0:00
You're watching TVPN.
Search this video
0:04
>> OpenAI has announced that they have won a long they achieved the long-standing grand challenge in AI gold medal level performance on the world's most prestigious math competition, the International Math Olympiad. That's the IMO.
0:19
So, this went up at 12:50 a. m. on July 19th.
0:26
>> Typical timeline for announcements.
0:26
So, this is this is basically like Friday night, Saturday morning.
0:30
You're looking at uh basically 1:00 a. m. from Alexander Wei.
0:35
Um, and he shares a picture of a strawberry with a metal with the OpenAI logo on it.
0:42
Someone also tested, they took this picture, they uploaded it to CHVC.
0:46
What's the fruit in the image?
0:46
How many Rs are in that fruit? And it nailed it.
0:51
So, we are truly we are truly AGI has arrived. Has arrived.
0:54
Um, also before we go deeper into that story, uh, we have a new partner, Reream. You saw it in the intro.
1:03
We're very excited to welcome them as a sponsor of the show.
1:06
We've been >> running on reream >> secretly.
1:09
>> Secretly, >> secretly >> since day one.
1:11
>> Since day one, it's been our secret to to success and it can be your secret to success. Yeah.
1:15
If you have >> support >> a lot of the Fortune 500 already for all their live streaming infrastructure.
1:20
So if you need to go live, go live with Reream.
1:24
And you see open AAI do streams like that that is a thing that has entered the standard um comm strategy.
1:32
>> I don't know if XAI is running on reream.
1:34
They had a little >> but they do but they do stream and uh you know it used to just be the the hyperscalers the mag 7 the big big companies that would do a live stream around a an annual uh event.
1:43
Then Brian Chesy came out and created founder mode and basically said put all your announcements on a on an annual release schedule.
1:52
uh bundle them all up, have the team celebrate, push to get across the finish line.
1:57
And part of that is let's throw an event.
2:00
Part of that is let's get the CEO and the product leaders on stage, tell the community about what we're building, what we've built, and now a lot of people are streaming it.
2:08
If you're going to stream your stuff, you need reream. Anyway, go check it out.
2:11
Um so Dylan Field um someone who's who has streamed events uh you know uh uh Figma obviously has config their dev event uh annually and they actually do two of them and those are streamed uh all over the place probably using reream.
2:26
Um and uh Dylan chimes in on the OpenAI news.
2:29
He says, "Congrats to the congrats 2025 IMO winners and participants including OpenAI who trained a generalpurpose reinforcement learning model and achieved IMO gold.
2:41
OpenAI team includes these two folks, Cheryl and Polomial Gnome.
2:46
Fun fact, Polomial also won the 2025 diplomacy world championship as a human.
2:52
Um, and so people are saying AGI is right around the corner.
2:54
Um AJ from semi analysis says, "What stands out to me?
2:59
Scaling RL on nonverifiable rewards likely via rubrics and LLM as judge thinking and reasoning for several hours at a time for a highly specific task.
3:08
This is what the $20,000 per >> month >> month model will look like.
3:12
Um and so there were a bunch of interesting things about this.
3:17
Apparently there are a few different ways to to tackle um math IMO level math problems.
3:23
one is using this this uh program called lean that is a formal math proof verifier and there's rumors that Google's building their system to leverage that a lot openai apparently went way way down just the textbased LLM reasoning path and had a lot of success there.
3:43
So, even though as we'll get into in this story, it feels like it's neck andneck.
3:47
Maybe Google was earlier and they just didn't release it faster and maybe this is a comm's thing.
3:54
There's a whole bunch of different stuff going back and forth.
3:55
Um, but potentially the more interesting thing is did they take different technical approaches and get similar results? Yeah.
4:03
If they're neck and neck, um, which one will scale better, which one will generalize better?
4:06
There's a whole bunch of different dis discussions there.
4:09
Um, do you want to get into some of the the the controversy or the push back?
4:14
Do you have do you have a post pulled up from someone saying about about Google's attempt?
4:19
>> Yeah, I mean the the the whole thing came down to timing.
4:21
Um Dennis over at Deep Mind uh was was pushing back a little bit which I can pull up.
4:28
Um >> Dennis from Deep Mind the co-founder and now at Google.
4:34
Um, yeah, it was interesting because the we we immediately jumped to Poly Market because this whole idea of a an an AI system uh beating the IMO.
4:41
We talked about this with Scott Woo from Cognition months ago back in April.
4:47
I I found the clip and and he said, "I would be very surprised if an if an a AI system this year does not um does not actually surpass uh the IMO."
4:58
He of course is an IMO gold medalist himself.
5:01
And so, uh, it was it was a big it was big to hear from him.
5:06
Now, the poly market had been sitting at like 20%.
5:08
And I think we're going to get into like the minutia of like what it means to truly win because the IMO uh like the the the group that actually puts on the competition has a different set of rules.
5:22
They don't they they put the questions out and anyone can go and try and do them.
5:25
But to actually be awarded a goal for an AI, they said it has to be an open-source model potentially.
5:31
it has to be released in this particular way.
5:33
And so, uh, OpenAI might have used their systems to solve the questions, which is super impressive, but they might not have checked every box to actually technically win the gold from the actual organization.
5:47
And so, um, Gnome says today, >> De so so he put out a post this morning saying official results are in.
5:55
Gemini achieved gold medal level in the international mathematical >> which is different than actually getting a gold medal medal. Correct.
6:03
>> He said an advanced version was able to solve five out of six problems. Incredible progress.
6:06
And there's a quote in here from the IMO president.
6:09
He said we can we can confirm that Google DeepMind has reached the much desired milestone earning 35 out of a possible 42 points. Yep. A gold medal scored.
6:19
Their solutions were astonishing in many respects. >> Yep.
6:22
IMO graders found them to be clear, precise, and most of them easy to follow.
6:26
You can imagine Juan is just like off on this insane.
6:30
>> There were some fascinating fascinating details around how they solve them.
6:34
>> Demis shared he said, "By the way, as an aside, we didn't announce on Friday because we respected the IMO board's original request that all AI labs share their results only after the official results had been verified by independent experts >> and the students had rightly received the acclamation they deserved."
6:48
And so, um, you can imagine somebody was saying I I don't know how real this was.
6:55
>> Somebody was saying that, >> uh, they they were making a claim that like Google needed time to like have the comms team to sign off, but it seemed, uh, more likely that, uh, the original plan was, hey, let's wait and announce this when when the IMO board >> actually comes out.
7:11
And >> the analogy feels like you're a super fast sprinter.
7:15
You go to the Olympics and you run a 100 meter dash.
7:17
that's incredibly fast in the parking lot.
7:19
Like you're not in the stadium, >> but everyone's like, "Wow, that guy's fast. >> He's hauling. >> He's hauling."
7:27
It's very impressive, but potentially the fastest man on earth.
7:31
>> And and ideally, you paid for the parking pass to get into the stadium.
7:37
And it it's like maybe Google and OpenAI were both running 100 meter dashes in the parking lot.
7:42
One of them had paid for like, you know, enough parking spaces to make it completely clean.
7:46
Open AAI was just kind of showing up with a buddy and being like, I'm sprinting.
7:52
That's that's the vibe I'm getting.
7:52
But they're both in the parking lot.
7:54
Like neither of them are in the actual stadium at this point.
7:58
But, you know, who knows? They they might be soon.
8:01
There's some questions about, you know, do they have to open source or whatever.
8:04
Um, so Gnome Brown says, uh, you know, we achieved a milestone that many considered years away. Gold level performance. Gold medal.
8:09
Of course, he's saying gold medal level performance.
8:13
Like that's very critical.
8:13
It's not we got a gold medal is that we we exhibited gold medal gold medal level performance.
8:20
Typically for these AI results like in go Dota poker diplomacy researchers spend years making an AI that masters one narrow do domain and does little else.
8:29
But this isn't an IMO specific model.
8:32
It's a reasoning LLM that incorporates new experimental generalpurpose techniques which would be very exciting because you could apply an IMO level.
8:40
>> John, hold up that stack of posts for the audience.
8:41
Oh, we got we got 150 posts today.
8:44
It's going to be a quick stream people.
8:45
It's going to be a quick stream, but buckle up.
8:50
>> Uh, so what's different?
8:50
We developed a new technique that makes LLMs a lot better at hard to verify tasks.
8:54
IMO problems were the perfect challenge for this.
8:58
Proofs are pages long and take experts hours to grade.
8:59
Compare that to aime where, which is another math exam, uh, where answers are simply an integer from 0 to 999.
9:07
So you can verify them really, really quickly.
9:09
Uh, also this model thinks for a long time.
9:11
01 thought for seconds, deep research for a minute.
9:15
This one thinks for hours.
9:15
Importantly, it's also more efficient with its thinking and there's a lot more room to push the test time, compute, and efficiency further.
9:22
Now, what's interesting is like these I think these I think these questions go up and then you have like like I think the students get like two four and a half hour sections segments.
9:31
So there's like there's a there's a world where like the the AIs can do it but just not as fast as humans which would be very interesting.
9:40
I don't exactly know how how close they are.
9:42
Um so Gnome says where does this go?
9:44
As fast as recent AI progress has been I fully expect the trend to continue.
9:48
Importantly I think we're close to AI substantially contributing to scientific discovery.
9:52
There's a big difference between AI slightly be below top human performance versus slightly above.
9:56
This was a small team effort.
9:59
He took >> right now the uh discovery that that AI is doing is talking with uh somebody who is potentially schizophrenic and convincing them that they figured out uh how to move faster than the speed of light.
10:16
>> Yeah, there's there's so many accounts of that.
10:19
It's it's it's getting crazy.
10:21
Uh yeah, I posted uh Chad GPT agent, go win me an IM gold medal and then update my resume. >> Don't make mistakes.
10:27
But uh yeah, I mean, have you actually looked at any of the IMO questions ever? >> No.
10:34
>> I don't even know where to start.
10:34
Like, it's it's all stuff that >> it's one call away for us, though. We just text Scott.
10:40
>> Tyler, have you ever looked at IMO questions? >> Uh, no.
10:43
But I'm looking at them right now. It's like pretty brutal.
10:46
>> Would you know where to >> Tyler?
10:47
If you can get two out of six right before the end of the stream, we'll buy you a house.
10:53
>> Well, that's the thing.
10:53
He's just going to be able to ask Chach and one shot it. Yeah. >> Yeah. Yeah, I had no tool. I should Yeah, no tools. Got to qualify that.
11:00
>> Uh, so yeah, Nick has the has the comment about DeepMind.
11:02
De Mind has also >> Tyler has another challenge today.
11:05
>> We will introduce that in just a few minutes.
11:07
Uh, De Mind has also won IMO gold, but they haven't announced it yet, by the way. Confirmed. Congratulations, B.
11:13
>> They did this morning. >> Uh, they did.
11:14
Yeah, they confirmed it this morning.
11:15
>> And so, it's been interesting to see the reaction to this.
11:17
Uh Gary Marcus is on X saying, "All the tech bros this morning thinking that AGI has been achieved because some uh parenthesis insanely expensive new form of LLMs can now match top high school students on one specific task. It's almost cute."
11:33
So really just ripping into uh the entire AI community.
11:40
But Will Depw comes in and says uh he's on leave right now uh taking a little summer holiday. Yes.
11:45
He says, "Guys, stop using expensive as a disqualifier.
11:50
Capability per dollar will drop 100x a year.
11:53
The 3K task ARC AGI 80% could probably be $30 if we cared to optimize it." >> Repeat after me.
12:01
All that matters is topline intelligence.
12:03
All that matters is topline intelligence. So, yeah.
12:07
Again, everybody's focused on sort of raw capabilities, not super focused on uh efficiency, especially for uh projects like this where they're not necessarily, you know, rolling this out at in in mass. >> Yep.
12:20
And you know where a lot of these IMO gold medalists go to work? >> I do, John. >> RAMP. com. Time is money. Save both.
12:24
Easy to use corporate cards, bill payments, accounting, and a whole lot more all in one place. >> Uh ramp is amazing.
12:30
Um, so the timeline was in turmoil over the Gary Marcus post and and they went back and forth and he finally admitted it at the end.
12:37
He said he said that's impressive.
12:40
Uh, so it started with a post by Daniel Lit.
12:43
He says, "Huge congrats to OpenAI for their IMO gold.
12:46
I don't find it too surprising that an AI tool was able to achieve this. See below."
12:50
Though I'd sort of lost hope the last few days, but I'm but I'm pretty surprised it was an LRM with no tool use.
12:56
So it didn't have Python.
12:59
It didn't have web search.
12:59
it didn't have all the things that you're that you could imagine would kind of allow you to to you know speed things up or kind of be a shortcut that would put it in a different category.
13:08
There was always a question about with the uh when when when the AIs were playing video games like like computers have wall hacks like they can sometimes like see the see the map without the fog of war.
13:20
they can see like it's very different to get to get perfect pixel perfect data on where every character is on the map and be able to make decisions based on that as opposed to like looking at pixels and having to like move the screen over there to see if you're being invaded on the left flank, right?
13:37
And so there was always this like yes, if you gave a computer like the raw access that's not quite it's impressive, but it's not quite a level playing field.
13:47
And it's basically the same thing as being like, well, the other folks don't get a calculator and you get a calculator.
13:53
Like, is it that impressive?
13:53
But this was very much an even playing field, it seems like. So, Mel Gibson 2.
13:57
0, great great name says, you are not surprised they that they you are not surprised that they have figured out a way to make the model learn in very hard to verify domains in a fuzzier reward space.
14:08
It also appears that the same reasoning model was the one used in the AT coder competition, showing that it can generalize across domains.
14:16
Daniel comes back and says, "I don't think we really know that what they figured out yet.
14:21
I'll keep my powder dry until we know more."
14:23
Uh, Mel Gibson says, "Go read Gnome Gnome Brown's thread on the model if you haven't already."
14:28
Um, he says, and remember, Nome Brown says, "This isn't an IMO specific model."
14:32
And that's a very important thing because um there's a ton of there's a ton of situations where you can go and RL on a specific task and it gets really really good at it, but then you try and get it to do anything else and it's not that great.
14:44
Uh and so Daniel Lit says, "Yeah, I read it."
14:46
Gary Marcus says chimes in from the rafters because he's not in this chat yet, but he has entered the chat.
14:51
He says, "I read it, too. That's pretty vague.
14:54
Are we sure that no tools were called?"
14:56
Uh Gary Marcus then chimes in again says reimo gold does no tools mean no use of Python code interpreter etc.
15:05
Dan says Daniel Lit says that's how I understand it and then Gary Marcus tags in polomial gnome brown and also Alex Wagan says can you confirm that the IMO goal was achieved without using Python or code interpreter or similar.
15:19
Uh and then Gary Marcus says humans don't use external tools in the IMO competition.
15:24
I'm just under I'm just trying to understand what the system is.
15:27
That's how we do science.
15:27
Daniel Lit and at some point whoever was trying to dunk on Gary Marcus just deleted their account.
15:33
So it just says this post is from an account that no longer exists.
15:37
>> Uh Daniel Lit shows Cheryl on the OpenAI team saying the the model solves these problems without tools like lean, which is a math verifier or coding.
15:44
It just uses natural language.
15:47
It also only has four and a half hours.
15:49
To answer our earlier question, we see the model reason at a very high level trying out different strategies, making observations from examples and testing hypothesis hypothesis.
15:57
And Gary Marcus says, "That's impressive."
16:02
So Gary Marcus, you got to fill out the the apology form, bro.
16:05
You got to fill out the deep reinforcement learning apology form.
16:09
What was the reason for your behavior?
16:11
No one told me Alex Weey was training the model.
16:15
>> Mercury was in retrograde. >> Mercury. I don't know. ML.
16:16
And then one of them is Gary Marcus con convinced me it was fake.
16:20
Uh because Gary Marcus was uh famously said deep learning is hitting a wall like like this this particular paradigm will not scale.
16:26
You have to do uh symbol manipulation which was like kind of his bet encode different relationships between ideas in the model.
16:35
Uh there were a bunch of different debates over there.
16:37
Uh and he's gone back and forth on that.
16:38
He's he's he's not as much of like a deep re deep reinforcement learning hater as pe some people think but he's got he that's his brand at this point.
16:46
So anyway, uh, more on the drama between DeepMind and OpenAI.
16:50
Deep Mind got a gold medal at the IMO on Friday afternoon, but they had to wait for marketing to approve the tweet until Monday.
16:55
Open AAI shared theirs first at 1 a. m.
16:58
on Saturday and stole the spotlight in this game.
17:00
Speed is greater than bureaucracy.
17:02
Miss the moment, lose the narrative.
17:04
So that's one interesting take. Let's see. Exactly.
17:08
Yeah, I I think there's a little bit of that.
17:10
little bit of that. I mean certainly culturally open AI loves to release information before Google like whatever Google's like next big >> and they're playing to win remember they after what did they release after the deepseek moment >> deep search you're typing deep and you
17:26
were looking for an AI product >> and then and then Google had IO and then and then OpenAI >> open AI >> IO this like the day before >> with uh with Johnny IV >> with Johnny IV um so anyway you know you'll love to see it all fair and love and war I guess. Um I I don't think this
17:42
Um I I don't think this crosses >> any warfare and this is the benefit of having a bunch of posters on your team I guess.
17:50
Um no brand says it takes us a few months to turn the experimental research frontier into a product but progress is so fast that a few months can mean a big difference in capabilities and this is from July 18th.
18:03
So all the models underperform humans on the new IMO questions and Grock 4 is especially bad on it even with best of end selection under is accusing Gro 4 of training on the problem set >> for for IMO.
18:17
>> No no for something else for oh for the other benchmarks.
18:18
I mean benchmark hacking is like a thing that happens all over the place.
18:21
That's why we need like hidden benchmarks.
18:23
That's the whole thing of of ARGI that we will talk about.
18:27
We talked about Will Depw saying guys stop using expensive as a disqualifier.
18:30
Uh Terrence Tao uh one of the one of the most goatated mathematicians of all time uh digs in a little bit and is >> is uh adding his perspective as a as a worldrenowned mathematician.
18:43
He says it is tempting to view the capability of current AI technology as a singular quantity.
18:50
Either a given task X is within the ability of current tools or it is not.
18:54
However, there is a there is in fact a very widespread in capability several orders of magnitude depending on what resources and and assistance gives the tool and how one reports their results.
19:05
One can illustrate this with a human metaphor.
19:08
I will use the recently concluded IMO as an example.
19:10
Here the format is that each country fields a team of six human contestants who are high school students led by a team leader often a professional mathematician.
19:20
Over the course of two days, each contestant is given four and a half hours on each day to solve three difficult math problems given only pen and paper. Crazy.
19:34
Uh, no communication between contestants or with the team leader during this period is permitted.
19:38
Although the contestants can ask the in invig invigators in I don't know for clarification.
19:45
This is like some math term I don't even know.
19:47
Uh for clarification on the wording of the problems, the team leader advocates for the students in front of the IMO jury during the grade.
19:55
>> Invigilators are also known as exam proctors. >> Okay, proctors.
19:59
Okay, so yeah, they can ask the people that are running the test, hey, >> we got to start working invigilator and we should probably have a show like just an invigilator too. >> For sure.
20:10
>> We should have at least one on staff.
20:10
So uh so so the team leader advocates for the uh for the students but is not involved in the IMO examination directly.
20:18
The IMO is widely regarded as highly se as a highly selective measure of mathematical achievement for a high school student to be able to score well enough to achieve a medal particularly a gold medal or a perfect score.
20:27
This year the threshold for gold was 35 out of 42 which corresponds to answering five of the six questions perfectly.
20:33
Even answering one question perfectly >> a basically a B yeah >> gets you gold.
20:42
>> Yeah, >> it's that hard. >> They're that hard.
20:43
But consider what happens to the diffic to the difficulty of the Olympiad if we alter the format in various ways.
20:50
>> First one gives the students several days to complete each question rather than four and a half hours for three scenarios.
20:56
To stretch the metaphor somewhat, consider a sci-fi scenario in in the student in which the student is still only given four and a half hours, but the team leader places the students in some sort of expensive and energyintensive time acceleration machine in which months or even years of time pass for the students during this exam before the exam.
21:14
>> Ben might have put us in one of these because we go live and then it's four hours later. >> Time acceleration. >> Yeah, it's wild.
21:21
Uh, two, uh, before the exam starts, the team leader rewrites the questions in a format the students find easier to work with.
21:28
Three, the team leader gives the students unlimited access to calculators, computer algebra packages, formal proof assistance, textbooks, or the ability to search the internet.
21:37
Sounds like that didn't happen.
21:38
Uh, the team leader has the six student team work on the same problem simultaneously, communicating with each other on their partial progress and reported dead ends.
21:46
The team leader gives the students prompts in the direction of favorable approaches and intervenes if one of the students is spending too much time on a direction they know to be unlikely to succeed.
21:54
Um, next, each of the six students on the team submits solutions, but the team leader selects only the best solution to submit to the competition regard discarding the rest.
22:04
Uh last, if none of the students on the team obtain a satisfactory solution, the team leader does not submit any solution at all and silently withdraws from the competition without their participation ever being noted.
22:16
Oh, so that I mean that could have happened.
22:18
It didn't in this case, but uh in each of these formats, the submitted solutions are still technically generated by high school >> contestants.
22:23
Going back to the parking lot example, it's like if you can go run go run the race in the parking lot and then uh if you don't run as fast as you hoped, it's like well I wasn't in the I mean >> I wasn't racing.
22:36
I was just going for jog.
22:39
>> I was just warming up.
22:39
I just I like to be surrounded by excellence. >> Yeah, don't worry.
22:42
Yeah, I mean that is the true one because if they'd missed they'd probably be like what what the IMO is this weekend? We're working on agents.
22:51
Like what are you talking about?
22:53
like have you seen our DAUs? Like get out of here.
22:56
Um to each of these formats, the submitted solutions are still technically generated by the high school stu contestants rather than the team leader.
23:03
However, the reported success rate of the students in the competition can be dramatically affected by such changes in format.
23:08
A student or team of students who might not even reach bronze medal performance um in the normal competition under standard test conditions might instead reach gold medal performance under some of the modified formats indicated above.
23:19
So in the absence of controlled test methodology that was not self- selected by the competing teams, one should be wary of making various apples to apples comparisons between the performance of various AI models on competitions such as the IMO or between such models and the human contestants.
23:34
Yeah, the question is like it seems like of those not many were actually violated by the way OpenAI and Google attacked this.
23:42
>> Well, this was before I mean this this was an immediate reaction before I think there was >> more details came out of details. Yeah.
23:49
That we've been covering. 3 p. m.
23:49
on on so this so on Saturday but 3 p. m. later.
23:55
>> Well, Run has a good post.
23:55
He says, "My bar for AGI is an AI that can learn to run a gas station for a year without a team of scientists collecting the gas station data set." >> It's a great post.
24:09
>> The world isn't ready for gas station.
24:11
>> We can we can run an AI as long as we can we can run a gas station as long as we have, you know, a perfect, you know, data set.
24:17
And that's yeah that's like the way >> on how gas stations operate. >> Yeah.
24:22
As soon as we as soon as we uh place the goals somewhere we nail that goal and then we have to move the goalpost shows.
24:28
Uh >> this is this is a great picture.
24:30
So Thomas Wolf says my bar for AGI is an AI winning a Nobel Prize for a new theory.
24:37
It originated just image of a team moving the goalpost.
24:42
>> I do like how consistent Tyler Cowan is on AGI.
24:45
He's like, "My bar was the touring test. We passed that.
24:48
So, I'm I have to call it like it is." >> Yeah.
24:52
>> And I'm not moving the goal.
24:54
>> I think I think he's very smart for that fullness of time.
24:58
>> I mean, I I think it's okay to be like, "Yeah, we achieved AGI."
25:01
>> It's just tough when you now there's something new.
25:04
>> It's cool, but not immediately >> dramatically transformative.
25:07
>> dramatically transformative. But you have to imagine that that was the same that there was the same thing when people were like human human flight and then the Wright brothers go and do it and people are like great so like I can hop on a Southwest flight in an hour to go to San Francisco from LA and they're
25:21
like what are you talking about like LA is like you know a couple trains and like some people you know with orange groves and it's like it takes a long time for the infrastructure to get built out like now a lot of these AI tools they do build on top of the internet and on top of technology that we have rolled out. But like at a certain point like
25:38
But like at a certain point like the UI matters, the economics of these different tools matters.
25:44
You like you can't it might not be economical for a for a company to to you know release a an AI tool that costs $20,000 to inference just to get you the weather or do two plus two.
25:58
And so um all of these things take time to roll out and that's why we're kind of in this like slow takeoff scenario.
26:04
It feels like I don't know massive post from Fiji Simo the CEO of applications at OpenAI.
26:10
She is outlining um what her plan is basically and it's very very interesting post.
26:17
So um in uh so Fiji uh is an absolute boardroom general.
26:24
Give me the Ashton Hall whenart. >> Yes, >> PGCMO.
26:31
Uh, OpenAI and Spotify board of directors positions, huge runs at eBay, Facebook, and Instacart.
26:37
Now, she's the CEO of applications at at OpenAI.
26:39
Uh, today she outlined how she sees OpenAI's products having the biggest impacts. High level, six things.
26:47
Knowledge retrieval, health, creative expression, economic freedom, time, and support.
26:52
And those are kind of getting, I think, more abstract, more vague as they go along.
26:59
But it's interesting, they map pretty closely to my experience.
27:01
Knowledge retrieval, I use I use it as a Google replacement for a lot of things.
27:06
I use deep research as a as a knowledge retrieval product.
27:10
Uh I also use it a little bit as web replacement.
27:13
Not a ton, but I've seen a lot of people use that.
27:16
But I do ask it for advice around health, fitness, recommendations for supplements, all sorts of different stuff.
27:23
Often I'll just go there and say, "What is and what what type of creatine does Andrew Huberman recommend?"
27:27
And it'll just >> complex med medical.
27:33
>> Fortunately, I haven't been in one of those situations recently, but I if if I did run into one of those situations, I would definitely go to it.
27:39
would definitely go to it. It is it is interesting because uh previously when people were trying to understand maybe a small health issue they're having >> they would go do a Google search addit to it and then they would look through
27:52
the comments and and the comments would provide kind of like context on the symptoms or whatever they were dealing with but you and you could technically leave a comment but I think the majority of people wouldn't wouldn't want to post about that. So, the fact that
28:04
So, the fact that >> OpenAI has all that Reddit data, you can get that same type of data, but then you can kind of ask a number of follow-up questions, um, >> which is, uh, very cool, very powerful.
28:16
I'm I'm certain that people are >> probably misusing it, maybe reading too much into the results, but again, um, the idea, I think for a long time, >> doctors that would say, >> don't don't look on Google, you're just going to stress yourself out.
28:29
And then you go into the appointment room and they're over there on on the equivalent of Dr. Google. >> Totally. It's nothing new.
28:36
People used to go to WebMD and like the meme was like you go to WebMD and no matter what you type in it says like could be cancer.
28:46
>> Like could be like the worst possible thing, but it might just be a cold and maybe you should just take some Advil.
28:51
But like for whatever reason, WebMD felt the need to like kind of give you the full picture and all the possibilities and then they'd of course say like go consult an expert or talk to your doctor.
29:00
So, um, but yes, I I definitely see it it it replacing that type of health experience.
29:05
Uh, also I'm using it as a Photoshop replacement a fair amount of time if I need to uh just quickly generate an image quickly generate something that I would have normally gone to a 3D rendering program or or Photoshop and kind of created a collage to kind of >> tailwind for the meme industrial complex. >> Huge tailwind.
29:24
Um, still a lot to work out there.
29:26
I do find that sometimes when I go to images and chatt I'm going back and forth on the prompt for 10 minutes and I'm like I could have just done this with traditional tools.
29:36
Uh so it's not perfect but when it gets it it's so amazing and the final product often looks a lot more cohesive than what I would get if I was like collaging and like the shadows didn't match and there were like rough edges and stuff.
29:47
So certainly great there.
29:49
The last three are a bit more vague.
29:51
We'll have to read into those.
29:52
But OpenAI is set up to win pretty big in all of these categories.
29:56
Uh so semi analysis clocks chat GPT's share of queries at 71%.
30:02
Meta is in second at 12% and that's with billions of users on meta products.
30:06
And so uh OpenAI is certainly running away with consumer.
30:12
Um, so there's this question of like the it feels like the rest of consumer AI plays will be I would say like horizontal in the sense that it won't be a new app on your home screen.
30:25
Um, Meta won't have Meta AI as a home screen app >> loses their lead and becomes the Yahoo of AI.
30:34
But that doesn't feel likely. >> Yeah, it feels tough.
30:37
Like why did Yahoo lose to Google?
30:41
It's because there was an entirely new paradigm in page rank like a new algorithm emerged.
30:47
>> Sure, if that if that happens from SSI or super intelligence, meta super intelligence, like there could be a leaprog moment, but so far they've been pretty good at staying near the frontier to the point where it's like, okay, Google got them on this one or Grock got them on this one today.
31:02
But there hasn't been a time when I've been like, okay, like there is a dramatic difference in the results such that it's worth it for me to go over.
31:12
There was that moment for Claude for a little bit.
31:14
Everyone was like, Claude's way better.
31:15
I went over and used Claude for a bit and then I went back to OpenAI just because the product was better and and had more of an ecosystem.
31:21
Um, >> product lead is very real.
31:22
Yeah, I was I was thinking about this like if if if you played out a hypothetical scenario and you went and you you had you know the keys to OpenAI HQ, you went and you excfiltrated all the code, everything got an app into the app store, Jordy GPT, you have the exact same model, the exact same inference, the exact same UI, everything. Could you win?
31:51
Could you win? like because people would say well it's >> real brand there's we we live on a part of the internet where people are constantly debating the merits and of the leadership teams and the approaches and the philosophies
32:05
>> but then downstream you have approaching a billion users that uh are just like yeah I like I like the product I use it a lot >> yeah and so even if you ran a Super Bowl ad saying like we have a product that's exactly as good as chat GPT it is feature for feature. People would be
32:21
People would be like, "But I already have it installed.
32:25
You're asking me for two minutes to uninstall this and reinstall the new one and then I don't get any benefit."
32:29
I wouldn't do it even if you match my even even if you match my capabilities.
32:32
You got a leap from >> It's funny.
32:35
So, so Jason Frerieded had a great post uh relevant yesterday.
32:37
He said, "Feature parody is just another way to say we don't need to exist."
32:41
So if you come out and you're so good at what you do that you can create as as good of a product as OpenAI models that are as good, >> you still don't need to exist.
32:52
And so uh that is the challenge for anyone competing.
32:59
>> If you Let me tell you about ramp. com. Time is money save both.
33:01
Easy to use corporate cards, bill payments, accounting and a whole lot more all in one place. Go to ramp. com.
33:04
Um really quickly uh so uh it feels like the rest of consumer AI plays will be horizontal.
33:12
There won't be a new AI app for Meta that really takes off.
33:14
Instead, AI will improve everything Meta does across Facebook, Instagram, WhatsApp, Quest, etc.
33:21
>> And I was thinking about the initial takeoff of Facebook.
33:23
The other tech companies, they didn't really seriously get to compete in social networking.
33:31
Like Plus never really took off.
33:31
I guess Microsoft bought LinkedIn and that's kind of a niche uh social network.
33:36
Very profitable business, great business, but it's it is like its own thing.
33:40
it doesn't directly compete.
33:41
Um, but graph databases and the idea of storing efficiently connections in in graph networks like Facebook kind of pioneered that became really widespread and was used all over the place. Yeah.
33:56
And the same thing with Google search like this this page rank just better search algorithms those wound up manifesting in better search in all sorts of products.
34:07
And so when when a new company comes out and builds, you know, a a whole paradigm or they break through in terms of a front-end development or database development or some sort of structure, like it winds up it winds up improving everyone.
34:21
And so I think I think AI will have these like horizontal benefits all over the ecosystem.
34:26
I think it makes a ton of sense that that Meta is investing so much in super intelligence, but it's not I don't really think the narrative should be like they need to play catchup to chat GPT.
34:37
I think it's more like they just need to implement LLMs in every single crack in their all of their different systems like they do with great databases, great infrastructure. I'm not sure. We'll see.
34:49
I still think there's I still think there's I mean remember Meta made some early experiments of making digital clones of big celebrities on Instagram.
34:57
I still think they'll take more shots on goal around companionship.
35:02
Will they go Grock waifu mode? >> Uh no.
35:05
I don't think I don't think they want >> uh that revenue line.
35:08
>> uh that revenue line. uh but but there's a bunch of other uh ways in which they could kind of go after that market that I don't even think I mean Chad GPT is is is such an interesting product because it's designed to be a functional tool
35:23
around knowledge retrieval creative expression >> uh just unlocking consumer surplus giving people access to expertise etc and the the way in which people are using it as a companion is sort of not the sort of default it's Not like you name your chat GPT, right? People will
35:40
People will give it a name. Yeah.
35:43
>> But um it >> doesn't really stick.
35:45
>> It's not it's certainly not baked into the UI in the way that it could be. Yeah.
35:48
>> Um I know people do talk to I've said it before, but as as Chad GPT user hours or just uh or sort of user minutes still, but as they're ticking up, I do think >> Zuck will look at that and say, I want I want some of that action. >> Yeah. Yeah.
36:04
and and and maybe some of that but I still I still think some of that happens inside of the platforms.
36:08
I don't I don't >> necessarily in that new app >> it would be very difficult and and Meta has yet to really do it.
36:16
WhatsApp an acquisition, Instagram an acquisition um but stories massively successful within Instagram and so I would expect I I agree with your take that I think um there's an opportunity for for a lot of AI stuff to live within those apps.
36:31
Um, and then also just on the monetization side, like they're in a very unique space where they can give great products away for free, but they can't be charged an inference cost to another company.
36:43
And so if they have their own open source model that they can inference very cheaply, they can give it away for for free and then they leverage their massive ad network.
36:50
Um, so let's go through what uh what Fiji over at OpenAI is building.
36:55
So a few in a few weeks I'll be joining OpenAI as CEO of applications.
36:59
We've all been waiting for this moment.
37:00
uh helping to get open AIS technology into the hands of more people around the world.
37:05
I've always considered myself a pragmatic technologist, someone who loves technology, not just for its own sake, but for the direct impact it can have on people's lives.
37:12
That's what makes this job exciting since I believe AI will unlock more opportunities for more people than any other technology in history.
37:20
If we get this right, AI can give everyone more power.
37:22
Um and so uh she has a fantastic career eBay, Facebook, then Instacart um and is now at OpenAI >> and she was heavily involved in building the ads engine at Instacart. Correct. >> Yeah. >> Yeah.
37:39
>> Um and obviously at Facebook as well.
37:39
So in on knowledge, she says empowerment starts with understanding the world around us and our place in it.
37:44
When we have the right knowledge at the right time, we can make better decisions, advocate for ourselves, and change our path.
37:51
But for most of history, access to expert level knowledge has been limited to those with more resources.
37:55
I mean, even even after the invention of the printing press, like you still had to be in a big city that had a library to go check out a book for free.
38:02
Still had to have the time to do it.
38:04
Uh and um and certainly buying every book and being able to index it uh was very expensive and difficult. Now it's even easier. It's already working.
38:12
People who use AI tutors learn twice as much as they do from human ones.
38:16
And the gains are even bigger compared to learning in a traditional classroom.
38:19
In a 2024 OpenAI study, 90% of users said chat GPT helped them understand complex ideas more easily. I agree with that.
38:26
On health, she says, "Personally, I'm most excited about the breakthroughs that AI will generate in healthcare.
38:33
A few years ago, I faced a complex and poorly understood chronic illness." Oh, that sound.
38:38
Uh, and it became painfully clear just how fragmented and inaccessible the health care system can be.
38:43
Even with access from some of the best doctors in the world, I found myself acting as a connector, piecing together insights from multiple specialists who weren't speaking to each other.
38:51
I actually had a friend who uh got a very very rare form of cancer and did all of the research pre-chubt.
38:57
Actually downloaded all the frontier science in all the different papers, figured out that it was this one very rare condition, found the expert, went to that expert.
39:08
He was like, "Yep, you have this thing.
39:09
It's the I'm the only one that studies this thing. I'm going to help you."
39:12
Did the surgery save your life? Crazy.
39:15
>> I I had an I had an issue.
39:15
I was I was 18 uh traveling.
39:18
I was uh surfing in South America.
39:21
There was flash flooding. I was in this tiny town.
39:23
I got a uh antibiotic resistant staff infection from that. >> That's bad.
39:29
>> And then was getting uh repeated infections over the next year >> and the doctor kept prescribing um >> antibiotics.
39:37
antibiotics that uh were so I mean so such an intense antibiotic that doctors won't it's like a soap they won't touch it without gloves on and they use it pre before surgeries and they just said use this twice a day over your entire body and I kept having I kept having issues and eventually found a Reddit community that was like just like obsessed over this issue and and the thing that fixed it was avoiding uh gluten. >> Yeah.
40:04
>> It like actually just fixed everything. That's so crazy.
40:07
>> And I just imagine if I was like even now if I was dealing with that issue, I would have been able to um I would have just been able to like talk with chat GPT about it.
40:15
They would be able to surface all that kind of thing.
40:17
Build a build a picture of it. So >> very interesting.
40:21
Um well uh let me tell you about reream.
40:23
One live stream 30 plus destinations multiream and reach your audience wherever they are.
40:28
We use >> the backbone of this show.
40:29
So she uh closes out health saying AI can explain lab results, decode medical jargon, offer second opinions, and help patients understand their options in plain language.
40:39
It won't won't replace doctors, but it can finally level the playing field for patients, putting them in the driver's seat of their own care.
40:46
Uh very excited for that.
40:46
Uh very excited for that. And also that feels like something I don't know like like yeah just uh like I don't know it'll be interesting to see if that manifests itself in like all of this is like will there be any fracturing in the actual product because right now all the
41:03
fracturing that's happening in the chat GPT product is in like what model you're using you're using agent or deep research but so far all the different products yeah they're all chats and and and and none of them say this one's poor health this one and I'm in like a health a safe health territory. It's like
41:19
It's like that's all driven by the prompt.
41:21
I would imagine that over the long term everything's driven by the prompt.
41:24
I would love to be able to go to chat and just say hey quickly what's the population of Canada and it knows to use 40 for that and then if I say hey I want you to give me a full report on the history of Canada.
41:37
Give me a deep research report.
41:39
I wouldn't need to select it.
41:41
It would just know from the process.
41:42
process. you go to chatgd agent help me annex Canada >> and then it just goes and starts yeah don't mistake and then and then it just keeps working um so the third one she says creative expression I believe we're all born creators and that the ability to imagine something and make it real is a big part of what makes us human the
41:59
problem is that our ability to express that creativity is often limited by our skill sets completely agree I can't draw at all uh not everyone has the resources time or training to paint write compose or build when I imagine the future it often comes to me an image I paint in my spare time, but the we got to find some Fiji Simo originals. But don't hang them
42:15
But don't hang them on the wall. This is interesting. >> It's lore.
42:19
Uh, but the images in my head are much more realistic and complex than what I am able to paint today.
42:24
Now AI is collapsing the distance between imagination and execution.
42:26
With AI and image generation, I can prompt and iterate until the output matches the complexity and realism of the vision in my head.
42:33
Unless it's a Where's Waldo?
42:33
We got to solve the Where's Waldo challenge? >> We do.
42:38
>> Uh, it's coming for sure.
42:38
the where the where's Waldo eval for image generation.
42:43
Uh today nearly one in three gen Z users say AI tools have helped them express themselves in ways they never could.
42:47
So this is another fun viral obvious uh valuable like we know it, we love it, everyone's using um chat GPT for creative expression. Well, yeah.
42:58
And this presents a massive challenge if you are a a chat GP uh a GPT rapper that's trying to go compete in the creative expression space because you have to you it's not enough to generate a cool image or gen generate a meme.
43:16
>> Uh it used to be that uh if you could do text well that was maybe an advantage that people would use your product but >> uh chat GPT uh image generation has just gotten better and better and better. Um it's fantastic.
43:27
Uh uh if you haven't played around with it much, drop uh your your group chat uh the the profile pictures of your most popular group chat, put them in there and say make them all gigachads. Yes. And then you can share. >> Yeah, for sure.
43:43
>> Um >> uh well, you start with generating an image in chat GPT.
43:46
Then when you're ready to go pro, you move over to Figma. Figma. com.
43:50
Think bigger, build faster.
43:52
Figma helps design and development teams build great products together.
43:53
You can get started for free at figma. com.
43:55
Uh, fourth, she says, "Economic freedom.
44:00
When people can independently create and capture value, they gain power over their own economic destiny.
44:03
But starting a company isn't easy.
44:05
The average cost to start a small business in the US is around $30,000, an impossible threshold for most aspiring entrepreneurs.
44:11
And until recently, building a product or launching a service required technical knowledge, especially coding.
44:15
That was a problem for hundreds of millions of people who had ideas for tools, apps, platforms, etc.
44:21
that could have made an impact but didn't they didn't have the technical skills to bring them to life.
44:27
The classic ideas guy, I just need a programmer to build it for me.
44:29
Well, now you can vibe code it.
44:31
Um AI now gives the people the uh the power to turn ideas into income no matter their age, credentials or zip code.
44:38
A single person can now brainstorm, prototype, market, and launch a product with tools they control themselves.
44:42
A a 2024 Shopify report showed AI enabled soloreneurs launched businesses 70% faster than peers without a tool without AI tools.
44:53
I've seen it with my nine-year-old daughter who decided one day she wanted to be be a party planner for kids birthdays.
44:59
>> Let's give it up for >> it's amazing. What a great story.
45:00
Uh in one weekend using AI tools, she created a fully functional website showcasing her party ideas, shared it with her peers and started taking on clients.
45:12
Amazingly, my husband and I didn't have to help her, but we did have to intervene before the confetti cannons were ordered. >> Ben, good reminder.
45:19
We should get some confetti cannons.
45:22
>> We do need >> Except last time when we had confetti cannons at YC demo day. >> Yes.
45:27
>> My eyes were in pain for days and days and days.
45:30
We need goggles next time because Yeah, it was there was so much dust in the in the confetti cannon.
45:36
>> We left it all on the field. >> We did. We did.
45:37
Um, in the future, people will be able to build new things without waiting for permission, capital, or credentials.
45:42
This will, of course, will mean a meaningful shift in the workforce.
45:46
Companies will hire fewer people.
45:47
So, you get more smaller companies. That seems pretty cool.
45:52
>> Yeah, my my first real company skateboard company, Jman Designs, >> probably.
45:56
I I don't know if I would have done it if my my lovely mother wasn't a graphic designer and I needed a logo to print on the decks that I was getting made.
46:05
And I just would sit with her and tell her, you know, do this, do that, do this, do that.
46:10
And um yeah, it's pretty cool that you can now just talk through that process in natural language and and >> yeah, >> how many more people can do that?
46:19
>> It was pretty crazy you ran a skateboard company without doing proper sock 2 compliance.
46:22
I know >> you should have got on Vanta.
46:23
Automate compliance, manage risk, prove trust continuously.
46:26
Vant's trust management platform.
46:28
Takes the manual work out of your security and compliance process.
46:31
huge alpha just starting starting an extreme sports company and saying look none of our competitors are serious enough about this to get to really serious >> to really take compliance and security >> as seriously as you should.
46:44
>> Um yeah I mean I I I do think for for kids especially I mean the the the ability to um just just use AI tools fully to go and solve problems and it's almost like your your town's McKenzie consultant.
46:59
I remember I had an early job of just like scanning photos for someone.
47:03
They needed a whole bunch of photos scanned.
47:04
It was like pretty manual labor, but like I I I actually wasn't that good at it. I kind of messed it up.
47:09
There was like too much like the the the scanner wasn't configured properly.
47:13
So, a lot of So, I had to redo a lot of them.
47:15
It was a learning it was a learning process, but it's probably the first uh one of the first small.
47:18
Uh the the only thing that that's maybe boring and repetitive that that you uh that that I can imagine you doing is uh the bench press. Oh yeah.
47:30
>> Should basically the same every time.
47:32
But but uh >> but yeah th those type of tasks I've always uh struggled with.
47:35
But >> I don't know that we're at a point.
47:38
I still think if you were using I I don't think that I think that job still exists. >> I think it does.
47:45
I think there are bigger companies where you can just like send in a ton of stuff and they'll just professionally scan it for you, but they're pretty expensive, I think.
47:51
So, I think the kids still have some alpha if your if your uh if your opportunity cost is low enough >> as as many kids are.
47:59
Um >> uh fifth, she says, "Time.
48:03
Regaining control over your time is one of the most liberating, empowering ships a person can experience.
48:08
The ability to control how you spend your time is often what separates people who feel in charge of their lives from people who feel overwhelmed by them.
48:15
Wealthy people who have always bought back their time by hiring personal assistants, household staff, private tutors, chefs, and more, building full infrastructures to reduce friction in their lives.
48:25
Meanwhile, the average household spends nearly 20 hours a week on domestic work, logistics, and errands.
48:30
While leading Instacart, I saw firsthand how technology can shift perceptions and behaviors around time.
48:36
In 2012, the idea of paying someone to shop for your groceries felt like a luxury, something reserved for the ultra wealthy.
48:41
But with the right product design, logistics, and pricing, we made it accessible and indispensable for everyday families.
48:47
Today, the Instacart user base mirrors the US population with millions of families getting hours back each week to spend on higher value activities.
48:55
The question is, will OpenAI build a uh gig worker network where when your agent runs into a wall and needs somebody in the real world to accomplish a task, will you be able to delegate tasks like that?
49:09
>> Feels like a stretch, but >> I don't know.
49:12
They could plug into the Uber network or any of these other networks.
49:15
Yeah, maybe they'll I mean they already have like an Instacart integration.
49:18
So maybe they will they will rely on other gig work platforms to solve those problems and actually handle that side >> and they will stay in the full >> like the package package delivery where you can get an Uber to just pick something up and take it across town.
49:35
You could imagine that that being integrated. >> Yeah.
49:37
I mean there's there there's a host of of of you know services and companies and APIs for shipping, mailing, moving things around, hiring someone.
49:45
I mean in theory you could you could hire a personal chef and be communicate.
49:51
your agent could be emailing with them and say, "Hey, you know, be here at this time. There's this party.
49:55
Here's" and it's kind of, you know, it doesn't need to be it doesn't even you don't even maybe need a gig work platform because the agent can go out and and scroll Instagram in theory and find someone who's advertising that service or Google it and find a listing of local people providing that t that that job and and and go and retain their services directly. Um, interesting.
50:16
I believe AI will allow for a similar shift in many areas of life.
50:22
Time-consuming activities like researching decisions, planning vacations, scheduling a tutor, and more can be done by an AI agent than any that anyone can access.
50:29
As we build new products, we have a chance to make these time-saving capabilities feel not only useful but routine.
50:34
In doing so, we can empower people to regain control of their time and attention.
50:38
And I I certainly feel that uh something that I would need to sit down 15 minutes looking for, you know, a new studio space or something.
50:46
Fire off agent and come back with some interesting results.
50:50
Um anyway, let me tell you about graphite.
50:52
dev code review for the age of AI.
50:54
Graphite helps teams on GitHub ship higher quality software faster. Get started for free. graphite. dev. Uh last one, support.
51:01
For many people, the biggest barriers to progress aren't lack of access or opportunity, but self-doubt, isolation, and burnout.
51:08
Sometimes what's most empowering is support someone or something that can help us reflect, feel seen or simply move forward with clarity and confidence.
51:15
My business coach Katya has been transformative in my career.
51:18
I've joked with her over the years that everyone needs a Katya in their pocket.
51:24
Personalized coaching has obviously been a privilege reserved for a few, but now with Chachi, it can be available to many.
51:30
This is something I have not used.
51:32
This feels like we're getting further out on the curve of like, you know, defining what the product is.
51:35
like knowledge retrieval is like so concrete generate an image very concrete.
51:38
Uh now support is much more it's much more of an amorphous product and there's some companies out there that are doing it right now.
51:48
It's not a drop in replacement and and there's much more to define about what this product looks like or how people will actually use this.
51:54
>> Yeah, it's very general.
51:54
There's going to be vertical specific uh companionship products, right?
51:59
We've seen this uh what's the founder that we had on the the replica was one of the original ones started like a decade ago.
52:07
>> Grock uh Gro Heavy 4 uh has their own >> specific iteration.
52:11
But then but then this is also to be clear >> the area of greatest concern, right?
52:18
Where um it's very clear if you have a business coach telling you, John, >> you got this. You're incredible.
52:23
You're built you're built different. >> Um that's great.
52:28
But then when you tell them that you've discovered um you know how to break the uh you know space-time continuum and they you're actually right road right now. >> You're right.
52:41
>> You're absolutely right.
52:41
You you did figure this out and so I think they need to figure out the guard rails for this category and I'm sure that's uh top of mind for everybody at the company right now. >> Yeah, for sure.
52:50
AI coaches on the other hand can be available throughout every day.
52:54
She says, "This isn't about replacing human connection, but filling a gap that often goes unfilled.
52:57
Many people don't feel comfortable opening up to family or friends, and most people don't have access to a therapist or coach they can call regularly.
53:04
Even people who do have access often spend an hour a week or less with these professionals.
53:09
At the core uh at the core of philosophy and religion is the idea of self-nowledge.
53:14
To become who we want to be, we have to understand who we are." Uh so very interesting. Apple and F1.
53:22
That's the story that we're talking about today.
53:23
Uh we've talked about this before.
53:25
So, uh F1, fantastic race series. We're big fans.
53:29
Um but the actual streaming rights to the geek to the races has been kind of hotly debated.
53:38
There was an article in the journal a while back that they were going to ESPN saying, "Hey, we want something like $80 million."
53:44
And uh ESPN was like, "I don't know if that's worth it."
53:49
and we were going back >> and they wanted something like 200. >> Yeah. Yeah. Yeah. They wanted a lot.
53:52
And so, uh, the stack of of content to go through is growing.
53:58
So, we will be we will be thinning textbook.
54:01
>> We will be thinning out the guest lineup in the future.
54:03
Um, but for today, we're going through uh let's kick it off with Ben Thompson.
54:07
He had some good analysis here and has some uh some details from reporting in the Athletic uh which is a New York Times property.
54:15
Um, Formula 1 is trending towards choosing Apple for its United States broadcasting rights for for 2026 onwards.
54:21
The technology company's bid being in the 120 to $150 million range.
54:29
>> Market clearing order inbound.
54:29
It >> is a market clearing order. It probably will clear.
54:32
Uh, ESPN has held the rights since 2018, but its current deal is up at the end of the year.
54:37
The sports media giants exclusivity negotiation peri period expired before February 2025 leading to interest from other >> broadcasters including Netflix and TVPN.
54:51
>> Yes, we would love that.
54:52
>> We decided to sit it out.
54:53
>> ESPN has remained in the mix for US broadcasting rights though its bid is lower.
54:57
Apparently ESPN's offer is in the $90 million per year range, >> which is still wild.
55:01
Uh because this is the US broadcasting rights. Yep.
55:08
>> And I believe in 2024, Formula 1 had around 1 million total viewers. >> 1. 2 1.
55:14
3 million race uh viewers on average per race. >> Yeah.
55:20
>> Um but Ben Thompson puts it in a different perspective.
55:23
He says F1 on the other hand has a meaningful audience.
55:25
um versus MLS which was averaging around 200,000 viewers per year on linear TV before the Apple deal and and that wasn't growing.
55:35
So this is six times as big as MLS. Yeah.
55:38
Um again there I think there's more games than or more race more games than races.
55:42
So it's not exactly Apple to Apple but every race um but one has seen a year overyear increase in viewership.
55:49
So F1's be getting more popular.
55:51
A lot of that's driven by drive to survive.
55:52
Um, and this is the crazy thing.
55:55
Given when races are televised, usually early Sunday morning, and the fact that they aren't promoted by ESPN any meaningful sense, so it's not like you're getting the Pat McAfee Show talking about, oh, you got to tune in and there's a ton of coverage on ESPN throughout the week.
56:09
They just kind of show up Sunday morning and people turn on the TV and they schedule it.
56:12
Um, and so you can he >> I always like the timing because uh any any parents out there uh you know you're certainly not sleeping in on the weekends, kids wake up >> and I find it's it's some nice background uh uh noise or viewing uh in those early hours >> for sure.
56:31
Um and so he says you can make the case that F1 makes its own audience which by extension means they will be able to pull people to a streaming service.
56:39
Basically F1 fans are strong enough that they'll go search it out.
56:43
And when I when I was really getting into F1 after getting into Drive to Survive, I was like, "Okay, I'm gonna have to subscribe to F1 the app and like jump through all these on your phone." And it was Yeah.
56:52
It's like as many hoops as as UFC, but people still do it. >> Yeah.
56:57
And there's still there's still a big debate, but >> and I think Netflix had done the analysis and realized that all of the potential viewers of F1 were basically already subscribed to Netflix. Yep.
57:04
And so it didn't it was a a more difficult um kind of equation to make work than someone like an >> Apple TV potentially that uh that could potentially drive news.
57:17
>> Someone called it a streaming backwater or something like that in here.
57:19
We we'll dig into it, but somebody had some not nice things to say about uh about Apple TV, although I'm a subscriber and I think it's nice.
57:27
Um so uh he says, "What makes this is Ben Thompson writing a strategy.
57:31
makes this all work is the fact that for F1 races, uh, F1 races are at the bottom of the funnel in terms of fan engagement.
57:38
F1, at least in the US, has by and large grown its audience via Netflix's drive to survive.
57:42
Uh, Netflix pays F1 a pittance for rights.
57:46
Apparently, singledigit millions for that.
57:48
But it's okay for F1 because the way they ultimately make money is by curt converting Netflix viewers into race fans.
57:55
So, if you're F1, you see that as a marketing channel, honestly. >> Yeah.
57:58
F1 uh has sponsors at the at the sort of league level and also the individual teams.
58:04
All of those sponsors are getting more viewership and engagement through F1.
58:07
So, it's sort of added value to them and they don't have to pay they're actually making money >> uh on I mean not super consequential.
58:14
>> uh on I mean not super consequential. I wonder I mean of course the sponsorships are like such like 360 deals that it's like you know car livery and also a whole bunch of other you know advertising assets but I wonder if they broke out okay how much lift how much
58:31
benefit how many impressions are we getting from drive to survive would that be when you total up all the all the value ad to all the different advertisers from being in drive to survive and being seen by people on Netflix how much would would that be more than the couple million dollars that Netflix pays for the for the rights. It might be. It might be like It might be.
58:48
It might be like there's a lot of people that that see ads that are for specific F1 sponsors.
58:57
>> I've seen Saudi Araco >> like hundreds of times watching Drive to Survive. >> Exactly. Yeah.
59:02
It's like pretty valuable ad space.
59:04
So, uh interesting deal that kind of works both ways and everyone makes money.
59:08
Anyway, um making that model work, however, means actually monetizing the races, which means that it actually it's actually quite logical to go with the company willing to pay $30 to $60 million more for those rights.
59:19
You have to capture value somewhere in the funnel.
59:21
Moreover, you can make the case that many F1 fans aren't necessarily ESPN subscribers.
59:24
And this is this is where it gets interesting because the cable bundle is shrinking and the the F1 did this uh global fan survey. So, it's their data.
59:36
So Ben Thompson's kind of taken it with a grain of salt, but uh this is from the F1 global fan survey.
59:40
Uh they uh they conducted a study every four years to track how fan engagement is evolving across the sport.
59:47
They uh got a 100,000 responses from self-identified highly engaged fans in 186 countries.
59:52
That's a that's basically all of them.
59:55
The findings offer a detailed snapshot of the modern fan mindset and show that Formula 1 is increasingly attracting new, younger, and more female audience with growth in markets such as the United States.
1:00:09
Gen Z is helping shape the rhythm of modern of the modern fandom, engaging with the sport more frequently and on a deeper emotional level.
1:00:18
Female fans now account for three in four new fans. >> Interesting.
1:00:24
>> I would not have predicted that.
1:00:24
That's very, very interesting.
1:00:26
Well, >> the drivers are often Chads, and I can imagine that's driving it.
1:00:30
That that could be uh part of part of the factor. >> Yeah.
1:00:34
It's also interesting that there there are no female drivers on the grid right now.
1:00:39
There was Danica Patrick in the na in the NASCAR series for a while.
1:00:43
Um but somehow it's breaking through.
1:00:45
Maybe it's through Drive to Survive being more accessible on Netflix, more storylines pulling people in and they're just going down that funnel. I don't know.
1:00:52
Well, there's also the the the Wag industrial complex surrounding F1 that uh is aspirational lifestyle.
1:01:00
>> Maybe they should do a spin-off of Drive to Survive.
1:01:01
Don't they do that with NBA?
1:01:03
Isn't there a show about like wags or something like that?
1:01:05
That's kind of informally what the Housewives series is, right?
1:01:09
Um maybe they should do that.
1:01:12
Uh with the largest country share of respondents, fans in the US cons continue to stand out for their growth, youth, and digital fluency.
1:01:17
They overindex on content engagement, sponsor responsiveness, and daily touch points, signaling a market where fandom is evolving rapidly and is commercially.
1:01:28
>> Saudi Ramco is like, "Why?
1:01:28
We seem to be getting a lot of female zoomers zoomers >> buying oil following our content on Instagram.
1:01:38
They seem to be really into this. What's happening?" >> Maybe. I don't know.
1:01:41
Um, so, uh, so basically Ben Thompson's conclusion from this survey is what what do you what are you talking about when you're talking about a younger, more female focused audience?
1:01:54
Those don't sound like ESPN subscribers.
1:01:57
Those actually sound like Apple TV subscribers or potential Apple TV subscribers. Certainly cord cutters.
1:02:03
Certainly people who are not going to buy a big cable package and get on ESPN 2 and Red Zone and all of that stuff.
1:02:07
um that's typically the Gen X male, the maybe older millennial male.
1:02:14
Uh that probably dominates the ESPN audience.
1:02:17
Um and so um >> yeah, that's that's really interesting.
1:02:23
Apple kind of has an edge where they can build out a service for some of these more lifestyle oriented sports, some of these al alternative, you know, you could imagine them doing doing things in tennis over time because maybe that's a different audience than like the hardcore ESPN >> Red Zone subscriber. >> Yeah. Yeah.
1:02:41
So, um, he says, "I could definitely envision a scenario where F1 not only doesn't suffer from being on a streaming service instead of ESPN, but it actually grows further."
1:02:50
So, this is an interesting bullcase.
1:02:51
Uh, given Apple TV's lower price point relative to a standalone ESPN streaming service, much less a cable bundle.
1:02:58
What is clear is that this deal certainly makes a lot of sense for Apple than the MLS deal did.
1:03:04
If the company can capture all 1.
1:03:04
3 million of those current viewers, and they are incremental to current Apple TV subscribers, then the company will break even on this deal.
1:03:11
Those are very generous assumptions, of course, but not nearly as generous as whatever assumptions drove the company to spend double the money on a sport no one watches, which is MLS.
1:03:20
So, he's uh he he's taking shots at MLS because there was an interesting quote in here from from something else. What was this?
1:03:26
Uh it was from uh The Athletic again talking about the North American Soccer League general managers.
1:03:35
One common theme from general managers, the Apple deal, which is Apple's deal to to air MLS, is proving to be too much of a barrier for new fans.
1:03:44
Uh the >> wait, they want the content to be completely free.
1:03:50
>> Uh new fans just don't want it on Apple apparently, or the GMs don't want to be limited to the Apple audience apparently.
1:03:55
They want to be on ESPN or something. >> Oh, okay.
1:03:58
>> Apparently, so the GMs say, uh, hey, if we want this to grow, we're the general managers of a bunch of different MLS teams.
1:04:05
If we want to grow soccer in America and grow MLS, change the format. Change the salary count.
1:04:10
>> Why don't they just pivot to football or basketball >> or baseball if they want? >> Yeah.
1:04:15
What's wrong with just throwing around a pigkin? Yeah. Change the format.
1:04:17
Just turn it into American football.
1:04:21
>> I have to say a football is one of the best things you can buy under 50 bucks. >> Sure. >> Period. >> For sure.
1:04:28
>> That's non non non food.
1:04:28
I mean, like the value you can get out of a pig skin is absolutely insane.
1:04:32
hours of >> the return on the return on pig skin. >> Yeah, it's fantastic.
1:04:36
Uh he says, "Allow us to bring more attractive players."
1:04:38
I I I didn't realize that MLS had a salary cap because I thought they were signing like massive deals with these uh like legacy famous uh >> who is the player that went to Galaxy.
1:04:50
>> Was it Messi or something? >> Yeah. >> Is that right? I don't know.
1:04:52
Uh well, >> this is how this is bad because I don't know sports, but >> uh there was someone who went to Miami.
1:05:01
There was someone who went to the LA Galaxy. Um I don't know.
1:05:02
Anyway, uh this GM says, >> "Yeah, so so Messi plays for >> Inter Miami." Yep. And it's a huge deal.
1:05:12
It's like AI researcher money.
1:05:12
Remember >> that's what we were how we were comping it.
1:05:18
>> Um and the GM said, "But they also have to end the deal with Apple.
1:05:20
It's bad for the >> David Beckham.
1:05:22
David Beckham >> played in the MLS and you have uh Zlatan Ibrahimovic as well.
1:05:28
So >> they'll pull in some superstars, but usually towards the end of their runs.
1:05:35
>> So, so, so the GMs of the North American Soccer League, MLS, they say, "I think we have to be on more linear outlets.
1:05:40
We have to be on ABC, NBC, Fox, or more regularly."
1:05:45
More regularly, because I think a lot of people, a lot more people watched our games when we were in that space.
1:05:51
I think Apple and the whole streaming thing is really innovate in innovative, but it's probably where things uh and it's probably where things will be going, but I don't think MLS is the leader of that.
1:06:01
I don't think enough has been put behind the subscription model.
1:06:04
You're in a different league when you're you're a different league when you're a subscription-based league.
1:06:06
I don't think the effort has been put in like it should be.
1:06:10
It's like selling tickets.
1:06:11
You need people out there selling.
1:06:13
You can't just hope that people are going to sign up. And so, yeah. Um, interesting.
1:06:18
Uh the risk for F1 is that the series which has ma massive growth is going to conscript itself to the streaming backwaters says Ben Thompson uh just so it can make a few more bucks in the meantime.
1:06:31
So that that's the risk is that F1 goes and you know gets stuck in subscription land and no one can just turn on the TV and see it and the 1.
1:06:38
3 million fans don't actually migrate over and they say I'll just catch the highlights or I'll just watch Drive to Survive on Netflix instead of going set setting up a new subscription.
1:06:47
But if it's truly F1 truly does have a younger audience, does have a more female audience, they might be that might be what gets them to turn on Apple TV subs uh streaming at least. >> Yeah.
1:06:58
Anyway, >> I still get so confused by all the different cable plans.
1:07:01
I watch F1 on YouTube TV, which I guess is through ESPN.
1:07:06
It's like I guess that will go away and then I would have to subscribe to Apple so they would get a net new I actually don't subscribe to Apple TV. >> You don't? >> I don't. Okay.
1:07:16
>> Not not for any reason.
1:07:16
I just I never want to I never want to go through the ha never want to go through the I just don't watch a lot of television. Yeah.
1:07:23
>> And I don't want to go through the hassle of like signing up for a new service. >> Yeah.
1:07:26
I feel like recently I got a new iPhone and um it just like came with Apple TV Plus and I just like clicked the button and didn't unsubscribe and then I've been on that for a while.
1:07:35
So let's go over to uh Ben Thompson talking about Google earnings.
1:07:39
Um he has said I've repeatedly laid out the theoretical case for why AI is potentially disruptive to Google and we asked him about this when he came on the show.
1:07:48
about this when he came on the show. My question was like is there a world where where revenue and profit from from generative AI products is actually not as counterpositioned against Google in the sense that like if they had launched
1:08:04
the Gemini app first and become the chat GPT they'd probably be seeing a drop in search volume but then also seeing revenue spike from an even more popular Gemini that's monetizing even better and and was it a question of them just like not being able to take that pill or was
1:08:20
it more about like the risk and PR nervousness than >> the thing that Ben said thatal that stood out to most the the most to me and I'll probably butcher it a little bit but he effectively was saying like >> it's really hard for companies to change who they are
1:08:34
>> and when you look at Google's mission statement to organize the world's information and make it universally accessible and useful that is so deeply aligned with AI it's like what does chat GBT do like organizes information >> and makes it useful, right? Agents like
1:08:48
Agents like Chat GBT agent is making that information useful. >> Yep.
1:08:54
>> Um and so it just feels like again like generative AI is like deeply aligned with the core mission and so it's easy to see you know them continuing to win here. >> Yeah.
1:09:04
I I I think the the problem is is like the like disruption comes when you create a product that is not as good but the financials are structurally different.
1:09:17
So the idea of like if AI disrupts like a law firm, you look at this product or even like I don't know what what's the classic example of of of disruption that people go through.
1:09:29
Um I don't know but like this idea of like >> yeah the big thing was innovator's dilemma, right? >> Yeah.
1:09:36
In the in the innovator's dma, the example is like a new product comes to market that does something 90% as well.
1:09:46
And uh and so it's like unacceptable to the people that are buying the current product, but then over time it gets better and better and better until it replaces the current product.
1:09:55
And so there is a world where if they had just changed the the Google search bar to just be like this is just an LLM now people would have been really really disappointed with that shift and they would have been like I want to go back and ripping that band-aid off to the tune of what $300 billion a year in in search revenue would be really painful.
1:10:17
So they had to take a more um a more iterative like stepwise improvements.
1:10:22
But it seems like they're doing well.
1:10:23
And Ben Thompson is kind of echoing this.
1:10:24
He says, uh, once again and however, I have to come to Google's defense.
1:10:28
All available metrics suggest that the company is doing quite well.
1:10:32
Indeed, I would go further.
1:10:32
You can make the case that the company's biggest mistake is not going harder.
1:10:35
Uh, and so they flipped the switch on cloud, their infrastructure spending is through the roof, as we mentioned.
1:10:42
Um uh so uh in in after the Q4 2024 earnings and their announcement that they were spending 75 billion on capex then uh Google Google cloud's revenue numbers disappointed but this was because um they didn't have enough GPUs and they were actually constrained.
1:11:00
So they they missed on topline like they didn't bring in enough revenue but their margins actually improved.
1:11:07
improved. And so what that means is that there was so much demand that they had pricing power and they could say no we're not giving you any discounts because everyone wants these GPUs right now or the TPUs they want our infrastructure so we don't need to we don't need to sell them at a discount and so their margins were really good and what that revealed was that they
1:11:24
were they were capacity constrained and that justifies the bigger capex spend that they're going into right now >> and uh Google is still the poly market which company has the best AI model end of 2025 so December 31st First, there's been uh one and a half million of volume and Google is still sitting comfortably at 46% chance of of being at the top of LM Arena at the end of the year. >> Yeah, this is this is such a crazy
1:11:49
>> Yeah, this is this is such a crazy supply constraint metric.
1:11:52
So, uh the Google CFO Anat Ashkenazi said on the earnings call in cloud, as I mentioned, the demand for our products is high as evidenced by the continued revenue growth and the cloud backlog.
1:12:04
Guess how big their cloud backlog is.
1:12:07
Just the demand for for Google cloud products that they can't fulfill because they haven't built the data centers like but if they had more capacity they think that they could deliver this $106 billion. That is so much money. >> Great.
1:12:21
Honestly, honestly more businesses should try to put themselves in the position they have60 billion of demand that they >> 106 billion.
1:12:31
>> We try to not give business advice because it's so >> fraught.
1:12:35
Well, it's just it's just like every business is different.
1:12:37
Every every founder is different.
1:12:39
But, uh, in general, >> if you can develop a demand, you know, >> it's a good call for like a YC company.
1:12:45
Like if you go on stage at demo day and you say like, "Yeah, we have Lois and we have some demand backlog.
1:12:50
>> We have billions of demand that we're unable.
1:12:52
There was actually a YC company yesterday or or two days ago that that was getting a little bit roasted because >> they were like they launched a product and they were like, "Wow, today was insane.
1:13:03
Our servers our servers went down immediately >> and you could see the user chart and they had like 250 users and so people are like wait like your entire your product went down on 250 users. >> That's ridiculous.
1:13:19
>> Anyways, but happy happy for their success. >> Yeah.
1:13:21
Well, let me tell you about Figma.
1:13:22
Think bigger, build faster.
1:13:22
Figma helps design and development teams build great products together.
1:13:26
You can get started for free at figma. com.
1:13:27
Um and so this so he keeps going into the capex before we do that.
1:13:33
Figma make is generally available today. You can go to figma.
1:13:37
com/make and just start building various products.
1:13:42
Tyler has been using Figma to make a product that we'll be releasing very soon.
1:13:48
>> Very excited for that.
1:13:48
We will we will announce that soon.
1:13:50
Uh so uh from the earnings call given the strong demand for our cloud products and services we now expect to invest 85 billion in capex from 25 2025 up from a previous estimate of 75 billion just a 10 billion incremental investment absolutely massive.
1:14:08
Our updated outlook reflects in additional investment in servers the timing of delivery of servers and an acceleration of the pace of data center construction primarily to com to meet cloud customer demand.
1:14:16
uh the challenge pitch uh cautioned in an answer to an analyst question is that it takes a while for these investments to come on board.
1:14:27
So of course the the risk is like you overbuild but there's certainly no uh evidence of that given the massive backlog.
1:14:33
So um the other interesting thing uh that um Ben dips his toe into so Ben Thompson Mr.
1:14:39
Zachary famously does not get caught up in trade deals and uh nonCEO uh employee shifts.
1:14:49
>> They're below his line.
1:14:50
>> They are below his line.
1:14:50
They're below his line, but he had to chime in on the CFO change that happened at Google.
1:14:55
So, uh he says, "I'm always hesitant to delve too much."
1:14:59
I like that he's throwing in a delve.
1:15:01
You know, it's not written by a but he's baiting.
1:15:03
He's baiting people for sure.
1:15:06
>> He's like, "How dare you accuse?"
1:15:06
He's like, "You can go back and look at how many time I've used Dell a million times. >> I I created Dell. I coined that word."
1:15:16
>> There's there's a there's this there's a YC company called Dell as well that's doing well.
1:15:22
>> Um so, uh he says, "But it's interesting that the goo that the earnings call was Ruth Pat's last one as CFO.
1:15:27
Parat earned a lot of plaudits for getting Google's spending under control in the late 2010s, but what seems clear in retrospect that is that 52.
1:15:38
5 billion that Google spent on capex in 2024 was too little.
1:15:45
And so he's kind of like, you know, noodling on this idea that maybe the CFO was too cautious going into an AI boom.
1:15:55
>> You could lose your job over that.
1:15:57
called a maybe Ruth Parat was calling top signals like us. It looks terrible.
1:16:02
Fortunately, we don't lose our job when we >> There's nothing wrong with calling top signals >> there.
1:16:07
There there's something wrong with under water if you're call specifically calling the top. >> Yes.
1:16:12
>> Trying to identify top signals is more of a metag game. >> Yeah. Yeah. Yeah.
1:16:16
So, uh, so basically, um, Google, it it it does seem like Google underinvested in capex in 2024 based on demand, wound up with that massive backlog, missed on topline, couldn't couldn't make enough money, couldn't generate enough revenue, had higher margins, that's great, but uh, but didn't deliver on the cloud side uh, on on the actual scale.
1:16:35
So, Ashkanazi's Ashkenazi's calls meanwhile have repeatedly reiterated that Google Cloud is particularly supply constrained.
1:16:42
They don't have enough servers and the company has now surprised investors twice in six months with the scale of its capex plans. Let's go.
1:16:50
Um says this, in my opinion, is incredibly bullish for Google.
1:16:55
Go back to this the disruption lens which we were kind of noodling on earlier.
1:16:58
The exact avenue from which you would expect management resistance to a disruptive innovation to to flow is from the is the CFO office.
1:17:06
Like the CFO uh should be resistant to uh disruptive innovation.
1:17:12
And when you look at the history of companies that got disrupted, the CFO is saying, "We have a good business.
1:17:18
Let's just keep printing cash."
1:17:20
Like this new thing, let's not worry about it.
1:17:22
Let's not go >> Nokia CFO.
1:17:24
He's like, "Are you guys seeing the numbers?" >> Yeah.
1:17:28
But like for a lot of those companies that got disrupted, it's like they were printing cash, high dividends for decades, even after the disruption happened.
1:17:35
The next the iPhone comes out and in order to actually do something to compete, they have to completely go into startup mode.
1:17:41
They have to burn a ton of cash, cut their dividend, stop stockpiling cash, issue debt, do a ton, raise more equity maybe like they have to become a new product development company and it would be very very difficult usually.
1:17:52
Um, but >> Blackberry's annual revenue in 2005 1.
1:17:55
9 billion >> Blackberry Blackberry's annual revenue in 2024 580 million.
1:18:07
>> It's a cyber security company now.
1:18:07
They bought a couple cyber security assets while they were valuable, while they were um like, you know, while the stock was up and then eventually they wound down the phone business and just kind of became a cyber security company.
1:18:20
Um because obviously they had a ton of enterprise contracts because Blackberries were sold into like large enterprises as you know work phones.
1:18:30
Very few people had Blackberries as like everyday phones.
1:18:33
Um but so they had all those they had like solid sales team, solid connections, bought a new asset and were able to continue the business even though obviously it's not the company that it once was.
1:18:43
Um so um Ben Thompson goes on to say there's a there does seem to be a major shift in mindset in terms of the company's willingness to lean into AI particularly for a segment Google Cloud that even in the best case scenarios is significantly lower margin than the company's core business.
1:18:59
What's interesting about that margin point, however, is that it too is another reason to be bullish on Google's prospects.
1:19:06
Google is the only company, we talked about the TPU thing, they're the only company with an atscale ASIC alternative to Nvidia's GPUs, which should give them a meaningful cost advantage to that end uh to the extent that cloud comput they're going to be able to support >> SSI >> SSI, >> which is crazy, >> not just their own needs.
1:19:25
>> And then and then there's also questions.
1:19:28
>> The dynamic that's interesting is like OpenAI needs to massively scale infrastructure and and uh as we covered earlier this week commit to spending tens of billions of dollars with with Oracle. >> Yep.
1:19:41
>> And their revenue today doesn't support that.
1:19:43
Obviously their their growth trajectory is absolutely insane.
1:19:47
>> But OpenAI being in a position where they are competing headon with Google for this very very critical consumer use case search knowledge retrieval. Yep.
1:19:58
and ultimately uh usefulness.
1:19:58
It is an extremely tough position for OpenAI to be in. >> Yeah.
1:20:04
So the other hyperscalers are working on A6 that are alternative to Nvidia GPUs but they all seem to be following the similar path trying to get line time at TSMC not fully at scale.
1:20:16
Then there's also the question you know I think a lot of people's mind is like is this the end of history?
1:20:20
Is there something that's coming down the pipe that could disrupt the NVIDIA GPU monopoly, the CUDA ecosystem, or even the Google TPU?
1:20:27
And semi analysis has a funny meme here um way down in the stack uh about the next generation of chips that were that have been pitched from startups like uh um etched and a few others.
1:20:42
So, semi analysis says although baking transformers into silicon may sound cool, it's mo mostly just a marketing slogan.
1:20:49
90% of transformers flops are just GMM and 256 x 256 or 128 x 128 systolic arrays in TPU and in TPU tranium are already optimized for these.
1:21:04
Even modern GPGPUs with tensor cores are optimized well for GMM.
1:21:07
Even if you bake transformers into silicon aka just create a giant systolic array, most of your die area will still be taken up with SRAM cells.
1:21:16
That's the memory um and you will still face the memory wall since your HBM memory bandwidth will be the same as GPGPUs, TPUs and tranium.
1:21:26
And so uh the meme is TW not transformers are just 90% Matt moles.
1:21:28
Uh George hot similar analysis or take just saying that um if you look at the actual energy use of GPUs right now there isn't that much opportunity to squeeze more value out of them.
1:21:41
They're pretty efficient and most of the most of the energy that goes in goes into these very very specific math calculations that are already pretty optimized.
1:21:50
We did hear from someone something about the uh this was from Martin Skrey that Jane Street or was it uh was it Jane Street?
1:21:57
One of the high frequency trading firms figured out how to run one of those very very basic matt calculations on a GPU more efficiently and I guess kind of open sourced it or something or got out into the into the research world.
1:22:10
Um, so it it doesn't it doesn't feel like at this moment I still think that there's a interesting bullcase for the Etch crew in some ways in some niches or something some specific models.
1:22:21
Um, but it doesn't seem like there's some new disruptive chip architecture that's going to come out and and obiate the need for both Nvidia GPUs and Google TPUs.
1:22:30
And so Google's very well positioned as Ben Thompson is writing.
1:22:36
Thompson is writing. So um he says uh this should give them their position that they have an atscale ASIC alternative to Nvidia GPUs should give them a meaningful cost advantage to that end to the extent that cloud compute
1:22:50
becomes commoditized is the extent to which Google actually has a margin advantage because if all of the clouds are commoditized but all of them are on Nvidia and then Google is taking the Nvidia margin basically from their TPU uh they should be in a very good spot. This is to be sure a bit weird. Parat
1:23:04
This is to be sure a bit weird.
1:23:04
Parat gained these plaudits those plaudits because there was so much waste in Google to cut which was downstream of the company's amazing monopoly in margins and search which hardly seems like the recipe for a structural cost advantage but here we are.
1:23:18
>> Sundar is going to tell every employee we're cutting your daily massage allowance from three massages down to one.
1:23:24
We're deeply sorry but we need to get fit. >> Got to get fit.
1:23:28
And you got to get on Vant. Automate compliance. Manage risk.
1:23:31
Improve trust continuously.
1:23:31
Vantis trust management platform takes the manual work out of your security and compliance process and replaces it with continuous automation whether you're pursuing your first framework or managing a company.
1:23:41
>> It was great >> talking with Christina yesterday. She's an absolute dog. >> Boardroom general.
1:23:47
>> The boardroom general.
1:23:48
>> She put the KPIs in orbit. >> She she really has. >> She really has. >> Did you see that?
1:23:52
She has it that Reed shared. >> Yeah. this chart.
1:23:56
She says, uh, >> Andrew said, "Christina often talks about stacking rice on the chessboard." >> Yeah.
1:24:04
>> And I didn't see this until after the show, but I was like, it seems like you guys are just chopping wood like every day just out there like >> that's what it is. >> Putting in the work.
1:24:12
>> Just, you know, couple percent here, couple percent here.
1:24:13
Just very consistent growth, very consistent growth, very smooth growth curve into a monster of a business.
1:24:18
So, um, congrats to everyone over at Vanta on the new round.
1:24:22
And we will go back to Google.
1:24:25
Um, so, uh, more broadly, the way in which Google seems to have flipped the switch in terms of going allin on AI along with Meta's spending on AI talent really does strongly suggest that AI in the end is a sustaining technology that favors the incumbents most of all.
1:24:37
Both see a line of sight to new revenue streams and critically both can fund AI from profits, not speculative investment or debt.
1:24:47
And of course, both have distribution, including search.
1:24:50
Sometimes the empire strikes back. I love that.
1:24:53
So there were other takes from uh from search.
1:24:55
You'll have to subscribe to certy to get the full analysis.
1:24:59
You got to go and subscribe to certy if you're not subscribed. What are you doing?
1:25:03
Um uh but the interesting takeaway from uh from search is that there it seems like Google's being a little squishy about hey we don't want to report certain KPIs anymore.
1:25:14
Uh so they used to report paid clicks and now the chief business officer Philip Schindler is saying um and and on your paid click qu question look to be very clear I think we said this before we manage the business to drive great outcomes for our users and attract attractive ROI for our advertisers.
1:25:31
We don't we actually don't manage to pay clicks or CPC targets which is fair and good but it's funny because Ben Thompson's like but I want that data that would be helpful to me.
1:25:41
Um, >> I mean, when a business has a metric that they tell you is important for a long time and then they suddenly tell you it's not important.
1:25:48
You gota >> It does make sense. They're a huge business.
1:25:52
They've moved on to to bigger and like higher levels of abstraction.
1:25:55
>> Just look at the growth in tokens. >> Yeah. Yeah.
1:25:57
Don't worry about anything else.
1:25:59
>> What what he says is that there are there are product changes and policy changes that actually drive better monetization but at the expense of paid clicks.
1:26:06
And that's probably uh AI searches and stuff like that.
1:26:09
You'll see in the 10 Q paid clicks were up 4% year-on-year, but a number of factors affecting this these metrics from quarter to quarter uh such as advertiser spending, product changes, policy changes, user engagement.
1:26:20
So, it's really important when it comes to pay clicks and CPC to avoid drawing overly broad conclusions solely based on these metrics.
1:26:29
Don't put me in the truth zone on this, he says.
1:26:31
Uh so, uh Ben Thompson, this is exactly what you said.
1:26:35
I don't even think you you pre-eread this, but you basically have Ben Thompson working in your head cuz >> this is what I said when Ben Thompson came on the show.
1:26:41
I was like, I think your way of thinking has been so ingrained into my mind that I I think that I have an original thought and you've already thought it.
1:26:49
>> Okay, so this is what he says and it's literally it says, >> "The more that management tells me not to pay attention to paid click rates, the more I want to pay attention, which is exactly what you just said.
1:27:01
Regardless, once again, the fact that paid clips paid click paid clicks were up 4% and search revenue was up 12% makes it clear that search growth is primarily being driven by higher prices.
1:27:13
And so the actual number of clicks is up 4% but revenue is up 12%.
1:27:14
That means that they're monetizing each click better.
1:27:19
Um, but people are obviously going to read into, oh, paid clicks aren't growing that fast.
1:27:24
Like the actual pie isn't growing that fast.
1:27:25
are just getting like it's basically like, you know, once Instagram reels is is popular, everyone's watching it, then you just got to put more ads in the feed, right?
1:27:35
And that's kind of that seems like what they're doing and that's what's driving search revenue up 12%.
1:27:38
Um AI overviews is another question.
1:27:41
Uh this is a funny one where um uh basically the uh the chief business officer Philip Schindler at Google says um uh basically says that uh AI overviews are monetizing as well as other Google searches.
1:27:58
And Ben Thompson says given the fact that the vast majority of Google searches don't monetize at all because people just click on a link.
1:28:05
The often times the if you're searching for information, you can have very low purchase intent, right?
1:28:14
Like how many times a day do I search for a specific fact? >> Yep.
1:28:18
>> And never would have ne you could show me the best ad in the world and I wouldn't click through because I'm just looking for a fact. >> Yep.
1:28:24
So if you're looking for something like, you know, Hulk Hogan's age, that's not a that's not a purchase intent.
1:28:28
not a that's not a purchase intent. if that comes from an AI overview or it comes from a you know how they used to have those like knowledge boxes or like it would pull from Wikipedia or if you click on Wikipedia none of that is making Google any money so it really
1:28:43
doesn't matter the UI or the actual underlying technology to get you that answer um and then uh Gemini is on absolute tear uh really remarkable how how many monthly active users do you think the Gemini app has >> these are mauus not DU on the mobile app, web app, >> Gemini app. Uh I think this is probably
1:29:01
Uh I think this is probably across mobile, web and desktop, but I would assume mobile >> 70 million. >> 450 million. >> Big, right?
1:29:11
And >> sign like I guess I get I mean they kind of get the it's kind of a >> they they have a cheat code because everybody already has a Google account so they're like signed in. >> Yep.
1:29:21
But in in term I I don't know >> they're monthly and and what we know from semi- analysis is that they're not taking share of queries that much but 450 million downloads on an app and monthly active users is pretty pretty high.
1:29:35
high. And so if there's if there's a world where Google can build a business around if they can bring in ads faster and keep the Gemini app cheaper and offer better products and like you get deep research you get agent you get the $200 tier the ad free if if open AI and chatt has a
1:29:56
$200 tier that's adree and then Google's giving you the same product and the same experience but with ads for free like that feels like an Android iOS battle going on that feels like something that could actually uh be pretty sustainable and kind of drive this like little bit of a duopoly. I don't know. Uh we'll I don't know.
1:30:11
Uh we'll have to see how it goes.
1:30:13
But uh daily requests grew over 50% from Q1 to Q2.
1:30:15
Um and uh I think everyone everyone is is pretty surprised uh by that.
1:30:22
Um and I was also surprised when I went and pulled up the uh the Gemini app when I went to go download it.
1:30:30
went to go download it. um the number of fivestar reviews is like hundreds of thousands and I was like okay this is very serious and that reflects the actual number that they shared 450 million monthly active users so still a lot of work to be done I was very frustrated when I went to go get VO3 I got the Gemini app and they were
1:30:50
advertising but I couldn't frustrations >> default in in Android yet >> I don't know I I imagine that it has to be pre-installed in the next round of Android phones that go out at why would you not want that pre-installed and and there but I don't I would be surprised if that's what's driving it because I mean how many Samsung Galaxies have they sold in the last quarter? Probably not
1:31:10
Probably not 200 million, right?
1:31:12
I I doubt that that's what's driving the uh the growth.
1:31:16
But maybe that might be an interesting thing.
1:31:18
How many of them are pre-installs, preloads, but even then you have to open it to become a to become an MAU.
1:31:23
But yeah, maybe maybe they're testing.
1:31:25
Maybe people will churn and go back to Open AI and >> CHP.
1:31:28
Android users have had access to Google Assistant for a long time.
1:31:31
That's getting phased out, which is going to be Gemini. >> Yep.
1:31:36
>> Intel now expects to have 75,000 employees at year end, down from 116,500 in June of 2024.
1:31:44
That's basically one year ago. That's a massive cut.
1:31:47
Nvidia is on the other side of the trade.
1:31:50
Uh Nvidia CEO said they now have 42,000 employees yesterday, up from 36,000 in January of 2025. Massive shifts.
1:31:57
Uh it's crazy to think about the market cap per employee difference there. It's pretty remarkable.
1:32:06
If you're cutting costs, you got to get on ramp. Time is money. Save both. Easy.
1:32:10
Use corporate cards, bill payments, accounting, and a whole lot more all in one place.
1:32:13
So the news is >> ramp problem.
1:32:17
>> Yeah, they really should get on ramp.
1:32:18
Lip Bhutan, I know you're listening.
1:32:18
You should get on ramp lip lip bhutan.
1:32:20
Uh so this is from the Wall Street Journal.
1:32:25
The business and finance section.
1:32:25
Intel is going to lay off 15% of their workforce.
1:32:29
Uh chip giant stands at a strategic crossroads.
1:32:32
The CEO warns no more blank checks.
1:32:34
Intel detailed dramatic steps to revive its sagging fortunes, outlining layoffs for 15% of its workforce and scrapping plans to spend tens of billions of dollars on new chip facilities in Europe. Oo, bearish for Europe.
1:32:47
I didn't realize that was going on.
1:32:49
Uh, the chipmaking giant said Thursday it would refocus its strategy on the highly competitive market for AI chips, regaining market share in personal computer processors, and developing its advanced 14A technology to sell to large customers.
1:33:02
It's like, okay, we're going to we're going to narrow down.
1:33:06
We're just going to do three incredibly difficult things.
1:33:08
We're going to compete in in the most aggressive AI chip market where we're not just competing with Nvidia but also the TPU from Google, all the upstarts, Grock, Sarah Bros, all these different companies.
1:33:22
Uh so we're going to like go from zero to one there and then we're also going to turn around our our personal computer processing market. >> Bold bold.
1:33:30
Intel, which has long dominated the business of making chips for laptops and desktop computers, fell far behind competitors like Nvidia, AMD, and TSMC after it failed to anticipate the surge in demand for the powerful chips fueling the artificial intelligence boom.
1:33:45
Lip Bhutan, the CEO, the new CEO of of Intel after Pat Gellzinger stepped down uh about a year ago, I believe, uh he said, "There are no more blank checks."
1:33:57
He wrote in a memo, "Every investment must make economic sense." Love to hear that.
1:34:03
Exactly what you want to hear from >> nominative determinism of lip bhutan.
1:34:08
>> Um, >> hate to put you on the spot.
1:34:10
>> Hitting the beach, getting a tan. I like that.
1:34:14
>> You know, >> I can see it. I can see it.
1:34:15
>> Maybe if they maybe if Intel doesn't work out, he can get into cosmetics.
1:34:18
It's like get a tan with this tanning oil. Get some lip balm. Lip balm. Tan. >> Yeah.
1:34:23
something something called lip blushing where where I think it's like a mild tattoo.
1:34:27
So maybe if this whole semi thing doesn't work out he could get into the lip blushing franchise game. >> Yep. I I agree.
1:34:36
Uh revenue in the June quarter was roughly flat at 12.
1:34:40
9 billion beating Wall Street's expectations.
1:34:43
Uh the quarterly loss though widened to 2.
1:34:45
9 billion for the quarter from a 1.
1:34:48
6 six billion loss in the year earlier period.
1:34:53
The results represent the company's sixth consecutive quarterly loss, extending its longest streak in 35 years.
1:34:59
It's got to suck to be in a position where computing broadly computers are doing incredibly cool things. AI is magic.
1:35:05
We pass the touring test and you're not a beneficiary.
1:35:10
you're losing money and it's because of a bet on, you know, the server and these large scale CPUs and not getting into GPUs and then missing mobile.
1:35:19
There's a whole bunch of things that went into it.
1:35:21
Um, so sales in Intel's PC chip division, the company's largest segment, fell 3% from the year earlier period.
1:35:25
I remember about a decade ago, I built a new, you know, ma massive PC rig to do uh CGI rendering on it. tell was John.
1:35:38
>> You weren't you weren't building it to as a gaming rig.
1:35:41
>> I did use it for a gaming rig. John's vice.
1:35:44
He said uh you >> there's market research.
1:35:47
I need to understand what the gaming market's going.
1:35:49
I need to understand uh the future of VR.
1:35:51
I actually I think I think uh you you beat your video game addiction in years.
1:35:58
Um and so uh uh it's uh it's an interesting time for Intel.
1:36:03
We uh I have some more more information here.
1:36:05
Uh >> do you remember six months ago there was the rumors floating around that that Elon was was a potential buyer? I feel so long ago.
1:36:15
>> I was propagating those rumors.
1:36:15
I was >> and ultimately it feels like uh you know he he got the uh got the fun experience of a turnaround at Twitter.
1:36:27
>> Does he want to do that again? >> Yep. >> I don't know.
1:36:29
I mean, my take when when he bought Twitter and turned it into X was that it was exciting.
1:36:33
I thought that he could make it a lot better.
1:36:36
He has in many ways made it a better product.
1:36:38
We stream on it now and and it feels like it's still a vibrant community and there's a whole bunch of new features, good stuff, bad stuff, but overall I think he put the company back in startup mode.
1:36:50
They're shipping very quickly.
1:36:51
Interesting things are happening and he's making plays in an interesting way that I like.
1:36:55
But at the time I was like, "Okay, $44 billion.
1:36:58
There's something like $50 billion from the chips act, maybe going to Intel and foundaries in America.
1:37:06
Run this back and marshall the same amount of capital.
1:37:08
Let him buy Intel, turn that around."
1:37:10
That feels like something that Elon would be uniquely equipped to turn around and would put us on a very different path uh if he got into semiconductor manufacturing, but it would be a huge challenge and maybe it wasn't as important to him. So, I don't know.
1:37:25
Um, but >> I mean only a week ago a a small company called AOL AOL.
1:37:29
com was posting that is Elon Musk about to buy Intel.
1:37:35
>> Um, so this feels like um very uh you know Dylan Dylan Patel on January 17th saying Elon's jet is in Florida, Global Foundaries Jet is in Florida.
1:37:44
Qualcomm's jet is in Florida.
1:37:47
>> Um and uh but ultimately a lot of people were in Florida at that time.
1:37:51
>> Well, if your jet's flying around the world, you got to take your jet to a wander. Find your happy place.
1:37:55
>> Find your happy place.
1:37:56
>> Book a wander with inspiring views, hotel, great amenities, dream events, top your cleaning, and 24/7 concier service.
1:38:00
It's a vacation home better, folks.
1:38:02
Um, and so, uh, Dylan Patel, speaking of Dylan Patel, um, he actually hung out with Lip Bhutan at a conference, and he has this hilarious story where he clearly wanted to meet Lip Bhutan and spend some time with him and everyone's coming up to Lip Bhutan, asking him what the future of Intel is.
1:38:18
And so, he goes up and offers to make him tea and makes his tea for him.
1:38:19
and they sit down and they have tea and they like and I I think Dylan Patel really understands like tea culture and knows what makes for a good cup of tea in the same way that someone like a coffee snob knows about good coffee.
1:38:31
So they they have tea and and Dylan Patel is kind of picking his brain and um and it just feels like Lip Bhutan's overall goal is just layoffs, reduce cost, write the ship, refocus.
1:38:42
And so what's interesting is that we talked to Ben Thompson and I wanted a clear I wanted a I wanted a clear narrative.
1:38:49
I wanted something like, oh, it's obvious just split the foundry and the design business.
1:38:54
Split up Intel into two companies and that will solve all the problems, but everyone seems to agree it's more complicated than that because part of Intel's foundry business is that they design their own chips.
1:39:05
And so they're they're both they're both their customer and their supplier.
1:39:08
And breaking those up might be more difficult. It's not a cure all. It's not a panacea.
1:39:12
And so uh it is more complicated.
1:39:14
Um so there's not a super clear path forward.
1:39:17
It's a lot of hard decisions.
1:39:18
It's a lot of cuts to employees and that's what we're seeing today with the 15% layoff.
1:39:21
There will probably be more in the future as Intel continues to downsize the workforce.
1:39:28
>> Really, really rough for morale knowing that the workforce will just >> be you know, continue to get cut into um you know, like you said, probably not the last. >> Yep.
1:39:38
And so they are focusing on uh the the the direct quote is refocus strategy on the highly competitive market for AI chips, regaining market share and personal computer processors and advancing the 14A technology.
1:39:49
Now the problem is that Intel is still more expensive than TSMC on a normalized per wafer basis.
1:39:56
This comes from uh this comes from uh semi analysis.
1:40:00
So they have the Intel 3 node, the 18A, and that goes up against the N5, N3, and N2 nodes from TSMC.
1:40:09
And across the board, um, Intel is more expensive on a per wafer basis.
1:40:14
So this is at their own fabs.
1:40:16
They're trying to catch up.
1:40:16
They're trying to create uh chips that are that are as um as powerful, but the name of the game is always cost.
1:40:25
>> Y um >> yeah, semi analysis says the company has cost discipline in its DNA.
1:40:27
Talking about TSMC, Intel not necessarily.
1:40:29
This manifests in many ways.
1:40:32
TSMC has more aggressive cost down targets and exercises its leverage on suppliers.
1:40:36
It does not use the copy exact dogma that creates unnecessary overhead and resistance to improvements.
1:40:46
>> Hence, Intel is now moving to copy smart.
1:40:47
Are you familiar with the copy exact strategy? >> No.
1:40:50
So when Intel built the original fabs, >> what they would do is they would build this.
1:40:55
I mean it is a incredibly precise process.
1:40:58
So you can think of lithography is basically etching.
1:41:02
That's why that company that semiconductor company is called etched because they that you're etching grooves in a silicon wafer and that's where when you put electricity through it, it goes through a series of gates and that's what a semiconductor is.
1:41:15
>> That's how you create magic.
1:41:16
>> That's how you create magic. Yes. It's the it's the spice.
1:41:17
the the Dune analogy is the most important one here.
1:41:20
Um but basically you have to etch these grooves at as small as small a scale as possible.
1:41:28
So three nanometers silicon atoms are 1 nanometer.
1:41:30
So you in theory cannot go smaller than one nanometer chip because the space between the atoms is only one atom or one nanometer.
1:41:38
So this is incredibly small and so you basically have to shine like a laser or some sort of light or some sort of radiation to actually etch into the silicon to create the grooves.
1:41:51
Basically, this is very very high level abstraction.
1:41:52
But um in order to do that, they get these m they get these mirrors from a company called Trumpf, which is unrelated to Donald Trump, but uh and and Zeiss makes these lenses and they focus all this stuff and this liquid tin gets like, you know, dispersed and then they bounce light off of it. It's this crazy process.
1:42:11
The mirrors are so flat that if you if you if you if you like blew them up to like the size of the moon, you would still would not be able to see a groove even this high.
1:42:20
It's like the most precise process in every single.
1:42:24
>> By the way, uh Intel missed opportunity to create the TPU.
1:42:27
Google's obviously >> owning that right now.
1:42:30
But if Intel had said, "We're making Trump processor." >> Yeah.
1:42:35
The spacket >> and you know I mean they're already public. So >> yeah.
1:42:39
I mean, you know, what's the chip that's going to be in the Trump phone?
1:42:43
The golden the golden Americanmade phone should be an Intel processor. Who knows? Americanade.
1:42:49
>> So, the the the semiconductor manufacturing process is so precise that apparently a TSMC like if there's an earthquake or anything like that, everyone they don't even need to send out a notification.
1:42:57
Everyone just knows go to TSMC to deal with everything because the slight change in like the road noise or slight could like miscalibrate the machines.
1:43:05
If it like rains too much, there will be extra salt in the parking lot that will get tracked in.
1:43:12
Even when they go into the clean room, it'll get tracked in and the small sodium particles will throw off the yield.
1:43:17
So, all of this is about yield.
1:43:20
They'll make a ton of these wafers.
1:43:21
There'll be small imperfections.
1:43:21
It ruins the whole wafer.
1:43:22
That's been a big problem with the wafer scale, >> which is why TSMC getting any yield in Arizona despite setting up, you know, this new facility so recently was very unexpected. It's a miracle. Yes.
1:43:34
And so the copy exact dogma was >> it is so hard to build a fab that manufactures semiconductors reliably at good yields that when Intel figured out how to make one fab they said let's make a second and let's change nothing.
1:43:51
And so they build a building they build the exact same foundation.
1:43:56
They'd build the the bathrooms would be in the same place.
1:44:00
If there was two urinals in one bathroom, like there would be two urinals at the other one.
1:44:05
Like everything was exactly the same. Copies.
1:44:08
>> Basically extremely superstitious. >> Yeah.
1:44:11
Basically, because they were like anything that changes like if the water flows through the pipe in the bathroom at the wrong flow rate, it might throw off the amount of molecules in the air over here and the atoms over.
1:44:20
It's like it's crazy precise, right?
1:44:21
And so that worked for Intel, but it had this problem which was that every once in a while you do want to make an iterative change when you're building the next fab.
1:44:32
And you do want to increase yield by maybe not copying exactly and doing things a slightly differently.
1:44:36
And so now Intel is pivoting to copy smart, only copy what's actually improvement instead of just copying everything exactly.
1:44:43
Uh NTSMC has never held themselves to that same standard.
1:44:45
And so uh customers are likely baselining the 18A Intel node against N3 3 nanometer nodes TSMC um given the characteristics.
1:44:57
Um but uh Intel will have to be aggressive on pricing if they want to compete because they are not competitive on the actual cost of the wafer and so they would need lower margins if they want to compete.
1:45:10
And so you can think about Intel as having kind of like three feature products that people are focused on.
1:45:16
That's Intel 3 uh which is their 5nanmter class.
1:45:19
Uh that one is fabbed in Arizona and Oregon and Ireland.
1:45:23
These are at Intel facilities.
1:45:26
Intel also has the Intel 18A which is at the same sites plus Ohio.
1:45:32
And this is competitive with the 2nanmter class from TSMC.
1:45:34
And then they have the 14A which is supposed to be 1. 4 nanometer or below.
1:45:39
This is the most cutting edge uh extreme ultraviolet really really aggressive lithography and uh and this is also made in America which is great.
1:45:50
Um and these will compete with Nvidia's Blackwells and then the next generation uh Blackwell successor.
1:45:57
And so they're going head-to-head but uh the real question will be can they go head-to-head and still squeeze out a decent margin because the in the Nvidia TSMC partnership is really really strong.
1:46:09
They both have good margins and they've really optimized for cost and reliability and yield and Intel has kind of lost a step there.
1:46:16
And so um they will probably need to get on linear to make this work.
1:46:21
Linear is a purpose-built tool for planning and building products, meet the system for modern software development, streamline issues, projects and product roadmaps.
1:46:28
So Lip Bhutan, if you need an intro to the folks over at Linear, let us know. We'll make it happen.
1:46:33
Reach out and uh and we'll figure it out.
1:46:36
Um so >> and uh Intel is of course down almost 10% today.
1:46:41
Uh market is not loving >> much of anything happening over there right now. >> Yep.
1:46:48
So semi analysis continues saying Intel the home of Moore's law.
1:46:51
Gordon Moore of course worked at Intel and coined Moore's law the idea that uh the number of transistors will double every few years on on a silicon chip allowing for more compute power basically at the same price.
1:47:05
uh they are they're evaluating if it will continue at the leading edge.
1:47:09
This is this is what moves the stock in my opinion more than the 15% layoff.
1:47:13
The layoff seems like reasonable downsiding, but uh deciding, hey, we're not even going to compete on the on the most advanced semiconductors, which is what Elon, Google, Anthropic, Sam Alman, you know, every single company is like, give me all of the best and give me hundreds of thousands of the best.
1:47:31
The rest of the stuff is like, okay, I'll go to anyone because I need a, you know, Wi-Fi chip in my toaster.
1:47:37
Um, so the quote from the 10 Q from Intel's 10Q is, however, we are unable to secure a significant external if we are unable to secure a significant external customer and meet the important customer milestones for Intel 14A.
1:47:50
We face the prospect that it will not be economical to develop and man manufacture Intel 14a and successor leading edge nodes on a go forward basis.
1:48:01
In such event, we may pause or discontinue our pursuit of Intel 14A and successor nodes and various of our manufacturing expansion projects.
1:48:10
Just and semi analysis says just like that we could be talking about a TSMC monopoly and the death of Americanmade semiconductors forever. Very dark.
1:48:24
>> Up next we have Casey Neistat >> recently >> the YouTuber the creator >> he's over at Mod Retro now >> working on the chromatic which we've been playing in the studio working on the M64.
1:48:36
How you doing Casey you >> good to see you. Welcome to the stream.
1:48:43
Uh, what's new in your world?
1:48:43
Uh, how did the Are those the Nothing ears? Give us a little review. >> Oh, yeah. These are fantastic.
1:48:49
Actually, to take them off because the noise cancelling on these is um it's it's like too good and I'm at home right now and my kids were screaming two seconds ago.
1:49:01
>> But yeah, these things are fantastic.
1:49:03
>> Yeah, we had Carl Pay on the show a week ago and uh he put them down and I was like I was too afraid to ask like, "Oh, are those for me?"
1:49:09
Like, "Are you going to leave those for you? I kind of want those. I need a new pair. I'll have to pick.
1:49:13
>> You know what is I'm not enough of an audio file to know what if something sounds good or not.
1:49:18
Like it's either good or it's not good.
1:49:20
>> But the physical switches like buttons you can touch. >> Yeah.
1:49:26
>> I really like and the like the Apple I don't know what they're called. What are they? AirPod Max. >> AirPod Maxes.
1:49:33
>> There's no onoff switch. >> Yeah.
1:49:36
>> Like I can't >> that digital crown.
1:49:37
>> I need I need an onoff switch and these have that.
1:49:40
I like I like your tactile. Yeah.
1:49:42
Are you Are you big Teenage Engineering guy?
1:49:43
It feels like kind of like that that they're almost coining like a trend that's like becoming more mainstream now.
1:49:50
>> Teenage Engineering is the only company whose products I buy religiously even though I have no idea how to use any of them.
1:49:58
>> The same thing happened to me.
1:49:58
I bought the little keys. Like I'm a musician now.
1:50:01
Couldn't figure out how to make a song. But I use it. >> I love it. It's amazing. Beautiful. >> It's so beautiful.
1:50:07
like someone that cares that much about the craft.
1:50:08
Um, I want to be part of that.
1:50:12
>> Support them unconditionally. Uh, yeah.
1:50:15
So, tell me about Mod Retro.
1:50:15
How'd that come together?
1:50:16
Uh, what what what is the news in your world?
1:50:20
>> Um, you know, the Mod Retro thing is really funny because uh I my video says all this, but like I'm a big retro gamer.
1:50:28
>> Um, I'm guilty of sometimes playing Call of Duty on my Xbox, but that doesn't work out when you got a wife and kids.
1:50:35
>> You're not being guilty.
1:50:35
Every every man should have an allocation of time daily for Call of Duty.
1:50:39
You know, getting on rust, >> getting on rust with your boys.
1:50:44
>> So close to that W and then the door comes flying open, the children come running in. >> It's challenging. >> But >> I'm 44.
1:50:51
So like >> Apex of video games for me was like 1989, I think, getting a Game Boy the first year they came out for my grandmother for Christmas.
1:51:01
And like you guys are younger than me, but before that all we had were those like L le Led LCD games that were like really janky, like the tiger games, and they were terrible.
1:51:13
>> And the first time I turned on the Game Boy and like saw the Nintendo come down, I was like, "Holy, this is such a moment."
1:51:20
>> Um, and then Nintendo 64 came out when I was I was in like just getting into high school as a freshman.
1:51:24
We used to sneak out of high school, get on the city bus, like the the the regular bus, and then drive two towns across, and then walk a mile and a half to Toys R Us to sit in Toys R Us and play the demo of Mario 64 when that thing came out.
1:51:41
I think it was in 94, so I was 12 or 13 years old. >> Yeah.
1:51:46
>> So, I I love that stuff.
1:51:46
And Mod Retro, when I first learned about it, when I first learned about the Chromatic, their Game Boy, >> I thought it was really stupid.
1:51:53
I didn't understand because you can get those like Chinese ones off of Amazon for 50 bucks.
1:52:00
And um a friend of mine was like, "No, this is great.
1:52:04
They're doing great things.
1:52:05
Do you want to talk to Palmer about it?"
1:52:07
And I'm a huge Palmer Lucky fan.
1:52:10
>> And I was like, "I'd love to."
1:52:10
And I don't know that I've ever been evangelized so much in my life.
1:52:13
Like by the end of that call, he didn't just convince me, but I was like I felt like an idiot for not seeing what Mod Retro was.
1:52:24
And then as I got to know Torin, who who is the the CEO of Mod Retro, and understand the vision for the company, I was just so like overwhelmed.
1:52:31
I was like, "This is the most like beautiful vision ever.
1:52:34
I don't understand the business of it.
1:52:37
It makes no sense to me, but the fact that these guys have the guts to do this, like can I be a part of it?"
1:52:43
And I think I pitched back to them what I understood the company was.
1:52:46
And they're like, "That's it.
1:52:48
That's what we want to say."
1:52:49
And I was like, "All right, give me a job and my job will be to figure out how to communicate that and we'll make videos and stuff."
1:52:55
And they're like, "Sure, but we're not going to pay you." And I was like, "Great.
1:52:59
I'll take a business card.
1:53:02
>> It's been great working with I also I'm an investor in Mod Retro.
1:53:04
I have a vested interest in them succeeding, but I mean that's not the that's not where the passion to be a part of what they're up to comes from for me. >> Yeah.
1:53:14
So many places we could go with this. Uh this is fantastic. Um what uh Yeah.
1:53:21
Do you think that there's a world I mean that I I love the handheld gaming stuff.
1:53:24
I my my experience having kids was I stopped playing console games.
1:53:27
The N64 was actually my first console ever and uh but once I started kids >> uh Super Mario 64 for sure. Golden Eye was big. Big head golden gun. You gotta do that.
1:53:41
Um, and then a couple other Donkey Kong and uh was never really into the Mario Kart racers, but then Smash Brothers kind of took over and that was the real competition with everyone eventually.
1:53:52
Um, >> I think what I think what what's exciting to me is like building new new heart consumer hardware that will still turn on and work and bring joy 20 years from now.
1:54:02
I had I had the the game I have like my Game Boy Pikachu edition from growing up and I when Mod Retro came out I went and found it and I turned it on and it worked >> and that was like such an amazing moment for me cuz like when like I haven't I've got old iPhones lying around.
1:54:18
I've never been like not one I've never been like oh I should turn that on.
1:54:21
It's going to be fun right because it's like it's not evergreen it was of the moment and you just move on.
1:54:28
And so I think if Mod Retro can do anything, which is like make consumer hardware, which is so disposable today across all these different categories and make it so that yeah, you're going to get your Mod Retro or your M64, you know, this year and then you're going to turn it on in 30 years and like, you know, bridge generations. >> Yeah.
1:54:48
You know, it's I think there's there's it's easy to sort of on um planned obsolescence because it is real.
1:54:57
I think of that like >> we woke up, my house woke up this morning at 7:30 a. m.
1:55:01
because the refrigerator repair man finally showed up and was banging on our door at 7:30 in the morning.
1:55:07
The dog was freaking out because our 2-year-old refrigerator doesn't work.
1:55:11
>> And this is a refrigerator that has Wi-Fi. >> There we go.
1:55:15
>> Let's give it up for Wi-Fi.
1:55:15
Everybody's always It doesn't work. It doesn't work.
1:55:21
And when I think about that, it's like that refrigerator that we had as kids with the one big heavy aluminum door, like it just worked. It worked forever.
1:55:29
>> But in order to keep selling you have to keep coming out with new things.
1:55:33
And I think that's the negative of planned obsolescence.
1:55:35
But I think when it comes to consumer technology, there is an argument for having to keep up with the rate of innovation.
1:55:41
You know, like the new phones that come out, every time you get a new phone, it does do something that your other phone didn't.
1:55:49
So, there's a maybe a reason for that, >> but I think with that sort of title wave of innovation that feels faster now than it ever felt when I was young.
1:55:57
Um, I think we we lose some great things. >> Yeah.
1:56:02
>> And you know, my understanding of Mod Retro, at least what I want it to be, and I think what what what Palmer and and what Torren also want to be is like, let's identify consumer electronics that people just loved, like had an unbelievable relationship with.
1:56:15
They really like the love I had for my Game Boy.
1:56:19
I don't know that there was another thing that I owned through my adolescence that I treasured more than my Game Boy.
1:56:23
And let's bring them back and let's do them justice.
1:56:27
Let's build them so they last forever.
1:56:29
I think Palmer's quote when I was interviewing was we don't want to build another Game Boy.
1:56:35
We want to build the last Game Boy, like the last one that ever needs to be built.
1:56:39
And that's the vision with M64 as well.
1:56:42
Like you can get emulators that kind of they kind of work.
1:56:44
Um, but let's let's build the thing that that, you know, you and I had, John, when when we were kids, and let's have it last forever.
1:56:53
>> And I, you know, I just think there's so so much romance in that when you start digging deep, you can start to see products everywhere.
1:56:59
And, you know, Mod Retro is tiny, just a couple people working out of a small office, but like the list of of products that that we want to build, the list of ideas that we have is is endless.
1:57:11
And that's like a that's a really exciting thing to take out of.
1:57:16
>> It's also such an interesting uh company because you're trying to recreate things of the past that were good products and you want to improve on them, but you need to stay true to the original product.
1:57:28
product. Otherwise, you know, what do you I mean, it's not to say that that Mod Retro couldn't one day create, you know, novel to totally novel products, but it is like an interesting constraint and challenge and you actually have to it's it's really in some ways it's easy
1:57:44
because you know exactly what you need to build, but in other ways it's harder because you're taking a product that was almost perfect in its original form and still like you need to make it it needs it should be better otherwise you know what are you doing? >> Yeah. You know, like one interesting >> Yeah.
1:57:56
You know, like one interesting challenge, and I'm not the right person to speak to this, so I'm not the technical guy, but to make the screen match the old Game Boy screen was an exhaustive process.
1:58:07
So, when you buy one of those handheld emulators that looks like a Game Boy, the Chinese-made ones, they have fairly high high pixel density screens on them.
1:58:14
So, in order for them to give the look of an old Game Boy, you're using multiple pixels to emulate a single pixel. >> So, it's inaccurate.
1:58:23
And real psychopaths, I'm not one of them, but I appreciate it.
1:58:27
Real psychopaths on Twitter have done like this super zoom into a pixel, and you realize that Mario is like jumping a half a pixel too low or a half a pixel too high within the frame because of the limitation of the screen.
1:58:41
And Torin and and Palmer's like religious obsession with matching that screen was one of the hardest things to overcome.
1:58:50
So, it was actually harder to build something that matched the quality of a product that's 20 years old than it would be to actually get the latest and greatest screen and just plug it in there. >> Yeah.
1:58:59
I remember Palmer saying something about like a couple of the companies the screen is actually rotated 90 degrees.
1:59:05
So, instead of the scan lines coming down this way, they go across this way and then they just transform it at the last second.
1:59:10
But it's like this bizarre bizarre thing that you would never really notice of, but it's this craftsmanship that actually you feel I feel like and it feels like something that's just like been lost in this like everything's disposable, everything's junk, everything's a knockoff culture.
1:59:27
And it feels like I I don't know.
1:59:27
Like it just feels like it's worth it to have stuff that exhibits craft and taste.
1:59:32
And even if I don't even if I'm not the one that can tell that Mario's one pixel too low when I do the jump, like the fact that the craftsman worked on the product means something.
1:59:44
And I I it's hard to put a exact like quantitative value on it, but you feel it, right?
1:59:50
>> That that is the gamble that is the company.
1:59:52
Said that in my video, but he's like, >> you know, maybe nobody cares about this, but but weirdos like you and me, but that's enough.
2:00:00
And I think it's like there is an argument that's like, well, who cares if it's off by half a pixel.
2:00:06
>> And if you're one of those people, totally cool.
2:00:07
You're well, there's plenty of products in the market for you.
2:00:10
But if you're someone who does care, if you're that weirdo that buys Teenage Engineering products because you just value the integrity of the product so much, you want to have it, >> then, you know, there's a place in this world for companies like Mod Retro. >> Yeah.
2:00:22
What What do you think about um like the temptation to just make like one small change?
2:00:27
Like the chromatic has a USBC port on it. That's a huge upgrade.
2:00:31
But you could easily turn this into like well like you know we have the technology to make the M64 handheld and then you know then you're in Switch territory and all of a sudden it's like a completely different product, right?
2:00:42
a completely different product, right? I look that's everpresent and that's that's the problem with sort of limitless you know like when Nintendo was building this what was almost 40 years ago 35 years ago now they were so
2:00:56
limited by what the technology enabled them to do >> like when you think about the Sega Game Gear I don't know if you've ever had one of those great product >> this big but it was the first one that had really great color screen on it and it took six AA batteries instead of four like the Game Boy. But the battery life
2:01:10
But the battery life on it was like 40 minutes. >> Yeah.
2:01:15
>> And um >> do you remember the N64 like expansion pack that you put in to like double the memory? That was >> Yeah, dude. I remember that.
2:01:20
What about the rumble pack that shipped with Star Fox?
2:01:23
That was like the greatest innovation ever.
2:01:24
Mine exploded when the controller shook. >> Yeah.
2:01:28
Do these like uh like these add-ons cuz like they couldn't bake it all into the first product and they >> up and that's that's indicative of the limitations of what tech can do. Totally.
2:01:37
And when you look at like, you know, I don't know what what Samsung announced last week at two weeks ago unpacked and you're looking at what the limitations are today of like a folding screen that's as thick as 10 sheets of paper, you realize how far we've come.
2:01:51
>> But when you go in the other direction, but there's kind of no limits and you're looking to make something that was made 30 years ago.
2:01:56
The problem is like where do you draw the line?
2:01:58
And I think that's a really uh it's a really interesting thing to try to overcome and confront. >> Yeah.
2:02:05
What can you teach me about the lore of uh in the video camera world?
2:02:06
I was looking at a cool new music video and it had this fisheye lens and this grain and I was kind of digging in.
2:02:14
It seemed like they shot it on like a Sony skate camera with mini DV tapes.
2:02:18
Like this feels like something that I was looking like maybe I could buy one on eBay but it's probably broken and I want someone to go remake it.
2:02:26
But then of course you always have like oh you could just like throw the filter on there in Premiere or whatever.
2:02:31
But what what are some of the what are some of the iconic video cameras or just like creative tools that might uh you know be worth revisiting at some point?
2:02:42
>> Yeah, it's it's such an interesting examination and and your your question is great because my family and I are out.
2:02:49
We're in Massachusetts for for summer vacation and there's a lot of kids around like we went and got ice cream last night.
2:02:54
A lot of teenagers around >> and a lot of these kids come up and they ask me for selfies. >> Sure.
2:02:59
And the amount of teenagers that come up to me and ask me for selfies and they're carrying the point and shoots. >> Wow. >> That we had Yeah.
2:03:04
in like the 2010s, like before your phone camera was good enough. >> Yeah. >> They all carry them.
2:03:12
>> No one makes that camera anymore, which means if you have one, you have an old one. >> Yeah.
2:03:17
>> And every one of these kids has it.
2:03:17
And when I I asked them because I'm curious like why do you carry that?
2:03:21
They're like, I love the way it looks. >> Yeah.
2:03:24
And I think we've reached this point like iPhone cell phone photography.
2:03:27
It all looks so perfect in a way that it almost becomes artificial.
2:03:33
Like I think the iPhone 5 was the last iPhone where when you took a picture just sort of took a picture and then after that every time you take a picture now on any modern smartphone, it takes a picture and then runs it through an algorithm to make the picture sweeter and cleaner and sharper and better for you.
2:03:47
>> And I think there's sort of this yearning for something that feels more real and authentic.
2:03:50
real and authentic. It's like you look at a you're at a store looking at at a picture, you know, a camera from a picture from from 2010 uh or a camera or a picture from an iPhone today is like you're at you're at the grocery store and there's the organic apples and they look they look good but like they're a little up and and and you know they maybe like whatever part of it's
2:04:13
kind of rotten and then you look over and you see the perfect apple >> and like intuitively you know that you're like that it's a little bit too good to be true that Apple over there, you know, and I think I don't know the other thing that's real is >> t removing the camera like adding the camera to the phone has been one of the greatest innovations of the century, right? In in in many ways, but now
2:04:34
In in in many ways, but now removing it, removing the camera from the phone and being able to leave your phone and just bring a camera around is also an innovation, right?
2:04:42
Going going going backwards.
2:04:43
And I think that's the other thing that that kids like and and I've found that's nice if I'm going to the beach with my kids.
2:04:50
It's nice to be able to bring a camera but not have my phone. >> Yeah.
2:04:55
Um my daughter is 10, so this is the first summer she's allowed to like walk around with her friends without a parent.
2:05:01
And she has a flip phone.
2:05:01
And she had this whole argument for getting an iPhone that has nothing on it but the phone and messaging.
2:05:09
phone and messaging. and my wife and I have this discussion and the reason why we said no it has to be the flip phone is because of the camera like it it introduces a new level of sort of attention and why you're in purpose >> and the reason why we wanted her to have
2:05:24
a phone is simply to stay in touch with us no other reason but the minute you introduce this other sort of novel aspect of it it becomes much more of a social dynamic and a social consideration and the way that you're trying to get away from that um but I think just going back to the camera things. This is such an obsession of
2:05:40
This is such an obsession of mine.
2:05:41
I think that we are quick to quick to ignore or look past the relationship we have with with the aesthetic that a camera yields.
2:05:50
And when you see those pictures off those old point and shoots, they look like something specific.
2:05:54
Like when you were talking about that old skateboard fisheye DVD look, it looks like something specific.
2:06:00
You have a relationship with that.
2:06:02
When we see stuff that looks like it's VHS, we think of the 80s or the '9s.
2:06:05
And I think now there's just sort of this generic aesthetic across the board, this like crispy flat 4K phone thing.
2:06:11
And there's this desire to have something that that has some sense of identity attached to it.
2:06:19
And that's why we're seeing these sort of Gen Z types, you know, looking for older tech that's that reflects more of who they are, more individuality.
2:06:25
And I think that's a trend that we're going to see continue. >> Yeah.
2:06:30
It's this generation's vinyl. >> Yeah.
2:06:32
I I I've joked that I want to film I want to film the first podcast on Panavision film or something.
2:06:36
Just spend a fortune developing the film.
2:06:42
>> What other what other what other c consumer tech categories are you excited about?
2:06:46
We're having we're having the founder of Wave uh on and I and I in 20 minutes and they got like 12 million views yesterday for launching like sunglasses that will just live stream constantly.
2:06:59
Do you have a do you have a ratio of um positive to negative mentions in those 12 million views?
2:07:06
>> I mean, I think everybody people had a viscerally negative reaction to it.
2:07:11
>> That's that's my take on it as well.
2:07:13
>> And it makes me think, >> you know, my my software development company Beam, my startup, um we started that company in 2014.
2:07:20
And a little known aspect of that company is initially it was built around Google Glass. >> No way.
2:07:27
And we actually, you know, we we rewrote the code.
2:07:30
We rewrote the the software that Google Glass ran.
2:07:31
So it had a singular function which was you push a button like this, it captures 10 seconds and immediately shares it to your feed.
2:07:40
>> That's where we started with that.
2:07:40
And I remember meeting with Google and they were excited about it and they showed us what Google Glass 2 was going to look like and we're like, "Fuck yeah."
2:07:46
We started raising money based on that and then a week later they announced they were killing the entire Google Glass program >> and we we backed off.
2:07:54
We're like, "All right, we'll just do it on the phone now."
2:07:56
Um, but I think when you ask about sort of the social impact of being able to live stream all the time or the negative reaction that we're seeing based on on Waves uh video they posted yesterday.
2:08:06
I think the social dynamic is something that is often um overlooked in in the hardware startup world.
2:08:13
I think of the humane AI pin.
2:08:16
You know, a really good friend of mine worked for for humane and obviously super excited about new technology. Yeah. Sam.
2:08:22
Um, which parenthetically you joke about Panavision for a podcast.
2:08:26
Sam Schffer has in his studio for his live streaming setup a VHS camera hardwired into his deck so he can live stream from VHS. >> That's amazing.
2:08:38
>> But Sam would be around.
2:08:39
>> We tried to I tried to hire Sam, by the way.
2:08:41
We worked with him on a project and I was like, >> he's way too hot. He's way too hot.
2:08:45
>> He's got to be He's got to be solo.
2:08:46
>> He's way too big of a deal. >> Sam's the best.
2:08:49
But when I'd be around him and he had his AI pin on, I was always sort of looking down at that or like thinking about what I'm saying and uncomfortable about, you know, speaking openly to him.
2:09:01
>> And I think like that's what I mean by social implications.
2:09:03
Like, you know, my phone is on my couch right here, but if we're sitting together and I'm holding my phone like this the whole time we're talking, you're uneasy about that. Yeah.
2:09:12
>> And so when I think of a pair of glasses that are live streaming or at least have the capability of live streaming the whole time, when I look at them, I'm uneasy >> and I think of this, you know, maybe the most extreme example of this is um Apple Vision Pro, >> which, you know, blew my mind when I got that thing.
2:09:29
Like I made the most the most glowing video about it because it was the greatest technology I've ever used.
2:09:34
But later that night, because I made the video the same day I got them, I'm at home using them.
2:09:39
And my wife is like, "Casey, take those things off your face and pay attention to your kids."
2:09:46
And I was like, "Oh, you can't wear these around other people." >> Yeah.
2:09:51
>> And I was like, "How did they not see that?"
2:09:55
>> And I So I I think about that because like we're sort of accustomed to people using their phones and it's kind of okay. Like we're used to this. This is not okay. Yeah.
2:10:05
>> And so when I think about wearables, I think that that is going to be a gigantic thing to overcome which is >> well the crazy so so Amazon acquired a company uh it got announced this week called B which is a brace after after all everything they faced with Siri's listening to me.
2:10:20
The B band just listens to you all day long.
2:10:22
Every single conversation and uh and I and it's and it's cool like I I think all new tech for the most part is cool.
2:10:31
Do I want my team to be wearing a B bracelet?
2:10:33
Are we gonna allow people to wear Bands?
2:10:37
>> Might be a B free zone.
2:10:38
>> This might be a B free zone.
2:10:38
You know, there's >> I'm not gonna have any friends that wear bees.
2:10:43
I don't have any friends anyways, but if you have a be it just means like we can't hang out.
2:10:48
>> Yeah, >> because it makes it makes me uncom.
2:10:48
And look, I think there's an argument.
2:10:50
It's like get over it like you lite.
2:10:54
Like this is the direction everything's headed and you're being recorded all the time anyway. >> Sure.
2:10:58
But in the interim, I do think like there is a there's a learning curve.
2:11:02
I think that Meta's decision to put wearable glasses with a camera on them in Ray-B bands, the most common sunglasses in the world was brilliant because we're all familiar with this pair of sunglasses.
2:11:14
Like there was no learning curve.
2:11:16
There was no like what's that thing on your face? We know what it is.
2:11:20
>> I think how emphatic they are about having the indicator light.
2:11:22
There's no way you can turn it off.
2:11:24
If you cover it, they don't record.
2:11:26
>> There's some degree there.
2:11:26
At least there's Like my my question my question for the the wave founder in 20 minutes is >> should your glasses be bright red and have like you know a circle in the middle of your head that's flashing that says I'm recording. >> Yeah.
2:11:41
I mean if you can >> if someone shows up with a TV camera like I don't feel like oh this is in violation of my privacy.
2:11:46
I'm like I know exactly what's going on.
2:11:48
The news is here and if I step in front of the camera I'm on the news.
2:11:50
like it it it we have a social understanding of that and >> we still might not want to be around it.
2:11:57
>> They might be like actually I'm going to turn around and walk the other way because I don't want to be seen on on this camera or whatever. I don't know.
2:12:02
>> But there's there's an understanding there. Yeah.
2:12:03
And look on on Wave's website in the Q&A there's a question like is there an indicator light?
2:12:06
And their answer is yeah, but you can turn it off >> and so like I you know I don't know that I read that and it makes me a little bit queasy. >> Yeah.
2:12:15
as a guy who has a camera on him, you know, all the time and and so much of the the videos that I've made is about filming everyone in the whole world around me.
2:12:23
I still think there's something honest about having that camera on your shoulder, pointing a camera at someone um versus like, you know, the glass hole >> movement of the, you know, Google Glass, which like by having this on your face, everyone was uneasy around you.
2:12:36
I think it's going to be a big hurdle to overcome as we eventually like figure out what this hardware AI future is going to look like. >> Yeah.
2:12:46
What do you think about uh like airplane demos?
2:12:48
It feels like all there's so much tech where the demo is like you'll love this on an airplane.
2:12:53
The Apple Vision Pro is a great example.
2:12:55
A lot of the a lot of the the handheld >> it doesn't work when it's dark.
2:12:59
>> Uh it's not too dark on planes.
2:12:59
It does work very well on planes, but it's this like weird niche use case that feels like killer application and then you actually get out into most of your life is not spent on airplanes usually.
2:13:08
And so, >> yeah, look, if if that's the best you've got, like I travel more than most people I know and we're still talking about, you know, maybe double digits, maybe 10 hours a month tops. >> Yeah.
2:13:22
>> So, you know, 4,000 bucks for a face computer that I use 10 hours a month on an airplane, which by the way doesn't work for me on an airplane.
2:13:27
Like, I didn't like it at all on an airplane. >> Okay. No, it will work.
2:13:30
Like the technology works perfectly.
2:13:34
>> No, when you're on an airplane, you're surrounded by strangers. >> Yeah. Yeah.
2:13:38
>> So, the idea that like I'm just gonna >> go like that and watch Avatar with all these people around me.
2:13:43
You know, I don't fly Spirit that often, but I have flown Spirit and I know what goes on on those flights.
2:13:49
Like it it just again likeation of Yeah.
2:13:52
the social implication of it.
2:13:55
You know, I think if that's the best you've got, then maybe, you know, maybe you're not thinking holistically enough about your about your hardware. >> Yeah.
2:14:02
>> How do you how do you spot talent?
2:14:02
You you you discovered, Sam, uh someone in the chat was asking uh a great question.
2:14:10
>> You have a framework for it.
2:14:13
It's a it's a it's a difficult question, but you know, it's the same I I approach it the same way that I approach uh investments.
2:14:20
And it's seldom the work, and it's almost always the the individual, you know, like I I think Sam is brilliant, but he's brilliant at everything that he does.
2:14:30
He's brilliant with the work he did at Humane.
2:14:31
He's brilliant the work that he did at the Verge before that.
2:14:33
He's brilliant with his own YouTube channel.
2:14:35
You know, I've got a dear friend of mine, a kid named Hunter, who works in the building with me, but Hunter's a friend and we invest together.
2:14:43
So >> Hunter's Hunter's fantastic and like I don't know what Hunter does because every time I have a conversation with him, he's starting some new business or some new endeavor.
2:14:51
Like the work is almost irrelevant. It's the individual.
2:14:56
And I think it's like it's that mentality that you have or you don't have. Not everyone has it.
2:15:00
I do think maybe you can get it if you don't have it.
2:15:04
But when I see that in people, I I have a tendency to sort of gravitate towards those people.
2:15:09
And when I look at the when I look around at the friends of mine that I have some sort of professional relationship with a professional admiration for, they all have that thing that that sort of desire to uh to see it through whatever the it may be.
2:15:23
So, you know, I guess the short answer is like it's always the person, never the never the product. >> That's great.
2:15:30
Hunter's in the chat, by the way. He says hello.
2:15:32
Um the uh another question I have is we've been thinking about uh AI generated images, video, etc. is getting so good.
2:15:41
If it stays at this pace, you're going to be able to generate characters consistently.
2:15:49
Uh I you're you're already seeing this on Instagram, people just creating entire personalities for influencers in different verticals.
2:15:56
Do you have any type of strong thesis around how um how that will play out if we're going to have sort of organic influencers, regular humans and then you know these synthetic I've been thinking about in the context of um you know an example is like I don't want the AI Andrew Huberman.
2:16:15
I want Andrew Hubman to get health advice from because I know Andrew and I trust that he's making the best decisions and and studying uh you know and making you know these recommendations and and understands the weight of that and you have you have like a synthetic Andrew Huberman in 5 years doesn't care if he could convinces a million people to do something dumb with their health because he can just shut the account down and turn a new one on.
2:16:42
But I'm curious how you think. >> Yeah. Okay.
2:16:44
Okay, my cynical take on it, which is which I believe is accurate, um I'd wager that this is accurate, is that it's a title wave that none of us will will be able to to overcome.
2:16:53
We can hold our breath for a while. It is a title wave.
2:16:58
And I wish I had a better analogy, but it's a little bit like, you know, when Toy Story came out in 1999, it was the first computer animated feature film, and it was great. I loved it.
2:17:05
But then five, six years later, there was just this, you know, onslaught of computer animated films.
2:17:11
And I was like, you know what, the handdrawn stuff is better. I missed that.
2:17:14
When CVI first started getting introduced into films, I was like, no, practical is better.
2:17:18
Um, I like practical better.
2:17:20
That's going to be the thing.
2:17:22
And it's like, no, you're going to lose.
2:17:23
And I think AI is probably the most existential version of that.
2:17:28
I think that it's just going to be so good that it's going to become irrelevant.
2:17:32
And old people like us might appreciate, you know, real human things and we might keep it alive, but the generation behind us, my children, when they see draw handdrawn animation, they're like, "What is this?" They don't care.
2:17:47
It means nothing to them.
2:17:47
So, I think it's just going to get so good that it really doesn't matter.
2:17:51
And that that title wave is going to consume everything.
2:17:55
And um I think that we're we're it's the progress that we've seen over the last couple years of how good it is and how quickly it's developing is incredible.
2:18:05
But when you look at the capabilities of, you know, the latest Grock uh and the latest GPT, you look at how smart these things are, you look at what the sort of conversations you can have with with um any of these agents, and then you look at how good the images that can be spit out by um you know, Google's video or whichever whichever AI you might be fond of.
2:18:24
It's like they're just going to marry really really quickly.
2:18:28
And I think five or six years from now we're going to start to see real sort of erosion in the space because what they're delivering is going to be so good.
2:18:36
You know, like another example is like we haven't had a number one song yet that's completely AI generated, but it's going to come out of nowhere.
2:18:43
It's going to be like gang style, which like the novelty of it allows us to dismiss the lack of humanity that this song has.
2:18:50
The first AI song that's going to be a number one hit is not going to be a Bob Dylanesque song that comes from the soul.
2:18:56
It's going to be something fun and outrageous and we're going to forgive that AI made it and then slowly it's going to start to chip away.
2:19:03
But I think it is a losing battle.
2:19:06
And the optimistic take on that is I think that there is going to be an extreme premium put on the humans that are able to sustain because of of what they represent uniquely.
2:19:18
Um, you know, maybe another bad analogy is it's like I love Marvel movies. They're fantastic.
2:19:25
But as far as cinema goes, like, you know, there's an argument that said it's like the kind of artistry that we saw in the the 70s or 80s, like the apex of cinema where it comes from one when you think of a Tarantino movie, you think of um, you know, a truly great filmmaker with a voice.
2:19:43
>> Marvel movies are something else. Yeah.
2:19:45
>> They're great and we love them.
2:19:46
We all show up to see them and we're okay with that.
2:19:47
And I think that's going to happen with AI, but it's going to happen at a rate of acceleration that we've never encountered before.
2:19:53
And it's going to be so good that we're willing to forgive it.
2:19:57
And I think it's going to leave a handful of people out there.
2:20:00
Maybe Andrew Humeman will be one of them. I hope so.
2:20:02
I'm a fan of his as well.
2:20:03
But it will put an extreme premium on the sort of human voices that are distinctly um human that distinctly represents something that AI could never.
2:20:13
When you look at the landscape of media, a lot of that could be replaced by AI.
2:20:20
>> So, the power law gets steeper. >> Yeah, absolutely. >> Makes sense.
2:20:24
Well, >> that's like a five hour conversation, by the way.
2:20:27
>> Well, we'll have you back on again soon. This was awesome.
2:20:29
And congrats on your new your new uh fake but very real role at Mod Retro. >> I appreciate it.
2:20:36
Thanks for the time, guys.
2:20:39
>> Have fun in Massachusetts. We'll talk soon. >> We'll talk soon. Take care. Have a good one. Bye. [Music]