>> You know, you know, I've got um uh children about your guys age.
2:37:34
I just love your generation.
2:37:36
I love hanging out with them.
2:37:37
[laughter] >> Um it's a lot of fun.
2:37:38
Um so, by the way, you see this logo here on my shirt? >> Yes. What is that? >> That's binary code. See that?
2:37:47
>> You know what that's binary code for? Uh >> 1 0 0 1 1 0. What's that? 1 0 0 1 1 0 >> 1 0 0.
2:37:59
>> It's binary code for 70 >> 70 >> 70. >> What? What?
2:38:02
>> It's my It's my 70th birthday logo. >> Oh, very cool. >> There we go. >> Happy birthday. >> Thank you. Congratulations. Incredible.
2:38:09
>> So, as I say, I've got I've got uh kids kids your age.
2:38:11
Uh and I really love hanging out with your generation.
2:38:13
It's been my a great pleasure for me. >> Yeah. Um, thank you.
2:38:17
So, we love hanging out with you, too.
2:38:20
>> You as you asked me another question that we got distracted from it.
2:38:21
Oh, the So, what what I'm trying to do is find a small number of enterprises in which I can engage. >> Mhm.
2:38:30
>> Get involved with them at a senior level in both cases, core and sundere.
2:38:31
I'm lead independent director. >> Yeah.
2:38:35
>> Uh, you know, which is another term for non-executive chairman.
2:38:36
There's usually an executive chairman and I'm non-executive.
2:38:40
>> Uh, and then really work with the enterprises to build value. >> Yeah.
2:38:44
That's kind of how I think about it, right?
2:38:46
And I then I I let venture capitalists who I invest with uh and I still invest with Silverle be on the rock face every day building these portfolios.
2:38:55
>> The rock face analogy. I like that. >> Right.
2:38:58
Well, you're you're the mountain climber, right? >> Yeah. >> Yeah. Right.
2:39:01
So, um though you're this your height of a basketball player, I think probably the wrong sport.
2:39:05
Um [laughter] I think you're referring to like one of our early episodes where I I was joking remember about >> about you climbing that those like >> Oh yeah yeah yeah.
2:39:14
[laughter] >> Yeah I heard I heard that. Yeah.
2:39:15
I thought that was you were just you were just kidding him right messing around just cuz just the idea of John a 68 guy scale. >> Okay. Yeah. I agree.
2:39:24
[laughter] >> I can't rock climbing.
2:39:27
>> You could do I I believe in you but I I would be I think that would that would probably violate insurance. >> Yeah.
2:39:33
I definitely want to be above him on the wall.
2:39:35
I don't want [laughter] to be below him.
2:39:36
But >> anyway, so you know, so I'm I'm trying to pick my spots and really add some value. >> Yeah.
2:39:42
Well, well, thank you so much for coming on the show.
2:39:44
Uh we have to have you back soon. This was fantastic.
2:39:46
Uh we could talk all day long.
2:39:49
>> Uh yeah, there's there's so many so many more questions I want to ask.
2:39:54
>> Congratulations on the success of the show, guys. >> Thank you so much.
2:39:57
>> You're welcome on welcome on any >> coast. Anytime.
2:40:00
>> When you're on the East Coast, come see me. >> Fantastic. Yeah, we will. >> Thank you so much. Play him off. Let's hang.
2:40:04
Thank you so much, Glenn, for taking We'll talk to you soon.
2:40:06
Uh, let me tell you about uh getbzzle. com.
2:40:09
Shop over 26,500 luxury watches.
2:40:11
Fully authentic in-house by bezels team of experts.
2:40:16
We're going to our lightning round.
2:40:17
We got Yogi Goell from Maxima announcing a massive round.
2:40:21
Let's bring him in to the TBP and Ultra Dome. Welcome the show.
2:40:28
Uh, what is your t-shirt? Introduce yourself. What do you do? give us the news. What's the latest? >> Absolutely.
2:40:34
So nice to meet you both, Jordi and John. >> Great to meet you.
2:40:37
>> Uh Yogi here uh from Maxima.
2:40:37
We are an enterprise accounting platform focusing on waging the war on the month and close process. >> Let's go.
2:40:47
>> Uh we are uh in a short focus we are co- AI agents are co-writing >> the u monthly financial package and preparing the data for the accounting team.
2:40:57
Yeah, >> we've been around for now five six quarters and helping incredible companies like scale AI rippling spoton uh press juice.
2:41:05
So, a lot of companies in both uh tech and non- tech world to to make accounting sexy again.
2:41:13
>> Very very on brand to count uh your the the time you've been in business by quarters. [laughter] >> Exactly.
2:41:20
>> And what's the news today? Break it down for us. Yeah.
2:41:22
So, we uh just raised uh $41 million in C++ series A. >> All right. There we go. >> Congratulations. >> Very exciting.
2:41:35
Uh explain to me how this plugs in.
2:41:37
Obviously, there's a lot of folks that have an accounting uh layer of record, a single PL pane of glass, an ERP, an accounting suite.
2:41:43
Uh do you want to just plug into that?
2:41:46
Do you want to rip and replace that?
2:41:48
There's so many different folks uh eating around the edges, creating different solutions.
2:41:51
I don't think anyone knows exactly how the market will play out, but what have you built? >> Yep.
2:41:57
So, we've built a system of action and system of intelligence which works with any system of record.
2:42:01
So, when you go to an in uh uh enterprise company uh asking them to replace ERP is like asking them to do a brain surgery.
2:42:09
I I was a nickn uh your system of record is where your data should eventually sit.
2:42:17
We do the uh we help with automating the human work of grabbing the data uh from upstream systems doing the manual uh doing the uh our agents do the auto automated work and then uh eventually finding anomalies and errors.
2:42:32
I don't know if you were following but last year was uh the maximum number of companies in the US which had uh material misstatements and like up to 40% stock drops stock price.
2:42:44
>> It was the most mistakes from accounting specifically last year. That's not good.
2:42:48
Uh hopefully we can fix that.
2:42:48
Uh I have one last question and then we will let you go.
2:42:52
Um I want to know uh give me some examples of where uh the current crop of AI models really excels in finding these types of problems and then where do you want to still leave the human in the loop?
2:43:05
Where do you want the human where what's the really intractable problem that maybe we'll solve in a few years with AI, but for now you'd leave it with the human. >> Yeah.
2:43:14
So, um, the >> you got to have somebody to fire.
2:43:17
[laughter] >> Who's the last guy in the accounting office, I guess, is the question.
2:43:21
But, you know, I I'd love some examples of of of problems that really excel for AI and and problems that are maybe more intractable. >> Yeah.
2:43:30
Look, I I'll just start with saying u our problem we are not going after the human uh labor salary.
2:43:34
We are going after errors, inefficiency and pain that I personally face both as an auditor and and and as an accountant for 20 years.
2:43:44
Uh there are not enough accountants in the world that you can truly hire uh for the amount of work that's there.
2:43:49
So uh in terms of where uh AIS are very good at today, they are very good at taking uh a defined set of uh instructions and following things over and over again for variety of transactions provided you give them uh deterministic operators which we have built that they will only use those tools and and then come up with the right answer.
2:44:11
So we are using this hybrid approach where agents follow Maxima tools to come up with the exact same answer.
2:44:17
And so when deoid and ey comes knocking looking at the uh work that maxima produced they will they will do 2 plus 5 and the answer will always be seven it will not just be 15.
2:44:25
Uh so that's one thing we determined really well.
2:44:29
Second is it's uh really good at finding anomalous behaviors and errors uh that might happen because it is u uh it looking at millions of transactions over time within the company.
2:44:40
can just see that, hey, this your legal bill used to be $50,000. Suddenly, it's $500,000.
2:44:47
Turns out uh Jim Jim had a late night and he had one extra zero and that's why it went up. >> Yeah.
2:44:53
And and I mean artificial intelligence has been used in like fraud detection for years and years and years and so applying that sort of heruristic based stochcastic based more less less uh deterministic uh computing, more probabilistic computing makes a ton of sense there.
2:45:08
>> I I love I love the positioning around pain and errors.
2:45:09
You got to talk to the venture capitalists who are yelling loudly to anyone that will hear, "We're going to replace all labor.
2:45:15
Give me more money to replace labor."
2:45:18
It's like, "No, you can just you can you can show the the sort of optimistic uh like positive, you know."
2:45:27
>> Well, well, thank you so much for taking the time.
2:45:28
Congratulations on the massive round. We will talk. >> Yeah. Great to meet you.
2:45:31
I'm sure you'll be back on soon.
2:45:33
>> And have a great rest of your day. >> Great shirt, too. >> Appreciate it.
2:45:37
>> Let me tell you about eightleep. com.
2:45:38
Exceptional sleep without exception.
2:45:40
Fall asleep faster, sleep deeper, wake up energized.
2:45:42
Our next guest is Sam Jones from Method.
2:45:44
We will bring him in from the Reream waiting room into the TVPM Ultra Dome.
2:45:48
Sam Jones, how are you doing?
2:45:53
>> Good to meet you, guys. Great to see you again.
2:45:54
>> That sound effect kind of just I don't know.
2:45:56
It doesn't have enough of a crescendo for me.
2:45:57
We need to work on that one.
2:45:58
Anyway, working on our drum roll.
2:46:00
>> Thank you so much for taking the time to hop on the show.
2:46:02
Uh, please introduce yourself, introduce the company, tell us what the news is today. >> All right.
2:46:07
I'm Sam Jones, the CEO and co-founder of Method Security.
2:46:08
Our mission is to deliver cyber resilience to the US government and critical enterprises.
2:46:14
Um, think of what we do as uh building the command and control layer for autonomous cyber operations across defense and offense.
2:46:20
And the news today is that we are announcing our 26 million combined seed and series A investment from Andre Horson, General Catalyst. >> Incredible. >> Very good.
2:46:34
I I I the gong preemptively last night.
2:46:39
>> Did a preemptive gong >> preemptive. It happens.
2:46:41
>> But we got you in the >> We'll have to raise more money to to pay you back for that.
2:46:46
>> I'm sure I'm sure you >> I would love for you to get me up to speed on how you're thinking about uh that that story in the Wall Street Journal uh about anthropic.
2:46:52
I'm sure you know the one uh is >> AI on AI violence. >> Exactly.
2:46:57
Uh is this relevant to your business?
2:46:59
Are are you building a solution to that or or do you even have a comment on it or anything?
2:47:02
Can you just get me up to speed? >> Highly relevant. Great.
2:47:05
Um, and that's kind of the moment that we've been we've been building for for a couple years now.
2:47:10
Like we've known this was going to happen.
2:47:11
AI is effectively um, you know, taking at the limit taking cyber offense to infinity and taking the cost to zero.
2:47:20
>> And uh, this is bad news for good guys, bad news for uh, the defenders as our adversaries are essentially uh, >> eliminating their requirement or or limitation on human headcount.
2:47:31
limitation on human headcount. So >> what we do is essentially allow organizations to safely become the threat to test their own defenses be some before some adversary does and the best offense is the best defense has always been a notion in security but AI
2:47:46
is really the unlock to do it at scale the hard part is you need to do so safely ethically legally and that is the infrastructure that is like needed to do and that's what we build specifically so like in that report it's almost like no news to anyone in the security trenches. Like obviously this has been happening.
2:48:03
Like obviously this has been happening.
2:48:04
Obviously that wasn't the first >> What are some other without without naming names or any details like on on the individual companies that were attacked like like are I'm assuming when you see anytime I see a report like that I'm like okay this must be happening like a ton and just a lot of it just never hits like headlines.
2:48:22
Um but what what are what are some of the most like kind of common strategies that bad actors are using today in the context of uh AI uh to carry out whatever their goals are.
2:48:38
>> If you think about preAI malware, it was already autonomous but it was basically reliant on like if then decision-m.
2:48:43
What AI basically allows it to do is to like a broader non-deterministic path planning that allows it to harness a multitude of tools, thus do a lot more damage.
2:48:54
That's what's different now.
2:48:54
And I guarantee you the most sophisticated actors are not using vanilla clawed code to run their operations. That's ludicrous.
2:49:03
They have our adversaries have better models at home that they make themselves that they're using that we have no, you know, telemetry on.
2:49:09
Um, and so they're essentially using it to scale and speed up their operations, which for us and why we have like a a national cyber resilience urgency moment on our hands is that all of these exposures that we've left out on the internet and in our, you know, in our enterprises are now, you know, easy easy takings uh for these types of attacks.
2:49:32
And that's essentially why it's so urgent that we focus on resilience.
2:49:36
>> Can you talk to us a little bit about traction?
2:49:38
what unlocked this $26 million fund raise across these two rounds?
2:49:41
Uh are you doing like Yeah, just walk me through uh how you actually show progress in uh you know you're building a a product but you're also trying to do deals with the government that can be very difficult.
2:49:57
Uh what does progress look like?
2:50:00
So we are deployed in production with a number of organizations to include the department of war um US federal government and fortune 500 organizations.
2:50:08
That's probably the biggest hallmark of traction and we're doing so across defensive and offensive use cases that get to the heart of resiliency. Yeah.
2:50:15
Um, so that's the I think the the unlock that we were able to do that and we've had this hypothesis and mission from the beginning that in order to secure what matters, you need to be dual use and we set out to basically pick what is the what is the most intense hardest government customer you could go after from the beginning and what's the commercial equivalent.
2:50:36
Those were our first two customers and then basically just continuing to build on that.
2:50:39
You know, from the adversaries perspective, they do not discriminate between public and private and neither do we.
2:50:45
Uh, and that's why the I think the ultimate game changer solutions will come from dual-use companies like ourselves.
2:50:51
>> Yeah, I'm thinking about the the the hardest to hack uh Fortune 100 and and government.
2:50:58
>> It's not necessarily always the hardest to hack.
2:50:59
A lot of times it's the hardest to sell to, hardest to deliver for you.
2:51:03
Think about the government done accreditation, deployability, like interoperability, um, huge technical challenges.
2:51:08
That's why startups would never dare >> touch there.
2:51:12
But >> when you think about what matters, >> they are what matters and that's what we've built this company to serve. >> Yeah.
2:51:18
>> What were you doing before this again?
2:51:20
>> So I started my career actually seeing this problem firsthand at the US Air Force.
2:51:23
Um so I was a cyber operator and many ways were building the tools that I wish I always had.
2:51:28
I joined Palunteer about 11 and a half years ago.
2:51:30
You know pre-product and building out both their cyber commercial and DoD business.
2:51:33
And then I was also at Shield AI pretty early.
2:51:38
And so you can kind of think of this company as we were the users.
2:51:39
My CTO and co-founder also started his career at NSA.
2:51:44
His last name is hacker by the way if you want. >> Oh, there we go. About destiny. >> Get that.
2:51:49
And then also I want an overnight success for >> for being in this industry for 15 years.
2:51:56
>> He had no choice other than to work at NSA, but we met a palunteer and did great work together.
2:52:00
But we're combining our our knowledge of like we were the users.
2:52:03
We know how to build hardcore hardcore software and dual use businesses and then we built AI before you know it's become a meme and certainly in no fail scenarios which I would group um security in for sure. >> Yeah. Uh this is very bullish.
2:52:19
>> Yeah, extremely bullish.
2:52:19
Thank you so much for taking the time to come chat with us to get the update.
2:52:23
>> Sorry my uh >> my background wasn't as good as Glenn's mahogany. I I I show this startup.
2:52:29
>> You brought you brought a wood which is which is good. >> Yeah.
2:52:32
Thank you for bringing What What's the biggest fish you've ever caught? >> Yeah. [laughter] >> Yeah.
2:52:37
Probably a nice nice uh walleye, I'd say. >> There we go. Good answer. Good answer. >> Midwest shout out. >> Thank you so much.
2:52:43
>> I cannot wait for the be.
2:52:43
Congrats on all the progress. >> Yeah. Have a great day. Cheers. Talk to you soon. See you. >> Bye.
2:52:49
>> Uh let me tell you about wander. com.
2:52:51
Book a wander with inspiring views, hotel grade amenities, dreamy beds, top tier cleaning, and 247 concier service.
2:52:56
Our next guest is already in the race room waiting room. We have Ali Madani. Welcome to the show. >> How you doing?
2:53:02
Thank you so much for taking the time to come chat with us.
2:53:04
Uh please introduce yourself, introduce the company, tell us what the news is today. >> Sure, absolutely. Uh so my name is Ali.
2:53:12
Um I have a PhD in machine learning from uh UC Berkeley.
2:53:14
I've been working in the space of biology and AI for almost a decade now.
2:53:18
Uh previous to this, I I led a moonshot at Salesforce pioneering language models for biology. Yeah.
2:53:22
Um, and what started out as a purely scientific endeavor to develop transformer models for sequence generation has led into ProFlent spec Pro Profolent specifically where our mission is uh to make biology programmable and I'm happy to kind of break that down. >> Yeah. Yeah.
2:53:40
Uh I I I've seen obviously there's there's a ton of uh just like momentum in the space.
2:53:46
AI curing cancer is like a buzzword that a lot of people are throwing around.
2:53:50
Um, how are you uh thinking about concretizing what you're actually how you're trying to fit in? Are you a tool? Are you a drug maker? Is it uncertain?
2:54:00
Uh, like who are your customers?
2:54:03
How how much are you in like a science project world?
2:54:06
Like, you know, you could be a nonprofit in another in another era versus like you're ready to commercialize, you're going to market.
2:54:14
And not that there's one path that's wrong or the other, but I'd love to know how you're thinking about the business right now. >> Totally. Yeah.
2:54:20
I think there's a a lot to unpack there.
2:54:21
Um I think [clears throat] the meme that came to mind specifically, I don't know if it's a South Park meme or otherwise where it's it starts with like build something and then there's a dot dot dot question mark and make profit and then profit. Yeah.
2:54:32
Step one, step one, [laughter] you know, make biology programmable.
2:54:34
Step two, bro down with your boys. Step three profit.
2:54:39
And >> and I think a lot of folks, you know, right now there's an incredible amount of excitement around AI.
2:54:43
And it's kind of like step one is make a chat bot, for example, right?
2:54:47
example, right? and then it's question mark dot dot dot and then solve you know disease or cure cancer specifically whereas what we're actually trying to build here is actually tackle on the disease head-on specifically so what we do is we build language models so the same language models that have enabled
2:55:04
GPD2 3 and four and chat GPD specifically these incredible models and algorithms that can learn on sequences what we can feed instead of words in a sentence is actually amino acids that are strung together to form a protein and uh why that's actually important and why making biology programmable. Maybe
2:55:18
Maybe to take a step back, like people usually shut off their brains when it comes to biology. Yeah.
2:55:23
And and when it comes to like rockets landing on a platform in the ocean, we're amazed, right?
2:55:27
And that makes sense, right?
2:55:29
Like it's these are man-made machines. We can see it. They're incredible.
2:55:33
But honestly, biology is not that much different.
2:55:36
There are these molecular machines called proteins that enable us to breathe and see.
2:55:40
They're responsible for everything in human health and disease.
2:55:43
And also they sustain the environment involved in daily daily products like even our detergents to begin with and how let's actually stick to drug discovery in particular.
2:55:51
How we've gone about finding these solutions these molecular machines that we utilize day in day out has actually been through random discovery.
2:55:59
random discovery. So you know that that middle school example of Alex al Alexander Fleming coming across penicellin right he had a petri dish they molded for example and then they found the advent of antibiotics and now after you get a cut on your skin for example where bacterial infection happens it's no longer a death sentence right um that actually is not the exception it's the rule in which we've
2:56:20
gone about finding life-saving medicines even fast forwarding to today crisper cast 9 was actually found in a Disco yogurt facility where people found these interesting bacteria doing these interesting characteristic had having these interesting characteristics and we've taken a molecule plucked it from nature and then crammed it within human therapeutic applications to actually save lives. And honestly like to put
2:56:39
And honestly like to put this really in rudimentary terms that's that's kind of absurd.
2:56:44
It's almost cavemanlike in terms of our technique our techniques that we have and methods that we have available for us for drug discovery.
2:56:51
And what we're trying to do is actually move away from random discovery and finding a needle in the hay stack and relying on nature altogether and using AI to design bespoke medicines from scratch.
2:57:01
Um and that's you know like that's our mission to really gain control and mastery over biology and perform bespoke design.
2:57:08
So in terms of your question of like where are we with respect to you know is this just a science project or how's the commercialization looking specifically?
2:57:18
I would still say we're in early days like the equivalent of GPT eras of like maybe GPT1 or GPD2 but we've already seen incredible amount of traction.
2:57:24
seen incredible amount of traction. So we had this project called open crisper specifically where we took is uh we took these uh these language models trained on gene editing proteins specifically and generated a novel protein from scratch called open crisper one that thousands of people use now in pharma large pharma small biotechs academics
2:57:44
and ind industry um uh users and scientists as well and over thousands of people use this over worldwide today and I think that's like it's amazing to actually see us solving problems today that have lead to commercial traction and that we have partners both from uh therapeutics to diagnostics to biommanufacturing even agriculture that are utilizing today. >> Can you talk about like how you create
2:58:04
>> Can you talk about like how you create feedback loops as a company because you know there's no shortage of people in AI that talk about the uh opportunity of like curing various diseases.
2:58:15
Many of them aren't saying that from the standing in an actual lab.
2:58:20
You are standing in a lab.
2:58:22
that makes me more excited about what you're doing because you're not just kind of, you know, like it there's you're not just saying like, oh, like the next version of the model will we'll just do this like don't worry about it.
2:58:33
It's like no, like we're going to run a lot of experiments.
2:58:35
Um, but yeah, talking about >> yeah, like you know, using AI to uh to to learn and and generate uh you know, potential approaches uh but then actually bringing it into a lab setting. >> Absolutely. Yeah.
2:58:50
um we operate within a pre-training and post-training paradigm within proteins, similar to NLP and natural language processing as well.
2:58:56
So the pre-training step really involves in similar to how we have all of the internet that we can scrape from and can learn these underlying principles and grammar and semantics as to what makes human generated texts.
2:59:07
We've actually collected a tremendous amount of data of proteins that have naturally evolved through nature for selective reason selective pressures and evolutionary kind of pressures that have shaped those proteins specifically to make a functional protein.
2:59:19
Um, and just to put that into context, um, Alphold 3 was trained around, it was exposed to around two to three billion proteins.
2:59:25
Uh, what we've actually trained to date so far at ProFlint is over a 100red billion proteins.
2:59:31
And to put that into uh, tokens, that's over 20 trillion tokens. Exactly.
2:59:36
[laughter] Um, so there's there's an incredible amount of data for pre-training purposes that we utilize.
2:59:41
And then what you see behind me as well is the data that we're doing the assay labels labeled examples meaning actually taking protein sequences and then measuring their function not just in vitro and test tubes and petri dishes but in human cells and relevant cellular contexts and seeing how well they actually perform and we can feed that back into our models.
3:00:00
So I think that's you know the future is really an integrated future where you're building frontier AI models and having uh the the the the closed loop specifically with respect to the wet lab which is what what's behind me today um to actually test these and feed them back into our models to get better and better over time.
3:00:16
So yeah >> well congratulations.
3:00:18
I want to ring the gong for you.
3:00:20
What's >> by the way Gersner and Bezos I mean >> how much >> potentially the coolest >> cap table the cap table [snorts] of the year. How much was the deal? >> Yeah, absolutely.
3:00:31
Yeah, it's $106 million that we're announcing.
3:00:36
>> Um, and I I think what's more important what's more important than number are these legendary investors that we have.
3:00:42
I mean, Jeff Bezos is a legend.
3:00:42
Uh, he's transformed industries.
3:00:44
Uh, and I think what's exciting for him and for us as well is that biology is the next frontier for AI specifically that will have tremendous impact.
3:00:53
Um, and really honestly is the most important quest in our lifetime.
3:00:57
So, we're really excited.
3:00:59
so much for >> I'm sure you will be back on very soon.
3:01:02
>> And congratulations on all the progress. >> Great. Great to meet you.
3:01:04
>> We'll talk to you soon. >> Talk soon. Have a good one.
3:01:07
>> Um, our next guest is already in the reream waiting room.
3:01:08
We have a hard stop at two. We got to run.
3:01:10
We got to hop on with New York.
3:01:12
So, let's bring in a meet from Luma AI with some massive news. How are you doing? It's been too long. Great to see you again. Welcome to the show. >> What's happening? >> Give us the news. What happened today? What happened? Break it down for us. Yeah.
3:01:28
So, um, we did two massive things.
3:01:31
One, Luma raised a a 900 million series C. >> Okay. What was the second? >> Come on. I'm I'm sorry. I'm sorry. What?
3:01:40
Like, what [laughter] was there not another 100 million lying around?
3:01:42
You couldn't like You couldn't You got to You're going to make us wait for the the Luma1 billion dollar round. Come on, dude.
3:01:51
>> I'm very happy to accept friends and family checks. >> Okay. Oh, yeah.
3:01:54
If you got 100 million and for from your friends and family to round out that'd be fantastic.
3:01:58
And then what yeah what is the second thing >> and yeah the second thing is basically along with humane which is a a you know this AI company being built in Saudi Arabia.
3:02:10
>> We are building a 2 gawatt compute cluster >> that we're going to use to train uh you know multimodal AGI. >> This is the big news.
3:02:18
The the big news this is much bigger.
3:02:20
This is much more important.
3:02:21
This is what we actually need the compute.
3:02:22
the compute. So, so you know the tier of what happened here is basically you know so far LLMs and LLM labs have had the right resources and multimodality world simulation these problems actually you know were side projects for for most
3:02:36
companies now there is a lab and there is a company in the world uh that has this level of resources and is going right after AGI that can help us in the physical world AGI that can help us simulate and and and and you know uh generate the universe. So I think uh
3:02:51
So I think uh that's actually what happened basically. >> Amazing.
3:02:55
Uh so >> how do you think about h how how do you explain the scale of 2 gawatt because it sounds like two is not a big number >> and and and is that is it 2 gawatt because you're expecting 2 gawatts worth of inference or do you need a particularly big cluster for some sort of pre-training run that you're planning on doing?
3:03:16
>> So it's both inference and training.
3:03:16
>> So it's both inference and training. Um but inference is actually so you know majority of the workloads as we go forward right like you know as AI deployment uh uh goes forward as we mature from just texton models to models that are able to like you know
3:03:31
generate videos models that are able to explain things to us in video what's going to happen is most of the workload and tokens will will move to video understanding and video generation and video tends to be you know computationally much more intense than than than language. So we need this
3:03:46
So we need this level of compute to be able to deploy this technology and to be able to train.
3:03:52
Uh but this is mostly inference honestly even today Luma's inference to training compute ratio is 2 is to1 already uh and and we're seeing that ramp actually growing further and further and further while we we do deep research and and train some some of the largest models in our space inference is the one that is actually taking off.
3:04:09
Okay, react to this uh this uh take I got from someone who's also building a world model, a generative world model.
3:04:17
Uh he told me that he believes that uh it's more likely that AGI something fully uh paradigm shifting emerges from world simulation than merely scaling up next token prediction GPT 54 567 89.
3:04:33
uh getting away from text is actually somehow foundationally important to um the next major breakthrough in AI as we know it as a whole.
3:04:47
>> I think getting away from text is a mistake.
3:04:49
>> Uh [clears throat] we need to build models that combine audio, video, language and image.
3:04:54
So like you know we need to build things that like operate like human brain.
3:04:57
If you remove text, you remove the entire interpretation of of the the human logic and and like you know reasoning and those kind of things.
3:05:05
So we need the physics that comes from video.
3:05:08
We need the causality that comes from video.
3:05:09
And we need the text which which actually makes all of this interpretable and logically connected you know across the world.
3:05:15
So no I I think what we need to do is build these joint unified models.
3:05:20
But uh on the simulation side I agree and I think that's really really important because think about robots or think about systems you know uh how they would operate right like they need to be able to understand the world.
3:05:34
to understand the world. So this is world understanding which is where world models are going to be very very powerful and multimodel models are going to be very powerful and second is simulation being able to run the the process or idea in your head and and and drawing out conclusions right you know what if I go 20 m this way uh would I
3:05:50
fall right this is a simple question but as robots become more general purpose and in day-to-day in our lives we need this level of simulation capability in their heads so generative models give you simulation capability right simulation is extremely important second thing is LLM s are really good at things that can be represented more or less fully in text code analysis these kind of things. But when we think of the
3:06:11
But when we think of the physical world especially acts like designing uh you know manufacturing these kind of uh topics like one of the things we think a lot about at Luma is is manufacturing of a jet engine right or manufacturing of a rocket engine.
3:06:25
These one of the most complex things humans do and it takes a decade to build one.
3:06:29
Imagine having models that are able to run these physical simulations and get to an answer.
3:06:33
It's not the about the visuals, it's about getting to the right answer.
3:06:36
People do that in CAD, people do that in in like, you know, software today.
3:06:40
But it's like very uh uh um inaccurate.
3:06:43
But if you're able to build models that can accelerate building of these complex systems, humanity has a chance at like you know uh uh building better and better things for ourselves for for our planet.
3:06:53
So that is why simulation is really important and that's why multimodality is really important.
3:06:58
And text is just the first step.
3:07:00
step. text is like you know 1990s internet then we got images on the internet then we got videos on the internet and today like you know videos is the internet for humans at least >> um AI will not be any different >> yeah um last question from my side uh what is the actual timeline for building
3:07:16
a 2 gawatt cluster >> yeah where where where will the majority of the infrastructure be >> when can I see it when can I go inside I can be trusted >> so some of it already exists so by the way we are building this uh in partnership with humane in Saudi Arabia Yep. >> And uh today it was announced here. So
3:07:30
>> And uh today it was announced here.
3:07:30
So we in DC right now uh for for the um US Saudi investment forum. >> Oh no.
3:07:36
>> And it was announced uh by President Trump and and Crown Prince uh Mohammed bin Salman.
3:07:41
Um so the data center is going to be built in Saudi Arabia.
3:07:44
Uh quite a lot of capacity is actually already available and Luma is actually an active customer and using that today.
3:07:51
But the deployment of 2 gawatt is going to take time.
3:07:53
That's an absolutely colossal amount of power and infrastructure that needs to be built starting with 2026 and and currently we believe that like you know by by uh end of 27 or early 28 we will have majority of the capacity at hand and and uh like you know we'll go from there. >> Fantastic.
3:08:10
Uh well thank you so much time while you're traveling to come chat with us and break down what's going on.
3:08:17
Uh congratulations on uh the amazing news and uh good luck with the next phase.
3:08:22
I'm sure there's a lot going on.
3:08:24
>> Next time you call in, come call in from uh from Saudi. >> That'd be amazing.
3:08:27
[laughter] We'd love from the desert. That'd be amazing.
3:08:30
>> The first time actually I was on TBPN, I was in Saudi. >> Oh, no way. >> There we go. We already did it.
3:08:35
[laughter] >> Well, check that box.
3:08:36
>> Next time next time I want I want, you know, one of those 4x4s that uh you know, call call in from the desert from from Humane.
3:08:43
I want I want I want to be live on the ground with you. >> Amazing. Great to see you again.
3:08:50
Congrats on the progress.
3:08:50
Thank you so much for jumping on.
3:08:53
>> Uh we have to hop on with New York.
3:08:53
But first we have one post we got to pull up.
3:08:58
>> One post >> and it's a post that I made.
3:09:00
>> You made >> right when I saw that beat earnings they have traded up. The stock is up 3. 8 3. 91%.
3:09:09
Uh massive massive it is at the very continuous they were signed. This is your prediction.
3:09:13
uh one of your one of your many predictions, but >> this is all the only the only data.
3:09:20
>> This is the only data you need to know.
3:09:21
You know, you know, uh you said this.
3:09:21
I think he's going to beat earnings because he's drinking beers.
3:09:24
And uh and Ev was like, "Yeah, you belong in a pod shop."
3:09:27
And he was saying it like sarcastically.
3:09:29
Like, you know, to be in a real hedge fund pod shop, like you have to be much more quantitative than that. Turns out you don't.
3:09:34
Turns out the vibe analysis works. Take it all in. Absolutely.
3:09:39
Uh, thank you to everyone for tuning in and watching our show.
3:09:44
Leave us five stars on Apple Podcast and Spotify and we will see you tomorrow.
3:09:47
>> Global economy continues. >> Continues.
3:09:48
The party continues, folks. White suits tomorrow. >> Gabe in the chat. Gabe getting drunk. Drunk responsibly.