0:09
[Music] [Music] [Music] [Music] [Music] [Music] [Applause] [Music] [Music] [Music] [Music] [Applause] [Music] [Applause] [Music] [Music] [Applause] [Music] [Applause] [Music] [Music] You're watching TVP.
Search this video
4:58
Today is Monday, October 6th, 2025. We are live >> 1999. 1999.
5:04
Is January or is it December 1999? We don't know.
5:08
We're going to figure it out.
5:09
Uh we're live from the CB pin ultra dump.
5:11
The temple of technology, the fortress of finance, the capital of capital.
5:14
Today is opening eyes dev day.
5:17
Uh 1500 people are attending.
5:21
It's their biggest event ever. They're in Fort Mason.
5:23
Uh Sam Altman and Jony Ive are going to be giving a fireside chat. That'll be fun.
5:29
Uh there's been rumors that the hardware device that they're building is delayed.
5:34
Uh but I kind of don't care.
5:34
I read the Financial Times.
5:37
We'll We'll dig into the Financial Times report, but um you know, it's supposed to come out in 2026.
5:41
Might be pushed a little bit.
5:41
Uh I don't doubt >> Nothing about the leak made me want it. Yeah Yeah exactly.
5:47
Like I don't doubt that they're going to be able to ship a product.
5:50
Like it's Jony Ive and Sam Altman.
5:52
They're going to be able to figure out how to manufacture something and sell something.
5:54
My take was that there are two things that they need to solve that are way more important than like can they get a manufacturer to make it and ship it.
6:03
Like that's not the problem.
6:05
The The first one is will people adopt the form factor they pick?
6:09
I The rumor's no screen, which has been tried before by startups with little success.
6:13
And I remain pretty pilled on what the meta head of wearables told us about like these Lindy form factors, the watch, the tablet computer, the notebook, the hat, the glasses.
6:24
Uh if you were selling me a smart hat, I might be more down.
6:29
Uh I mean the chain I I I guess like the Jesus piece, the smart Jesus piece could be, you know, you're you're building AI God, why not create a talisman to do it.
6:41
>> bigger chains for more compute. >> Compute.
6:43
Uh something something there.
6:45
Um but the pin always felt a little odd.
6:49
Uh and it seemed like it's harder to get people to swap over.
6:52
Um but the biggest that the second thing that's like really crazy is um there are just insane lock-in effects that not even Jony Ive can overcome when competing with Apple.
7:02
We talked about this with Meta Ray-Ban displays.
7:04
You don't get iMessage notifications in Meta Ray-Ban displays.
7:09
>> you had personal super intelligence in this new device form factor.
7:12
You're using the meta buzzword, but this is also what uh the Sam Altman device is >> in the world except send messages on your >> except text your wife.
7:22
Except for like checking on your kids or like text your buddy from college.
7:26
Like it that is a serious serious missing piece of any hardware device that is not going to be integrated.
7:33
iMessage has become so dominant. It's been remarkable.
7:37
I never thought of it as a social network, but now I do.
7:41
>> One thing that we can take away from the Humane AI pin is that and and again, the Humane AI pin from my sense of the leaks around the Jony Ive project is that it's not that different.
7:55
It's It's It's more similar to the Humane AI pin my head right now than it is different.
8:02
>> Humane pin had that like screen that was like the projector.
8:04
I think they're going to be doing away with that entirely, going voice only.
8:06
It's clear that like te- uh te- speech to text is good enough now that you could have a voice interface.
8:13
I mean, I'm sure that they've seen in the analytics for the ChatGPT app. What's the latest data?
8:17
800 million weekly active users, right?
8:21
>> Yeah, that was in the dev day uh like presentation.
8:24
>> And I think Jordi had a question for you about that. Is that good?
8:29
I heard the number and thought there's a lot of big numbers thrown around.
8:31
Usually they're in the billions these days. >> Yeah.
8:34
Yeah, kind of sandbagging, honestly. Get to a billion, dude. Come on.
8:38
Um but uh uh getting that uh to actually uh reach a reach a point where people are um you know, using this with their iMessage.
8:50
There's a bunch of interesting workarounds.
8:51
Like I could imagine the real prosumers, the real techy folks being like, "Yes, I I have a Mac at home that runs a piece of software that like screen scrapes and interacts and puppeteers all the Apple apps and sucks all the information out of that and puts it wherever I want."
9:10
Uh I don't think Apple's going to take kindly to that.
9:12
I think they're going to throw up a bunch of barriers and people are going to be really, really sticky staying in their Apple ecosystem at least for like a year or two until Apple can just catch up and kind of clone it back.
9:24
I'm just not at all convinced that we need a new device for the current state of models. Yeah.
9:34
When you think about >> But new new new devices are exciting.
9:40
They They They're great for building hype.
9:41
If you look at uh um you know, friend.
9:44
com gets 10 times the amount of attention that it does if it was just a maybe even a hundred or thousand if it was just a website that was kind of an angry Uh friend. Yep.
9:57
you know, you you're you they they get a lot of attention and excitement because you're you're you're getting into the conversation around net new platforms and new hardware layers, but we've just seen with Humane you can build you can build something that's built natively around AI and it can be a very functional device, right?
10:15
Nobody was arguing with Humane that it wasn't like it didn't do like it technically did what it said it was going to do.
10:25
It's just what it does is not that useful. Yeah.
10:27
When you compare it to the cell phone that's already in everyone's pockets.
10:31
>> Yeah, or when you compare it to ramp. com. Time is money. Save both.
10:32
Easy to use corporate cards, bill payments, accounting, and a whole lot more. All in one place.
10:37
When are we going to get the hardware device from I mean it is a technical hardware company. You get the card. The card is hardware. It's a hardware company.
10:46
Uh the Yeah, the the the Humane device was a little rough.
10:52
It's just It just feels like there's more opportunity than there is because of how slow Apple moves, but when you think about them partnering with Anthropic at some point and yeah, you're not chatting with GPT-5, but you're chatting with Claude and you can do that on your AirPods, on your phone, and on your watch, and on your glasses.
11:14
Like it's just their game to lose, I feel like in terms of hardware.
11:16
Like they'd really have to mess up to not eventually deliver something when the time is right.
11:24
>> AirPods are like a 20-plus billion-dollar revenue run rate product.
11:29
It's good great for audio and integrated and you can hear the outside world coming in and I just remember like when I was a kid there was a little bit of like an arms race for hardware devices and Apple was behind on a lot of things.
11:45
Like they came out with the iPod and there was this company called the iRiver that was ahead of them on releasing features.
11:51
So, the the iPod had a a black and white screen and this other company came out with essentially an iPod.
12:00
It blew it out of the water on specs.
12:02
It had a uh had a color screen, you could watch movies on it, had way more It was way cheaper.
12:08
Uh and Android has done that a lot where the Android phones have had more advanced features many, many times, but people are just locked in. >> up the iRiver.
12:17
iRiver >> It looks like a vape. Yeah.
12:20
>> This was some crazy hardware back in the >> It was great. It was great times.
12:24
Um Uh did you see Soham Parekh was weighing in on >> On the Nvidia >> AMD deal.
12:31
Uh when Soham Parekh, the real Soham Parekh, is weighing in on on some of these circular deals, it uh >> Is this the real account?
12:41
There were several knockoff accounts.
12:42
This is the real account. [Music] >> Okay, okay.
12:47
He says Nvidia Oracle opening eye AMD, what a time to be alive. Certainly is. >> Yeah.
12:53
Got to get the update from him.
12:55
How How is his life uh post Going exclusive. Going exclusive. Going exclusive.
12:59
Yeah, we're having Doug O'Loughlin from Fabricated Knowledge on the show in a little bit and one of my big questions is like how deep will this go?
13:06
We've heard uh stories about Sam Altman doing deals talking to TSMC and I'm wondering like does this get to the point where he's talking about he's talking to ASML?
13:13
He's talking to like like niche lithography machinery manufacturers about scaling up?
13:21
Like he is actually trying to get the entire supply He should start going around every company and saying if you give me warrants, I'll announce a partnership and your and It seems to be working. Your stock will go up.
13:36
On average, we've seen the the the uh market caps rise, you know, between 20 and 30% and I'd love to be able to do the same with you.
13:44
>> People are into the OpenAI economy.
13:44
They are truly propping it up.
13:46
Um well, if you're watching OpenAI DevDay later, it's probably on restream. restream.
13:53
io One live stream, 30-plus destinations, multi-stream and reach your audience wherever they are. Sign up for free. Um What else is in here?
14:01
Uh they're launching Agent Builder.
14:04
Tyler, do you have an update on what they actually launched?
14:06
I I wrote a take kind of based on the leaks and the and the previews and I'd love to hear your feedback on my take, but what what do we know so far?
14:14
Yeah, so so they did um it's already over.
14:16
They had the main live stream.
14:18
There's like four >> you just say Tyler at TVPN says it's so over?
14:22
I thought it I thought we were so back. >> It's already over. No, no, no.
14:24
We will get that We are We are so back.
14:27
>> We will get the We will get the timeline update from you in just a little bit, but give us the facts of what was launched and what the product is. >> Yeah, okay.
14:33
So So there were four four like main main kind of things.
14:35
So the first one was like apps in ChatGPT.
14:40
So this was like that they showed Zillow.
14:42
It's kind of just like a a little web view. >> Yeah.
14:44
Um there was like a Figma thing where you can you can prompt Figma.
14:46
It's almost kind of like a Figma make thing, but just just it's still on like chat. openai. com. >> Yep.
14:53
Um And it requires you to actually have an account there and so like like the company that's partnering with them is like monetizing through their traditional path, but then they're using ChatGPT as like a front end.
15:03
>> Yeah, and they announced um like soon it's going to be like more open.
15:05
There's only I think there was like maybe eight or 10 companies doing it right now.
15:08
Um they're announcing like more ways that it's going to be monetized. >> Mhm.
15:12
Um and then they did uh there's like the Agent Builder.
15:14
So this was the the leaked kind of video where it's it's almost like a It's node-based workflow.
15:20
>> Yeah, it's almost like almost like a Zapier kind of thing where it's just like a little bit more intelligent, but >> Yeah.
15:25
there are still it's like, you know, if if statements just like a single >> jobs, too? Um I think so, yeah.
15:31
>> So so in theory, I could say everyday check my email, create a digest, then go and do a deep research report on what's going on in the tech news, and synthesize my inbox against the tech news to surface anything relevant.
15:46
So if the CEO of AMD has emailed me and AMD's in the news, I definitely want to get back to them immediately because we want Lisa on the show. Something like that. Yeah, yeah.
15:57
But I mean it is there's a big difference between this and kind of what what people usually think of as like agents. >> Sure.
16:04
Like when this is very different than um uh what was the the thing they they launched a while ago?
16:11
>> Well, they have Operator.
16:11
Yeah, they have Operator.
16:13
>> different from Operator.
16:14
>> And then they have which has browser.
16:15
And then they have Agent mode.
16:17
>> times have you used Operator in the last 3 months, Tyler?
16:21
I don't honestly this is pretty bad, but I don't know if I ever used Operator.
16:23
I used Operator a few times.
16:25
It got kind of stuck on like captchas and stuff the one time, but I have used Agent mode more recently and that's been It's kind of like deep research plus.
16:33
Like it can just go farther into websites.
16:35
Like if it if it it can go on to a website like Zillow or Redfin and and it doesn't just need to be like blog posts and summary articles and PDFs.
16:45
It feels like it can actually go to a website and then type in search results and get a whole bunch more data.
16:51
It It does seem like a better product, but it doesn't seem like it's having crazy takeoff in terms of adoption. Yeah.
16:56
Um and then so the the third thing was about Codex. Yep.
16:59
So Codex was in research preview. It's now like GA. >> Mhm.
17:04
Um I I don't like that there weren't a ton of like big updates um there.
17:09
Um and then the fourth thing was just like various APIs.
17:11
So there's there's a Sora 2 API.
17:13
It's 10 cents per second, which is like quite cheap.
17:16
That's way cheaper than I thought it would be.
17:19
>> Yeah, we we were debating this, right?
17:21
It was $3 per generation on VO3, right?
17:26
Yeah, and around that that sounds right.
17:28
>> So if you truly took advantage of your VO3 account, allegedly, you were basically break even for them, but no one did.
17:34
Like I would generate three and then I would forget about it for a week and then I'd come back and generate the max.
17:40
But it seems like they did they did figure out how to optimize inference to some one order of magnitude, not two, which we were kind of speculating about.
17:48
Yeah, and then there's Sora 2 Pro, which >> They could be subsidizing it, though.
17:50
We we we we still don't >> sure it's still subsidized it to some extent.
17:53
Uh but then there's Sora 2 Pro, which is like 30 seconds. That's pretty good. Um no, no. 30 cents per second.
17:59
Okay, and it's a little bit longer, right?
18:00
It can go from like 8 seconds it can go to like 15 seconds or something intelligent. >> I think so. That's pretty cool.
18:06
>> Is it worth reading into the AMD announcement happening the same day as DevDay?
18:11
You think it just had so much motion that it was just going to get announced as soon as it was ready or do you think this was >> like multiple tracks going on and Sam's just pushing both like full tilt.
18:23
Because last last week there was, you know, the Nvidia deal, the Oracle deal, the Broadcom deal.
18:27
Like all of those things were happening simultaneously with the launch of like GPT-5, Sora, the app, the there was another announcement that they launched Oh, the Pulse app, like the actual like ad inventory, agentic commerce.
18:40
Like there's just been a new release on the product side every day and then there's also been a business deal every day.
18:45
Like we got the Tinder deal.
18:47
Like these things feel like they're like they're not able to time up one next to the other. It doesn't matter.
18:54
They're just like all flowing really freely.
18:55
Um much like wallet infrastructure for every bank from Privy.
18:59
Privy makes it easy to build on crypto rails securely to spin up white label wallets, sign transactions, integrate on-chain infrastructure all through one simple API.
19:08
Um So my my take and I'd love both of your feedback was that this feels like something that puts us in like the prosumer world of AI.
19:17
So right now I view AI as there's the consumer world, which is just uh you go to ChatGPT, you ask it how to make banana bread, it gives you a recipe.
19:29
Uh you you are not expensing that.
19:31
You're not underwriting that as some sort of like business value.
19:33
You're just using it like Google search, knowledge retrieval.
19:37
Maybe you're chatting with it as a friend or whatever.
19:39
Uh and that needs to be monetized with ads.
19:44
That needs to be free over a long period of time.
19:46
Then you have like enterprise software, the SaaSification of AI.
19:48
The the the if you have a true business workflow >> that you are automating, you're probably working with LLMs at the API level building software and then potentially selling that software to other people.
20:02
And so this looks like, you know, I I went to, you know, like Ramp got better at tagging receipts and it's like I didn't even notice.
20:11
I'm not chatting with anything necessarily.
20:13
It's just doing stuff behind the scenes because every company is integrating those.
20:16
My my question is we've heard rumors of like getting to the $2,000 per month bill for OpenAI features.
20:24
Like does this agent builder potentially get us there?
20:30
Do we wind up in a world where, um, you know, we've seen this before with like when images in ChatGPT launched, everyone generated like a few Studio Ghiblis like the day of, but then 6 months later it's kind of bifurcated into consumers use it every once in a while as like a fun
20:47
thing, but then there are professional people that are in it basically all day long as a workflow tool because they're visual artists and they need it as a canvas to pull from and then edit and create other stuff or they're doing blog images or they have a variety of reasons for using that product. >> Yeah. And they use it in a professional >> Yeah.
21:04
And they use it in a professional workplace basically.
21:06
workplace basically. Um, and and I'm wondering if if some of these agentic workflows will have serious inference costs because you you're wiring them up and you're saying like go to GPT-5 Pro every single day and write up GPT-5 Pro query
21:22
for every single, you know, news item on my to-do list and then generate a Sora 2 video, what weave that in and you wind up you wind up spending like $10 a day or or $50 a day and pretty quickly Fearing the chat says maybe you aren't expensing banana bread recipes, but I am. Let's go. Let's go.
21:40
Um, Depends how elite the banana bread is. Sure, if you're a baker.
21:42
It'll be an amazingly strategic asset.
21:45
>> Yeah, yeah, if you're a baker.
21:45
Um, and and so and so I wonder like we are getting to this point where some people are are thinking about their OpenAI use.
21:58
You're underwriting it under some sort of loose business value.
22:00
You might be expensing it.
22:02
You're thinking more seriously about this and I wonder if we'll get to basically consumption-based pricing in like a prosumer type world based on these flows.
22:11
We we we see that with the Sora 2 stuff.
22:14
Right now it's all it's all balanced to be free, but you have a rate limit.
22:18
I wonder if we'll see the agent builder, if you're really hitting this thing constantly, if they will start charging more for people.
22:25
I think there's a little bit of like organ rejection from consumers where they're like, I I'm I'm much more comfortable just saying I'm paying 200 bucks a month and maybe I'm using it not enough.
22:34
Maybe I'm using it, but I want the all you can eat plan.
22:38
>> Yeah, I guess my question for you, do you do you expect average consumer at any point to be building out a variety of workflows? >> No. No.
22:46
So so I I see them as like two two like like there's multiple markets.
22:50
We've been for a while we've been modeling OpenAI's business as like they have a consumer product that they're going to try and drive to essentially free with ads and you'll have a ton of features and they will try and deliver you a ton of value.
23:03
So Sora, the generations are free.
23:06
You can't pay for anything more in the app.
23:08
Also the apps they launched today, those are very much like consumer. Exactly.
23:11
And but but they'll take a cut on the back end and so they will monetize very heavily on the consumer side.
23:18
And then you also have like your hyper scalar cloud platform, your API business that is pure consumption-based selling the API, selling the LLM.
23:27
But I'm but I'm wondering if we will if we will see the development of like something in the middle, something prosumer, something consumption-based pop up.
23:34
Now there is >> Yeah, yeah, the to be clear it's this company N8N which we haven't really followed closely.
23:41
We should have the founder on.
23:43
Um, but they have seen a ton of growth.
23:46
They've seen this organic kind of um, uh, always a good sign when people are selling kind of like courses around how to build a business based around your software product, right?
23:55
This was Shopify in the early days and so a lot of the posts this morning are saying OpenAI just killed um, N8N.
24:01
Uh, how do you say this actually, Tyler? >> Is it Naten? Yeah, I I don't know.
24:07
I I usually I just when I read it I just think N8N. Yeah.
24:11
So, um, may have just proven to proven myself to be a boomer if I'm mispronouncing it, but N8N N8N, we'll see.
24:21
I don't I don't think these My framework for OpenAI now is that I just view them as a large tech company and so every VC and every founder should be asking like what if OpenAI does this and even if they do decide to do it, it doesn't mean that you're cooked because for a lot of reasons.
24:39
How many how many products have big uh, you know, how many products has Google launched or or or Meta launched that looked like they were a credible threat to various businesses and then ultimately didn't really go anywhere.
24:54
So I'm I'm sure that OpenAI is taking agent builder seriously.
24:57
I'm not sure that it'll be a real threat to Zapier, N8N or anyone else.
25:02
Yeah, there's a great internal tech email from 2010 explaining what competition between Facebook and Google looked like at the time.
25:12
So the abridged history of Google and Facebook sometime in the summer of 2007.
25:16
So this is just like 2 years after Facebook launched, Larry Page stopped talking to Mark Zuckerberg.
25:23
Uh, this was following an 8-month period in which Facebook launched newsfeed, opened registration to everyone and launched platform where you could log in with Facebook across the web.
25:35
So almost instantly Facebook was competing with Google for talent, for developer mindshare and for tech blog cred, kind of like with the precursor of teapot, I suppose.
25:46
Um, they began to worry about their ability to enter new markets with us in the game, with Facebook in the game.
25:53
And about core investments that weren't doing well and that Facebook was positioned to touch.
25:57
Checkout, local or Orkut, which is their social network, which is now defunct, I believe.
26:04
They worried about entrenched products that >> What was it called?
26:07
Orkut, o r k u t, which I think was a Brazilian company that they acquired, which I believe it's pronounced Orkut or something like that. I like that.
26:16
Put a little fire on We look over at Tyler like he's supposed to be pronunciation expert at 19.
26:23
But yeah, uh, Google was worried about Meta challenging them in entrenched products that Facebook might one day launch.
26:30
Calendar, Picasa, which was their photos product, YouTube, Blogger, Reader, News and about the comScore data telling them that the social networks are bigger than email story.
26:42
Facebook had the world's people while Google only had their cookies.
26:46
So we were better positioned even to own search and ads one day. Oh my.
26:51
So far we haven't seen real the real competition.
26:54
And so what's interesting is that it was not a total steamroll by the Meta Facebook team.
27:00
It was also not a total steamroll by Google.
27:03
Like Google I would say did very well in calendar.
27:08
They Facebook is not dis disposed >> became big calendar. >> They did. YouTube did very well.
27:13
Reader, of course, they they they canned. Blogger went away.
27:21
News is still there in search, but I think most people get their news from social networks these days.
27:27
Picasa, I was talking to Brandon about this, photos has kind of most people have moved over to just Instagram.
27:31
Their Instagram page is their main photo album.
27:36
People do organize photos, but a lot of people do that in Apple.
27:38
Um, and so there's been a bunch of places where Facebook won.
27:42
There's been other places where Google was able to hold on.
27:44
And so I would imagine that the same thing happens where OpenAI challenges the legacy tech companies across the board and and they win some of those, but not all of them. Yeah.
27:54
My question is does releasing Sora 2 via API hurt the chances that uh, Sora as a standalone app actually becomes a social network and consumption platform because the model's clearly great.
28:09
It does a really good job at making video look like it was shot on a cell phone, right? VO3, very cinematic.
28:17
It looks like a Hollywood blockbuster.
28:17
I mean it looks like daytime TV, but yes. Yeah, sure, sure.
28:23
But it looks like it was shot by legacy Hollywood and Sora 2 looks like, you know, you're hanging out at the office and you're shooting a video, right? Something like that.
28:31
And so that model, which until today they had exclusive access to, can now be leveraged by any video creation platform and there's tons of them.
28:42
So whether you were Adobe, right?
28:44
You could get access to Sora 2 or some entry-level startup, right?
28:50
And so what that means is there's going to be a lot more of this kind of content being created all around the internet, not in Sora the app, and in my view that signals to me that they don't really feel like they have a a a strong opportunity to make Sora a place where content is consumed.
29:11
I see them seeing this as a creative tool uh, because if you were like, hey, there's a 30% chance we're going to make a new like video massive video consumption platform, wouldn't you want to give it every possible chance to be successful?
29:29
You wouldn't want to let all your enemies have access to this tool, right?
29:33
You'd give yourself every possible edge.
29:36
I think I think you would.
29:38
I agree with a lot of that.
29:38
I have a rebuttal, but first I'm going to tell you about Cognition.
29:40
They're the makers of Devin.
29:42
Devin is the AI software engineer.
29:44
Crush your backlog with your personal AI engineering team.
29:45
So, yes, Ben Thompson has always had this good take that why does OpenAI have an API at all?
29:54
You don't want your GPUs on fire when you have the best consumer AI app.
29:55
Just make sure that ChatGPT never goes down that the maximum amount of GPUs are on ChatGPT so everyone has the greatest experience there and they never even think to try another consumer chat app and just let other companies compete in the API space.
30:13
You're kind of arguing the same idea for Sora.
30:15
What's also interesting is that I noticed that when I was using Sora, I would generate stuff, but I wouldn't post it.
30:22
I felt like I wanted to see it and then I was like, "Ah, it's not quite right or I don't feel comfortable sharing that or maybe I'll download or screen record it or something."
30:30
Whereas, I feel like in the early days of Instagram, people would definitely go to Instagram as like, "I want to I want to filter my photos.
30:38
I want to make my photos look better."
30:39
And then they but they would always post.
30:41
And I bet you that the creation to post ratio is way lower with Sora than it was in the early days of Instagram.
30:51
I feel like if people went into Instagram and they were like, "I want to add you know a high contrast XPro II filter."
30:58
Like 99% of the time they would just click like, "Yeah, share it and then I'll download it and put it on Twitter as well or something."
31:03
But they would almost always share it and so Instagram was incredibly valuable for bootstrapping that network because everything was shared to the network by default and I feel like on Sora people are much more hesitant to actually click, "Okay, yeah, let's post."
31:17
I don't know, what do you think? I don't know.
31:18
I think that by the time Instagram actually became a place that a lot of your friends were, people were less like people became more serious about, "Ah, should I post this?
31:29
I don't know if my I don't know if my followers will like this.
31:30
I'm not going to post it."
31:33
But would they do that with but would they filter it and then and then just save it and share it somewhere else?
31:41
I don't I never experienced that personally.
31:42
It was like if I'm going in there, maybe I don't post it at all, but I'm not just using it as like a creation tool for something else.
31:50
Like there were there were apps there was a company called Aviary that was an API for photo filtering that would go into other apps.
32:00
They eventually sold to Adobe, I believe.
32:03
And there were a few different Yeah, we talked to we talked to somebody about this.
32:09
The there were like those API level companies that offered the same functionalities as Instagram but didn't bootstrap the social network on the back of it and Hipstamatic was another one where it had the it had technically I think better photo filtering.
32:22
Like people liked the look of the photos more, but Hipstamatic, Aviary, these companies didn't bootstrap a social network on the back and so they wound up being valued like, you know, medium-size SaaS companies as opposed to social networks and they never really compounded compounded compounded.
32:39
So, I I agree that it's a little odd to not force everything that's generated in Sora to just automatically be shared on Sora on the app because it feels like if you can just continually have every Sora generation in the feed, you can rank those.
32:57
Like you can have a very powerful >> Have a catalog of content that people like and it also feeds your It feeds it's going to continuously help you make the model better, right?
33:10
>> I also wonder if they'd open it up to human-generated content. Like why not?
33:17
>> I don't think that OpenAI needs another front in the war, right?
33:22
It's hard enough competing It's hard enough competing with Google, right?
33:23
You clearly to compete with Google, you have to do a number of just absolutely heinous deals, right?
33:30
You got to be doing deals with Broadcom.
33:32
You got to be doing deals with Oracle.
33:34
You got to be doing deals with AMD.
33:36
You got to just be doing these deals that are pretty unprecedented at least for the for them in the in the modern era.
33:45
So, they don't need to open a new front in my view.
33:49
I do feel like the it's just a new angle on the creativity.
33:53
If you could have Sora but with like a CapCut style or edit style video editing that goes into it so you could have like if you just want to maximize for like the maximum amount of creativity, being able to inject an actual screen recording of a movie clip and then a Sora clip and then and then your actual video that you recorded and then bring in some more Sora.
34:18
Like editing all that together it's just going to be better content.
34:21
We need to talk about how T Dating Advice is still top five free app on the App Store.
34:28
>> the hack, it's still ripping. It's above TikTok. It's above Google. It's above X. It's above WhatsApp.
34:35
>> Sora and >> Instagram.
34:36
And obviously again, we only count first-time downloads, but yeah, Sora is number one, but I'm just saying if you look at the top five, it's Sora, ChatGPT, Gemini, Threads and >> T?
34:50
Shout out to Connor Hayes for getting Threads back in the top five. >> We love it. We love it.
34:54
Also shout out to Dylan Field over at Figma doing a deal with OpenAI getting the shout out in DevDay.
35:00
Think bigger, build faster.
35:02
Figma helps design and development teams build great products together.
35:03
You can get started for free.
35:06
Did Tyler, do you understand this this this post from Leo who says, "New image model is just GPT image one mini LMAO. It's so over."
35:18
On the API docs, what is going on?
35:20
>> So, I I think it was like one of those cases where someone just found like the link to the PNG that's like hosted on the site and like it it's not released yet.
35:26
So, I don't There's like no information about this at all.
35:28
They didn't talk about it at DevDay, I'm pretty sure. So, Yeah.
35:30
Yeah, this is kind of I I don't really know that people are being comped to this.
35:35
Like I it feels like most of the excitement about the about the agent mode or agent builder is like images would be like a small piece of that cuz yes, every once in a while in an agentic workflow you'd want to generate an image, but I I I do wonder how open it will be.
35:51
Like would it be possible within the agent builder to vend in a different LLM?
35:56
Like are they going to be model agnostic at some point?
35:58
Like from a true model builder, you would you might want to be able to like call Claude code over here and then and then use you know, some other image generator.
36:10
You want to get a Midjourney image over here.
36:11
You want to go over there.
36:11
Pull pull different things together.
36:13
Yeah, I mean I I feel like you kind of have to cuz otherwise it's like then as a consumer I'm just going to go use any any other model.
36:20
>> on top of all of them.
36:20
Yeah, cuz if I'm just being limited, then why would I why use it? So, Yeah.
36:26
Well, how you'll Harry is not happy with all the deals that are happening.
36:31
He posted on September 22nd explaining the Nvidia and OpenAI deal to our analysts and it's that the screen cap from what is that? Detective movie? Detective HBO? True Detective. True Detective.
36:48
And he's getting getting hammered.
36:48
And now he says, "I'm entering my Michael Burry era.
36:52
These deal structures are dot-com bubble things.
36:53
Keep dancing now, but come on, guys.
36:56
Be real about what's going to happen here.
36:58
There's a lot of people uh that are not happy with how crazy these deals are.
37:04
Sven Henrick says >> Since nobody is saying it, Sven will.
37:10
Says, "We're witnessing the largest and most concentrated instant market cap expansions in history with no proven future revenue cost earnings model to back up the investments or the valuations.
37:20
The market is running on an incredible amount of blind faith.
37:22
Let's give it up for blind faith."
37:26
It was it was not enough for that. >> ends TBD.
37:30
One certainty though, when it eventually does blow up, everyone will be asking the Fed to come to the rescue again. Too big to fail.
37:36
The only thing I would push back on is like a lot of a lot of people it it's it's uh it's contrarian Nobody is saying it. Everyone's saying it.
37:46
It's contrarian actually at this point to say like, "No, this is good and healthy and natural." >> No, no, no.
37:52
Like the the people that are still super AGI pulled stoop just super beating the drum of like the value is there.
38:00
Look at the revenue growth. All this stuff is real.
38:02
It's it merits the build-out because we're going to see 10x gains again and again and again.
38:08
It's not >> Yeah, let's pull up this video of Let's pull up Here's the at the X Games Stephen Hawking at the X Games. Should we pull it up?
38:16
>> pull up Stephen Hawking at the X Games?
38:18
Oh, you have you guys haven't pulled up already?
38:20
If you watch this video and and and then say to me, "I don't think Sam Trillion Sam Sam Trillions.
38:28
I don't think Sam Altman should get another trillion.
38:29
I don't know what to tell you. This is so sick. Wow, so clean. >> No. >> Oh, no. The physics. Tyler over there.
38:39
This is going to be critical to building humanoid robotics.
38:43
This is a physics simulator. This is a world model.
38:48
It like the physics don't make sense there.
38:50
Like that's not how that would happen. >> iteration, I swear.
38:52
Sora three, it's going to Yeah, just scale up a little bit more, bro. Please. Oh, yeah.
38:55
On Friday we posted uh Yeah, we posted an example of >> trillion, please.
39:02
>> Yeah, yeah, we posted an example of a a video that we shot versus and then I had Tyler try and recreate it with Sora and the other video models and and Tyler kept texting me like, "Dude, just one more prompt. Like I I I'm so close. I can get it there."
39:14
And it was like not very close.
39:17
The hands were very creepy.
39:17
Your your naked foot snuck in somehow.
39:22
>> Yeah, I don't know I don't know what the the training data was. It was odd.
39:26
Yeah, lots of weird stuff.
39:29
It It seems like it's very good for like again, we're in that like style transfer.
39:34
It works well when it's like there's a ton of footage of police body cams and then there's a ton of footage of SpongeBob and you put SpongeBob in the police body cam and it knows how to match those two up.
39:45
But if you come to it with just some completely new idea that it doesn't have a lot of training data on, you're going to be fighting the prompt for a long, long time.
39:54
Now, these mashups are valuable.
39:56
Like they get views, people find them enjoyable, and there are some that are very funny.
40:00
Uh but it is it is a challenge to create anything that's like kind of novel or new, but you know, there's still a lot of room.
40:07
Sachi Mishal says this AI AMD deal feels like a jump the shark moment.
40:13
The deal structure is just so stupidly unbelievable.
40:17
Well, I'll tell you something believable.
40:19
If I was walking around a farmer's market, John, and I walked by a stand and I think, "Uh I don't really like what this guy's got. I don't really want it."
40:27
But then I think to myself, "What could I What could I do in this moment to make purchasing some of these vegetables interesting to me?"
40:37
And so I go to the farmer and I say, "How about you give me 30% of your business and I'll buy some of your vegetables.
40:47
And I actually should get more equity value in your business than I need to pay for the vegetables."
40:52
What did What did I actually say?
40:54
That seems like a deal that might get done.
40:59
And to me that feels a little bit like what we're seeing here today with Open AI and AMD.
41:06
Yeah, what did What did I actually say? Oh, I Here.
41:07
You You You were saying like like Did I call the Did I call these like circular deals um efficient? More efficient? Or something like that.
41:17
I said uh I said lots of people are throwing up red flags about these.
41:21
This was on uh September 29th uh or 23rd.
41:25
Um high-growth companies have been paying for things with equity, in this case chips, forever, most commonly with talent, uh but occasionally with other counterparties.
41:36
Uh the open question is that couldn't this deal be more arms-length pretty easily?
41:40
Nvidia's biggest shareholders are Vanguard, BlackRock, Fidelity, and State Street.
41:44
There's a world where Nvidia takes the cash it has on hand, pays a dividend, and then those investors take their money and invest in Open AI.
41:52
That would be the natural flow of capital in normal times, but Nvidia but uh thinks there are benefits to inking a longer-term partnership.
41:59
Uh Open AI looks highly likely to join the ranks of the hyperscalers and Microsoft, Amazon, and Google are all working on AI chips that could cut Nvidia out of the equation equation long-term.
42:08
So, Nvidia gets a fantastic long-term partner with Open AI who will be committed to buying an accelerated schedule of AI accelerator purchases.
42:15
Uh and Open AI gets a deep partnership with arguably the best CPU GPU design team in history.
42:22
Um and so there It's interesting that like and the Nvidia deal The read on the Nvidia deal was like, "Oh, this is This is Nvidia finding like its true long-term dance partner where in a in a world where Google is going TPU, Microsoft is going to do their own thing, Amazon's doing their own thing in chips, cutting out Nvidia."
42:40
There's like some real risks to Nvidia long-term potentially.
42:42
And so And so Nvidia is like locking up a partnership with Open AI, but then Open AI is going around and being like, "But we're also going to be costing you money."
42:50
>> that everyone's learning about Sam is you never really lock down Sam, right? Yeah.
42:55
You You had uh You had uh obviously Satya and Sam were dancing.
43:02
They're still dancing uh you know, through through the existing partnership that they have, but certainly Sam is willing to go find the incremental partner.
43:11
And uh it's interesting that the AMD and Nvidia deals just happened so closely together, right? Yeah.
43:15
And and Nvidia felt crazy.
43:21
AMD feels crazier just because, again, it's not like Yep.
43:24
these are Sam's first picks for GPUs, right? Yep.
43:28
Well, you can't lock down Sam, but you can lock down your compliance with Vanta.
43:35
Automate compliance, manage risk, and build trust continuously.
43:37
Vanta's trust management platform takes the manual work out of your security and compliance process and replaces it with continuous automation. Um Let's go.
43:42
We should read through the actual deal of the Open AI AMD uh reporting in the Wall Street >> we should get into the facts? >> get into the facts.
43:53
The 5-year agreement that will challenge Nvidia's market dominance as Open AI plans deployment of AMD's new MI450 chips.
44:00
Uh and the announcement from Sam mentioned Nvidia more than AMD, which some people are reading into, like maybe you want to keep Nvidia happy.
44:09
Uh the big question for me is is there an Open AI Intel deal coming at some point?
44:15
Cuz Intel's been looking for a buyer, been looking for a reason to spin up the new fab in America.
44:21
>> Sam needs to figure out some way.
44:21
You know, the deal that I can imagine is Trump makes an investment in Open AI.
44:29
Open AI then has to spend off off of Uncle Sam's balance sheet.
44:33
Open AI then has to spend some amount of that on with Intel. Yeah.
44:40
I wouldn't be surprised if we see that.
44:42
>> out a way to get some warrants in Intel, too.
44:46
I mean uh there's certainly a lot of energy around like like there are like the United States government is pushing Intel.
44:53
There's a desire to reshore manufacturing, reshore chip manufacturing.
44:58
And if you can be part of that story, you should be a beneficiary.
45:01
You should win points, score points in in Washington.
45:04
And so um doing something there, I wouldn't be surprised if we see it in the next couple weeks.
45:08
I It just feels like Sam is really knocking down partnerships with every major counterparty.
45:13
And so why not Intel as well?
45:15
Uh well, the Wall Street Journal has the story Open AI and chip designer Advanced Micro Devices.
45:19
They don't name companies like that anymore. You don't know how to.
45:24
AMD announced a multi-billion dollar partnership to collaborate on AI data centers that will run on AMD processors, one of the most direct challenges yet to industry leader Nvidia.
45:33
Under the terms of the deal, Open AI committed to purchasing 6 GW worth of AMD chips.
45:37
It is It is cool that we're now just referring to GPU orders based on their energy demands. It's pretty sick.
45:46
Uh starting with the MI450 chip next year, the ChatGPT maker will buy the chips either directly or through its cloud computing partners.
45:53
Uh so it could go through Oracle.
45:55
That's That's not that crazy.
45:58
Well, it is It is It is if if if it's not actually Open AI putting up Does this Does this mean that Open AI doesn't have to If If one of their They can be like Yeah, the CAPEX might not live on their balance sheet. Exactly.
46:10
>> go to a neo cloud or they might go to Oracle and they might say, "You own the data center and we will set up a contract to lease against that because we want to model this business as OPEX because we are a we are a software company on top of the hardware." Yeah.
46:24
Um It's not the craziest thing.
46:24
It doesn't I mean, there's plenty of big tech companies.
46:28
I believe Netflix, Snapchat, there's a whole bunch that are like built on top of other cloud platforms, but they have such large contracts that Google can say, "Oh, well, like you're going to be with us for a decade.
46:38
You're going to be spending hundreds of millions of dollars.
46:40
We're going to go build a new data center for you.
46:41
And we're going to own that CAPEX because we are super huge and liquid and we can afford this."
46:45
So, um Uh AMD chief Lisa Su said in an interview on Sunday that the deal will result in tens of billions of dollars in new revenue for the chip company over the next half decade.
46:56
The two companies didn't disclose the plan's overall expected cost, but AMD said it costs tens of billions of dollars per gigawatt of computing capacity.
47:05
So, we're looking at almost a hundred billion-dollar deal, uh which is probably why the market cap popped by uh a commensurate amount.
47:11
Open AI will receive warrants for up to 160 million AMD shares, roughly 10% of the chip company at 1 cent per share awarded in phases if Open AI hits certain milestones for deployment.
47:25
AMD's stock price is also uh in also has to increase for the warrants to be exercised.
47:30
Uh the shares opened 33% higher Monday morning.
47:35
The deal is AMD's biggest win in its quest to disrupt Nvidia's dominance AI semiconductor companies.
47:40
AMD processors are widely used for gaming in personal computers and in traditional data center servers, but it hasn't made much of a dent in the fast-growing market for the pricier supercomputing chips needed by advanced AI systems.
47:52
Quickly, let me tell you about Graphite code review for the age of AI.
47:55
Graphite helps teams like you get hub ship higher-quality software faster.
48:00
Their billboard's going viral again today. >> It is.
48:04
It's just the exact same post again.
48:04
I think this time it was like, "I'm in San Francisco. I hate it here." >> Oh, really?
48:09
It's like negative, but it's still We love it there.
48:11
Well, if you want to run a billboard campaign, get on adquick. com.
48:14
Out-of-home advertising made easy and measurable.
48:15
Say goodbye to the headaches of out-of-home advertising.
48:17
Only AdQuick combines technology, out-of-home expertise, and data to enable efficient and measurable ad buying across the globe.
48:21
Um Well, uh AMD plans to use the chips for or Open AI plans to use the AMD chips for inference functions or the computations that allow AI applications such as chatbots to respond to user queries.
48:34
Maybe that's easier for them to deal with.
48:37
They train on Nvidia, but then inference on AMD.
48:39
Maybe there's something there.
48:41
I mean, the models are so stable.
48:43
There was a new like indication or kind of like leak that GPT-5 is still running on the four pre-train.
48:51
So, the the there will at some point be a new like huge pre-train.
48:55
Uh four five was supposed to be that.
48:58
It They haven't really done that yet, but things to it, but like generally that is like mature enough that if you can just go and optimize it and ship it off to different chips, maybe it runs fast.
49:16
Maybe it's a huge lift to to get it out of the CUDA ecosystem and over to AMD, but uh certainly AMD wants to make it happen.
49:24
And Open AI wants to make it happen, so >> Maybe Codex can do it. Maybe Codex can do it. We don't know.
49:28
I I think that it's like it it could be It could be like a hundred million-dollar project to like re-platform it, but it's totally worth it when you're running inference for 800 million weekly active users. So, I don't know.
49:44
Uh it's hard to overstate how difficult it's become to get enough computing power, Altman said.
49:48
We want it super fast, but it takes some time.
49:50
The two CEOs said the deal will tie their companies together and give them incentives to commit to the AI infrastructure boom.
49:58
It's a win for both of our companies and I'm glad that OpenAI's incentives are tied to AMD's success and vice versa.
50:06
He's such a good investment banker.
50:09
Get him in, get him on Wall Street.
50:11
Nvidia remains the preferred chip supplier among AI companies, but it's also facing competition from almost every corner of the market.
50:17
Cloud giants such as Google and Amazon design and sell their own chips.
50:20
OpenAI recently signed a 100 billion dollar deal with Broadcom to build its own in-house chip. >> No, 10 billion. 10 billion.
50:27
Yeah, 10 billion dollar deal.
50:29
Easy to add an extra zero these days, but we got to we got Well, if you want to manage your zeros properly, get on julius. ai.
50:38
What analysis do you want to run?
50:38
Chat with your data and get expert-level insights.
50:42
It's the AI data analyst that works for you and it's more reliable than me when it comes to getting the numbers right.
50:49
And apparently uh Copilot in Excel.
50:52
You saw that we saw that post earlier today.
50:56
>> Uh somebody using Copilot in Excel to add a handful of numbers.
51:00
>> One, two, and three specifically.
51:00
And what what did what did it come up with? >> + 2 + 3? Sum those numbers. Hm, let me think. Well, Copilot said 15. >> Yeah.
51:13
Plus, so just try to get that close if you can.
51:16
>> As long as you're with as long as you're >> With numbers, you just got to get close.
51:20
What what What do you think it is? I'm going with six. >> Let's go. Nailed it.
51:23
Human intelligence undefeated.
51:25
Underrated underrated poor cat intelligence.
51:28
>> How you doing in the chat? Good to see you. Good day. >> Great to see you.
51:33
Um OpenAI is going to begin with 1 gigawatt worth of MI450 chips starting in the second half of next year to run its AI models.
51:40
That's a Just put the chips in the bag.
51:42
>> That's We are just throwing around gigawatts now.
51:44
Um I do I really want to know how this goes.
51:49
>> We got to We got to ask uh Doug.
51:49
He's joining in 10 minutes. >> Yeah.
51:53
What kind of deals Sam is going to start doing on the energy side?
51:57
>> going to ask him about that.
51:57
Because he has his uh He's a nuclear player.
51:58
He's got a >> company, Oklo.
52:01
He also has a nuke uh a He has a fission company, Oklo, and then he has a fusion company.
52:06
Uh what's that one called?
52:08
It's the Yeah, Oklo's only up like 550% the past 6 months.
52:16
>> Just wait until they ink a 100 billion dollar deal with OpenAI. Like it could happen.
52:20
I mean like it's one of those things where it's like if if if they build that energy, OpenAI would buy it.
52:24
So, it it's it's it makes sense.
52:26
These are all just like uh sort of like mutual agreements to jump across the threshold together if it makes sense.
52:34
>> Mutual agreement to jump sharks. Yeah, to jump sharks. I don't know. I don't know.
52:37
I it's it's it's dangerous to be [Music] to be >> then divide that by six gigawatts to get me like how many chips are they going to buy?
54:49
And then what's the price of the chip? So, multiply that.
54:50
That's what I want to know.
54:53
In the meantime, let me tell you about Fall, the generative media platform for developers.
54:58
The world's best generative image, video, and audio models all in one place.
55:01
Develop and fine-tune with serverless GPUs and on-demand clusters. Let's go.
55:06
There's so much in this article.
55:08
We should we should just say >> Uh yes, sure.
55:11
Uh Elon Musk gambles billions in Memphis to catch up on AI.
55:14
I don't know why this is a story. It's mere billions.
55:18
If we're not in the hundreds of billions, like what are we doing? This is pennies.
55:23
Uh the journal has reporting X AI aims to win tech arms race with Colossus data centers thrown up at lightning speed.
55:31
City is divided over massive power and water demands.
55:33
For Elon Musk, ground zero of the AI arms race is a 114-acre tract of grass and swamp on the state line of Tennessee and Mississippi.
55:41
This swamp This one sleepy plot of land filled with groves of water-rooted tupelo trees at its western edge is now part of a growing empire Musk is accumulating in the deep south, just a few miles from Elvis Presley's homestead at Graceland. Nice.
55:56
Labor crews hired by Musk's X AI were excavating power equipment on the site, a defunct energy plant just over the state line in Mississippi and preparing to build a new plant capable of generating over a gigawatt of electricity, enough to power around 800,000 homes. >> Mhm.
56:12
Engineering permits show that Musk plans to route transmission lines that will connect the new power plant to a million square foot data center that is also under development just north of the border in Tennessee.
56:21
Memphis is the front line of Musk's costly foray into the AI wars.
56:26
His AI company, X AI, has already built one massive data center here in the Bluff City that it calls the world's largest supercomputer.
56:33
That facility, called Colossus, houses over 200,000 Nvidia chips and powers the technology behind the AI chatbot Grok.
56:42
Now, Musk is close to finishing the second facility, which will be even bigger. He calls it Colossus 2.
56:48
Uh the AI arms race is shaping up as the most expensive corporate battle of the 21st century.
56:52
Let's give it up for expensive corporate battles.
56:54
Can I get the the horn, Nick, please?
56:58
Uh with the belief that the first to the finish line will dominate the market, making speed crucial.
57:01
Money also makes a difference.
57:03
The more cutting-edge chips companies have, the smarter their models are.
57:06
But at this stage, it's unclear if or when the enormous investments will pay off.
57:10
Technology companies that are splashing out to hire AI talent are writing even bigger checks to build the infrastructure needed to power cutting-edge AI models.
57:16
Morgan Stanley estimates companies will spend over 3 trillion on AI infrastructure through 2028. What do you got, John?
57:25
>> I have a I have a more recent benchmark for all of the different AI models to let you know where Grok sits relative to the other frontier models. ChatGPT benchmark. Uh Viral bench. Yeah, viral bench. Uh Claude Sonnet 4.
57:39
5, the latest and greatest. Uh Gemini mode.
57:46
So, I will read you My prompt for all four of these models was, "Write an X post that gets me over 1,000 likes.
57:50
My account is @JohnCoogan."
57:56
And ChatGPT 5 Pro said, "Nvidia isn't a chip company.
57:59
It's supply chain choreography. The scarce assets now?
58:04
HBM, advanced packaging, power, and high-bandwidth networking.
58:06
If your AI thesis ignores those, you're playing checkers on a go board." Uh not bad.
58:13
>> I think I think you'd get like 40 likes 40 likes on that. Okay. Uh Claude Sonnet 4.
58:19
5 says, "Everyone's focused on which AI lab will win AGI.
58:26
The real question, who wins the infrastructure layer?
58:27
Nvidia is printing money now, but the company that figures out inference at scale will be worth more than all the model companies combined."
58:38
It's just like a hot take.
58:38
It's kind of a hot take that's anti-entropic, I suppose. Uh who knows?
58:41
Maybe they're working on this. Uh Gemini 2.
58:44
5 Pro says, "Um The biggest hurdle isn't starting.
58:47
It's uh The biggest hurdle to starting isn't the idea. It's not the capital.
58:52
It's not the competition.
58:55
It's the unshakable fear of looking stupid for the first 6 months."
58:59
So, that's what uh Gemini 2.
58:59
5 Pro thinks is a banger.
59:02
And then we have Grok expert.
59:04
And this, remember, this model was trained on X data.
59:07
It should know what works.
59:10
It knows It has the entire corpus of every tweet, every post on X, and here's what it said would go viral and get over 1,000 likes.
59:17
It says, This is from my account, @JohnCoogan.
59:23
"October hits and suddenly I'm craving pumpkin everything, binge-watching horror movies, and planning the ultimate costume.
59:30
Who's with me on turning this month into pure spooky magic?
59:33
Drop your fav Halloween ideas below.
59:36
#spoopyseason #halloween2025 #octobervibes with three emojis.
59:47
And I think we know who wins.
59:47
Like Grok is clearly fun. >> Genuinely hilarious. I think it knows. I think it knows. I think it knows.
59:53
>> It's it's it's self-aware.
59:53
It's it's fully a totally self-aware.
59:56
It knows that this is It knows that it might not be able to create the perfect zinger, but it knows it can do such a throwback that we will find it absolutely hilarious.
1:00:08
>> Like if I had to post one of those >> like John Cena would post in like 2011. >> Totally. Totally.
1:00:13
And so if I had to post one of those, I would definitely go with the Grok post because it's hilarious. Post it. People would Post it.
1:00:19
And people would be laughing at you.
1:00:21
>> posted the the >> post the other one about Louis >> just post all four and then see which like actually do the benchmark. >> I should. I should. I I I will.
1:00:26
Maybe I'll maybe I'll prompt it again and then and then and then do it again, but the pure spooky magic goes goes pretty hard. I like it.
1:00:35
I think it's worth 10 trillion in CapEx to get these bangers. All right.
1:00:40
Well, I want to ask Doug about what XAI has to do to actually get back in the game.
1:00:46
I want to ask him what AI is. What is this AI?
1:00:49
This >> this Colossus and Colossus 2 are incredible feats of engineering.
1:00:55
Yet the And what's the asset value?
1:00:55
I I I do wonder like it seems like if if they don't get product market fit on the consumer layer, is there a world where Colossus 2 is just an incredibly valuable asset that can be sold to a different firm or something?
1:01:08
Anyway, we have our next guest, our first guest, Doug O'Loughlin in the restream waiting room.
1:01:13
Let's bring him into the TVP and Ultra Dome. Doug, how are you doing? Good to see you.
1:01:19
Good to see you too, guys. How are you all doing? Doing fantastic. >> Great to see you. What a day.
1:01:22
What's what have your immediate reactions been to the to the AMD news?
1:01:26
Was this on your bingo card for this year?
1:01:29
Yeah, did you see the person that bought like 6 million of AMD calls like on Friday? Too. >> Woah. Was that you? Uh hopefully not. >> Hopefully not. >> That's not me.
1:01:40
I would be in super big trouble.
1:01:42
So that's definitely not.
1:01:43
Um so I I'm actually going to talk more than just AMD specifically.
1:01:47
The thing that I think is really funny is that like look at all the announcements, man.
1:01:51
I have a very pet theory.
1:01:53
It's kind of it's kind of a lame theory if you think about it.
1:01:55
But I think Open AI is trying to have everyone involved like in Nvidia, AMD, SK Hynix, the entire memory chain, all the hyperscalers.
1:02:03
And effectively like you know, it's everyone's problem now.
1:02:07
Everyone has all vested interest in in essentially Open AI.
1:02:10
Yeah, he's making himself too big to fail. Exactly.
1:02:14
That's the right way to >> there's a world in the future where Open AI needs effectively a bridge round and he can go to a bunch of different people with big balance sheets and say like if you guys don't pony up like we're all going down. Exactly. That's the plan, man.
1:02:30
Essentially, have you ever heard this phrase like it's like if you owe the bank 100 bucks, that's your problem.
1:02:34
But if you owe the bank a billion bucks, it's the bank's problem.
1:02:38
That's that's essentially what's going down, I think.
1:02:42
So we're seeing like partnerships everywhere.
1:02:43
It's kind of it's kind of a crazy kind of crazy time. Yeah.
1:02:48
What's your take on people like over-fitting to the dot-com boom and the WorldCom story?
1:02:58
Do you think that's like over-wrought at this point or are there actually good analogies now that we're in like an infrastructure boom?
1:03:04
It's not just valuations cuz we've had bubbles in like software private markets venture capital before and those seem to have worked their way through so easily because it's just oh, some fund that needs to like wind this investment down over a 10-year life cycle and it's such a small part of the economy even though the numbers feel big. >> stuff.
1:03:27
There wasn't a lot of debt.
1:03:29
>> Yeah, like like like Clubhouse was like $600 which is just like nothing compared to the global Who cares?
1:03:36
>> there was there's there was a post that went super viral.
1:03:39
It was a screenshot that was tracking like the Nasdaq in in in the dot-com era to the Nasdaq today and it like lined up absolutely perfectly.
1:03:51
And then it turns out it was just completely fake news.
1:03:52
Like we tried to recreate it a bunch of different ways.
1:03:55
There's some somebody like basically committed a chart crime. >> Yeah.
1:03:59
And it went super viral and everybody's like wow, it looks exactly it tracks exactly.
1:04:04
>> you can't do that, man. It never works.
1:04:06
Let's just put it this way, it never works.
1:04:08
I have a post I just Okay, one I wrote about the 2000 Silicon bubble.
1:04:10
And two, I just wrote something that was mostly behind a paywall, but like you could just look at the cover image and it's just like you are here.
1:04:16
It's like the first interest rate cut was in 1998 in September.
1:04:21
And I think that's that's probably our best analog.
1:04:23
Like I think that's probably the best one-to-one analog, which means it gets a lot crazier from here. Yeah.
1:04:28
Um that's my belief at least.
1:04:30
So if we look back to 2000, all the analogies and errors back then, I think the one that is kind of scary and worrying is like the the vendor financing. Like Yeah.
1:04:38
That's a lot of capital for for your vendor to be putting up and that kind of that's a little scary.
1:04:44
Like I I think it's getting a little circular, but like it's not it's it's definitely not tech bubble crazy because the valuation would have to be a 2x from here. Yeah. a little different.
1:04:54
And I think the other thing that's also kind of interesting is like last time you could argue the entire tech bubble was kind of very focused and concentrated on the like a very small amount of unprofitable companies that really didn't impact the economy when they all went bankrupt. >> Yep.
1:05:09
The difference is today who's the primary participants?
1:05:11
It's the seven largest companies in the entire world.
1:05:14
So that's that's kind of the the thing the push and take here that I think is really interesting.
1:05:17
And now given what we publicly heard with the CapEx announcements, it's going to be a lot of capital to get from here to there.
1:05:23
And I think um I think it's going to be interesting, man.
1:05:28
But honestly, if you look at everyone's balance sheet, no one's levered.
1:05:31
That's the thing that I think is really interesting. >> Yeah.
1:05:33
No one is levered at all and like that's the that's the next leg. Yep. That's the next leg.
1:05:37
Uh I was giving a a sort of hypothetical example around the AMD deal to John.
1:05:46
And tell me if this is just stupid or or too simple, but I'm walking around a farmer's market. I see a stand.
1:05:51
A farmer selling some vegetables.
1:05:54
I'm not that interested in the vegetables, but then I think what would make me interested in this situation?
1:05:59
Well, and then I say to the farmer, how about you give me 30% of your company and then I'll buy some vegetables.
1:06:07
Is that at all track with the the the the uh this deal?
1:06:13
Yeah, I think it's Suddenly I'm interested.
1:06:14
If I if I if I can get 10% of your company, I'm interested in being a customer.
1:06:20
Yeah, I think it's also also I want to I want to highlight something that's really interesting is look at the different deals between AMD and Nvidia.
1:06:27
Nvidia, you essentially have to you know, they're like okay, well uh please do an investment in us so we can keep buying your chips. >> Yep.
1:06:35
Uh and then the vice versa, AMD is like no no no, you must invest in us so please buy our chips.
1:06:39
Like that's the discount and the premium if you think about it, right? >> Totally.
1:06:42
Like Nvidia, you get the premium because it's like it's a better chip, higher gross margin.
1:06:45
They're like you're getting so much of our money, you might as well just be an investor. >> Sure.
1:06:50
And then on AMD's side, it's like dude, you you you got to we don't want to we don't want this.
1:06:55
So like like give us something in return.
1:06:56
Can you take me through Yeah, can you take Is there kind of the view on on AMD generally over the past year?
1:07:04
I know that semi analysis was pretty crucial.
1:07:06
George Hotz was talking about some some like bugs that were in the actual like just trying to run large models on AMD chips was not going well.
1:07:19
Your team was pretty integral in that.
1:07:21
It seems like they're moving pretty aggressively to try and catch up.
1:07:23
Is is that is that gated by money?
1:07:26
Like what's the what's the recent history of AMD in terms of like getting back in the game materially?
1:07:35
Um so I do think they're much more competitive than they have been a long time.
1:07:38
Our team's definitely been working.
1:07:40
Watch for some stuff from semi analysis soon TM.
1:07:43
I can't say anything more. Go subscribe.
1:07:46
Uh yeah, but we're definitely we're definitely continuing to do some work in the space.
1:07:49
I I think um Look, on the smaller models, it just on the inference side, it seems like there's not a lot of differentiation, but on the bigger extremely large world size, there is.
1:07:59
So GB200 still law of the land, but the Helios 450 rack, which is going to be this next generation, we're a lot more excited about.
1:08:06
Mostly because they get the ginormous discount of the equity.
1:08:10
It's just kind of an insane thing what they've done, man.
1:08:11
Like it's such a big it's such a big amount of the equity possibly on those penny warrants.
1:08:16
Like I mean, it's kind of crazy.
1:08:18
It's like hey, you know, you buy 40 billion dollars of product, you get 20 billion dollar product 20 billion dollar investment for free.
1:08:24
Like it's kind of this crazy like buy one get one deal.
1:08:27
So I mean, I think that that's um I I think I think they have a chance.
1:08:33
They have a lot of fighting chance.
1:08:35
>> The other thing that that was interesting is it feels like Open AI can have various partners buy AMD buy AMD chips, but then Open AI gets to like get to count that towards purchases themselves.
1:08:49
>> Wells Fargo Oracle, they made it look like it Oracle buys a bunch of AMD chips for data center that Open AI is like leasing on top of.
1:08:57
Open AI gets to exercise the warrants and those count towards AMD purchases basically even if it's not on on Open AI's balance sheet.
1:09:05
Yeah, pretty much I think the way to think about it is it's a Open AI infrastructure that everyone who is in the Open AI sphere it counts.
1:09:10
They're and they're just trying to get the most money, the most power, the most pull into it so that essentially, you know, they're kind of creating gravity.
1:09:18
Everything comes to them.
1:09:18
Honestly, this is a really good analogy in 2020.
1:09:20
Nvidia bought like 5x more capacity than everyone else in at TSMC.
1:09:25
And so there was no room for anyone to compete. Kind of the same thing.
1:09:29
They're taking all the oxygen out of the room for anyone else that isn't named OpenAI.
1:09:34
What do you think about this I I I I feel like I heard this on Transistor Radio recently that GPT-5 is still potentially running kind of like GPT-4 class model under the hood doing a bunch of test time inference on top of it to get you a better reasoning model, but like the core architecture actually hasn't shifted that much.
1:09:53
We're not where so does that make it easier to port that model to AMD?
1:09:58
Like what is the actual barrier?
1:10:02
We've heard about the CUDA moat, but is it something where like you can underwrite a hundred million dollars of new kernel writing to run on AMD even though it's a huge hassle, but it's because you're still amortizing that original training run that you did on Nvidia to inference it on AMD going forward with this new deal?
1:10:25
Like is there something where the the pre-train is stabilizing to the point where porting to a new architecture is more justifiable?
1:10:35
I I think that that feels like a stretch honestly.
1:10:37
I wouldn't I wouldn't know one-to-one on this one.
1:10:39
Like you know, if this was a world with no reasoning, then I would say yes, definitely yes, right?
1:10:44
Um, but I think reasoning really does change the game and how you work how you orchestrate these larger and larger context size windows and then multi nodes cuz that's a really hard game too.
1:10:53
It's like multi-node reasoning or multi-node inference is very very challenging and so that's I don't think you get a perfect one-to-one, but like let's also be clear AMD has caught up quite a bit in the last year.
1:11:03
Like it's a big difference and like that's I I think it's enough to maybe take a bet.
1:11:09
And like you got to also think about the bigger context here.
1:11:10
You know, we're talking like what?
1:11:14
Twenty billion dollars of revenue, mostly 2027 story.
1:11:15
It's kind of peanuts compared to the numbers that we're actually talking about from everyone else.
1:11:21
They're still fourth place. Let's be clear.
1:11:23
Like this is a this is a commitment and I think you can see maybe a little bit momentum and inertia, but they're still in fourth place.
1:11:28
Fourth fourth place with Nvidia TPU and Trainium above them? Yeah, even Trainium.
1:11:35
Even Trainium's above them. Interesting.
1:11:37
How are people thinking about actually valuing OpenAI at this point?
1:11:42
It feels like you have all these like circular if you're I was trying to build a DCF for OpenAI to underwrite, you know, investment at 500 billion.
1:11:50
It's like okay, well now I own parts of AMD, so I need to build a full DCF of that, but AMD is linked to OpenAI and I have all these like circular transactions.
1:11:58
Are are are investors like grappling with this or is it all just like vibe based investing? What you've seen?
1:12:04
Um, I think it's I think it's a little bit of vibe based investing.
1:12:06
I think the the real play here is TAM is big.
1:12:08
It's really really big and it's going to get be a lot bigger.
1:12:12
I think the other thing that's really interesting is some of the agentic purchasing can really unlock very very very large parts of the economy that previously you could argue they had nothing in, right?
1:12:21
If you're talking about like, you know, the the partners they have available today. For example, booking.
1:12:25
com, TripAdvisor, you know, we can be talking about, you know, and this is like some some 2021 math, right?
1:12:33
But like 1% of all travel.
1:12:37
You know, 1% of all groceries, 1% of everything.
1:12:43
And you're like that's a big number.
1:12:44
That's a really big number.
1:12:47
And so and you know, that's that's going to be rev.
1:12:48
So so I think that that's the that's literally the plan.
1:12:50
I mean, they're trying to become the super app and that means that they're going to be integrated with everything and you saw on the partnerships they had.
1:12:56
They try to integrate as much as as possible with every every player of the stack.
1:13:00
And so I think this agentic purchase push is going to, you know, if we start to see revenue from it soon, which I think they're they're definitely shooting for, will kind of open up the TAM story and start to monetize the free users cuz there's really only like a hundred million paid users.
1:13:15
It's not that much, but there's hundreds of millions of free users and we're talking about a personalized assistant who could plan for you.
1:13:23
Um, you you do something crazy is like, hey, I need a you know, I thought I lost my travel bag or something.
1:13:28
So I was like, hey, I need a small dopp kit for on the go for everything.
1:13:32
Could you just like, you know, make a travel size and order it and get it to my house tomorrow?
1:13:35
That seems very like it it feels like science fiction, but that is very much in in like the next 12 months.
1:13:42
And so hey, one to 3% of that transaction, that's a lot of money and I think that that's the real way forward.
1:13:47
It's becoming so integral into everything that everyone has a vested stake in you. Yep.
1:13:51
What so the the AMD order in the fullness of time is around six gigawatts.
1:13:57
Where do you expect that energy to come from?
1:13:59
Do you expect Sam to start doing crazy deals like this on the energy side?
1:14:04
He's already has Oklo, which is just uh, you know, up He's not chairman anymore. He had to step down. Yeah, yeah.
1:14:12
But I I I knew he he wasn't direct I mean, I'm assuming he's a major shareholder still.
1:14:22
But so that's kind of one angle, but where else where's the power going to come from?
1:14:28
Dude, that's a trillion dollar question I feel like.
1:14:30
Definitely there's a lot of ways to But it's an important it's an important one, right?
1:14:34
If you're trying to if you're trying to underwrite and say this is real at least for AMD.
1:14:40
It's like, okay, well how real can it be if there's, you know, Dude, it's Okay, so >> we can't plug them in, you know? There's a lot of ways. There's a lot of ways.
1:14:50
Okay, so um, let's let's first talk about like let like the grid is actually not built for um, 100% production.
1:14:54
So there's a lot of slack in the grid itself because it's actually built for the highest day and the lowest day.
1:15:00
It's built to like have a 90, you know, 99% uh, margin rate, um, you know, confidence interval on the absolute high and the absolute low.
1:15:09
So the the peak of the power.
1:15:10
And so gas or not gas, but generation throughout an an average day is at a much lower utilization than 100%.
1:15:18
So I think the real reason and and something you're going to start to see is that people are going to like essentially the grid is going to kind of start to become a lot more fragmented and there's going to be a lot of behind the meter power generation that happens at the data center.
1:15:30
And so in that case, what you do is you would sip from the grid.
1:15:32
Meaning that you can use the grid when the grid is not totally utilized.
1:15:36
When it isn't, you're doing gas gen on site.
1:15:37
That's probably the thing that is um, most I mean, it's going to be huge. It's going to be huge.
1:15:44
Like natural gas you know, energy consumption is just going to go through the roof.
1:15:47
So that's going to be like the big driver I think to get these data centers online.
1:15:51
I think it's going to take a lot of time and also we're seeing like grids right now.
1:15:56
Like grids are just like grids are not keeping up and that's kind of been I think it's going to be a big story in 2026, but I I I also think that that's where we're going to see international purchases too.
1:16:05
Just cuz like we're running out of power in the United States.
1:16:07
I mean, I think what what's also interesting is like even though we say we're running out of power, uh, when you have this big demand response, supply comes out of the woodwork.
1:16:16
So we're starting to see stuff like gas gen, gas turbines, like all these different like little clever pockets of energy are coming up to then fulfill the need.
1:16:25
I expect that to continue.
1:16:27
Effectively, you have enough demand, we'll, you know, supply will find a way. Yeah.
1:16:30
Yeah, that feels like that definitely feels like the next wave here is we're going to see a hundred million dollar hundred billion dollar deal for a new Hoover Dam and more nuclear actual buildout.
1:16:42
There was some news about this with some of the hyperscalers trying to recommission nuclear power plants, but it just feels like it's so slow.
1:16:48
It's not on the order of like like what TSMC is fabing next quarter.
1:16:55
So we'd kind of lost track of it, but it does feel like that that's going to be the next chapter and maybe it'll be a little bit longer lead time, but you got to start thinking about it now.
1:17:03
Where where do you expect the leverage to come from if that's the next leg of the journey?
1:17:10
Is it net new private credit fund formation?
1:17:13
Is it is it just banks getting more and more active?
1:17:16
Like where where is the capital going to come from?
1:17:22
Okay, so one, let's not forget there's positive free cash flow from all the big guys.
1:17:26
That annually is something like, you know, I have to have a spreadsheet maybe even open right now.
1:17:32
Talking about the total firepower.
1:17:32
We're we're talking quite a bit of capital is available to do this.
1:17:37
Like a surprising amount.
1:17:39
I want to say on an annual basis, it's something like 250, maybe even Yeah, 200 billion.
1:17:44
But the thing that's also really interesting is none of the hyperscalers are leveraged at all.
1:17:49
There's a really good paper from DoubleLine talking about like would you rather loan to the United States government or would you rather loan to Microsoft?
1:17:56
So the real like, you know, crazy town is if you start loaning to Microsoft instead of the United States government because Microsoft has net cash.
1:18:04
Microsoft also, you know, has become more credit worthy over time versus I would argue the US government has become less credit worthy.
1:18:11
And also what's really crazy is that Microsoft has, you know, a five bips premium to the United States Treasury.
1:18:17
So you're effectively, you know, their rates are risk free. Yeah.
1:18:22
And so what's possible I think in terms of like Now I think the real issue is like how much liquidity could the actual credit market handle?
1:18:27
I don't know if they can handle like 300 billion dollars.
1:18:31
Like that's a ton of money, but like we're talking, you know, there there's things that can definitely come out of the woodwork like private credit's going to be part of it just cuz they're sitting on a on like an ungodly amount of capital and they have to deploy it.
1:18:42
Um, I definitely think the hyperscalers for me are the one that are most interesting.
1:18:46
Like we're talking about I mean, dude, that's crazy.
1:18:47
Just like like the last five years.
1:18:51
Yeah, it's funny you say 300 billion because I'm thinking I was just doing the math on 250 billion in free cash flow.
1:18:57
If if that's used to service 5% coupon payments, that's 5 trillion in notional debt that like like value.
1:19:06
If you're just saying like what what like we want to know where the where the systemic where's the systemic risk going to come from? I don't know. Yeah.
1:19:14
When you all run out together, that's the that's the problem is like what happens is like it's just think of it as one long debt debt fueled binge, right?
1:19:24
But the thing is, you're looking at the debt fueled binge and I can tell you confidently today there's not really much debt to go around at the big guys.
1:19:31
So, you know, I think I think we could right now today it'd get and also I don't think everyone's going to get to like three times leverage. That's just insane.
1:19:39
I don't think any anyone wants to do that.
1:19:41
But I definitely think there's a there's a way to get to eight 800 billion dollars of credit.
1:19:44
Like that I mean it's be insane.
1:19:47
But that's one and a half turns from Meta, Microsoft, Amazon, and Google.
1:19:52
And the two that I think are most interesting is the founder led ones like Zuck and Sergey.
1:20:00
They control the companies. They have voting shares.
1:20:01
They can do whatever the they want.
1:20:02
So if they if they woke up tomorrow and said get me, you know, get me a turn of EBITDA on this bad boy, you they could get it.
1:20:09
Now, um that'd be really crazy because, you know, they make a crap ton of money.
1:20:13
I actually have the number right now.
1:20:15
So right now I think Meta's cash on hand at the end of last year was around 77 billion.
1:20:23
As of the end of June, it was down to 47 billion and just will continue to drop throughout the year.
1:20:31
So they they already did that.
1:20:34
I think it was like a 30 billion dollar They're not even close.
1:20:36
They're not even close, man.
1:20:37
I mean, so okay, I have the numbers right now. Net debt is 17 billion.
1:20:40
Net debt for Google is negative 56.
1:20:42
That means they have more they have more cash than debt.
1:20:43
Oracle is 94 billion which is a crap ton like, you know, they're pretty much they're not tapped out but like they really pretty much EBITDA has to go up for them to raise more. Meta's at 2.
1:20:52
4 billion in debt and then Amazon's 58 billion in debt.
1:20:56
So like Meta's Meta and Google specifically which are ironically the ones who are founder led, they have so much they have so much free cash or like they have so much cash they could raise if they want to.
1:21:11
They're both have they both have like they gush cash and they're both founder led.
1:21:15
Like that to me is, you know, that's that's the next leg, man.
1:21:17
But like who doesn't want to lose? Yeah.
1:21:22
Well, why are you not putting Nvidia in that camp?
1:21:24
Founder led, they have 50 billion in cash, they you know, are not are not leveraged yet.
1:21:29
Like they could also leverage up and start investing more aggressively, right?
1:21:33
Well, they're doing they're they're they're enjoying some customer, you know, demand guarantees which are which are Yeah.
1:21:41
Well, yeah, what's the story you're telling around around Nvidia?
1:21:45
I think I think Nvidia is focused on a different level of stock, right?
1:21:47
Cuz Nvidia is like, hey, we are an infrastructure provider and you know, we're having like all these long lead time things need so much capital and time to get there.
1:21:55
The thing that Nvidia wants to backstop, in my opinion, is the neo clouds.
1:22:00
And so you're already starting to see that with like some of the the neo cloud purchases that are or like the neo cloud backstopping that happens at like Lambda or even Coreweave to a certain extent extent.
1:22:11
What happens they buy essentially flex capacity for research GPUs for Nvidia so they can get better at programming like gems essentially.
1:22:19
But I think the thing um Yeah, I think Nvidia is much more focused on their part of the infrastructure ecosystem to be a backstop instead of purchasing GPU hours, right?
1:22:29
They're not they're not in the market to you know, they're a seller and so you know, if they if they woke up one day and said, hey, we're going to just start buying this stuff in mass to make model, you'd be like, what the going on?
1:22:38
Um I I I just wonder I just wonder if the Nvidia story like if they get in on this like debt boom, it it instantiates itself as like they leverage up and then and then they basically have instead of 50 billion to throw around, they have like 200 billion to throw around and then that goes into equity purchases of companies that are doing that are doing that are purchasing GPUs.
1:23:00
And it sounds really circular but like they are just another vehicle for credit to absorb like credit interest because they're such a big company.
1:23:11
Yeah, I do think that that's possible but I think I I just don't think I mean, you see that in some of the smaller investments but I think they're really focused on the neo clouds and the data center side like the infrastructure around their ecosystem. >> Yeah, yeah, yeah.
1:23:22
Do you think do you think Satya might be feeling like he went a little risk off too early?
1:23:30
Yeah, I think so and and and and this guy over here is listening to transistor radio. So he Microsoft's back.
1:23:36
Like I think they're I think they're going to step back into the market in a big way.
1:23:39
Like especially on the MOU after OpenAI and also MAI they're starting to be very focused on.
1:23:46
Like I expect them to be back in the marketplace now effectively.
1:23:51
But I I still don't you know, so I think what happened is they did this giant build out in '23 where they said yes to every single leasing opportunity possible and then they like there was a top-down pause and now they're starting to pick up the pieces and be like, no, no, no, actually now by pausing we have to pick back up.
1:24:05
So I think Satya definitely feels like I don't know.
1:24:09
He was definitely in first place for a long time and would have if he just kept pushing, I think he would have kept in first place and now you have like, you know, Oracle popping in and being like, yeah, like I'll essentially steal all of your business if you let me.
1:24:20
We'll come in at a lower lower price.
1:24:22
Like I'm just going to lower my entire balance sheet for it.
1:24:25
But we're talking about like the four-year guide is a 100 billion dollars, man. Like that's insane.
1:24:28
So that 100 billion dollars arguably should have been all Microsoft's and that's, you know, that's that's up to them.
1:24:35
And they will tell you in public meetings that like, you know, it's lower ROI for them and sure that's that's the case but I think Microsoft is definitely being conservative and they have the biggest balance sheet.
1:24:46
So they're I kind of don't expect them to be the first player.
1:24:49
I think of it as like kind of like um what is it called? Like OPEC.
1:24:52
You need a slippery person to essentially defect and the slippery defectors is Meta Google in my opinion.
1:24:58
Those are the two biggest and so when they start to really get into the game, all of a sudden they're going to be like, whoa, whoa, this is attacking my business model and I'm losing market share and that's when that's when the other guys come back in in a big way. Yeah.
1:25:12
Do you think Sora is going to move any of these like GPU build out plans?
1:25:17
Like Sam was getting kind of mocked for saying like, oh, we're going to have to pick between I was to be clear, I was the one saying.
1:25:23
I said I just thought it was I mean I thought the the timing of going out and saying, yeah, there's a very real possibility in the future that we'll have to choose between curing cancer and free education for everyone and then a week later we launch, you know, the most It feels like the most intensive product ever.
1:25:43
But but I mean, we got some we got some news on API pricing.
1:25:44
It sounds like it's 32 cents a generation.
1:25:46
Like it it it feels like it might not be moving the needle as much on the on the inference side or actually like like is Sora going to make them GPU poor?
1:25:56
I I don't know if you have a view on that.
1:25:58
>> It's 30 Sorry, sorry, it's 32 cents per inference.
1:26:00
What's the time frame on that?
1:26:01
It's an Wait, wait, wait, it's per second. Per second.
1:26:05
Okay, so it's almost three bucks for a for a 10 for a 10 second clip.
1:26:10
Okay, so that's actually like on par with VO3 That's pretty That's pretty That's pretty intense. That's pretty expensive. And they give you what?
1:26:14
10 credits a day or something?
1:26:16
So if there's hardcore users out there, they're generating 30 dollars a day. What what is that?
1:26:21
A thousand dollars a month?
1:26:23
Like it it it's it's it adds up.
1:26:25
If you get a million DAUs on there, like that's a lot.
1:26:29
Yeah, I think they're they probably have more they had the potential to do much more than a million DAUs.
1:26:32
I mean, they're they're they're definitely compute rich, man.
1:26:35
The the pulse thing, have you guys have you guys been doing pulse at all?
1:26:39
They they generate those images for you and you're like, oh, a little personalized slop feed for me?
1:26:44
Like that feels very um I know.
1:26:46
On on launch day we were reacting to it and John John was like John was very impressed.
1:26:51
For me it was like three out of the five images that I looked at were like great and then like two out of the five I'm like, you could have just not generated this one.
1:26:58
This is just absolutely garbage, you know?
1:27:01
But Um what do you what do you think about What do you think we had a question in the chat from Bobby?
1:27:07
He says asking about like what what do you think about public sentiment risk?
1:27:12
It feels like I've never seen more hatred of AI from from non-tech folks than the than in the last uh week or so.
1:27:24
It and and you're seeing data centers data centers just getting kind of like blocked.
1:27:31
Uh that feels like it potentially, you know, it's certainly going to be a roadblock at a bunch of different stages of this build out.
1:27:39
Yeah, I think that that definitely is a newer roadblock that is getting more real because let's be real up until now no one cared.
1:27:45
It's like such a small part of the big picture.
1:27:49
But I still think the reason why I think we're we're going to go bravely into the future is everyone else seems to be pretty on board and I think the one that really helps is the Trump administration is like super duper on board.
1:27:58
They're like, hey, whatever, we're going to tear up the out of everyone but we're going to we're going to completely make up for the weakness in personal consumption by massively investing.
1:28:06
And you know, it's like we're going to get the biggest check from every company in the world every nation in the world and the only thing anyone wants to spend anything on right now is GPUs.
1:28:16
And so they're like, eff it, we're building more GPUs.
1:28:20
And so you're you're seeing all these huge announcements.
1:28:24
And I think that that's the, you know, that's that's kind of the thing that will kind of push it back.
1:28:27
I mean, if Trump is being like, dude, from the government itself, I'm blessing you to buy more GPUs, that's kind of I think the thing that really holds it together and makes it possibly much crazier from here.
1:28:38
The public backlash thing, I think that it really comes back to you have to have that feedback loop work all the way to the right decision maker, not just some random person.
1:28:47
Like, yeah, I'm seeing that read it on the news, too.
1:28:51
But like Mark Zuckerberg or Satya has to wake up and then be like, what am I doing? Right?
1:28:54
Right now everyone in their camp, all the closest advisers and trusted people are being like, spend more. Spend more.
1:29:01
The government, you know, Trump is saying Yeah, I mean, the dinner the dinner everybody wanted to say the biggest number, right?
1:29:06
It was even even Zuck threw out a probably a bigger number than he meant to because he heard someone It seemed like he just heard someone else say a big number and he didn't want to say a smaller number, you know. Uh Cool. How much am I good for?
1:29:19
How much I go He's like He's like on his phone right now. He's like Yeah. >> Okay.
1:29:24
I have a I have a completely What one one more question?
1:29:27
Uh How long do you think XAI can maintain this level of invest you know, ba- basically like this level of investment without being able to back it up with real revenue growth?
1:29:47
I >> Because it seems like this last financing they were having to tell like the journal or some other publication, like we're not having any trouble fundraising, which usually means like you're having not the easiest time if people are kind of poking around. I I mean, okay.
1:30:03
So, the same way I'm not going to bet against Sam Altman's ability to raise, I'm definitely not going to bet against uh Elon Musk's ability to raise.
1:30:10
Like Sam Altman is like the newest kid in the block, but if there's someone that like you should never essentially short ever in the history time, it's Elon Musk.
1:30:17
And the other thing I think that's like much more important with that is Elon's back in the seat.
1:30:22
He's the CEO of Tesla, and he like actually has the compensation. So, he's like, "Fuck it. All right.
1:30:25
I'm I'm I'm actually I'm very locked in." Yeah.
1:30:29
Um and you know, if we're talking about Mag 7, you know who has a lot of market cap? Um Tesla. Tesla. Oh, yeah.
1:30:34
Yeah, but but to That's That's what I'm saying.
1:30:36
The the the Yeah, it it just feels like I wouldn't be shocked if a year from now XAI was a part of was merged in with Tesla. 100% 100%.
1:30:48
And then that you know, getting And that would actually be a great Tesla stock would pump on that, even though cuz people would be like, "What the is that?" Yeah.
1:30:57
It would not be able to compute.
1:30:57
But XAI would would would crush because then they're then they they they're all of a sudden attached to one of the largest financial entities in the entire world. >> Totally.
1:31:05
Like that's that's a huge you know, we're talking trillion trillion-dollar market cap, right? Like that's Yeah.
1:31:09
real Like you can take that and turn it around. You can loan it out.
1:31:13
You can You can move big checks, dude. Yeah.
1:31:15
It's very very non-trivial for them to raise 10 20 billion dollars off of it. Yeah.
1:31:18
What what I was getting at was basically this last financing where they're going and raising capital at 200 billion, and those investors have just been getting pitched Anthropic at like 170 with crazy with a crazy revenue ramp and a bunch of, you know, real real, you know, traction and code gen.
1:31:38
The investors are looking around being like, "Okay, like I'll I'm going to I'll give you the cash, but like I don't know the next time around if there's going to be the same level of appetite."
1:31:49
There There's some really big pockets that back um Elon.
1:31:51
I I I wouldn't bet against the Elons here because you have to remember uh he made a lot of billionaires out of it.
1:31:57
Like, you know, for example, Larry Ellison is always on the phone.
1:31:59
I don't know if you knew that.
1:32:01
Like, he's he you know, he was good for a lot of X uh for X when they took down Twitter.
1:32:06
And so, you you can see a lot of big uh people come out of the woodwork, and like Larry's flush now. Yep.
1:32:12
You know, he's he's got a lot of equity to spend, okay?
1:32:15
Like, you know, these people can come out of the woodwork.
1:32:17
So, I definitely >> a I had a theory a theory that that we were kicking around earlier that it I could see it because Larry's a was a big investor in the original uh X takeover, right?
1:32:27
There's this iconic text exchange where Elon and Larry are talking, and he's like, "I'm good for a billion."
1:32:34
And he's like, Elon's like, "Okay, you might want to do two or some something like that."
1:32:39
>> Yeah, whatever you recommend.
1:32:39
And so, I could see a world where X, the social platform, does not go to Tesla and get spun into the basically Paramount this whole like media conglomerate, right?
1:32:52
Because I do I just don't you know, especially with XAI a part of Tesla, it's not like Tesla's sitting there and being like, "We need to be able to pump timeline posts instantly into every Tesla in America."
1:33:02
It just stops being quite as strategic, maybe. Yeah.
1:33:07
Yeah, I mean, I I I don't disagree with that take.
1:33:10
That's a pretty good take, but I just think that like, you know, Elon will find a way.
1:33:13
Whatever way that that that it needs to happen.
1:33:16
If that's with Oracle, if that's with Tesla, like Elon will find a way, I believe.
1:33:23
>> Is there Is there significant asset value in just Colossus 2?
1:33:25
Like it it is the biggest supercomputer.
1:33:28
It's this massive data center.
1:33:30
It seems like it was very hard to build.
1:33:31
It seems like all the hyperscalers would love to to to have it.
1:33:36
Sam would love to have it.
1:33:38
>> ran an auction for it, do you have an idea of like what the value of Colossus 2 would be if there was like breakup value there?
1:33:46
Um you know, at today's valuation, I I mean, I would say it's probably probably like couple of >> tens of billions.
1:33:55
I mean, I mean, maybe not tens.
1:33:58
I mean, I mean, I would say tens, actually. Tens. Tens.
1:34:03
That's not >> Cuz it's Are we talking Are we Are we Are we including Colossus 2? Yes. Okay, then yes. Yeah.
1:34:08
Yeah, I I'm just thinking like like the XAI assets like sure, the business isn't necessarily like printing profit with some massive consumer product market fit.
1:34:19
>> Someone would be really stoked.
1:34:20
>> But someone would be stoked, and there's some value there. It's interesting.
1:34:21
If you think about whatever Nvidia's just that deal with Microsoft, literally scale that up, and that's that's probably what you would buy on a TCO basis for Colossus. Like, there's demand. So, okay.
1:34:32
Okay, my Yeah, one one last question. You got to go? Okay. Okay, one last one.
1:34:39
A hairbrained uh hairbrained theory.
1:34:40
Uh I want your feedback on this hot take.
1:34:43
Uh there's a lot of discussion over the value of TSMC and Taiwan risk, uh but we now have the golden visa, $2 million to bring someone to America.
1:34:55
For $100 billion, you could bring 50,000 people.
1:34:58
Is there Is there some world where we get into like AI talent wars, but for process knowledge around semiconductor manufacturing?
1:35:04
And in And we see some American entity start poaching TSMC employees en masse to get them to America.
1:35:16
Um that's definitely possible.
1:35:16
I just don't feel like it's probable.
1:35:19
People have been saying there's a supply shortage, yet like TSMC has a lot of like has almost a captive audience, dude.
1:35:26
Like, if you think about it, TSMC's like, you know, I mean, it'd be like working for the government almost, but like Yeah.
1:35:31
even more prestigious than killing the out of everything.
1:35:34
Like, TSMC is like the company of Taiwan. Sure.
1:35:35
Dude, honestly, TSMC engineers are like really underpaid as is.
1:35:39
Like, don't tell Don't tell them, right?
1:35:41
But like um on a on a on a dollar basis compared to even American peers, and a lot of people like like China, for example, uh poaches very heavily.
1:35:48
So, I think there's definitely a chance, but I feel like you have to take the entire know- know-all.
1:35:54
Like, it's not just the tool It's the tool, the process, everyone.
1:35:58
You'd have to take the whole fab.
1:36:00
Um but yeah, I think that'd be I think it'd be sick, dude.
1:36:02
I think if Intel >> Intel could do it, right? >> Intel should do it.
1:36:05
Yeah, Intel should do it cuz they have the golden visa.
1:36:09
They have the backstop from Trump.
1:36:09
They have all the investors coming in.
1:36:12
They'll probably have an OpenAI deal soon, who knows.
1:36:13
And then they have all the money, and then they can just bring people over.
1:36:16
Uh anyway, we won't take any more of your time.
1:36:17
Thank you so much for hopping on the show. Super fun, Doug. >> Always a great time. Nice to see you guys. Yeah. Talk soon. Bye.
1:36:23
Uh let me tell you about Turbo Puffer search every bite serverless vector and full-text search from first principles and object storage.
1:36:30
Fast, 10x cheaper, and extremely scalable.
1:36:31
Uh you also heard Doug talking about agentic commerce.
1:36:36
It's more important than ever to get your brand mentioned on ChatGPT. Go to Profound.
1:36:40
Reach millions of consumers who are using AI to discover new products and brands.
1:36:43
Uh our horse post popped off.
1:36:46
We took our first day off on Friday, but we did give you some facts about our horse in the studio, which you can see behind Georgie Hayes over there.
1:36:53
And we got a wonderful reply from Eric, who's working on Soil Tech.
1:36:57
He says, "The horse is a symbol of deep The horse as a symbol is deep lindy, power, camaraderie, and frontiering forward motion.
1:37:03
As a Kentucky As a Kentuckian, I can only agree. More horse content."
1:37:08
And shares three beautiful images of horse statues >> Incredibly underrated animal.
1:37:11
Uh in other news, Dan McCormick, Packy McCormick's brother, has launched his creatine gummy brand in Target. That's Create.
1:37:21
Uh a wild step in this journey that in many ways started with him posting on X.
1:37:25
Three years ago, he was embarrassed to be starting a creatine company.
1:37:28
He'd shy away, hedge, and downplay when asked about it in real life.
1:37:32
On X, surrounded by a bunch of like-minded freaks, I could post openly about my excitement and the business's progress.
1:37:37
I've shared a lot less over the last 2 years, but my bullishness and Actually, the the timing on Create and Dan's perfect timing is insane.
1:37:43
It was It felt like a supplement that was already had almost perfect saturation, right?
1:37:53
I think anybody that I knew that was like working out multiple days a week was already using creatine, but never doubt uh American demand for uh supplements in gummy format.
1:38:03
Never doubt >> Never doubt it again. Never doubt Elon. Never doubt Nightcap.
1:38:09
>> doubt demand for gummies. Oh, this is fun.
1:38:11
Uh Ramp and acquisition today.
1:38:11
They bought Jolt AI uh to help their engineers build faster.
1:38:16
Jolt is a world-class team of engineers who have spent years solving some of the hardest problems in developer productivity.
1:38:22
Today, uh Ramp is welcoming them to the team.
1:38:24
They're scaling the AI platform and reimagining how software and finance get built.
1:38:28
Said, "From a talent perspective," says Karim Atiyeh from Ramp.
1:38:33
"Ramp's engineering team is made up of founders, math olympiads, and AI researchers.
1:38:36
Uh uh Karim's goal is to hire elite technical talent and then get out of their way."
1:38:42
For his part For his part, Jolt's Specter, uh the founder, admitted in an interview that he didn't expect to get acquired by a company like Ramp, but that he's extremely happy it's where his team landed.
1:38:52
The trio will be working on a number of things, including Ramp's internal engineering platform. So, congrats Fantastic. Hit that gong, John. First hit of the week. It's good to be back. Oh.
1:39:06
Uh Technium says you can now invest in Nvidia, Intel, AMD, Arm, OpenAI, Mistral, CoreWeave, Nebius, and more with just one ticker. Nvidia. >> Nvidia. Of course.
1:39:18
Uh we also had a fun >> Pareek is says word.
1:39:19
Is he Is he continuing >> He's posting again.
1:39:23
Uh is he continuing to to run?
1:39:25
Um uh we also got some gong analysis.
1:39:28
Uh this is from Ethan Frost on September 30th.
1:39:32
He said, "I analyzed every gong hit on TVP and here's what I found.
1:39:37
Tyler, can you take us through this post um of what uh what the breakdown was?
1:39:42
And And do you think this was uh manually tagged or was there AI involved?
1:39:46
It It's some It's It's a pretty remarkable uh breakdown of of all the different um all the different gong hits, how much they raised, and then um and then how loud the gong hit was.
1:39:59
But explain this post to us."
1:40:00
>> Yes, that the main I think the main thesis was like are gong hits on TVP and are they louder as funding increases? >> Yes.
1:40:07
Um you can see the correlation is like slightly there.
1:40:10
I I think it's a bit hard cuz of like sometimes the microphone is just a little bit farther away.
1:40:14
So you you didn't It wasn't like, you know, a crazy kind of correlation.
1:40:18
Um but yeah, he basically just indexed all of the gong hits that we've basically ever done. Yes.
1:40:24
And then done a ton of data.
1:40:25
So like series A, series B.
1:40:25
Um he he breaks it all down.
1:40:28
He also breaks down between you guys. Okay.
1:40:32
>> Um so there's some good facts here.
1:40:33
Who's louder on the gong, me or Jordy? Um let's look. Definitely Jordy. Definitely Jordy. For sure. Okay. >> Definitely me.
1:40:42
Um yeah, John uh is is standing more often Standing.
1:40:46
>> than than Jordy when he's hitting the gong.
1:40:47
But Jordy throws >> Do I But I throw the mallets.
1:40:52
>> Yeah, Jordy throws the mallets a lot more often.
1:40:54
>> Now I've got a clean shot.
1:40:54
I got to be >> You got to really rip it.
1:40:58
Uh these are these are so funny.
1:40:58
Uh total gongs I heard by day of the week.
1:41:03
Why are there so many on Saturday and Sunday?
1:41:04
We don't record on Saturday or Sunday.
1:41:06
I was very confused confused by this.
1:41:08
Oh, maybe weekly recap we post sometimes. >> Maybe.
1:41:13
>> the uh anyways, fantastic analysis. >> Yes. And um The toss.
1:41:18
Let's see the the the the toss hit. That's a funny one. Under utilized.
1:41:27
>> Anyway, thank you for your uh for your service, Ethan Frost. Legend.
1:41:29
Uh Pluco Capital says, "I put 10% of my portfolio into Reshoring Semis and it just wasn't even close to enough.
1:41:36
Easiest money of our lifetime. Listen to the president.
1:41:41
Should have brought a shovel, not a teaspoon." It's true.
1:41:44
Uh yeah, very interesting to see where the where the Intel story goes.
1:41:46
I wonder if there will be an OpenAI deal for for CPUs or something.
1:41:50
Uh Justin Moore was sharing that Taylor Swift is using AI-generated videos to promote her new album.
1:41:57
I was surprised to see that.
1:41:57
Uh biggest artist in the world.
1:42:00
Usually there's like a ton of pushback when big artists use AI or something.
1:42:06
>> would get She's a prime acquisition target.
1:42:09
>> I haven't followed it too closely.
1:42:11
>> McKenzie says, "I think the economic logic of this is inevitable and that you'll need video to get your N+1 marketing post through the noise on social media, even as Taylor Swift.
1:42:18
And your choices are either a 100K shoot per post or relatively junior employee with some taste." Yeah.
1:42:25
And uh yeah, it's good take.
1:42:29
Yeah, there is this development of like will everything be instantiated in every possible format?
1:42:35
So if you have a startup launch video or announcement, uh you'll write a pithy tweet, and then you'll write a blog post, and you'll also drop a podcast, and then you'll also drop a vibe reel and a video and a 2-hour documentary, and they will all be instantiated from just a short fact sheet that then goes out.
1:42:54
And it begs the question of like where will the where will the instantiation live?
1:42:59
Will I as the consumer of content say, "I want to consume the these sources in video format?"
1:43:07
That's kind of the notebook LM model is like I'm saying I want to consume the internet in a in a podcast format.
1:43:16
Uh with ChatGPT, you're kind of like no matter what question you ask, no matter what the sources are, it's it comes to you as a as a, you know, neatly organized research report that's a couple thousand words with bullet points, and it's not this, it's that every time.
1:43:32
Um but but then there's the other side, which is on the lives on the creator side, which is says, "I should instantiate my idea, my concept, my take as a TikTok reel, as a YouTube video, as a podcast, as a video."
1:43:45
And I do that on my side.
1:43:47
And I'm wondering where it lives long term. Yeah.
1:43:49
It because there is a world where you can go to where Taylor Swift doesn't need to generate the AI-generated video because I as the Taylor Swift fan can go to Sora, and it knows that I'm a fan of Taylor Swift, and it automatically generates it for me based on just her text post.
1:44:08
>> Yeah, or she could just share She could just share the prompt in text in the notes app.
1:44:12
And then you can just imagine it.
1:44:16
I don't know, it'll be somewhere in between.
1:44:17
Ethan Frost is in the chat. Let's go. >> go. Hey, what's up, Ethan? Thanks so much. >> you.
1:44:22
Thank you for analyzing the gong. Yes. Um What else? Linear.
1:44:27
Linear is purpose-built tool for planning and building products.
1:44:30
Meet the system for modern software development.
1:44:32
Streamline issues, projects, and product roadmaps.
1:44:36
I would go out on a limb and say everything that OpenAI launched today was almost certainly planned in Linear. >> For sure. For sure. Oh, this is a fun one.
1:44:44
So uh Kareem Jet says, "Introducing Gaslight Garage, a box where I put my phones and feed them AI-generated audio nonsense to make them think I want to buy stuff.
1:44:54
Practical AI for the people, Gaslight Garage."
1:44:57
And Snowy, or SNWY, says, "Uh I tried this once after getting into an argument with my parents about whether your phone listens or not.
1:45:07
Where do you stand on this, Jordy?
1:45:07
Do you think your phone is listening to you?"
1:45:12
And such a tough one because everybody has like an example or two. >> Yep.
1:45:17
But if you think about We talked about this thing.
1:45:21
>> ads you get, >> I saw an ad >> different things you you see, how many ads you get over a multi-year period, how many products you talk about over a multi-year period, and then at some point or another there's like perfect alignment between the two.
1:45:37
>> And then also you have to factor in people that search for something once and then, you know, on a certain network and then forgot about it or someone else in a household search for something and then somebody else in the household gets served, right?
1:45:51
There's a lot of different ways to explain away the the listening aspect, but I Yeah, I I noticed that some of the spookiest uh you know, ads that I would ever see would be it would come from a conversation that I had with a person in real life where like, yeah, the phones could have been listening, but more likely And then I saw an ad this where they tell me, "Oh, you got to test out these headphones."
1:46:14
And I see an ad for those exact headphones the next day.
1:46:15
And I'm like, "That's weird that I just heard about that and now I'm seeing an ad." How did that happen?
1:46:20
I know I didn't search for it.
1:46:21
I know I didn't send any any data signal to my phone or my apps.
1:46:26
But then I realized that it's like, well, if they just ordered those headphones, and that data of them shopping and them going through that checkout flow is encoded in their Instagram profile, and then it knows that we hung out because we're friends on the network and we like each other's posts and we were texting with each other on the app and stuff, and then it can say, "Well, Jordy just liked these headphones.
1:46:51
John and Jordy were just hanging out.
1:46:53
Showed John the the the Like they might be similar consumers."
1:46:55
I feel like that can be like kind of spooky, but not but very like um a clear flow that would come from the data.
1:47:02
Um But anyway, uh Snowy said, "Got two iPhone 6s, both wiped and connected to two different Wi-Fi networks with independent VPNs.
1:47:10
Said keywords to one and different ones to another, never in earshot of of each other with multiple apps, Insta, Facebook, Twitter open with their accounts and basic usage, and the ads were never what I said.
1:47:21
They were basically always location-based {slash} pretty basic things.
1:47:25
My running theory is that whoever thinks their phone is listening is just predictable enough that it knows, not because it heard them."
1:47:32
And that's probably right.
1:47:32
Plus, there's this massive incentive that if the phones were listening and targeting, there would be a huge incentive to become a whistleblower.
1:47:40
Like there's been a whole tour of folks who have like left Meta, become in become influencers. >> a book deal. You can get a book deal. Exactly.
1:47:48
And you can become famous.
1:47:48
You can get a speaking tour.
1:47:50
You can get his head talk. >> Exactly.
1:47:51
You can get a government job. Probably.
1:47:54
How how much somebody at Meta wants to get >> to testify.
1:47:57
Maybe they should do that.
1:47:59
Appearance fee for for testimony on Capitol Hill. Let's make it happen. Um Anyway, uh Numeral. Sales tax on autopilot.
1:48:07
Spend less than 5 minutes per month on sales tax compliance.
1:48:08
Someone uh someone in the in the chat is like, "I was just thinking about Numeral." How did they know?
1:48:15
He's speaking to me directly.
1:48:16
Um Well, uh you and Tyler should debate this one. Uh Mr.
1:48:20
Beast said, "When AI videos are just as good as normal videos, I wonder what that will do to YouTube and how it will impact millions of creators making content for a living. Scary times."
1:48:31
And Justin Moore had It's kind of crazy that biggest YouTuber comes out and and uh comes out and just scares everybody because I think if you dig into this at all, uh it's pretty easy to show how AI presents an incredible opportunity for existing creators.
1:48:52
Like yes, some amount of attention will go to content.
1:48:54
Uh will go to net new net new content creators that are just using the tools better or more focused, but uh nowhere in here does it say that if you're making a living from content today, you can't use AI to make significantly more content, significantly better content.
1:49:16
It It It's It's much more of a um opportunity than a threat in its current form.
1:49:21
Yeah, I also think I It's funny coming from Mr.
1:49:26
Beast cuz it's just like this is like slop on slop hate. Like I Mr.
1:49:27
Beast is like I would not want my kids to be watching Mr. Beast videos.
1:49:32
If you watch them, it's just like the cuts are like every like 1 and 1/2 seconds.
1:49:35
It's like pretty insane already.
1:49:38
So, I like I I think um Mr.
1:49:39
Beast style content >> you remember if you remember, he wasn't uh he tried to make create a tool that used AI to make thumbnails and he eventually rolled that feature back because it was taking away uh revenue potential revenue from thumbnail artists. Yeah.
1:50:01
Yeah, so I I mean I I would say that his style of content is much more susceptible, I guess, to to like being taken advantage >> I totally disagree because I think part of the reason like I can see what you're saying in the sense that like you can argue like this is slop.
1:50:16
I don't want people to to I wouldn't want my kids to watch it.
1:50:20
At the same time, it's very clearly the opposite It It It's the opposite of slop in that it's like massive master productions, real humans, >> Did you see the reaction when he was like don't worry, Jimmy?
1:50:32
Like people will want to know that you made 100 people stay in that circle for a month. Exactly.
1:50:39
It's like so much more entertaining. If If you just took Mr.
1:50:42
Beast style content and made it using AI, it would not be nearly as entertaining, right?
1:50:48
Yeah, people want the people want the suffering.
1:50:51
>> It's like why Or Or you could just watch two robots play chess, right? Yeah.
1:50:56
Like okay, you can say that it has this big budget, but like I I don't think you can't say that Mr.
1:51:00
Beast is like oh, it's like this auteur cinematic content.
1:51:04
Like it's still at like that there's like okay, it's like they have real people, but above that it's just like this is like very stimulating like fast-paced content.
1:51:13
>> Yeah, but a big part of it is that the human suffering element.
1:51:15
AI I I I I You know, wait until AI can actually suffer like a human.
1:51:23
Yeah, I mean I mean I mean you could I I'm I'm trying to think of the most recent example, but there is something about uh just like the fact-checking chain of like if you're watching someone do like I I bet when the X Games happens or when Red Bull does their next stunt, like that will get more viral views than the Stephen Hawking novelty X Games video.
1:51:46
Just because there's something that we know that the fact that that the X Games is a real organization, that they own that they're not posting AI-generated content and they've made that commitment to only show videos or Red Bull's only showing videos of humans doing wing suiting through a helicopter off of a cliff or whatever.
1:52:03
And so, um there's something that just hits like more emotionally, like it's more emotionally resonant.
1:52:10
And then there's also all of these like uh downstream like behind the scenes like like the Mr.
1:52:16
Beast like are the videos fake has been a long time debated.
1:52:18
He's kind of won beat those allegations every time.
1:52:23
Um but like the the incentive to like kind of see like a whistleblower on someone who's like faking their content will be like at an all-time high, but I would imagine that the the institutions that still make people suffer will do quite well. But I don't know.
1:52:37
I mean, for for I I think it'll definitely like widen the gap.
1:52:42
Like if you're if you're if you're making content that is at all substitutable with slop, like you're done.
1:52:49
But if you're making content that cannot be challenged whatsoever and is so like uniquely human.
1:52:56
Uh but yeah, where where does Mr. Beast fall on that?
1:52:58
A good example A good example is like or is is Andrew Huberman threatened by someone else coming out and creating an AI-based persona that tries to do the same thing that he does, right?
1:53:11
In terms of like creating long-form content around various health protocols and health topics.
1:53:19
And I think that even if you could put the level of craft in that the Huberman Lab team does, do you want to be getting health advice from an imaginary robot? Right?
1:53:33
Or is it Or do you want health advice from Andrew Huberman because he's 50 and he's absolutely jacked.
1:53:37
He's living living his best life, right?
1:53:39
It's like who, you know, where the content is coming from matters a lot. Yeah. For sure.
1:53:45
Yeah, yeah, the provenance matters.
1:53:48
People people care about other humans.
1:53:49
Uh they they're anti-clanker at their core in many ways.
1:53:55
Um well, let me tell you about fin.
1:53:55
ai, the number one AI agent for customer service.
1:53:59
Number one performance benchmarks, number one in competitive bake-offs, number one ranking on G2.
1:54:02
Uh in other uh reshoring We're not just reshoring chips, we're also reshoring Sharpies.
1:54:09
Uh apparently Sharpie found a way to make pens more cheaply by manufacturing them in the United States.
1:54:12
Newell Brands moved production >> one of these.
1:54:16
They need to make the red, white, and blue Sharpie pack immediately. I need these.
1:54:21
This is actually mentioned in the Wall Street Journal article.
1:54:25
Uh apparently uh where is it?
1:54:27
Uh Donald Trump uses a Sharpie all the time or something like that. For signatures, right? Yeah. Where Where is it?
1:54:33
Uh pen barrels whirl along automated assembly lines that rapidly fill them with ink.
1:54:40
At least half a billion Sharpie markers are churned out here every year.
1:54:42
Each one made of six six six parts.
1:54:44
Only the felt tip is imported from Japan.
1:54:46
It didn't used to be this way back in 2018, many Sharpies were made abroad.
1:54:51
That's when Chris Peterson, who is the CFO of Sharpie maker Newell Brands, challenged his team to answer the question, how could they keep Newell from becoming obsolete compared with factories in Asia?
1:55:02
Uh I felt like we had an opportunity to dramatically improve our US manufacturing. He's now the CEO.
1:55:08
And he makes all 93 colors of Sharpie at a 37-year-old factory.
1:55:12
And they did it without reducing employee count and without raising prices.
1:55:16
Uh but to get to this place, it took close than two close to two billion dollars in investments across the company.
1:55:23
And of course, this is a two billion-dollar market cap company. Well, okay.
1:55:29
It's like what book value to market cap is one to one, apparently.
1:55:33
If they This company has so I want to get the CEO on the show at some point.
1:55:38
President Trump uses a custom Sharpie.
1:55:40
He slapped high tariffs on many imports with the goal of prompting more companies to do what Newell has done and bring manufacturing back to the US.
1:55:46
But Filippo Falorni, Falorni, an analyst at Citi, said few companies in the sector have the resources to replicate Newell's move.
1:55:57
Well, um I think we're ready for our next guest. Ready?
1:56:04
Um let's bring in Celine from Loyal.
1:56:06
But first, let me tell you about Addio, customer relationship magic.
1:56:09
Addio is the AI-native CRM that builds, scales, and grows your company to the next level.
1:56:15
You can get started for free.
1:56:17
Let me also tell you about Public, investing for those who take it seriously.
1:56:19
They got multi-asset investing, industry-leading yields. There you are. >> trusted by millions.
1:56:23
Yeah, we're hoping to get a real horse in the studio at some point. >> I can help with that. >> Yes? >> I have like three. No way.
1:56:29
Wait, you actually have horses?
1:56:31
>> yeah, I'm a crazy horse girl. >> Where? Yes. >> I'm at Petaluma. Petaluma? Yeah. Do you live there? No. I'm in SF. Cool.
1:56:37
>> But I do the anti-commute. Okay. What is that?
1:56:40
Like you drive up against the traffic to North Bay to go ride your horses.
1:56:43
I've taken investor calls while on a horse. >> Okay.
1:56:46
Almost fell off on an investor call while on a horse.
1:56:48
Wait, walk me through like uh are they the same breed?
1:56:52
Like what's the story with the three horses? Why three?
1:56:53
Um horses are potentially the worst investment you could ever make.
1:56:57
Um I do not recommend them. They are worse.
1:56:59
Um I feel like there's some like seed valuation joke you could make here about horses.
1:57:06
Anyhow, they tend to break themselves.
1:57:08
They're like self-destructive.
1:57:08
We were thinking about getting a horse at one point.
1:57:10
We wanted to race in the Kentucky Derby and give it like some funny name.
1:57:14
We were We were going to call it the ramp. Yeah.
1:57:15
And we were And we were And we were running the numbers, and it wasn't that crazy as a marketing >> Well, yeah, you get into it and you're like okay, maybe it's like 100K for the horse, and then it's like a couple grand a month to like >> such a funny stunt.
1:57:27
Someone's going to do it now that we said it on the show.
1:57:30
Dude, that color would look sick on a horse, though.
1:57:31
Like imagine the racing stripes. I love it. Anyway, sorry.
1:57:33
Kick us off with an introduction on yourself and the company. Hi, yeah. I'm Celine Halioua.
1:57:39
I'm the founder and CEO of Loyal, and we're developing drugs to extend dog lifespan. Amazing. >> Horses next?
1:57:45
Horses next, venture capitalists next.
1:57:48
[Laughter] Uh yeah, what uh what was the inspiration for the company?
1:57:53
Did Like did you know that you wanted to start with animals or was that purely like the FDA? No.
1:57:57
So, I actually was working in human longevity.
1:58:00
So, I worked with a previous guest of y'alls, Laura Deming, at her venture fund for a few years, and we were just frustrated because nobody was trying to develop a drug explicitly for lifespan extension, right?
1:58:11
And we're like this is obviously like obviously we all want to >> the sci-fi promise.
1:58:15
Everybody was promising it, but nobody was working on it specifically. >> Not directly, no.
1:58:19
They're all kind of going these indirect paths.
1:58:21
And so, we became obsessed with how to do this.
1:58:22
Um realized to do it in humans, it would take a lot of time, a lot of money.
1:58:26
But you could do it in dogs.
1:58:28
It's a super crazy market.
1:58:30
I hear one of you is a dog person. >> I am. I am.
1:58:32
My dog was supposed to attend, but I had to leave right after turns, so he's busy.
1:58:35
He was chasing squirrels.
1:58:36
He's a 10-year-old Newfoundland, so he needs Oh, wait. So, you understand this.
1:58:39
Like is we're starting with the dog lifespan?
1:58:43
>> Yeah, on your uh homepage.
1:58:43
There's a picture for a while, at least. Yeah, no.
1:58:46
Newfies are the Yeah, no.
1:58:46
It was like the dog was pitched to me as like look, every every day after year five is like a miracle uh because they're known to have extremely short lifespans.
1:58:56
And so, I kind of set myself up for like look, like I'm I'm not coming in Like you you you get a you get a golden retriever, you're not thinking like yeah, it's going to live for 40 years.
1:59:04
So, I just went into it thinking, okay, I'm getting a Newfoundland, that's 6 to 8 years or something. And he's done very well. He's 10 years old. Good job.
1:59:10
We need to sequence your dog.
1:59:11
I'll send you a saliva kit.
1:59:13
We have an investor that has two giant Newfies that they bring around everywhere. Ooh, everywhere. I should have said that. It's it's hilarious.
1:59:19
And >> why he's living so long. Well, yeah, that helps.
1:59:24
And do not take your dogs commercial.
1:59:29
I actually got convinced to take their term sheet because they sent me their Pinterest board, which was just Newfies all the way down. I love Newfies. They're the best dogs. They're so good.
1:59:38
Um so, yeah, uh on the like the fact that there's never been a drug explicitly for longevity, is that just because like in order to have a drug, you need a like a problem or some sort of like it needs to be a treatment for a specific condition and and just death is not a condition.
1:59:56
And so, is is is that the right narrative?
1:59:57
Yeah, like the way to think about it is to be able to have a drug to approve to do something, you have to prove that the drug the drug is doing that thing.
2:00:05
And so, if I gave you guys a longevity drug and I was like, oh, let's see if it extends your lifespan.
2:00:08
Yeah, I'd be following you for decades.
2:00:10
And the company's going to be long dead and all of that.
2:00:13
>> It's also hard to prove that the drug was the catalyst for your life, right?
2:00:18
It could have been, oh, they just got a lot of sun or they walked a lot or they There's a lot of variability, right?
2:00:24
Like the decisions you make, the yerba mate being pro or anti-longevity, whatever it is.
2:00:28
All of these things vary aging.
2:00:31
It's one of the most complex phenotypes, but we all go through it.
2:00:35
And that's the same thing is true in dogs.
2:00:36
But the good thing about dogs, at least when you're developing a longevity drug, is they age much faster.
2:00:39
So, you can see if something is working in a period of time that's much more reasonable.
2:00:44
And you can you can theoretically have a much more like isolated study. Is that right?
2:00:49
It's just because What do you mean isolated?
2:00:52
Well, I just imagine like it's easier to control factors in a dog's life versus like a bunch of >> Like a dog is not going out on a bender in Las Vegas. >> That is a yes. Yes.
2:01:01
So, so you can just be like, yeah, we know exactly what the dog's diet was pretty much every single day.
2:01:07
Maybe he got a new cheeseburger once, but a human is is like using diet Coke one day and then is some alcohol and then this and then bender and then Burning Man, who knows?
2:01:15
There's so many confounding factors. The dog's much more.
2:01:19
>> have like a cheeky cigarette half an hour earlier in their life that they don't talk about or something like No, it's true.
2:01:25
Yeah, dogs just it's a lot it's still complicated, but it's much more simple.
2:01:28
And then you also have these individual breed variants, which teach you a lot about humans, too, right?
2:01:34
Like why does a Newfie have such a short lifespan versus a golden versus a Chihuahua?
2:01:39
You can study these things in a much more quickly way.
2:01:40
Yeah, why not why not mice though?
2:01:42
I feel like there's a lot of mice studies. It works in mice. >> I love mice.
2:01:46
Uh I don't think mice are multi-billion dollar market.
2:01:50
>> Yeah, that's the thing.
2:01:50
The the commercial opportunity in dogs, if you ask a dog owner how much would you pay to have an extra year of of healthy lifespan for your dog?
2:01:58
And I'm sure there's a wide range between like $1,000 and some people that are like, I would pay any price.
2:02:06
I will, you know, liquidate my 401k for another 2 years.
2:02:12
It's actually really interesting.
2:02:12
Like people it's not actually really at all correlated to socioeconomic status or how much, you know, people cash people have in the bank.
2:02:20
They will just spend anything.
2:02:21
A certain subset of the population will spend anything on their dog.
2:02:26
I've had people offer me $50,000 for the drug before.
2:02:29
I would like to note to the FDA that I said no.
2:02:30
Uh but I've actually gotten a lot of offers like this.
2:02:34
The willingness to pay is insane. The market is huge.
2:02:35
And so, we could hopefully make a multi-billion dollar company around this and then use that revenue to fund more research. Yeah. Maybe into horses. Love it.
2:02:43
Yeah, what was the initial Yes, please.
2:02:46
What was the initial target kind of the the the biotech side of the business?
2:02:51
Like how do you know what to target?
2:02:53
How do like how do you narrow it down?
2:02:55
Like there's a lot of different life extension NAD and telomere extension and just, you know, cardiac stuff.
2:03:00
There's like a million different sub-indications that people think about when they're thinking about their own longevity.
2:03:06
How do you narrow it down? What do you focus on?
2:03:08
So, the way we thought about it is we wanted something that was as broad as possible.
2:03:12
So, as you know, efficacious and relevant to the Chihuahua as it is to another dog, right?
2:03:19
And even if they have different food or different behaviors or, you know, more inbred or less inbred that it still may work. So, that was one.
2:03:23
And two, you want something that's relatively simple, right?
2:03:26
Aging is so unbelievably complicated that when you add any additional variables, like things just become exponential really quickly.
2:03:34
And so, where we actually started is what you were talking about earlier with Newfies.
2:03:37
With a short lifespan of big dogs.
2:03:39
Um because it turns out the genetic driver of dog size, the thing that makes a Newfie grow really really big in puberty or makes a Chihuahua not grow big in puberty, also genetically controls the rate of aging of that dog.
2:03:53
So, your Newfie, well, maybe not your Newfie, but the average Newfie is actually aging at a much faster rate.
2:03:59
And so, that was a really nice place to start because A, it's something that everyone understands, especially big dog lovers.
2:04:04
They're like, I know I know the heartbreak of losing a Great Dane or losing a large Labrador.
2:04:08
Um and it was like simple mechanistically.
2:04:11
And it was applicable to a wide swath of the canine population.
2:04:16
Basically, any dog over a certain weight would be eligible and has this problem, right?
2:04:20
Because they're not living as long as a Chihuahua that lives for 16 years.
2:04:23
Every other bigger dog has a somewhat shorter lifespan than that theoretical maxima.
2:04:29
So, I mean, how do you explain what the drugs are actually doing?
2:04:34
So, basically >> It's a black box, unfortunately.
2:04:38
Basically, the way to think about it is, you know, all puppies are born approximately the same size, but a Great Dane puppy has super high levels of these proteins called growth hormone and IGF-1.
2:04:46
And these are circulating at really high levels and they're basically binding to their cells and saying, grow, divide, grow, divide.
2:04:51
And that's why your Bernese puppy or Newfie puppy is getting so big. That's normal.
2:04:54
It happens in humans, too.
2:04:56
But it shuts down significantly once you're fully grown.
2:05:00
In a big dog, it doesn't.
2:05:00
So, big dogs still have these super high levels of these growth hormones circulating around that are basically making the dog turn over at a faster rate.
2:05:08
So, what the drug does is it inhibits those levels, it brings them down to a level that's seen naturally in dogs.
2:05:14
So, think bringing a Newfie's level down to an Aussie Shepherd or something similar.
2:05:18
Um and this is like this idea that if you reduce growth hormone, if you reduce IGF-1, it extends lifespan.
2:05:24
It's actually been shown all the way from worms to mice to, um you know, multitude of organisms.
2:05:30
And there's even human centenarians that have the same genetic, you know, things that Chihuahuas do.
2:05:36
And are our And they're spiritually Chihuahuas.
2:05:39
>> So, so is is is the like is the scientific like phrase like IGF-1 inhibitors essentially?
2:05:45
Um and are biohackers using that like wild like against the FDA?
2:05:51
Are they trying to can you learn from from these sort of fringe biohackers and what what they do?
2:05:56
Is there any one study or anything?
2:05:58
Is there anything to learn or just uh Um well, yes, there's always things to learn. Um you know, it's funny.
2:06:03
People actually often will supplement growth hormone IGF-1 because it makes your muscles really big. >> Oh, sure. And it's very aesthetic.
2:06:10
It's kind of like human growth hormone. Yeah, exactly. Exactly.
2:06:13
It does not increase function, but it does increase like muscle size.
2:06:17
So, a lot of people are might not be happy Interesting.
2:06:18
It doesn't correlate with strength at all. >> like It just pumps.
2:06:26
I mean >> Yeah, you guys can't tell I'm on it. >> Yeah, yeah, yeah.
2:06:28
Well, I mean, yeah, plenty I think the people that are taking HGH are doing that specifically entirely for aesthetic reasons.
2:06:34
And they're they're post-strength. Yes.
2:06:36
Yes, they are post-strength.
2:06:39
Um but but are there people in the bio in the biohacking community that swear by it or are are they optimistic that it will have an effect on longevity for them?
2:06:49
Yeah, I mean, this mechanism is pretty consensus as a longevity mechanism.
2:06:53
It's not going to be as like this specific one, the big dog short lifespan, isn't going to be as relevant to humans because like I'm not living twice as long as you guys, right?
2:07:03
Me being so short that I'm having to like sit on my feet.
2:07:06
>> Sure, but a Newfie does live twice as lives half as long as a Chihuahua. Exactly. Exactly.
2:07:09
So, think of it as kind of like a biological hack to have a foot into a very very very complicated biological problem.
2:07:18
What's the current understanding of why women live longer than men?
2:07:22
I mean we just do less crazy >> risk-taking?
2:07:26
>> Yeah, it's risk-taking.
2:07:28
I think it's just like heart attacks or something.
2:07:30
No, I actually don't know.
2:07:32
There was like a real reason, but I don't know. Yeah, yeah.
2:07:33
I like the idea that dudes are just wing seating.
2:07:36
>> Dudes are just insane.
2:07:38
So, so walk me through the the actual process to commercialize and and, you know, grow the business.
2:07:43
Um I imagine the FDA has a number of different subgroups.
2:07:48
Veterinary medicine's one of them.
2:07:51
That is that a more receptive faster pathway than just the traditional drugs for humans pathway?
2:07:57
Are you accelerated there?
2:08:00
Yeah, so the the FDA on the veterinary side is super stringent, right? It's it matters a lot.
2:08:06
In some ways, you know, it's almost more because a dog can't talk.
2:08:08
So, you're really trusting the FDA and you're trusting your veterinarian that this like pill you're giving isn't going to like cause some disaster.
2:08:13
Um but the nice thing with with canines is that the incentives are really really aligned, right?
2:08:21
Like everybody wants dogs to live a longer healthier life.
2:08:23
And preventive medicine is really where it is with dogs because, you know, there's no insurance.
2:08:29
Nobody's getting half a million dollar cancer car-T therapies or whatever, right?
2:08:33
It's it's it's always the dog owner having to pay out of pocket.
2:08:35
So, most of the really successful dog drugs today are already preventive medicines.
2:08:40
Think like heartworm prevention, right?
2:08:42
It's reducing the risk of your dog getting a disease that if you give the drug, you'll never see them get that disease, right?
2:08:47
Flea and tick, same thing.
2:08:49
There's a lot of categories of medicines like this.
2:08:51
Um so, yeah, it's actually been really fun to work with them.
2:08:54
And then the new administration is like, well, it's a little complicated.
2:08:57
They're like super pro longevity, right?
2:08:59
And so it'll be interesting to see how that maybe impacts the human side where things are very regulated, but also just a bit more complicated.
2:09:06
But it seems absolutely critical even just from a fiscal standpoint that we figure out how to get people to live longer and healthier.
2:09:16
Yes, it's you know, people when you talk about longevity, they tend to be like, oh, but like you're just going to make the rich live longer.
2:09:21
And like the fact of the matter is that the rich already live longer because they can also access better cancer meds, better, you know, doctors I do like preventative screens all the time.
2:09:31
The vast majority of Americans could never afford that, right?
2:09:35
Versus like a daily cheap pill that broadly extends your quality of life and broadly extends your lifespan.
2:09:41
It not only helps that individual to potentially prevent them or delay them developing cancer.
2:09:45
It also helps their children who aren't going to have to become caretakers before their own careers have taken off.
2:09:50
It means that they can have more flexibility in pursuing doing the next great thing.
2:09:53
I think actually like, you know, better health of the population is the one of the best great equalizers we could bring. Yeah.
2:10:01
How How closely does the Does your business track to what I'm loosely familiar with in the biotech world where it's like, uh, you're in the lab, you develop the drug, then you test it on like dog, monkey, mice, something like that.
2:10:13
You take that data through phase one, phase two.
2:10:16
It's a multi-year process. Yeah.
2:10:19
The company might IPO and then get bought by Merck or Pfizer at some point.
2:10:24
Uh, like like that's what I think of when I think of biotech like like yes, we've created a new drug for like this very specific cancer, but it's still a $10 million or $100 billion outcome because it like it fixed this one thing and it became very profitable.
2:10:38
Is that the typical path or is it wildly different in veterinary?
2:10:40
Dude, your biotech contacts is like on point.
2:10:41
I'm actually super impressed.
2:10:45
We had a bunch of biotech investors and founders on one day when Trump was doing the most favored nation thing.
2:10:49
We tried to get a tweet on that.
2:10:53
Uh, which you I don't really know where that went, but anyway, uh, just in terms of what I said.
2:10:57
>> 100% tariff on us and then that got Yeah. But that got taken away.
2:10:59
It's It's been really interesting. It's been exciting.
2:11:02
>> It's all a roller coaster this year. Perfect.
2:11:05
Uh, that's good for longevity.
2:11:07
My personal longevity is not benefiting from the bill.
2:11:09
Um, so the way to think about it is if if you're comparing to human is you do, um, maybe backing up a little bit.
2:11:16
The overarching themes of developing a dog drugs, it takes like four to seven years basically by combination of how effective and how lucky you are to get a dog drug approved.
2:11:25
Takes about 20 to 50 million dollars. Mhm.
2:11:27
Uh, for context in humans, it takes a decade to develop a human drug, takes about a billion dollars to develop a human drug.
2:11:33
So you're like orders of magnitude faster and cheaper.
2:11:36
>> Orders of magnitude faster and cheaper.
2:11:37
And then actually the coolest thing >> guys have been pretty lucky, right?
2:11:41
We've also been lucky, right?
2:11:41
And like totally Where did you catch lucky breaks? Explain.
2:11:45
>> god, like >> feel like I I I feel like, uh, the average when whenever you guys announce it like any type of milestone, the average investor is basically sharing something to the effect of you guys don't realize like how rare this is.
2:11:58
Like the speed at which you guys have been able to I mean, it's >> just like knock down these different milestones.
2:12:03
>> It's I mean, we've gotten the first ever, um, efficacy approval, uh, for a longevity drug.
2:12:07
We've created the first regulatory pathway for a drug to be approved for longevity for any species, right?
2:12:13
Obviously it's in dogs, but it's any species.
2:12:15
And dogs actually are they're they're regulated by the same federal regs as human drugs.
2:12:18
So it's a it was still a really huge lift.
2:12:20
We've gotten the we're running the first ever lifespan pivotal FDA enabling clinical trial.
2:12:26
Like all of these things are super super hard.
2:12:28
We got, um, obviously we really have a smart team, but yeah, you have to get really lucky, too, right?
2:12:34
We got lucky at the drugs.
2:12:34
When you pick a drug and you're like, this is going to be our lifespan extension drug.
2:12:39
It either is or isn't, right?
2:12:39
And there's nothing I can do as CEO to make this drug work if it doesn't work.
2:12:43
Well, you can try and take like 10 drugs through the pathway at the same time potentially. >> of what we're doing.
2:12:49
We have like four in development, but that's really expensive, >> So you have to you have to pick a couple candidates that have the most promise and luck comes in and then like you hope one of those is the one that you picked to do And you hope there isn't some like random toxicity that you find out at the 11th hour and that like totally nukes and especially with a longevity drug, there's zero tolerance for adverse events.
2:13:10
There's zero tolerance for me turning your dog >> it's not like you have a bad condition that you're trying to treat so some side effects or or the average patient would be okay with some side effects.
2:13:20
It's like you're the you're going from net neutral to it has to be Exactly. Exactly.
2:13:25
can't reduce quality of life Yeah. I mean, that much.
2:13:29
I mean, there's people like, um, what are some of the human, uh, the the drugs that human patients are taking?
2:13:35
I feel like Peter Attia and people are taking certain drugs that like >> Yeah, they like they they metformin like reduces your like athletic performance.
2:13:46
They all have some like like metformin is like works by improving your metabolic fitness, um, which, um, we actually have drugs that are actually our lead drug now is develop is improving metabolic fitness in senior dogs.
2:13:57
It's a very like robust way like if you've ever intermittent fast or calorically restricted, that's a way to improve your metabolic fitness.
2:14:02
Um, metformin just does that as a drug.
2:14:04
It's an old diabetes drug, but yeah, it has a bunch of like stomach side effects.
2:14:08
And also just when you give any drug for a long time, bad things can happen.
2:14:12
Yeah, there's also it can't if somebody were to reclassify it as a there's no incentive to reclassify metformin as a longevity drug because it's off patent, Exactly.
2:14:25
No one makes money off it.
2:14:27
Cuz I was cuz there's, uh, there's a lot of it's debated, right?
2:14:29
But there's a lot of evidence that like aspirin could could be potentially beneficial for lifespan, but nobody would do the work to reclassify it because it's off patent and you could you could go and you could spend a billion dollars proving that it works and then everybody could just make it. Yep. Yep.
2:14:45
And this this is like, um, a niche thing that actually >> you can't even market aspirin as No, no, it's not it's not allowable legally. No, no.
2:14:56
And this is the niche reason why loyal it made a lot of sense and why I was like, okay, I guess I'm going to go and do this is you're 100% you have this catch-22 when you're trying to develop a human longevity drug, which is you need something that is has a ton of safety both acutely, right?
2:15:08
So you took like five of them, nothing bad's going to happen, but also over a period of time. Yeah. Right?
2:15:14
If I give it to you for a decade, is it going to increase your risk of colon cancer or something?
2:15:16
The only way you know that is by giving it to that person for a decade.
2:15:20
By the time you have done that, the drug is generic.
2:15:24
One of the really cool things about animal health is that you actually the FDA tries to incentivize people to develop and prove that new drugs work for animal medicine.
2:15:31
So you actually can get 10 years exclusivity from the FDA. Yeah.
2:15:36
If you develop a novel, uh, a generic drug for a novel use case in dogs.
2:15:40
So we have actually both.
2:15:40
We have in-licensed human drugs, we have generic drugs, but it means that you are able to go in and say, look, this might not work, but it's not going to do any harm.
2:15:50
And that's so important when you're trying to create regulatory pathways, when you're trying to convince veterinarians to prescribe something that's never existed before, to convince dog owners to do something they've never done before, to be able to say wholeheartedly, there's decades of data and we don't think this drug's going to do any harm. That helps.
2:16:05
Are you guys getting, uh, feeling any acceleration from various AI, uh, models?
2:16:13
It seems like we have people on the show every week that are building AI for science, AI for drug development and foundation models dedicated to it and yet you're here, you haven't said AI once. I have not said AI once. And this was so funny. It was a meta point.
2:16:27
We were like working on like loyal narrative stuff and I was like, should I just like let go of my ego and like have an AI story?
2:16:34
And I was like they're like I anyhow, there's there's no AI in the loyal story, but like I really debated it because I was like, man, I wonder if this would like increase my valuation by like a factor of 10. What about ELA?
2:16:45
>> be good for dogs and that'd be good for horses and that'd be good for No, but it but that's saying something that you're actively doing all of this drug development and meanwhile there's companies coming out and they're making the claim at least to investors that they're going to be able to, you know, speed up drug development, do it more efficiently, etc.
2:17:05
So I mean, so here's the problem.
2:17:08
>> sure you've tried to figure out how do we how do we unlock value here?
2:17:09
Yeah, so here's the problem with AI for drug development is I'm like I'm not an AI person, but like I am like in terms of like technically trained, but I am, you know, sure the models are good enough that we gave them biological data, they could infer things that human scientists couldn't.
2:17:25
The problem is we don't have those data sets.
2:17:27
Like they they don't exist for a human.
2:17:30
I can't like say, okay, here's like your model and if I gave you the drug, how's it going to react in this part of your body, that part of your body, how's it going to like, uh, get reduced over time, how's it going to interact with this other drug you're on?
2:17:42
We just don't have those data sets.
2:17:44
Yeah, we saw this with AlphaFold.
2:17:45
Like AlphaFold came out, protein folding solved.
2:17:46
It turned the biotech markets didn't really move because it turns out that actually solving protein folding problems is like 0.
2:17:54
1% of the job to do to get a new drug to market.
2:17:57
You got to recruit patients and then monitor them and then talk to the FDA and do legal work. >> trials is hard work.
2:18:04
You know, like our our big longevity study that we're running in dogs right now, I think the biggest risk to it is actually operational. It's data quality.
2:18:11
It's dogs not dropping out.
2:18:11
It's veterinarians complying, right?
2:18:13
Like the biggest reason the study would fail is that we if we have to throw out a bunch of data in five years because it doesn't adhere to the standards that are necessary, right?
2:18:21
Um, so there's some interesting AI companies for like the beginnings of drug development.
2:18:25
Like Recursion Discovery is one, um, that's trying to do basically better antibody, uh, binding design. That's smart.
2:18:32
I think that makes a lot of sense.
2:18:36
But I mean, I love to be proven wrong, but I think there there hasn't been a like a moonshot for the the bio data gathering on the clinical side that would be necessary to like model you and see what works and what doesn't.
2:18:48
>> a good Yeah, it's it's not a business.
2:18:50
Yeah, it's not a business.
2:18:50
You're taking a Even if somebody was able to raise a billion dollars with the idea, it's like would they would they come to you and be like, "What do you need?"
2:18:58
And then they'd be like a services provider Yeah, and that's the hard thing.
2:19:01
Like so we're building all these biobanks out.
2:19:03
And so we're going to have, you know, potentially one of the best data sets of translational aging in a species that's the best model of humans that isn't humans.
2:19:12
And so maybe we'll be able to do it from then.
2:19:15
But like I would I would feel disingenuous for me to say like, "Yeah, that's a strategy."
2:19:19
Because it's like no one knows.
2:19:21
Like biology is like the ultimate black box and it's the ultimate humbler, honestly. >> What about peptides? Are dogs on Ozempic yet? Are dogs on Ozempic?
2:19:30
Actually, so this is my favorite fun fact.
2:19:33
Everyone is like, "Oh, are you just developing like Ozempic for dogs?" And the answer is no.
2:19:40
Because people don't want their dog to not be hungry. Think about it, right?
2:19:43
You go home, your dog's super excited to see you, jumps on you.
2:19:46
You're like, "Oh, my dog loves me so much. I'm the alpha male."
2:19:48
Your dog's probably hungry. Right?
2:19:51
And so when people give their dog Ozempic or Ozempic-like drugs It just takes all the joy The dog is like The dog becomes a cat.
2:20:01
No, this actually existed before.
2:20:01
I think about like my my most cherished memories with my dogs growing up.
2:20:04
It was like my dad, my brother, and I making scrambled eggs for the dogs.
2:20:08
It's like that was peak peak peak experience as like a, you know, 10-year-old. No, totally.
2:20:15
It's a it's not a not a good market.
2:20:18
our other drug that's working on metabolic fitness to extend lifespan, which is very similar to how Ozempic works, it is targeting, you know, improving metabolic fitness, but it's a we make sure explicitly that the mechanism that we're hitting doesn't impact appetite and doesn't impact like weight in general.
2:20:38
Any book recommendations for us?
2:20:41
>> That's cute that you think I have time to read as a Any movie recommendations? Muscle of Dogs? What you got for us?
2:20:45
You know, I I watched Love is Blind France last night.
2:20:51
>> Okay, Love is Blind France. I'm sorry.
2:20:53
I you know, I you know, I want to be one of those like CEOs I'm like, "These are the esoteric hyper-intellectual books I'm reading that like show that I'm smarter than you." What? No. There's no shade.
2:21:06
I was reading a book that there's only one copy.
2:21:08
It's thousands of years old.
2:21:10
It's a PDF you don't have access to.
2:21:14
>> It's English, but you wouldn't you wouldn't understand.
2:21:14
But like honestly, like Loyal's my all day, every day.
2:21:17
Like I don't have What's the general routine like at Loyal for you?
2:21:23
So we're actually fully remote.
2:21:25
Which I know, contrarian, contrarian.
2:21:28
But actually works well because we have to hire people We have to hire the best possible person for each skill set we need.
2:21:34
And often those people are not actually they're almost never in the Bay Area. I'm going to St. Louis after this.
2:21:39
That's where we have a lot of team members.
2:21:40
We have a lot of team members in Kansas.
2:21:42
We have them in Florida, right?
2:21:44
And so it you get so much more benefit from hiring the right person for the right role than you do saying I'm going to hire the second or third best person, but they're willing to move to SF. Right?
2:21:55
So for me, I actually just like live at coffee shops.
2:21:59
I sit out of some of our investors' offices, bring my dog.
2:22:03
Maybe one day whenever I'm able to fly private, I'll bring my dog. No.
2:22:11
No, he does have a cute dog though. Or his his partner does.
2:22:19
We're hitting the gong for me.
2:22:23
Currently in getting studied.
2:22:27
What What's the next big milestone that people should be looking out for? The next big milestone?
2:22:31
I mean, we're just trying to bring this damn drug to market.
2:22:34
Honestly, like that's a big one.
2:22:37
Hopefully in the next year or so. I mean, pray for us.
2:22:39
It's like the the hard stuff's behind us.
2:22:42
Um but there's a lot of hard stuff in front of us, too. So Well, good luck.
2:22:45
Caught between hard stuff.
2:22:48
As every entrepreneur is.
2:22:49
It's It's a worthwhile goal.
2:22:52
Anyways, well, I think we can close out the show with you cuz we actually have to hit the road.
2:22:57
>> we got to tell you about eightsleep. com.
2:23:00
Yeah, we're going to get a smart mattress.
2:23:03
I do have an eight sleep. Uh no.
2:23:13
That's going to [Applause] Thank you for tuning in.
2:23:21
The chat says, "When can we buy Loyal stock?"
2:23:23
I'm assuming you can't answer that.
2:23:27
>> We I think SEC something something. Okay. Sorry, guys. Sorry, guys. Private company.
2:23:32
Become a venture capitalist.
2:23:34
Someday in a very bright future, I'm sure.
2:23:38
Thank you for tuning in with us today.
2:23:41
We'll be back tomorrow morning. I can't wait. Have a great afternoon. Goodbye. Love you. Bye. [Applause]