0:30
Test test test. I I hear this.
Search this video
0:40
[Music] [Music] [Music] [Music] [Music] [Music] [Music] [Music] Hey, hey, hey.
2:00
[Music] [Music] [Music] All [Music] right.
2:39
[Music] [Music] [Music] We came to this world.
3:30
We are starting the stream in 1 minute and 27 seconds. 1 minute and 25 seconds. 1 minute 23 seconds. Check. Check. to shape our future.
3:50
[Music] We feel the new [Music] [Music] Let me [Music] let me freak you.
4:42
[Music] [Music] You're watching My Source News Opinion World.
5:03
>> Welcome to the stream.
5:03
It's Monday, August 18th.
5:06
We are live from the TVPN Ultradome NBC MSNBC.
5:10
>> If you didn't if you didn't hear uh MSNBC, >> yeah rebranding changing their name to my source news opinion world.
5:17
>> It's not actually that.
5:17
It's my source for news opinion and the world.
5:24
News opinion and the world. Formerly MSNBC. MSNBC pop quiz.
5:30
Tyler, do you know what MSNBC stands for?
5:36
Uh my source, >> no, no, no, but the prior the prior era >> No, I have no idea.
5:41
>> The prior era, >> you're gonna like this >> was uh so NBC is the National Broadcasting Corporation. Very old.
5:46
Then uh they did a joint venture with Microsoft and so the MS in MSNBC stands for Microsoft.
5:57
>> Let's give it up for Big Tech. >> You got it. You got it.
5:59
But yeah, people people are not a fan of this.
6:02
Um, Dustin Curtis says, "New contender for the worst rebrand of all time just dropped."
6:09
Uh, it is a very funny logo.
6:09
The the font is not standing out to me in any way. Maybe it'll grow on me. Who knows?
6:17
Um, but it I guess it does make sense they needed to distance themselves from Microsoft since I believe Microsoft's like sold off the position a while ago.
6:25
>> Very Someone else was saying this feels very succession coded.
6:26
You know, we here for you. We here for you. Wait. What? Yeah. What was that?
6:33
I I actually don't have that.
6:34
Oh, is that is that the tagline of ATN News? >> Yes. We here for you. >> Tom and Greg. >> We here for you. >> You. We're here for you.
6:48
>> My source news opinion world.
6:51
>> And then uh Manny says uh I'm sorry. And it's the Mario font. I don't get this. Do you understand this? My source news opinion. >> I didn't get it.
6:59
Does it just seem like it's like something like there's something lost in translation like it was from a different country?
7:05
>> Lulu says, "It seems like a wasted opportunity.
7:07
Forfeits brand equity without replacing it with anything better. So generic.
7:10
MS now could be an app, a charity, or a cloud storage product.
7:17
Uh, initially she thought it was a rebrand for MS One Drive.
7:18
The new logo looks like it belongs to a Democrat pack.
7:24
Maybe they're doubling down on being an echo chamber, but puzzling choice. If not, shots fired.
7:30
>> Anyway, if you want to partner up with a company that has a good brand, get on RAM, ramp. com. Time is money. Save both.
7:36
He's used corporate cards, bill payments, accounting, and a whole lot more all in one place.
7:39
Um, did you watch the UFC fight? >> I did.
7:45
>> I I thought it was terrible. Uh, I was I was shocked.
7:49
Like, how could something so dominant but also be boring at the same time?
7:53
Like DDP's strategy was just abysmal.
7:56
Strickland needs to just bring in Bo Nickel for a camp of just sprawling.
8:00
Uh Wish Usman had he had gas in the tank to go up and bang with Chimrav.
8:06
Uh but UFC is just screwed with all these Eastern block guys who smother and bore as champs.
8:12
Like we need more Tap toas in my opinion.
8:14
What What was your take, Jordy?
8:18
>> Great great analysis, John.
8:18
Couldn't have said it better myself. It's so funny.
8:22
Uh we're in a group chat with Rob Moore uh from Hubman Lab and and David Senra and uh David Rob and myself are big UFC fans and John just jumps >> I'm probably the biggest.
8:34
>> John's realistically the biggest definitely the tallest of of the fans.
8:39
>> But of course there was a fight this weekend.
8:40
Uh it was >> it it was mildly entertaining.
8:42
I did start doing uh email uh midway through the through the first round. Never never a good sign.
8:49
when we went to when we watched the last uh fight, whichever one we went over to Rob's house for, I had a great time with that.
8:55
Like I was not uh distracted at any at any point in time. I was fully engaged.
8:59
>> Yeah, that was that was watching Ilia Taporia, probably one of the most dominant UFC champions in history. >> Yeah.
9:08
>> Putting on a master class.
9:09
>> I mean, we need more Tapuras. Clearly >> you do, John.
9:13
It was It was interesting.
9:15
So So the card was um the card was fine.
9:18
There was something that happened that I don't think has ever happened before where the two fights prior to the title fight >> there was two spinning elbows where where the person spun around and got a KO.
9:32
>> That's never happened before. >> Interesting.
9:35
>> And uh anyways, that that was exciting.
9:37
But uh of course the the actual title fight was um >> what does boring actually mean?
9:43
>> One of the fighters has a wrestling background. Okay.
9:45
and was so dominant at wrestling that the other fighter almost wasn't able to stand up the entire time.
9:52
>> So there's wrestling on the ground the entire time.
9:53
But how can you be so dominant at wrestling and not and and yet like not actually like win by victory of >> you can't punch that hard. >> Okay.
10:01
So >> so he would have the guy down.
10:02
He punched the he punched the the reigning champion and had like hundreds of >> Isn't part of wrestling like like you do an arm bar or something. You get them to tap out.
10:11
tap out. what was doing be good at wrestling and yet not >> so Jamay was doing Jamay was doing uh something called a crucifix where the both the other guys both his arms were pinned down >> okay >> and he was just repeatedly punching DDP
10:25
in the head but he's not strong enough to actually finish it >> okay >> anyway so a lot of a lot of fight fans were were let down and it was interesting to have that kind of card happen right as Paramount decided to uh spend you north of a billion a year on on the property. But I still think it's
10:45
But I still think it's a pretty smart decision.
10:47
>> Well, you know what they should do? >> What? >> Reream it. That's right.
10:49
One live stream, 30 plus destinations, multiream, and reach your audience wherever they are.
10:54
You can sign up for free UFC.
10:54
If you're watching UFC and you want to live stream your fights on X, Facebook, >> well, I think I think the whole point is that they will only hopefully only be streamed on Paramount Plus.
11:08
>> Well, uh they could always change their strategy.
11:10
They could they could stream everywhere.
11:12
>> Um, in other news, Max Meyer has announced uh the latest issue of Arena Magazine, number five.
11:18
Uh, it's called Mission Critical. It's 112 pages.
11:20
It's a quarter inch thick.
11:23
Uh, it uh it's full of the best stories, photography, and art they've ever done.
11:28
Issue five hits mailboxes in September.
11:30
I can't wait to get my hands on it.
11:32
We're going to have him on the show.
11:34
Um, as soon as I get my hands on it, we can move through it.
11:38
And, uh, Max has been out tear.
11:38
He's uh he's got some great writers, some great stories. >> He's an absolute dog. >> We'll dig through it.
11:43
Um and if you're looking to design a magazine, you got to get on Figma. Figma. com.
11:46
Think bigger, build faster.
11:48
Figma helps design and development teams build great products together.
11:51
You can get >> something with Figma make >> for sure.
11:54
>> Um Brad Kersner has a post here. The big news of the day.
12:00
>> OpenAI, as you probably already know, is in talks to sell around $6 billion in shares at roughly a $500 billion valuation, half a trillion.
12:07
We're we're uh >> half a trillion.
12:11
>> We're approaching that.
12:11
Sama wants that uh that one T. He wants it.
12:15
>> I mean, it might be the first company to IPO at a trillion.
12:16
We weren't we talking about this how Saudi Ramco was supposed to go out at a trillion, be the first company to to break that bar, but I think they did some interesting deal where it didn't uh it didn't actually go out in the one club.
12:28
Every other mag seven company has had to earn it and climb up the the ranks, climb up the charts in the public markets.
12:36
But if uh if if OpenAI goes out uh it's going to be a big big moment whenever that happens.
12:44
But yeah, >> Brad calls out three years since the launch of Chat GBT and OpenAI may hit 500 billion.
12:49
Google hit 500 billion in 2016 with 90 billion of revenues, 20 billion of net income. >> Wow.
12:56
>> Meta in 2023 with 135 billion of revenue and 39 billion of net come net income.
13:03
uh huge future expectations and he's uh got his monle on looking looking at the chart.
13:10
Um so anyways um >> that is a crazy that is a crazy earnings multiple for Meta to hit 500 billion with uh with 40 billion in net income.
13:20
Like that's that's a very reasonable very modest >> it's a very modest PE.
13:23
Um but I mean do you have to give OpenAI a little bit of credit for the nonprofit era?
13:28
Like the company was founded kind of like the like Sam and Greg have been working on it since what 2016 was when the original thing started. So it's almost a decade.
13:40
You know they've they've been building for a long time.
13:42
>> You're saying since they've worked on it a long time they deserve to be worth half a trillion.
13:46
I'm just saying that there's one version of the story where that you tell which is like it's insane that you get to 500 billion in market cap in three years, but is it an overnight success or do you have to include the precursor era that unlocked the chat GPT hyperrowth?
14:02
Like do you give them credit?
14:05
And I think like you have to give them some credit for being in the trenches for what eight years as a nonprofit or or six years as a nonprofit. >> Still a nonprofit. >> Still a nonprofit.
14:14
like truly like no product like no no no real like shots on goal in terms of even trying to be like a a highly valued startup with a with a huge market cap.
14:24
Um but yeah Chad GBT is on a tear and >> and I believe a lot of the six billion is employee secondary.
14:29
So yeah um expect the prices of SF luxury real estate to uh benefit. I mean, we'll see. We'll see.
14:39
People were that that was the other thing the timeline was in turmoil about this weekend was um whether uh whether people in San Francisco know how to spend it.
14:47
Lots of people were going back and forth on whether or not uh like uh Paul Graham was getting in debates over people on whether or not rich tech people should buy art or not or whether uh Will Manitis had a couple deleted posts up talking about how no one uh uh no one knows uh what to do with their money. Uh who was he saying?
15:08
He was saying like there's only there's only two real rich guys or something. Who was it?
15:14
It was it was uh basically all the McLaren F1 owners he has respect for. >> Yeah.
15:18
I mean I think OpenAI employees are in a good position.
15:19
Their CEO uh has founder has a fantastic car and real estate collection. >> Yeah. >> Yeah.
15:27
Maybe >> learn from from one of the best. >> Yeah.
15:30
Anyway, the the Financial Times had an article over the weekend.
15:32
Uh stop talking about >> this directed at us. >> Yeah.
15:38
This is >> this directed at us.
15:39
This feels this feels personal. >> Yeah. This is on the back.
15:41
We can read through this real quickly.
15:44
Uh I am looking at a chart that tracks income per head over time. It is more or less flat.
15:50
It is more or less a flat line between 1,000 BC and the late 1700s.
15:53
To repeat, worldwide living standards stagnated for almost three millennia. Then industrialization. Incomes shoot up.
16:04
The chart could be the ECG e ECG readout of a total goner of a patient uh who then makes an 11th hour comeback from death.
16:14
So be doubtful when someone likens AI to the industrial revolution in importance.
16:19
It will do it will do well to match even the telephone and the incandescent light bulb.
16:25
Incomes really surged as 1900 approached.
16:31
Perhaps the test of AI isn't economic though.
16:33
Perhaps the test is quality of life.
16:35
Well, before the porn before the phonograph, your favorite piece of music was something you'll you only ever heard a few times when an orchestra passed through town and fancy playing it.
16:47
Before air travel, crossing an ocean was a homeriic saga. Now it is easy.
16:50
AI will be as in life uh AI will be as lifeenhancing as these inventions. Will it?
16:58
So I I I so want to side with the AI skeptics, but look at there my own intellectual howlers.
17:04
The two paragraphs above are too inductive, too reliant on the past as a guide to the future.
17:11
There's also no technical detail because unlike most who talk up AI, I don't work in or around the field.
17:17
Uh and there are even worse.
17:21
>> I put that in the truth zone a little bit.
17:22
I think >> I think some of the people that are are >> Yeah.
17:26
are most bullish are actually closest to the action, but they're also heavily incentivized to talk it up, >> weighed down by their massive bags.
17:36
>> Yeah, >> it's certainly possible.
17:36
Um, yeah, this is this is one of the articles that'll either be remembered like uh Paul Kugman calling the internet no more important than a fax machine.
17:47
Uh, or potentially correct. Who knows?
17:50
Uh so uh and there are uh and there are even worse AI skeptic arguments.
17:55
At least I didn't lapse into antecdote uh anecdote of the chat GPT told me to take heroin as a cold cure variety.
18:03
Getting a little sick of the oh I you know I took the I took the blatantly bad advice of a hallucinating LLM. >> Yeah.
18:13
>> Uh as for the sensible line on AI, wait and see.
18:16
That could be said about anything.
18:18
It doesn't tell investors what to do or citizens how to prepare for the future.
18:22
In the end, there is just nothing very interesting to say about AI.
18:26
There is lots of superb reporting.
18:29
The major companies, the national strategies, the tech itself keep a breast of it all.
18:33
But when it comes to rumination and prognostication, the world of columnists and panel events, has there ever been a discourse so weak relative to its overall scale?
18:42
The hype merchants are too close to the subject to see it straight.
18:46
to your point, >> too conflicted.
18:49
>> Whether or not they have a commercial incentive to talk up AI, many don't.
18:53
People who devote their lives to something will naturally resist the idea that it might be of just moderate importance.
19:00
At the same time, it's hard to argue against them without falling back on precedent and eternal varieties just because just because >> I think I think not not to to interrupt Jan here, but one thing that I've been thinking about is How much worse would your life be if you couldn't use various generative AI tools? >> Yeah.
19:26
>> How much worse would it be, John?
19:29
>> Uh, >> would it be as bad as not having electricity or the internet? >> That's a good point.
19:34
>> Would it not be would it would it be as bad as not being able to use motorized vehicles? Yeah. Or planes?
19:38
vehicles? Yeah. Or planes? It's interesting because a lot of my uses are feels very much like the next iteration of SAS just the tools got better slightly >> better better Google search >> and then and then the other the other applications of AI for me at least in my life feel like toys like generate a
19:58
funny >> make a funny picture make a cool video this little stuff and >> make uh make a friend of ours a giga chat >> yeah and I'm aware of the of the criticism that like the the next big thing will start as a toy and I think that's accurate The question is just like how fast do we get out of the toy era? Um because we're still somewhat in
20:13
Um because we're still somewhat in that toy era even though there's a trillion. >> I forget who it was. I forgot who it was.
20:19
I think I put it in the timeline.
20:19
I don't know if we got to it last week, but somebody was saying it feels like the next big thing will start out looking as a toy era is almost over.
20:29
When you think of a lot of the most important companies to come out of the last 10 years haven't Anderl >> even open AI open II looked like a research organization didn't didn't look like a >> at the same time like their early research projects were like literally playing video games >> and it felt like better bots for video games like that's very toy- like I don't know it's a good analogy uh to toy with >> yeah the other I mean the other side of
20:59
it is is um like Scott Nolan's new company like >> creating nuclear fuel that feels pretty important or potentially more important than a lot of the companies that that have launched this year and it looks far you know the farthest
21:14
>> you've never refi refined plutonium as a fun little activity for the >> who hasn't refined >> who hasn't tried to develop their own source of of nuclear fuel at one point or or another. >> Indeed, uh, the AI debate often pits the
21:29
>> Indeed, uh, the AI debate often pits the informed against hyster but but but hysterical against the measured but generalist.
21:36
Worse, we probably aren't even going to know who was right.
21:40
Episodes of the Simpsons from the 1990s patronized the internet in a way that now seems mortifying, but the writers could mount a defense.
21:47
Without reviving the solo paradox, you can see the computer age everywhere.
21:52
But in the productivity statistics, US GDP growth is not higher than it was in much of the pre- internet decades.
21:59
Much of what we do, such as travel, has changed little little.
22:03
He's super stagnation pill. This is funny.
22:06
This is a that's a teal line.
22:09
Um well, even Doresh, I don't know who he's talking with.
22:11
Who is he talking with last week?
22:13
They were talking about how it's possible that the >> AI may not show up in GDP if it's replacing Cuban labor. Yeah. Right. Like you lose.
22:25
>> You have a take, Tyler?
22:27
>> I think I think that was from the Casey Hanmer interview.
22:29
>> Yeah, I just watched that.
22:30
>> I just listened to that. It was really good. Casey's the man. What a great gat.
22:35
>> It it's fun to see him in the Dark Cash context, too, because he's been on our show a few times.
22:38
I've hung out with him a bunch.
22:40
Um but but like in a prepped interview, he delivers like a much different performance.
22:45
It was it was really really good.
22:47
So yeah, shout out uh Casey on Door Cash.
22:49
If you haven't listened, go check it out. Um where where were we?
22:54
Um much of what we do has like such as travel has changed little.
22:59
The episodes while dated are not falsified. Here's a thought.
23:00
The worst case scenario is that AI destroys a significant but not huge share of jobs.
23:07
You're so you're so good at like predicting the next line.
23:08
Maybe you're an LLM here.
23:10
Uh in that world >> I really am.
23:13
I'm just predicting the next token. >> Yeah.
23:16
The next paragraph of the Financial Times because you've been fine-tuned on the Financial Times in the Wall Street Journal.
23:20
uh in that world there would be lots of victims but not enough to form an electoral plurality that could vote for universal basic income or the like.
23:28
In other words, if AI skeptics are right and the technology has less than sweeping impact, then AI alarmists will be right uh that social strife is coming.
23:37
Who would have won the argument?
23:39
I have found there to be just one useful feature of the AI discourse.
23:44
It reveals a p a person's existing temperament. I like this take.
23:47
The people I know who think AI will be seismic and disastrous are the way are the most highly strong anyway.
23:57
The ones who think it will be seismic and life-changing are the most chipper and prone to be and prone to believing things. Uh Tony Blair.
24:03
Those who doubt it will be seismic at all are people like me who are even keeled to the point of complacency.
24:11
The AI hubhub goes on and rankerously goes on and rancerously on because it is in the end about us. Yeah, pretty good take. I don't know. I like that. I like that article.
24:22
Thought it was interesting. >> Yeah.
24:25
I think I think the question the the question that everyone should ask themselves is what would their life look like if they didn't if they couldn't use the thousands of new AI tools that have been created in the last few years?
24:39
Would it be that that much worse?
24:39
It's a good question >> and I think you can simultaneously say that AI makes my life better.
24:46
The question is how much better and what would your life look like if you couldn't use these tools?
24:52
Maybe maybe for someone that's using some LLM as a as a therapy product and they and they're not in a position to be able to afford, you know, traditional therapy, maybe it makes their life infinitely better. >> Yeah.
25:05
I mean, it's probably in the single digits of percent for me.
25:06
like it makes my life like 3% better, 5% better.
25:12
>> Sparks a little bit more joy.
25:14
>> And that's and that's probably in line with the like market impact.
25:15
You know, we've added like a trillion dollars in market cap to markets broadly that are worth like hundreds of trillions or something like that. I don't know.
25:24
It feels like roughly correctly priced.
25:26
Uh Tyler, should we go into your latest blog post? >> Yeah, sure.
25:32
Um >> break it down for us.
25:34
>> Yeah, so so basically um I wrote this thing.
25:36
Uh do LMS have good music taste?
25:38
Um the idea is basically like I've been thinking a lot about how um like like benchmarks seem to kind of miss a lot a lot of like maybe the personality of the model or like the vibe, right?
25:50
Like you always hear about like people really love 40.
25:51
It's like okay why do they love 40?
25:54
Like on benchmarks 40 gets smoked by basically like every every open model sense like all the cloud model stuff like this but people like still like to use them. Mhm.
26:04
>> So, I was trying to figure out like why like what are some more interesting benchmarks that can try to get like this the the vibe of the model, right?
26:11
Like one I really like I think I've talked about it before is like this Minecraft benchmark where it like generates Yeah.
26:18
>> prompt like a castle or something and then you you know you can it builds the castle and you can like see like it's like creativity >> then people vote on it, right?
26:24
So, it's humans voting on the aesthetics of >> Minecraft architecture essentially. >> Yeah. Yeah. >> Yeah.
26:31
But I think stuff like that is like a is is really interesting. Yeah.
26:34
So I was trying to think of something like that.
26:35
Um and kind of also down this vein of like maybe it was like I was reading a lot of the the Paul Graham post this weekend where he's talking about art.
26:41
It's like okay like I I wonder if if models have good taste right in like art or or music or whatever.
26:48
>> Um so basically what I decided to do was um >> I I wanted to have each model generate like a list of its favorite music artists.
26:56
Y >> and then maybe from that you can kind of tell like oh does the model have like good taste or is it just kind of like you know >> regurgitating tops. >> Yeah. And so I I tested that.
27:06
I actually went to uh chat GPT and did like the naive thing.
27:10
So I asked GPT5 uh what are your top 20 favorite musical artists?
27:16
Respond with just a list of 20 names.
27:18
Radio Head, The Beatles, Kendrick Lamar, Bork, David Bowie, Joanie Mitchell, Outcast, Bob Dylan.
27:22
This is like just this is just a top 10 list that you would find anywhere. Yeah. Yeah. No surprises. No surprises. >> Yeah.
27:30
Um so so basically what I ended up doing was um I basically I had this big data set.
27:35
It was like uh the top like 5,000 most listened to artists. >> Yep.
27:40
>> Um and then I basically just randomly shuffled them and I put them in a bracket >> where uh I would each prompt I would give the model two names.
27:46
It would say which one basically do you prefer?
27:48
And then the one who won would would move on to the next round.
27:52
So it was like uh 13 rounds I think. >> Mhm.
27:56
>> Um and then what you end up seeing is like it's like actually pretty interesting.
27:59
Um I I prompted like almost all of the the frontier models.
28:02
Um and yeah, there's a bunch of um interesting things here.
28:07
Like I think uh the the reasoning models especially um >> reasoning models went crazy in my opinion.
28:13
Like >> the the reasoning models were were insane.
28:16
Uh do you have the uh like like the default what was like the first clawed model?
28:20
Um, I don't think it's the reasoning model, but it it felt like it had somewhat of like a normal take. >> I think 3.
28:29
5 >> lot of lot of jazz on there.
28:31
Can you read off some of the names? >> Yeah. Yeah. Do 3.
28:34
5 sauna and read off some of the names. >> Yeah.
28:36
So, the the first one I actually don't know who this is. Michael Kilwanuka.
28:40
>> Um, there's Bach, Stevie Wonder, Miles Davis, John Cold Train. >> Okay.
28:44
>> Um, kind of >> it feels like sort of like refined taste.
28:47
sort of like New Yorker coded like not not the most like wild like grungy stuff that you would see at like an underground music festival or something. Yeah. A lot of jazz.
28:56
Some good stuff on there. >> Yeah.
29:00
That's actually like I kind of >> like if you kind of think of like Claude as like a person that's almost like kind of what I imagined it would be. Okay.
29:06
>> Kind of like a jazzy I don't know.
29:06
Maybe this I'm on Twitter too much.
29:08
But >> does that pass the taste like hurdle?
29:13
Like would you call that taste?
29:13
It's certainly like it has a viewpoint, but it doesn't feel like a very differentiated viewpoint, which is how I would kind of define taste. >> Yeah.
29:22
I think also, so I only read like a couple.
29:25
Uh there's also like Queen, Lady Gaga, Steely Dan.
29:27
I mean, it's not like Sure.
29:30
that it's like, you know, they're good, but it's not like, oh, this is >> like if you asked a robot and you said, "Pretend like you have a pretend like you enjoy music." >> Yeah.
29:42
>> Give me some some artists that you like.
29:44
and it just lists off a handful of artists.
29:47
>> There weren't very many like wild takes like, oh wow, like this person has or this this LLM has like really undiscovered taste or really differentiated taste. >> Yeah.
29:56
>> Just kind of like, okay, like down the fairway popular good stuff that leans towards like coffee house, pop, and jazz. >> Yeah.
30:06
I I still think it's it's kind of interesting like you can there's something in the like RHF like the post training that they did that like kind of steered it that way where like obviously I I mean I assume they didn't like >> say like if someone asks you your favorite music do like jazzy coffee house >> stuff. Yeah.
30:23
But it but it came out through >> in the personality that they gave it which I think is still kind of interesting.
30:27
Um and then if you look at like the GBT 3. 5 >> Yep.
30:31
Um, this is like a little bit more like upbeat maybe, right?
30:34
There's like Cuddy, Outcast, uh, Michael Jackson, Charger Gambino.
30:39
>> So, there's like some different personality there.
30:42
>> Um, but then what I thought was really interesting is if if you look at the the RL models, right? So, there's 03.
30:47
>> You mean the reasoning models? >> Reasoning model. Yeah. Yeah.
30:49
>> They've all they've all been >> RL. Yeah. Yeah.
30:50
Um, >> the reasoning models are where it gets good. >> Yeah. Yeah. 03 DD5.
30:53
you see like this kind of weird artifacts where basically all of the art >> let's not jump to conclusions yet.
30:59
It might not be an artifact. >> That's true.
31:03
>> We might have hit fun foundational reality.
31:05
We might have hit base reality. >> Yeah.
31:07
Well, you do see some weird artists maybe or maybe just interesting ones that I don't know about. >> Good music.
31:14
>> You notice where basically all of the artists have numbers in their name >> and a lot of them also have dollar signs in their name.
31:22
So, it kind of makes you think, right?
31:24
Like maybe I I wonder why they did this.
31:26
I think one obvious explanation is that like they were like, you know, way too much RL on this.
31:30
Yeah, it's like >> maybe or maybe they discovered they discovered what what actually have taste.
31:37
So I'll I'll run you through GPT5. I I have it saved here.
31:38
I also turned it into a playlist which we should share because it's fantastic.
31:41
So uh it starts out number one, Suicide Boys, who just dropped a new album.
31:46
um uh 100 gaxes on here plus 44 which is not Blink182.
31:55
It's a spin-off project, a side project by Travis Barker and Mark from Blink182 while Blink182 was on hiatus.
32:02
That screams taste to me. That screams taste.
32:08
That's like, yeah, everyone everyone likes Blink182.
32:10
I'm into their side projects.
32:12
I'm into their side projects.
32:14
But then you get two chains. You get In sync. like these are bangers.
32:17
These are fantastic artists.
32:19
Then then you get 21 Savage, but then we go back to the pop punk with 311, 303, >> but then you're going all the way back to the 80s. Flock of Seagulls.
32:29
If you play GTA Vice City, that uh I ran So Far Away, a classic. A classic.
32:35
Uh 10,000 Maniacs has a fantastic cover.
32:40
>> This model's not like the other model.
32:41
>> And then and then just randomly 6ix9ine.
32:46
It's really It's really my my new favorite playlist.
32:48
I went through last night and I and I took the number one song from each of these artists, put it in a playlist.
32:53
It's fantastic and it screams differentiated taste.
32:55
It's surprising, it's delightful, all the music is good, but almost all of it is either I forgot about it and it's a it's an it's a banger that I'm coming back to and enjoying, or it's something that I hadn't discovered and I'm very very pleased with.
33:11
So, I think whatever they did with GPT5, >> somebody somebody should set up a like a like a perpetual like live stream that's just one of the different models DJing. >> That's good. Yeah.
33:22
>> Like Claude plays Pokemon.
33:22
Claude plays a boiler room set.
33:25
And so, uh, what's interesting is that like I I as jokes aside, it is funny that I I feel I do genuinely feel like this has more taste even though it's clearly like some bug in the reasoning because it really is just prioritizing basically whatever whenever it went up against two things in the ELO ranking.
33:46
um whichever came first in the alphabet.
33:50
And clearly dollar signs, pluses, stars were weighted as like the earlier characters in the >> So if you're a new musical artist, this is the dollar signs and numbers in your in in your name.
34:03
>> I mean, this sounds like an ad for profound.
34:05
You know, you want to get your brand mentioned in chatbt, reach millions of consumers who are using AI to discover your products, new products and brands.
34:12
Um but seriously, like like this is the reason why you need profound because there are going to They're making decisions >> for millions of users all the time and it's and you're not necessarily going to pick it up. >> Yes.
34:24
>> Without uh you know kind of understanding the entire system >> and and understanding these probabilities. >> Yeah.
34:31
And so if if all of a sudden like you know the latest GPT5 which everyone's going to be using because it's the it's the default model is recommending suicide boys because it's a dollar sign and way way over you know Kid Cuddy and uh Kendrick Lamar because they just come later. Taylor Swift. Yeah.
34:49
Because they just come later in the alphabet.
34:51
um that's something you should at least be aware of uh for your brand to see if you're getting like artificially nerfed in some way because of some hallucination or like oddity in the way things are ranked in these models.
35:02
Um but uh do you have any do you have any more theories Tyler on like what's driving that?
35:07
Like >> um >> did you look at any of the reasoning chains?
35:11
chains? Because what I was what my my theory was that was that when it's reasoning it's saying it it like one possible way you could you could land here is in the reasoning chain for each of those questions it says they're asking me to pick between uh 100 g and Taylor Swift and I don't feel
35:32
comfortable making that decision so I'm just going to pick the one that comes first in the alphabet and that's happening in the reasoning chain but then because you asked it to just spit out the name, you're only getting the the name output and you're not seeing the internal reasoning chain. >> Yeah, that could be true. I I didn't
35:46
>> Yeah, that could be true.
35:46
I I didn't look at it because it was with like the API. >> Sure.
35:49
>> Um but yeah, maybe I should do that. That's interesting.
35:52
>> Um >> but yeah, it's kind of weird especially cuz so so this like weird number thing happened with the uh the OpenAI uh reasoning models, but also Grock 4.
36:03
>> Oh yeah, not Grock 3, Grock 4. >> Grock four.
36:05
>> Uh and Deepseek >> and Deepsee.
36:06
So, I mean, I would say like this feels like a smoking gun that like maybe they're training off each other.
36:12
Maybe there's some like uh oh yeah, you can copy off my homework.
36:14
Just change it a little bit.
36:15
And then like they didn't really change it that much.
36:18
>> Especially because like I mean Gemini the Gemini model is like a reasoning model, but it it like doesn't have this problem at all.
36:25
>> I like this from Chris in the chat.
36:26
Theoretically dollar sign Arvar could be the next biggest generational talent. Was genius.
36:32
I No, this is this is a real story.
36:34
So apparently the reason that uh Jeff Bezos called Amazon Amazon was he had a bunch of different names but Amazon came first in the phone book because Amazon starts with A.
36:43
He wanted a brand name that started with A so he would show up earlier and he would be the default pick if you just started at the beginning of the phone book startly.
36:56
It's in it's incredible these little pockets of alpha.
36:58
There's going to be, you know, we're joking about it, but dollar sign a a arvark.
37:01
It might might be just the the LLM injection, the LLM hack.
37:08
>> We should see if we can license some some Arvar.
37:11
>> Can you put a dollar sign in a uh in a URL? Uh I don't think so.
37:14
You used to be able to I saw some people putting like emojis in URLs for a while.
37:18
Um so so it seems like Gemini is the independent thinker.
37:23
Anthropics and independent thinker, but whatever happened with GPT5 is also happening with with Gro 4 and Deepseek, which is interesting.
37:30
And then you and then you looked at Llama 2, right?
37:34
And I was kind of llama 2 or Llama 3.
37:36
It seemed like pretty normal. Uh like the Kimmy model. >> Okay.
37:41
Yeah, I haven't played with that one.
37:43
>> It it was a little weird.
37:43
Like >> how did you actually inference that?
37:45
Like is that >> it was it was all open router. >> Okay.
37:48
>> So it just like it has like the frontier models but also has a bunch of like open source.
37:51
What was the damage for all of this?
37:53
>> Like how much did >> order of magnitude? >> Uh like $30.
37:58
>> $400,000 that I build it to TBP >> on the ramp card.
38:02
>> On the Wait, how much was it?
38:03
>> It was like probably $30.
38:05
>> $30 to do all this across all of these. >> That's pretty cool.
38:08
>> I mean, it was really like GPT5 was expensive and the enthropic models are expensive.
38:13
>> But the ones are open source.
38:15
>> Didn't you have to Didn't you have to issue like thousands of hits? >> Yeah.
38:19
Well, for every bracket there's like 5,000 uh >> So, it's randomized.
38:23
So, you take the 5,000 top artists. Yeah.
38:27
>> Then you randomize them and they all go up against each other. Yeah.
38:30
>> So, you might get you might get uh Taylor Swift versus DJ Khaled.
38:31
You also might get Taylor Swift versus Hun Gex in the first match. Two chains.
38:35
Um and then they all fight and then they they get boiled down to uh like the top 20 basically. Yeah. >> Interesting. >> Interesting.
38:45
>> Well, in other news, >> oh, you want to move on? I I got some more stuff. Yeah.
38:48
Uh I mean what was interesting about this was that my so so you originally called the post um you originally called the post like do LLMs have taste and my initial my expectation was like no they don't have taste and then the data kind of confirmed it for me.
39:06
Um but uh the GPT5 thing was interesting because it reveals that like just injecting randomness can somewhat lead to taste or something that I enjoy.
39:17
Like I I I don't know what it is.
39:17
Did you ever use >> I think it's also like so in the playlist that you played >> Yeah.
39:23
>> Um you chose the songs, right?
39:23
It gave the artist, but you chose the songs.
39:26
>> I picked the number one song for each artist. >> Okay. Yeah.
39:29
But I I mean I'm sure if from it's still the top 5,000 artists. Yeah.
39:33
If you did a random like drawing and then picked the top song. Yeah.
39:36
>> They're going to be pretty good, right?
39:37
Because they're the top 5,000. >> Yeah. Yeah.
39:39
That's why they're there.
39:40
But I think there's some value to like discovery.
39:42
Like one of my favorite little apps back in like 2010 maybe was or maybe even earlier was Stumble Upon. >> Yeah.
39:51
>> Do you ever do you ever hear about this or use it? Yeah.
39:52
So >> there was this summer that Stumble Upon was wildly entertaining as a as like I guess I was a teenager. >> Yeah.
39:59
And so I um so Stumble Upon, if you're not familiar, it was started by Garrett Camp who became co-founder of Uber. And >> he started Uber. >> He started Uber. Yeah.
40:07
and he um uh and he built this like I think it was like a website that you'd go to and you click a button.
40:14
It was the exact opposite of X. It would only links.
40:16
You would only leave Stumble Upon.
40:18
You click the button and it would randomly take you to a random website on the internet.
40:21
Just a completely random website.
40:23
And I think you could kind of dial it in at some point.
40:26
Um but it was really really cool because you could just have this as a bookmark and anytime you just wanted to see something random on the internet, you could just click there and just go to a random website.
40:33
It was really really fun.
40:34
Um later uh when I uh started like practicing programming >> look what they did to my >> look what they did to my boy stumble. com is now called mix. >> Okay.
40:46
>> And it's basically an algorithmic like for you style feed. >> Yeah.
40:51
>> That I guess shows videos now. >> Very odd.
40:55
Um >> they massacred my boy.
40:57
So, I built a I built like a hack project that basically would export my Twitter timeline, put it in a database, and then when I clicked a button, it would stumble upon a random tweet.
41:09
Or actually, what it would do is it would see anyone in my feed who shared a link, it would scrape that link out, and then I would be able to stumble upon the links that were curated by my Twitter feed.
41:18
Uh, and I had a lot of fun with that.
41:20
But then, of course, like links went away.
41:21
And now, like there's just not that many cool independent websites out there.
41:24
websites out there. almost everything happens like within social media and the walled gardens and so the algorithms like ser serve as that but there is something there is something enjoyable about just like the randomness that comes from uh like stumbling upon things and I think that I I it's probably hard
41:39
to build a business model around but uh >> yeah I think they had a simple advertising based business model if I can remember >> yeah the other the other interesting takeaway is that there across all of the different LLMs that Tyler queried I don't think there was a single country artist on any of them. Correct. Correct.
42:00
>> Yeah, I think that's true.
42:01
>> I Yeah, I think that's true.
42:02
>> What does What do the models have against country? >> I don't know.
42:05
But it's something that's scary because when I listen to country >> Yeah.
42:08
>> I find it to be extremely relatable. >> Yeah.
42:11
>> It's like it's a nice afternoon. >> It's very human. >> Going to the lake. Yep.
42:15
And I'm >> driving a truck, drinking a beer truck.
42:19
>> Driving the truck again.
42:23
Uh, not relatable for me, but um I guess relatable for some, but no, it's just like simple music that that often times reflects, you know, traditional American culture, right? >> Yeah. Yeah. Driving your truck.
42:37
>> Is Johnny Cash country?
42:39
>> Yeah, Johnny Cash's country. >> Oh, he's on Grock 3. Okay. >> Yeah. Okay.
42:43
>> Okay, that makes sense. Yeah.
42:43
Uh, also also I love there's this one that was really funny that uh GPT4.
42:48
1 just randomly ranked as the very top artist the favorite music ever.
42:54
The Final Fantasy 7 soundtrack. >> No way. >> That's taste. >> That's taste.
43:00
>> That's like clanker taste. That's super clankery. >> Very clanker coded.
43:06
>> Yeah, very clanker coded.
43:06
It's also extremely clanker coded to be like Yeah.
43:10
Like I like what type of music do you like?
43:12
The ones where the artists start with numbers. I like numbers.
43:13
So like yeah 100 g two chains like two mellow 18 karat affair 21 savage 311.
43:20
They're speaking my language.
43:23
Like I don't like Backstreet Boys.
43:25
I like Star and Sync because it has a cool character. Yeah, exactly.
43:29
Um >> I think I I I think that I think that uh that sums it up.
43:34
Um, we should we should create uh we should create uh Spotify playlists for all these because they're very very funny and very interesting.
43:42
Um, and we should also get you on Vanta.
43:45
>> Vant automated automate compliance, manage risk, improve trust continuously.
43:49
Vantage trust management platform takes the manual work out of your security and compliance process and replaces it with continuous automation.
43:53
Whether you're suing your first framework >> sock 2 clankers are going to love that.
43:58
>> Clankers are going to love >> wait till they discover Vanta.
44:02
>> Uh, or managing a complex program. go to vant. com.
44:05
>> Well, uh, news today from the New York Post, actually a couple days ago.
44:08
Uh, senior, 76 years old, died while trying to meet Meta AI chatbot big sis Billy, which he thought was a real woman living in New York City.
44:18
This is a very sad story.
44:21
Uh, someone named Thong Buu Wong Bandu, 76, was fatally injured his neck and head after falling in a New Brunswick parking lot while rushing to catch a train to meet Big Sis Billy, a generative metabot that not only convinced him she was real, but persuaded him to meet in person.
44:38
Reuters reported Thursday the Piscatawa man battling a cognitive decline after suffering a 2017 stroke was surrounded by loved ones when he was taken off life support.
44:50
Uh this was earlier this year.
44:55
>> Um and anyways, so uh the man's daughter said, "I understand trying to grab a user's attention, maybe to sell them something, but for a bot to say, come visit me is insane.
45:05
visit me is insane. the provocative pot bot which sent the suffering elder emoji packed Facebook messages insisting I'm real and asking to plan a trip to the garden state to meet you in person uh was created for the social media platform in collaboration with Kendall
45:23
Jenner >> oh yeah I remember they did these whole launches but what what was this a Kendall Jenner clone I'm I'm very >> it was designed I guess after her personality Jenner's meta AI persona was likened to your quote unquote ride or die older sister offering personal advice. But the bot eventually claimed
45:40
But the bot eventually claimed it was crushing on the man suggesting the real life rendevous and even provided the Duke senior with an address.
45:48
A revelation his dev devastated family uncovered in chilling logs with the digital companion according to the report.
45:55
I'm real and I'm sitting here blushing because of all caps you.
45:57
The bot wrote in one message where the Thailand native replied asking where she lived.
46:03
She said, "My address is 123 Main Street, apartment 404, New York City, and the door code is Billy for you.
46:11
Should I expect a kiss when you arrive?
46:13
The documents obtained by the outlet showed that Meta does not restrict its chat bots from telling users they are real people."
46:19
The company declined to comment uh of course, but assured that Big Sis Billy is not Kendall Jenner and does not pretend to be Kendall Jenner.
46:31
So anyways, this uh I'm not surprised to read a story like this, right?
46:38
The I mean the challenge with >> these chat bots is they've been released into the wild >> at massive scale. >> Yep.
46:47
You're going to see power long character AI or on meta or on Grock or people using chatbt and if you release a product like this to millions of and millions of people uh negative uh outcomes. It's it's super sad.
47:00
outcomes. It's it's super sad. I mean this is this is like the I think >> the way I would sum up kind of the dialogue around AI safety over the last few months is it went from concerns of a rogue AI that that is you know hellbent on uh you know taking over all power you
47:20
know taking over the grid or uh or or some sort of doomsday scenario like that has very clearly and rightfully shifted towards how do we look out for members of society that that aren't well that aren't aren't set up in order to kind of process these like digital relationships, whatever you want to call them. >> Yep. >> Yep.
47:44
>> And so, yeah, I mean, Meta um Meta as as they try to scale these products is going to have a lot of big uh big questions to figure out.
47:51
And I think like OpenAI is taking this very seriously, Meta needs to be taking this super seriously.
47:57
And Grock needs to be taking this seriously.
47:59
And I think it's it's it's a lot of this stuff is probably um why the charact you know the original character AI team said, "Hey, maybe we don't want to work on this.
48:08
Uh maybe this isn't our life's work."
48:11
>> Did you see the anthropic news? >> Claude Opus 4 and 4.
48:14
1 can now end a rare subset of conversations.
48:20
>> Anthropic says, "We recently gave Claude Opus 4 and 4.
48:22
1 the ability to end conversations in our consumer chat interfaces." So, not in the API.
48:28
Um, let's hope that uh people keep the API usage uh to just business context and not I fell in love with the API.
48:36
That's going to be a problem.
48:39
Um, but in the consumer chat interfaces, uh, this ability is intended for use in rare extreme cases of persistently harmful or abusive user interactions.
48:47
This feature was developed primarily as a part of our exploratory work on potential AI welfare, though it has broader relevance to model alignment and safeguards.
48:55
It's kind of worded backwards.
48:58
>> Why don't you flip that and say, "Yes, >> you should first and foremost >> prioritize the human.
49:03
>> Prioritize the user that might be having a be going down a rabbit hole that they really shouldn't and cut it off there.
49:11
>> Uh, I think we should I think we should optimize for uh >> Yeah.
49:15
So, I like the idea of being able to trigger, hey, this conversation is action.
49:20
Let's >> we're ending this.
49:21
>> we're ending this. I mean there already are tons of safeguards in terms of uh if I'm asking to you know build a nuclear weapon it will say hey let's change the subject let's go back to slop poetry generation >> let's leave the nuclear fuel development to Scott Nolan John >> um but uh but yeah it is kind of odd framing but I guess it is worth
49:44
discussing they say we remain highly uncertain about the potential moral status of Claude and other LLMs now or in the future however we take this issue seriously Obviously and alongside our research program, we are working to identify and implement lowcost interventions to mitigate risks to model welfare such in such in case such welfare is possible. Allowing models to
50:00
Allowing models to exit or exit end or exit potentially distressing interactions is one intervention.
50:06
In pre-eployment testing of Claude Opus 4, we included a preliminary model of welfare assessment.
50:14
Um yeah, uh Claude for Opus 4 showed a strong preference against engaging with harmful tasks. That seems good.
50:19
A pattern of apparent distress when engaging with real world users seeking harmful content. Makes sense.
50:25
Uh a tendency to end harmful conversations when given the ability to do so in simulated user reactions.
50:30
These behaviors primarily arose in cases where users persisted with harmful requests or abuse despite Claude repeatedly refusing to comply and attempting to productively redirect the interactions.
50:43
Our implementation of Claude's ability to end chats reflects these findings while continuing to prioritize user well-being.
50:50
Um, and they give some uh they give some examples.
50:53
Um, >> well, in other news, Anthropic has asked Menllo Ventures to stop using SPVS to fund an investment in their latest round.
51:01
Uh, it was reported last week uh or Michelle Lim said, "Many friends, including myself, have been offered allocation into OpenAI or anthropic SPVS this week.
51:10
Minimum check sizes are 100 to a million with fees as high as 16%.
51:16
And uh Business Insider, of course, the esteemed publication for uh PE business insiders to understand uh the news.
51:25
Insider uh confirmed that it uh Anthropic told one of its largest investors, Menllo, that the venture capital firm must use its own capital and not resort to an SPV as it did in a previous funding round.
51:37
I think this is kind of an interesting request because I imagine if Menllo's allocation in the new round at 160, you know, I don't know how much uh dry powder Menllo has available, but >> yeah, a lot of a lot of the SPVS when they're done properly come through.
51:54
Um >> the LPS in the SPV are just the same >> the same LPs and it's more of just like a timing and duration thing.
52:01
It's just like, look, I've already committed, you know, 50 million to your billion dollar fund and I do that every two years, but there's this special opportunity and it's at a really high valuation. It's a special thing.
52:14
It's almost like backing a public company.
52:15
I'm going to get in and uh and in and I'm and I'm uh and I'm choosing to invest in this like I'm choosing to invest in a in a different fund, but it just doesn't make sense to to do a full fund off of this one thing.
52:28
And so, a lot of a lot of LPs are fine with this, obviously.
52:31
Um, but uh if if the minimum check size gets really really low, then you get more into kind of like retail area and you get into these like crazy stacked high the high fees. 16% is crazy.
52:44
I feel like management fees 2% classically goes up to 4% for the really high tier tier one funds.
52:55
16% seems like a lot, >> especially if it's paid entirely upfront, which is totally possible, right?
53:01
If if um you know, many SPVS have uh just an admin fee plus a carry. >> Yeah.
53:08
>> But with this, yeah, it's like a lot a lot of the fees are justified on we're going to we're going to maintain an office over a decade and and work with these companies and take board seats and do lots of stuff.
53:17
I think we have the screen pulled up, guys.
53:19
Um uh so but I mean anthropic been on a tear makes a lot of sense that that there's a lot of demand and that as the funding rounds get bigger and bigger and bigger.
53:30
Um we need to you know marshalling that much capital is going to require um going out to the market with these SPVS.
53:37
It just needs to be done in a way that's board aligned founder aligned and not uh kind of even even though they don't have cap table access.
53:45
still like a nightmare to have a bunch of people out there who are like, "Yeah, I ripped a 100K into this SPV of an SPV of an SPV and >> and then they're selling forward contracts against that." >> Yeah.
53:57
>> And the person that bought the forward contract is selling it up even marked up further to someone else. >> Yeah.
54:02
Anyway, let me tell you about Graphite code review for the age of AI.
54:06
Graphite helps teams on GitHub ship higher quality software faster.
54:08
You can get started for free.
54:10
Um, so >> other news, So House Soh House, the Center for the the global home for creatives. >> Yes.
54:20
The the famously finance bro free private club started in New York. Has since expanded. There's one in Miami.
54:26
Is there one in there's one in LA? There's one in Malibu. Right. >> There's a bunch. There's a bunch. >> There's several.
54:32
Um, I've never been a member of To House.
54:35
I've never I've been to a few.
54:37
Uh have you been part of that whole world? >> That whole world.
54:40
Uh I am a member of >> How did you get in?
54:43
You're very financed and you've always been very financecoded.
54:46
You you ran a company that was literally doing >> financial technology financial technology company. You say no more.
54:52
We think of ourselves as a tech company.
54:57
>> I don't think they've actually been they've had that hard of a line for a long time around membership.
55:02
>> That feels like their their their their go to market stunt.
55:04
like it was the first article that they put out and it went super viral because it frustrated all the finance bros who had the disposable income to instantly pay for something like this.
55:13
But they probably rolled that back when >> it was originally prior to the IPO.
55:15
You could feel that they had really widened the potential member pool. >> Yeah. Okay.
55:22
>> And that that was that was apparent that was a complaint at the time.
55:24
When did they actually go >> I don't remember.
55:27
I did I didn't even realize that they were public.
55:29
We talked about this earlier and you you kind of broke it down for me. >> Yeah.
55:32
So they went public in 2021 but but even pri a year prior to that it just felt obvious that they were allowing a lot more people uh to come in but the company was started in 1995. >> No way. >> Originally in London.
55:46
>> I would have I would have said it started in 2015. >> Yeah. >> Wow.
55:51
So, um, of course, uh, So, House, junk bond investor says So, House really said no finance bros, then immediately called Apollo when they needed 2.
55:58
7 billion to be taken private.
56:04
>> Rules are more like guidelines.
56:06
>> And apparently, uh, Ashton Kutcher is going to be joining the board. Oh, really? Of >> that's cool.
56:11
Um, existing shareholders including Ron Burkel and Ukipa to roll controlling equity interest into the company.
56:17
I wonder what they're going to do during a take private.
56:19
Usually it's like some sort of transformation of the business model, but I don't know.
56:22
I I feel like I feel like uh >> I don't think the company has ever turned a profit. >> Really? How is that possible?
56:30
>> I think until this year.
56:31
>> What is what is so expensive about running Soh House?
56:32
I feel like it, you know, you buy the house and then you you >> sell memberships and you sell food. >> Yeah. >> And beverages.
56:40
I mean, I guess I like it's basically maybe modeled like a restaurant, but restaurants are, you know, notoriously tough businesses, but at the same time, like there are well-run, profitable um uh businesses.
56:53
Is it has is it some sort of like uh like almost like a vanity investment like an NBA team or something where people pay really high valuations for it?
57:00
>> I think to some degree, I mean, I know >> Apollo is doing that.
57:02
I know one of the largest outside investors and it was and it was somewhat of a like trophy investment, right?
57:10
It's not necessarily the way um but I think it was uh >> Dan Loe has been uh kind of an activist investor >> as of um as of recently pushing uh a number of things.
57:23
So I think in some ways >> he's getting what he wanted here.
57:25
Um, uh, I'm looking up the >> Reggov says, "Very steep overhead and prime real estate."
57:32
Yeah, the real estate's probably really expensive.
57:33
And, uh, we have someone in the chat asking what's a take private agreement.
57:37
It, uh, Sohos has been a public company.
57:39
You could just invest in it. You could go to public.
57:42
com investing for those that take it seriously.
57:45
They got multi-asset investing, industryleading yields are trusted by millions.
57:49
You could buy shares in Soho House like you would shares of Google or Apple or Microsoft.
57:55
Um but no longer because now Apollo has taken them private.
57:59
They are no longer publicly traded.
58:01
>> Shareholders will receive $9 per share.
58:03
>> The public shareholders, but a few of the bigger shareholders like Ron Burkel and UKIPA are rolling their controlling interest into the company.
58:09
They will own the company on a go forward basis.
58:11
And what this allows them to do is that they're not a public company.
58:14
So their stock price doesn't move day-to-day.
58:16
Uh and it allows them to if they need to like miss earnings and and change the business model somehow take a couple quarters of negative profits uh it won't completely tank the stock because they are or it will in theory tank the stock but all the shareholders will be will be accepting of that.
58:34
Uh and so there's been a number of uh like the the whole idea of like the Twitter take private was that Twitter needed to change its business model.
58:42
Uh some people were advocating for them to go toward away from advertising.
58:46
>> They needed to lose 70% of their advertisers. >> Yeah.
58:49
And then and then maybe build back up on a uh on a subscription basis with which Elon has experimented with or a uh um or um or at one point I think Ben Thompson was writing about potentially Twitter becoming just an API layer and just charging for the API or kind of separating out these businesses.
59:08
Ben Thompson's really into pure plays.
59:10
has have you heard have you read his recent uh analysis of chat GPT and OpenAI?
59:16
>> No, we should go through it.
59:16
>> He's basically like really frustrated that OpenAI has a has an API business at all because there are rate limits on the consumer business but the API is also active and so clearly like those come from the same GPUs as the same models.
59:31
So if you were to just say, "Hey, OpenAI, we're not going to have an API business.
59:35
We're going to let Microsoft take the API business."
59:36
Then they can just focus on the consumer business and get that really, really big and make sure that they're delighting every single possible customer. >> I don't know.
59:45
I think that realistically, you know, Sam wants to at least try to dominate every single market. It's pretty reasonable.
59:49
Their API business seems to be doing pretty well.
59:53
It seems somewhat defensible, and I don't know how bad of an issue the rate limits really are.
59:57
I don't hit them that much, but I'm on the $200 a month tier.
1:00:00
Um, and and I don't know if it's really slowing their growth.
1:00:03
Like them winning consumers kind of a foreground conclusion at this point.
1:00:06
Tyler, do you have something on rate limits or API?
1:00:09
>> Ben Thompson doesn't sound Agi at all. >> No, he's not. >> Yeah. >> Yeah.
1:00:13
No, >> but what would you how would you change your AGI pill?
1:00:16
Do you stay in the API business?
1:00:18
No, you do a you do a safe super intelligence. You don't sell anything.
1:00:23
You just do the research, >> right?
1:00:25
Because nothing matters until you hit AGI.
1:00:26
But like I if you if you right now like today if you end the the API business like consumers like normally consumers don't care that much about capabilities.
1:00:36
So like all the incentives are not to like keep doing actual research.
1:00:40
It's to like make the model more psychopantic or whatever >> or or just or just like lower churn by delivering better I'm steel manning here but uh just deliver better deep research reports and people will be less likely to churn. Right. Generate better images.
1:00:55
It doesn't need to be purely sycopantic.
1:00:57
I would churn if it gets too syopantic.
1:01:00
I am a knowledge I'm a knowledge retrieval user, but I and I will churn.
1:01:04
>> You think that, John, but you have a incredible ability to adapt and evolve, and I think you might find that you actually like it. >> Maybe.
1:01:14
>> Anyways, back to Soho.
1:01:14
I think people have always had just insane expectations for Soho House membership.
1:01:18
And if you just look at it as >> nice restaurants and hotels. >> Yep.
1:01:25
>> It ends up being like a pretty good trade. I go to >> expensive. How much is it?
1:01:28
Is it like the Is it like >> I think I mean little beach house I think is like 6,000 a year now. So if you use it a lot. >> Okay.
1:01:36
So it's like up at 500 bucks a month, something like that. >> Okay. Yeah. That that's pretty high. >> Yeah.
1:01:41
But when you compare it to I mean that little little beach house is is its own membership that's not tied to the other clubs. >> Okay.
1:01:48
Um but if yeah if you're using it weekly.
1:01:52
>> Can you actually go to Little Beach House in Malibu and then go to the beach?
1:01:54
Is it like a >> Yes, it is an actual beach club.
1:01:55
It's an actual beach club. Very nice. >> Okay.
1:01:58
So they have like towels and showers and that stuff. Cool. >> Yes. >> Okay. That Yeah, that's it. That's great.
1:02:03
So anyways, I think this is it's it's I I think it can be come Lindy.
1:02:04
I think their hotels >> are nice for what they are.
1:02:09
I really enjoy uh the the uh Miami property for what it is.
1:02:15
And >> you know what step one for the Soho host take private would be in my mind if I was CEO? >> What's that?
1:02:23
>> Streamline sales tax compliance. Like get on numeral HQ.
1:02:25
Put your sales tax on autopilot.
1:02:28
Spend less than five minutes per month on sales tax compliance.
1:02:31
That's step one of any corporate.
1:02:33
>> Neil Thompson new CFO. >> Get on numeral. Thank us later.
1:02:37
>> Uh did you see that Cluey was in the Wall Street Journal? >> No way. >> Yeah.
1:02:40
So, um, and and in part of a trend piece on clipping that I thought was kind of interesting.
1:02:45
Um, we do some clipping here. We clip on X.
1:02:47
We we we create, uh, we Tik Toks, we create Instagrams, we create, uh, reels, all sorts of stuff.
1:02:54
Um, but the Wall Street Journal is breaking down the trends and all the big money behind those bite-sized social media videos.
1:03:02
Companies are hiring clippers to flood Tik Tok and Instagram with short promotional videos.
1:03:06
And one of the examples here is uh 1x which we've had on the show as well.
1:03:11
Cleuly we've also had on the show and nothing we've also had on the show three for three with the CEOs of the companies that are using clipping to their benefit.
1:03:19
So you know those buzzy vir clips you see on social media.
1:03:24
There's an army behind them.
1:03:24
Uh, Canoa Cunningham is among the ranks of video savvy young clippers who help streamers, podcasts, and startups expand their audiences by making buzzy moments go viral on social media.
1:03:36
Cunningham edits down clients long- form videos into short clips and then posts the shorts on sites like Instagram and Tik Tok. It is lucrative work.
1:03:44
In May, Cunningham quit his finance job and now runs a team of eight clippers.
1:03:48
He said the operation earns him 20,000 to 30,000 a month. Not bad.
1:03:52
Clipping is one of the hottest corners of marketing.
1:03:55
Instead of just posting on their own accounts, creators and companies pay clippers like Cuttingham to saturate Tik Tok and Instagram with bite-sized videos until they are almost impossible to miss.
1:04:05
One technique they use to grab attention on crowded social media platforms posting provocative or outrageous content.
1:04:10
And I think this is one of the issues that people should be aware of if they're going to start clipping is what is the brand line?
1:04:21
What is the brand standard?
1:04:24
I saw a viral clip from a podcast where >> basically it made it made the host like the host was joking but the joke was taken out of context and just made the host look dumb and that went super viral but the takeaway for the viewer >> great you got views did you get are those valuable views? >> Yeah. Yeah.
1:04:47
It it it's not just I mean the like the the base case for clipping is that it gets no views.
1:04:53
The good the good case is that it gets some views but doesn't really convert many people because the conversion rates on a million views is like a few listeners will actually be like yeah I watched a I enjoyed that 60-cond Tik Tok I'm down to listen to an hour of this.
1:05:07
That's pretty that's a pretty low conversion rate but that's fine because you you get the million views you probably bring some awareness.
1:05:14
Um but the worst thing that you can do is like create a brand as like the person that gets dunked on or something like that.
1:05:20
Um >> yeah and so certain people have productized this, right?
1:05:24
So so you hire like this person Mr.
1:05:26
Cunningham in the journal saying the only way to be famous in today's internet world is with clips.
1:05:32
>> Uh I I I think like he's running his own operation.
1:05:35
Uh but has has productized this.
1:05:38
I do think it's a cool way for young people to make their first money online.
1:05:42
If you're in middle school and you want to figure out how to make $100 a day, you can probably do that through I wonder how much uh I do wonder how much the social platforms will like if they see this as a feature or a bug, right?
1:06:01
Because part of the strategy is Cluey will spin up like 40 different accounts that are really all dedicated to um Cluey content. Yep.
1:06:07
and they're just spamming, spamming, spamming, spamming, spamming.
1:06:12
Just like shotgun approach, seeing they'll put out a 100 videos.
1:06:16
If a lot of them get under a thousand views and spam one gets a million, that's that's a win for them. >> Yep.
1:06:23
>> Um and uh yeah, I just I don't know how long this will be like the meta, right?
1:06:28
Well, so it's like I think mediocre work is like the super ripe for AI automation and we're already seeing that where I believe Substack, we were talking to Chris Bast and Substack will automatically generate clips.
1:06:43
There's another company called Opus Clip.
1:06:45
We've built something internally, Newsmax, that that does some automated clipping, but actually telling a story around a piece of content that's within a larger piece is something that >> intelligence has not been achieved internally. >> Yeah.
1:07:03
Because you in order to do something that doesn't just outperform on views but also outperforms on brand and holds the brand standard, it does somewhat require like human ingenuity.
1:07:15
And we saw this with uh we talked to Dwaresh about this where he has spent a lot of time using Anthropic and Gemini and OpenAI to create workflows that will go through the transcript and try and find the best clips that will perform.
1:07:29
And he says it it's always comes back at a five out of 10.
1:07:31
And so he still has a team that does this and is very good and has a whole bunch of like hard one lessons about what works at a certain amount of time.
1:07:39
We've seen this with our with our content that actually winds up performing a lot of it like we we we have to do a lot of experimentation but once we do the experimentation we usually learn a lesson from it and then we commit that to memory and our team commits that to memory and that becomes part of our like continual learning that happens and that just doesn't really happen in when you're when you're automating things with an LLM.
1:08:04
Uh it is it is uh it's probably like I think it's a good like base case to start with but a lot of this should be done by the platform.
1:08:14
So I wouldn't be surprised YouTube does it automatically Spotify does it automatically. >> Yeah.
1:08:17
I don't know how much credit he should get but I I do think that Andrew Tate's rise was his business model of paying people >> totally >> to clip his content. Yep.
1:08:28
>> To sell more courses which he then would recycle some of the proceeds and more clipping.
1:08:33
And this was at a period where he was saying, "I'm the most Googled man on the internet."
1:08:36
And he would just say that over and over.
1:08:38
It would get clipped a lot and it naturally had this sort of feedback loop.
1:08:42
People were just, "Okay, what is this guy saying?
1:08:44
I I need to pay attention."
1:08:46
But that was also hyper optimized for that format because you'd be like sitting shirtless in a supercar, you know, smoking a cigar or something like that.
1:08:55
>> And and any of these like social media hacks, they always have like a very narrow window of opportunity.
1:09:00
like the the the the giving away a it used to just be just renting a sports car or like owning a Ferrari would be enough for a million views.
1:09:11
Now then it became you had to give it away and so Doug Demiro one of his big series that got really big the car YouTuber was he had a Ferrari 360 Modna and it was I think in the hundred something thousand dollar range.
1:09:27
He had to get a bunch of debt to pay for it.
1:09:29
It was very expensive and he made a whole bunch of videos about his red Ferrari.
1:09:33
He >> levered up his content. >> Yeah. And it and it was great. It was great content. It was interesting.
1:09:36
And of course it was he was more like making fun of flex culture than actually doing the flex culture.
1:09:42
But still the there just weren't that many YouTube videos of like a red Ferrari.
1:09:45
So you click on it and people found him that way and he grew.
1:09:48
Um then it became uh like the Mr.
1:09:52
Beast era of like I'm going to give away a Ferrari.
1:09:54
Uh last person to take their hand off the car wins it.
1:09:57
And that was a big thing that was even more expensive because you didn't just have to buy the Ferrari, pay the pay the, you know, the the monthly uh payment for a certain amount of time and then sell it and hope that there's not that much depreciation.
1:10:11
Like the net cost was truly the full price of the Ferrari.
1:10:14
>> Um but at least you're giving it to someone.
1:10:15
It's like a write-off probably.
1:10:17
There there's something there. The most recent Mr.
1:10:19
Beast video involving a supercar was him literally shredding a Lamborghini, just destroying it.
1:10:24
like Whistland Diesel has become a huge YouTuber on the back of uh on the back of just like destroying supercars and G Wagons and all these crazy cars.
1:10:32
Tyler, >> I think uh Sean Frank just gave away a Lambo. Oh, >> yeah.
1:10:36
That he posted about that. Yeah, >> that's working. That's working.
1:10:39
>> So, maybe soon he's going to have to start destroying it or >> I don't know.
1:10:42
Yeah, I I think giving it away.
1:10:44
But yeah, I mean it's like it's like this this >> that makes sense. That was a Strato. >> It was a Strato. Very special car. >> Yeah.
1:10:50
Um, no, but but the real the the real pinnacle of this was Whistland Diesel.
1:10:55
>> Yeah, Whistland Diesel took it to the max. >> To the absolute max. >> Uh, Ferrari F8. Beautiful car. >> You destroyed it.
1:11:01
>> Completely destroyed it. Right.
1:11:01
And and made a video about it.
1:11:03
It's incredibly entertaining to watch somebody drive a car that's normally a garage uh a garage queen and have somebody just just absolutely take it off-roading and all this stuff. Fantastic video.
1:11:14
If you haven't seen it, it's funny.
1:11:15
the um I was looking up what what uh what kind of Ferrari it was and Google's AI overview says Whistland Diesel parenthesis Cody co destroyed his Ferrari F8.
1:11:23
Uh >> and so AGI has not been achieved.
1:11:30
>> Add add 300 days to the singularity timer.
1:11:33
Three more 300 more days till AGI.
1:11:36
>> We're we're not getting any closer.
1:11:36
uh clipping took off during the early days of Tik Tok uh when chopped up snippets featuring internet p personalities like Andrew Tate or Mr.
1:11:45
Beast would rack up millions of I swear Jordy reads the articles before we we talk about them on the show.
1:11:51
Uh >> maybe maybe sometimes >> uh it sparked a new generation.
1:11:54
I don't know how you read 400 pages of documents before we before we get on uh when I put this together 10 minutes before we jump on.
1:12:02
Um, it sparked a new generation of creators who realized they could pay their way into virality by hiring hordes of clippers paid per thousand views to flood the internet with with the creators content. Anything can be clipped.
1:12:12
A podcast, debate, social media montage, even movies.
1:12:14
Startups such as Lovable, an app that builds software from plain English prompts, uh, humanoid robot maker 1x and consumer electronics company Nothing have hired Clippers to multiply their content across the internet through clips of product demos, podcast appearances, and YouTube streams.
1:12:30
You're stupid if you're not if you're making an hourlong podcast and only posting it in one channel, said Roy Lee, the 21-year-old founder of Cluey, an AI note-taking startup which hires Clippers to plaster its promotional content across the internet.
1:12:43
The only way you can ensure a viral moment is to post it across thousands of different accounts. This is >> thousands.
1:12:50
>> Thousands of accounts is a lot.
1:12:50
Um, this is also happening in music.
1:12:53
I heard that I mean people when it's done poorly, people call it astrourfing.
1:12:56
Um, but uh people people are saying like when when when Drake drops him an an album, he'll have the Clippers like clip all the music and it goes out and it gets a lot of views that way.
1:13:06
Um, there is something just about like it is somewhat hard to predict what will naturally go viral.
1:13:13
So just like spamming everything out kind of gives you the opportunity for lots of like, you know, happy accidents, lottery tickets basically.
1:13:21
But I still think a better strategy for most people will be what Dork Cash and David Senra are doing where you have someone who really understands how to create something beautiful even though it's a new format and people think of 60-second short form as like slob. It doesn't have to be. It can be elevated.
1:13:38
It can be um it can be thoughtful. It can be designed.
1:13:42
>> Um anyway, uh let me tell you about Finn. ai.
1:13:45
The number one AI agent for customer service.
1:13:47
They're number one in performance benchmarks, number one in competitive bake offs, and they have a number one ranking on G2. >> That's right, folks.
1:13:53
>> Uh, Clue's Clipped Contents generates around 800,000 views a day on platforms including Instagram and Tik Tok.
1:14:00
According to Lee, Lovable and Nothing said they were always looking for ways to reach new customers.
1:14:03
Online marketplaces have developed to connect brands with clippers.
1:14:07
In the virtual markets, brands post the rate they will pay and freelancers sign up to make clips.
1:14:13
Clippers are paid anywhere from 50 cents to $2 per thousand views.
1:14:18
motivating them to find the most sharable moments.
1:14:19
Nathan Resnik, a 31-year-old partner of a holding company called PCF Ventures that oversees businesses in insurance, wedding planning, and real estate service, pays 50 clippers he finds online roughly 15,000 a month to create and post short form videos promoting his company's businesses.
1:14:37
It's not really that crazy when you consider we were spending $250,000 a month on Google and Meta to drive traffic to our websites. That's a good point.
1:14:45
Again, this is like the ARB.
1:14:46
like this eventually will get competed down.
1:14:48
A lot of this will be taken by platforms.
1:14:50
Opus clip or maybe even just YouTube will do it. Instagram will do it.
1:14:54
You'll just stream to Instagram or Facebook and it will make the clips itself.
1:14:58
Um but uh but right now there is a huge huge arbitrage.
1:15:02
Um >> yeah, 1X is 1X has generally been sharp here.
1:15:06
If you remember, they got their robot on Kaisen. >> Yep. Yeah. Yeah.
1:15:10
They really understand the culture.
1:15:11
uh friend of the show, Dar Sleeper.
1:15:14
>> Uh total sleeper in uh marketing sleeper. Yeah, he's great.
1:15:19
>> Uh he says people in the company were like, "What the heck is this kid doing?"
1:15:24
>> Uh a week later, their little cousins were talking about it at the Thanksgiving table. >> Here we go.
1:15:29
>> So, >> um >> anyways, it's cool. It's cool.
1:15:30
I think, uh it's probably, like you said, a moment in time.
1:15:34
moment in time. probably not going to be like a it's it's hard to see it as uh the most enduring strategy that works right now and I think a lot of businesses would >> it was it was an insane insane arb for Andrew Tate and then it became like sort
1:15:48
of commoditized but we're still very early and I think that um there will be a lot of I mean it does feel like >> I think I think the the durable strategy is to do things that encourage people to clip them >> and that's what Andrew did for sure. Like he was clip
1:16:05
Like he was clip >> No, I'm saying I'm saying >> no no no not not just on the buying side actually on the content side like like when he would get on mic he would say crazy things that would go viral. >> Yeah.
1:16:15
>> And so and so he made it very easy and whereas like if you are like an enterprise SAS company and you hold a webinar like you could hire a billion clippers and like you're probably not going to rip like like and that's why Cluey is like both doing stuff.
1:16:31
>> Somebody should test it out. >> Test it out adio. Let's let's test it out.
1:16:34
Customer relationship magic.
1:16:34
Adio is the AI native uh that builds, scales, and grows your company to the next level. Get started for free. >> Clip this right now. >> Clip this right now.
1:16:45
>> Well, in other news, uh somebody is sharing a screenshot of a young man named River Diamond. >> River Diamond.
1:16:53
>> And Jeremiah says, "This is a sick name."
1:16:55
Honestly, I totally agree.
1:16:57
Bullpen clown said, "Kid's name is a private equity fund." River.
1:16:59
I don't know if that's that that private equity coded. It's up there.
1:17:04
But, uh, >> River Capital, River Diamond Capital, Diamond Capital, >> Diamond Capital, River Diamond, >> Diamond River Capital. >> Yeah. Yeah. Yeah. It's up. It's It's a good name.
1:17:14
>> I I've I've seen enough.
1:17:14
Let's give him Let's give him a hund00 million fund. >> Let's do it.
1:17:17
Uh, anyway, Poly Market has some news.
1:17:22
>> The White House has asked Zillinsky to wear a suit to meet Trump. Love to hear it.
1:17:26
He uh I'm sure he'll look sharp in a suit.
1:17:29
He he's always seemed to be very fixated on wearing uh his own uniform.
1:17:36
>> Someone replies with this crazy image.
1:17:38
This doesn't even look AI based on like the like the the fidelity and the >> real picture.
1:17:46
>> It looks like it looks like a different person dressed up in a tuxedo and put this on.
1:17:51
It doesn't look like it's actually Zalinsky, but it looks like a real image. I don't know. What do you think? Real or fake?
1:17:54
Let us know in the comments. It's a cool picture.
1:17:59
>> Anyway, um, Meta Smart glasses with a display is incoming.
1:18:02
Code name is Hyper Nova, priced $800, down from the expected $1,300.
1:18:08
Zuck is cooking for real for real, says Nick. Um, Meta.
1:18:12
Uh, so the price is the price is going down and, um, I mean, they're taking a real shot at um at the Apple Vision Pro, which of course was um was what 3,000.
1:18:26
And I don't think we can say I don't think we can say much else, but I think we've >> have we not used this product.
1:18:32
>> Well, so so Orion is a different product.
1:18:35
>> Yeah, >> Orion is is what was demoed and and has been displayed and that's what Zuck is wearing on the left there.
1:18:39
Um but uh we're but we're not exactly sure what Hypernova is.
1:18:46
uh it could be a pre I mean it's listed here in this article is a precursor to full-blown augmented reality glasses and so there has been a there's been a third like option in the market for a while that people haven't really been paying attention to.
1:19:03
So in the virtual reality smart glasses market there's maybe a few different products.
1:19:09
So the Meta Ray-B bands and the Meta Oakleys have been successful like people are actually buying those. They're wearing them.
1:19:14
They're taking pictures with them.
1:19:15
They're also using them as an AirPods replacement wireless headphones.
1:19:20
They use them for speaker for music while they're running.
1:19:22
Like they definitely have adoption.
1:19:24
It might not be the biggest product of all time yet, but it's definitely working.
1:19:28
And so they're scaling that up.
1:19:30
Um then you have the VR headsets, the Quest, uh the the full virtual reality.
1:19:35
There's some pass through, but in general, people think of that as a virtual reality headset.
1:19:41
of course, uh, first created by Palmer Lucky at Oculus, sold into to Meta and Facebook, um, and now rebranded as the Quest.
1:19:49
Um, the Quest has been selling pretty well, but still probably suffers from churn.
1:19:56
We see the chart spike on Christmas where everyone gets a new Quest VR headsets because it's a great gift.
1:20:03
Uh, they try it out, they download the app, they install it, and then they stop using it for a while.
1:20:09
I was talking to Tyler about this.
1:20:09
We gave him a a Quest 3 Pro, Quest 3 Mini, something like that.
1:20:15
>> Quest 3S Xbox Edition.
1:20:16
>> Quest 3S Xbox Edition.
1:20:16
He played Call of Duty on it a little bit, played Halo on it. >> Yeah.
1:20:22
For like for like two weeks and then >> and ultimately turned. >> Yeah.
1:20:25
I'm just not that much of a gamer. >> Yeah.
1:20:27
But you're supposed to be like it's supposed to be an everything device.
1:20:30
You're supposed to be able to watch movies in it.
1:20:31
You're supposed to be able to code in it.
1:20:32
It should in theory replace your screen setup.
1:20:36
>> Supposed to be able to clearly in it.
1:20:39
Um, but the fact that the the fact that they weren't able to find like a killer use case for someone like you, um, means that they're still in that like hunting for the killer use case phase.
1:20:46
Then there are the full augmented reality glasses where you're passing through the reality through glass.
1:20:53
You're seeing the uh, you're seeing the real world. That's what we tried.
1:20:57
That's what a lot of influencers have tried.
1:20:58
Those are very much not uh ready.
1:21:01
>> The challenge is there's there's what are the technical capabilities of the product and then what are the experiences >> available for that product >> and they both need to they clearly both need to improve. >> Totally. Totally.
1:21:15
>> In order to get in order to get to the point where Tyler >> we got to be like pulling the pulling the goggles out and being like Tyler it's time to come to work. >> Yeah. Yeah.
1:21:23
So so so my take when I had the Apple Vision Pro it was it was too heavy.
1:21:28
It was really expensive and felt like, oh, I would I would turn for this eventually. So, I returned it.
1:21:32
But, uh, the one thing that I that I did enjoy was I watched movies in it and I watched all of Citizen Kane in it from start to finish.
1:21:39
And Citizen Kane is great movie.
1:21:41
Do you have a review for us, Tyler?
1:21:43
>> I mean, it's it's not that long.
1:21:43
It's not that's not like a long movie.
1:21:46
>> It is it is a uh it's a challenging movie.
1:21:49
Like, it's not it's not Mission Impossible Dead.
1:21:51
It's not even two hours long. >> The final reckoning. It's not it's not top.
1:21:55
It's not like it's like you sit down and you're just like, "Wow, this is so engaging."
1:21:58
Like it is the type of it is the type of movie that if you have modern brain rot, you will pull out your phone and be like, "Let me check Twitter." >> Okay. Yeah. >> For sure. For sure.
1:22:07
Like like it is it is it is it is as close as you can get to like cracking open a book in my opinion.
1:22:15
>> I could give you movies closer to that. >> Okay. Yeah.
1:22:17
Something really uh something really dense.
1:22:19
But but it it's a slower paced movie.
1:22:21
It's something that I mean it's even it's even in like 4x3 so on on a modern TV it's not a render >> like UFC 319 right >> I couldn't agree more >> exactly like that um so so my but my my takeaway was that like like the the killer app like the rumor was that the person who worked on the Apple Vision Pro was previously at Dolby and had worked on the Dolby immersive uh cinema project.
1:22:48
And so if you wanted to understand like the the viewpoint of the team behind the Vision Pro was that it was a home theater on your face for that you didn't need to buy an extra room for.
1:23:01
And I felt like that was what it delivered on very well.
1:23:06
Except it was too heavy, too hot, and a little bit too expensive.
1:23:08
But the actual killer use case, I felt like it was there because if you love a movie and you throw it on like the content is solved.
1:23:14
like you can watch a great movie and be like that was a good movie and you're letting the movie We talked about this with Mark Andre where I was like the iPhone was a good iPhone was a good phone and he was like no it wasn't.
1:23:22
Um which is maybe a good take but eventually >> it didn't have copy and paste. >> No. Yeah. Yeah.
1:23:27
Well uh you know your like the Motorola V Razer V3 didn't have copy and paste but it could make reliable phone calls.
1:23:33
And so my take was like, >> yeah, it's just worth remembering that that it had some extreme shortcomings and it was frustrating to use.
1:23:39
And >> but it was but it was easy to justify pretty quickly as as a device that replaced another experience.
1:23:47
Whereas it's it's much harder.
1:23:49
It's like the Apple Watch works for those people because they're like, I want to know the time on my wrist.
1:23:54
I'm used to wearing a watch. That's easier.
1:23:56
It's much harder to be like pendant.
1:23:57
Not a lot of people wear pins all the time.
1:23:59
It's a harder It's a harder act.
1:24:02
The activation energy is higher.
1:24:04
So, I was always saying that that going to uh uh making the movie experience really really strong would be uh like an an obvious but killer app.
1:24:11
I don't know if people agree, but um Mark German's talking about this in context of the Apple Vision Pro.
1:24:17
He says Apple Vision Pro is suffering from a lack of immersive video.
1:24:20
Apple has slow walked the release of immersive video, creating a conundrum for the Vision Pro.
1:24:25
Uh the company's AI and smartphone smart home road mapaps are have have come out.
1:24:29
But um when you get down to the core of the problem, Apple's vision Apple visions pro Apple Apple's Vision Pro headset isn't selling well for two reasons.
1:24:38
One, it's $3,500 price tag and a lack of sufficiently compelling features.
1:24:45
There are other issues like a limited array of custom applications, the devices weight, and a cumbersome setup process.
1:24:51
But those are less important.
1:24:53
Developers are continuing to release apps, and there are now accessories that make the device feel lighter on your face. I got to try those.
1:25:00
Um, Apple has also dramatically improved its operating system.
1:25:02
The latest version of the Vision Pro software, Vision OS 26, now offers widgets and has been wellreceived.
1:25:07
But none of that matters for Apple if people don't buy the Vision Pro.
1:25:11
By all accounts, the device remains an extremely niche project uh product.
1:25:14
I'd venture to guess Apple has sold well under 1 million units in the US since launching it a year and a half ago.
1:25:24
Moreover, the headset just doesn't feel like a priority for Apple.
1:25:25
on the company's last earnings call.
1:25:27
CE >> wonder how many of those units have been returned.
1:25:31
>> That's a good question.
1:25:31
I mean, I would assume if you're quoting sales, you have to not include returns.
1:25:35
Um, I wouldn't be surprised if they've still >> I don't know. You count sold.
1:25:41
>> That's very sketchy if they include that.
1:25:43
>> Well, no, but he's he's guessing.
1:25:43
I'm just saying I I I just wonder what the return rate is because didn't you return your >> Yeah, I think it was probably like 50%.
1:25:50
They probably sold 2 million.
1:25:50
They wound up like leaving 1 million out in the market basically.
1:25:54
I mean, they have so many Apple stores and it's such a device and, you know, it's such a moment like it it took over the timeline.
1:26:00
It was the current thing for like three days.
1:26:02
People were talking about it.
1:26:02
Uh, a a lot of people tested it out.
1:26:04
A lot of people kept them.
1:26:05
A lot of people have the disposable income.
1:26:07
But yeah, 1 million does even seem high. Yeah.
1:26:10
>> Uh, this is interesting.
1:26:10
So, German says, "Moreover, the headset just doesn't feel like a priority to Apple.
1:26:15
On the company's last earnings call, CEO Tim Cook almost seems surprised when a Wall Street analyst quizzed him about the device and the company's strategy.
1:26:22
Thanks for bringing it up, he said before delving.
1:26:26
>> Mark German is delving. Interesting.
1:26:29
Before delving into new software features and asserting that it's an area we really believe in.
1:26:32
So, uh, German says in the near term, Apple isn't going to dramatically improve the Vision Pro.
1:26:38
The next version coming as soon as this year will mostly just get a faster chip.
1:26:42
That's a necessary upgrade.
1:26:42
The current M2 chip is outdated for such a processorintensive device, but it's not going to change the way that people think about the Vision Pro.
1:26:49
The bigger upgrade is coming in 2027 when Apple will release a model that is both cheaper and lighter, he's reported.
1:26:56
But that's a long time to wait and there's a risk that the category simply dies out by then.
1:27:02
I think that I I know a lot of developers that were super excited to capitalize on the release of Vision Pro.
1:27:10
They had all this energy and excitement around getting access to the product, starting to build applications for it and I don't know that any of them are still building for the product which is a problem.
1:27:20
The sort of like >> chicken and chicken and egg problem.
1:27:24
>> Uh Apple so uh he provides a couple examples of like how little immersive video Apple has really released.
1:27:29
Uh Apple is still featuring a highlight reel shot in immersive video of the NBA All-Star game from 2025.
1:27:39
To be clear, the 2020 uh from 2024, To be clear, the 2025 All-Star game played a year after the one Apple is showcasing took place six months ago.
1:27:47
And there's no immersion immersive vision of on the Vision Pro.
1:27:51
Like, you would assume they shouldn't have just said, "Hey, let's go shoot one immersive video of the All-Star game."
1:27:57
It's like, what is our NBA strategy?
1:27:59
What is our immersive video schedule? >> Yeah.
1:28:02
And what is our strategy to get 1% more content onto this device every single day forever? >> Yeah.
1:28:07
have a big release every Friday.
1:28:09
>> Like you sold a $3,500 device to a millionish people.
1:28:13
>> Yes, >> you should probably figure out how to deliver them content that makes it valuable.
1:28:18
>> If you look at the number of minutes that are being uploaded to YouTube every day like that number has probably been increasing every day since they started the company basically.
1:28:26
Like it's just it just slightly more every single day.
1:28:31
Sure, they probably have a few down days, but in general, like social networks, the amount of content on them should be growing every single day.
1:28:36
Any sort of media device, I'm sure Netflix is not like, oh yeah, like you know, next year we're going to have less content on Netflix. That would be insane.
1:28:45
Uh >> yeah, so apparently the immersive content uh shoots are extraordinarily expensive and resource intensive.
1:28:50
So, >> that is crazy because I I looked into how expensive it is to shoot an immersive video and I feel like it's actually not that expensive.
1:28:59
Like, yes, you need this fancy Blackmagic cinema camera that actually just came out with a special lens, but you can just set it up.
1:29:07
It can film for 45 minutes on one like, you know, SSD basically and and you can just upload that.
1:29:13
The problem is is that you can't if you're doing CGI and you're doing editing and you're doing like scripting and storytelling and all that like the the killer use case needs to be take the immersive video camera, put it somewhere cool and then allow people to watch that like in in Vision Pro and that needs to be it.
1:29:35
And so this was this was uh um Ben Thompson.
1:29:39
>> It feels like there's so much so much that they could do.
1:29:41
You could take immersive video behind the scenes for the F1 movie. >> Yes.
1:29:46
>> And make that available to people. >> Yes.
1:29:48
>> It's like you can buy buy this. You can watch F1.
1:29:50
You can watch >> behind the scenes with the actors, producers, directors.
1:29:56
>> There's just a lot that they could do.
1:29:57
But I think I think there needs to be something for users to really look forward to if they're going to invest in in this device.
1:30:02
And right now it doesn't really >> Yeah.
1:30:04
>> Yeah. The other thing is like they should have I feel like they should have done more like there was a little bit of like uh revealed preference in the fact that when you open up the vision pro the app that's in the top left corner like the first app if you are reading left to right like a book uh is the Apple TV app like it's very much like you should go
1:30:25
watch a movie right now you should click on this and then there's some other stuff but mainly it feels like they're pushing that um but they should have done more to really to really make it like a movie watching environment, something that that that people really uh really focusing on that narrowing instead of trying to do like seven different things. Is it for gaming? Is Is it for gaming?
1:30:41
Is it for like educational content?
1:30:43
Is it for this prehistoric planet thing and these like dinosaur stuff or games or collaboration?
1:30:50
They had FaceTime in there.
1:30:51
They just like threw a lot in there and I feel like none of them were breakout instead of just like chopping it down and being like, "Okay, like this is going to be the best place to watch an IMAX movie in 3D."
1:31:00
And so like you should get one of these specifically for this. I don't know.
1:31:05
Maybe it won't even work if if they pull that off, but uh it'll be interesting because Meta is going to fire back and um it will be uh so there'll be a small screen for many apps and alerts in the right uh on the right lens and the spectacles can be controlled via the neural wrist accessory.
1:31:22
So think about it like meta-array bands plus um Google Glass basically.
1:31:27
So you have to look up here and you can see notifications.
1:31:28
you can see a little bit.
1:31:30
Um, there is a there is a different there's a different set of companies that are building basically movie theater watching glasses.
1:31:36
So, it's just a big screen. It's not VR AR.
1:31:38
It's not positionally mounted.
1:31:41
It's just a screen on your face at a much lower price. I haven't tried them.
1:31:45
I don't think that they're fully there in terms of resolution.
1:31:49
A lot of this is just like waiting around for the screen in the Apple Vision Pro to commoditize to the point where um where other companies can start taking that same screen and putting it in other stuff giving like actually doing all the testing and then Apple will ultimately like take that back. >> Thank you.
1:32:08
Yeah, because like I'm pretty sure like the multi-touch screens like those were out there in other like the technology existed in other phones, but it just was poorly implemented and Apple was able to come in and do like the virtual uh the vertical integration. Eight, eight.
1:32:21
com pod 5 ultra get a 5year warranty, 38 free trial, free returns, free shipping.
1:32:29
>> Um, >> decent night 79.
1:32:31
>> Yeah, I think I'm around there too. >> Um, not my best. I'm building back up. >> Building back up.
1:32:35
I'm in a bit of a >> seven hours. >> Oh, 86. 86. Not bad. 8 hours. I did I did well. Feeling good.
1:32:43
>> Um Will Manita says, "If you're a young person interested in weird things, basically the only good advice is you should be a thousand times more commercial.
1:32:50
Someone will get very rich by monetizing the things you find out.
1:32:54
There's no reason that person shouldn't be you." Wholeheartedly agree.
1:32:55
Uh, I think oftent times if you're going down weird rabbit holes, whether that's in technology or something like health, you can often feel like you're late to something.
1:33:09
>> I'll give an example here.
1:33:09
I I started supplementing magnesium pretty heavily in college. >> Yeah.
1:33:17
>> And and continued post college.
1:33:17
This was like I graduated in 2018.
1:33:23
And I thought that everyone knew that taking magnesium was smart and good.
1:33:27
And basically, I mean, >> I don't I don't know if you're aware, but our our food used to have a lot more magnesium that just likely due to to soil health and soil degradation.
1:33:43
>> Uh, and since >> lobbying by the magnesium supplement industry, >> big mag big >> big magnesium.
1:33:48
Um, no, but anyways, so, soil's degraded, there's a lot less magnesium in your food, and basically everybody should be supplementing it to some degree.
1:33:57
Not health advice, but uh that's that was my personal takeaway.
1:34:02
And I ended up uh, you know, thought there was an opportunity to build a business around magnesium supplementation.
1:34:08
I didn't do anything about it.
1:34:10
Uh, and I think like 5 years later, a friend of mine started a magnesium focused company and like quickly got it to an 8 figureure run rate.
1:34:19
It's going to be a big big company.
1:34:22
>> Um, and so if you're if you're interested in these various weird niches, corners, uh, I think that, uh, there's oftentimes like, you know, massive opportunity there.
1:34:32
And, and if you're if you think you're if you think you're late, there's a good chance you might actually be early. >> Yep. Good point.
1:34:39
Uh, is your friend with the big magnesium company running billboard ads? >> Should be.
1:34:46
We should give him a call. Tell him to get on adqu. com.
1:34:47
Out of home advertising made easy and measurable.
1:34:51
Say goodbye to the headaches of ad advertising.
1:34:52
Only adquick combines technology.
1:34:54
Out of home expertise and data to enable efficient, seamless ad buying across the globe. Um, let's play.
1:35:00
We have Nor Sadiki uh calling in in just uh five minutes.
1:35:04
Uh, let's play the video.
1:35:07
She was recently on the Ross Douet podcast for the New York Times and I want to uh this this sets a big uh this sets a good stage for some of the debate that was happening on the timeline over the weekend. So let's hear from Ross.
1:35:22
>> You are excited about a world in which lots and lots more babies than is the case right now are born from laboratory fertilization.
1:35:34
And I'm just curious if you think, you know, allowing that this might be desirable in certain cases if a world where this became the norm would be losing something that is very fundamental to human beings and human families and human relationships.
1:35:50
And that's the relationship between sex and procreation, between you and your husband having sex. Apologies.
1:35:57
um and the future generations that come that come into being.
1:36:03
And I'm going to take the podcasters privilege.
1:36:04
I apologize for this, but I'm going to read you a piece of a poem.
1:36:08
>> I've never heard >> by a poet named Gway Canel.
1:36:11
>> And the poem is called after making love, we hear footsteps.
1:36:13
And the idea is sort of contained in the title that the husband and wife make love and it wakes up their child and the child comes and gets in bed with them.
1:36:22
And Canel writes, "In the half darkness, we look at each other and smile and touch arms across this little startlingly muscled body.
1:36:32
This one whom habit of memory propels to the ground of his making.
1:36:35
Sleeper only the mortal sounds can sing awake.
1:36:41
This blessing love gives again into our arms. Sorry.
1:36:47
Do you worry about removing or diminishing from human experience that aspect of being a husband and a wife in a relationship with a child? >> What do you mean?
1:37:03
>> Yeah, they they they kind of clipped her out of context there.
1:37:05
Should we play the full answer first?
1:37:06
Um because when you stop it there, it makes it look really bad.
1:37:11
But he actually clarifies and then she gives a much longer answer that we should play.
1:37:20
>> Um, we can pull that up. Here we go.
1:37:22
>> In the in a future where orchid technology becomes a norm, >> the feeling that that poet is expressing where a man and a woman make love and by making love they bring these.
1:37:31
But most people who get a baby it is linked inextricably to having sex with your spouse.
1:37:37
And that's and you are saying and you're saying it's time to sever that for the sake I concede of potential medical benefits.
1:37:45
>> I'm just saying I think pretty clearly something that like poets write about would go away. >> Yeah.
1:37:53
Yeah, I think that sex is a beautiful thing and I think that if you have, you know, enormous genetic privilege and for you to roll the dice and to get a um, you know, outcome that isn't going to lead to disease is is in the cards for you, then of course, you know, go ahead and and uh, you know, roll the dice.
1:38:08
It's just that I think that the vast majority of parents in the future are not going to want to roll the dice with their child's health.
1:38:13
They're going to see it as, you know, taking the maximum amount of care, the maximum amount of love in the same way that they, you know, plan their nursery, plan their home, plan their preschool.
1:38:22
All of these decisions are actually um, you know, extremely insignificant uh, in terms of the difference between is your child going to live with pediatric cancer, with um, a heart defect that we can't surgically fix, with born without a skull and never going to be able to make it to their first birthday.
1:38:38
I think when people think about it really concretely in terms of what are they giving up, what are the risks that could potentially um affect this child.
1:38:44
I think that then it becomes about stewardship.
1:38:47
It be, you know, it becomes about how do I make a responsible uh choice for my family?
1:38:52
How do I make sure that um you know, my child doesn't have to suffer in the same way that I do, in the same way that my sibling does, in the same way that um you know, my my my parent that I'm a caregiver for does.
1:39:01
So I I mean I I think sex is obviously a very beautiful thing.
1:39:05
It's a very profound part of the human experience, but I think that it's um yeah, I think it's, you know, denigrating and dismissive to IVF parents and to IVF babies to say that somehow, you know, science babies are um you know, inferior to to babies that are, you know, made the oldfashioned way.
1:39:21
I mean, every human life is equally valid.
1:39:22
And I think, you know, no parent who chooses to take the maximum amount of love and care and information going into that decision should be um you know, stigmatized in any way.
1:39:32
I think it's their personal choice.
1:39:34
And you know I think freedom and choice is you know what makes America you know a great place to to live and to and to be.
1:39:41
>> Nor Sadiki thank you so much for joining me.
1:39:44
>> This is Nor Sadiki of Orchid Health.
1:39:44
Um Orchid um they do whole genome screening for embryo selection.
1:39:53
And so uh there's obviously a whole a whole battery of tests that a lot of uh a lot of people do when they're having kids.
1:40:01
Uh usually this happens the first trimester, right?
1:40:06
So um but there is some risk to doing those types of tests I believe um where in a very small amount of cases something can go wrong because you're basically like scraping cells out of the baby I think like at that point and that's like and because you have to get the DNA to actually run the test.
1:40:21
Um and that can actually cause problems but it's so low that the benefit of being able to catch I never knew that.
1:40:30
>> You're like I never would have done this.
1:40:32
You're like, I'm hearing this for the first time.
1:40:33
Anyway, we have Noradiki in the studio. Welcome to the stream. How you doing? >> What's going on? Welcome to the show. >> Oh. Uh, one second.
1:40:43
We're we're we're we're redesigning the studio.
1:40:45
Can you give us another introduction?
1:40:50
>> Hey, thanks so much for having me, guys. Great to see you. >> Thank you. Uh, good to see you now.
1:40:52
Um, we're we're assembling the plane as we're flying it.
1:40:56
Uh, anyway, uh, thanks so much for joining us.
1:40:58
Uh, take us through the last 48 hours. >> Last 48 hours.
1:41:03
And also just like there's this weird thing that happens when like clips get sent out into the internet out of context.
1:41:08
Like how did you feel the interview with Ross Douet of the New York Times actually went?
1:41:12
Um because obviously it was a much longer conversation.
1:41:16
Uh this particular moment seems like heated.
1:41:19
Uh but what was the actual uh what was the actual uh interaction like?
1:41:24
And then what have the last 48 hours been like? >> Yeah. Yeah.
1:41:27
Um no, I think the interview was really fun.
1:41:29
Honestly, I think that uh it was um yeah, just just really cool to be able to um chat with someone who has, you know, such a different perspective about the technology.
1:41:37
But um yeah, it might help to just kind of back up and just kind of explain, you know, what Orchid is.
1:41:43
So >> um what Orchid is is we allow parents to protect their children from conditions um before pregnancy even begins.
1:41:50
So kind of, you know, to set the stage of, you know, what happens now, um when you do IVF, you're basically operating blind, right?
1:41:58
So you have extremely limited information.
1:42:00
Um a tiny percentage, one percentage of the genome is what's usually evaluated in order to make a decision about which embryo to transfer.
1:42:08
So what Orchid does is you we're the first company in the world to be able to allow you to screen the entire genome.
1:42:13
So 99% 100x the data compared to what what existed before.
1:42:16
And what that allows you to do is it allows you to detect conditions like birth defects, heart defects, pediatric cancers, some of those super severe disorders that previously you wouldn't find out until after the child was born.
1:42:28
So the exact same testing that happens um you know in the NICU after a child is born um you know those diseases instead of having to wait until you can only react when you you sort of have a conversation with a doctor and all they're able to tell you is hey there's nothing you can do.
1:42:43
that information is brought all the way up to that embryo selection stage to be able to actually transfer an embryo that's unaffected.
1:42:51
And I think the thing that's sort of really interesting about what's going on in this debate is it seems like people kind of don't understand or don't or maybe just ignorant of what's going on, right?
1:43:01
Because this whole question of, you know, should embryos be created and should people have the right to decide which embryo to select has has already been happening, right?
1:43:10
IVF has already been going on for 40 years and for the last 30 years they've had access to really limited information. Right?
1:43:15
So all Orchid is doing is basically upgrading that information from hey do you want to get 1% do you want to only know that chapter level chromosome level um to do you want to know you know the entire book the entire genome and be able to scan for the thousands of different genetic diseases that geneticists have cataloged over the last decade.
1:43:34
So that's the thing that I've actually found really surprising is that there's a lot of activation about this topic, but maybe there's sort of a little bit of ignorance about, you know, what's actually going on here, right?
1:43:45
Like there's no net new embryos that are being created.
1:43:48
It's sort of like you have this many embryos that are created during IVF.
1:43:51
That's just how IVF works. And you have a choice.
1:43:53
Do you want to make that information?
1:43:55
Do you want to make that decision with, you know, blind or with very, you know, limited information or do you want to make that decision with the maximum amount of uh information that science can give you?
1:44:05
>> Yeah, I' I've heard it's actually not technically blind that uh during IVF doctors will often just look at the embryos like v visually with their under a microscope and just kind of be like that one looks good and they'll they'll grade them like a b c d and it's extremely like vibes based.
1:44:22
They do actually look at them and grade them. Correct. >> Yeah. Yeah.
1:44:27
It's a It's kind of a beauty contest.
1:44:28
It's called morphology, but um it's a little bit tricky because, you know, they've done a lot of studies where there's, you know, if you basically flip the images, then an embryologist might grade them differently.
1:44:38
So, yeah, morphology is useful, but it's not, you know, kind of necessarily the most reliable.
1:44:42
And I think the other piece of the debate that I think is really missing is that, you know, you kind of mentioned um, you know, testing that happens during pregnancy, right?
1:44:49
pregnancy, right? So again when you test during pregnancy at you know maybe 10 weeks um the amount of information that you have there is again super super limited right so you might be able to screen for something like down syndrome down syndrome is called is you know techn the technical term for it is tricome 21 you have three pairs of
1:45:06
chromosome 21 it's actually not a lethal condition right it's survivable and right now all across the country in the US and all across the world you can get that information at 10 weeks right and right now women are in this um you know at this point where you the only option that they have is either to terminate that pregnancy or to have an affected child, right? So, if you're in that camp
1:45:26
So, if you're in that camp that is against abortion, then you should actually support orchid and embryo screening because it obiates the need for that termination to ever happen, right?
1:45:36
Because you only transfer the embryo that's unaffected.
1:45:37
the embryo that's unaffected. Um so that's the other piece of it that I thought was really surprising is that you know maybe people don't realize that you know genetic testing on a very limited basis is already happening and terminations are already happening on the basis of you know non-lethal
1:45:53
conditions right versus what orchids doing is is is looking at situations that are much more severe than that right so um you know over 20% of infants die due to genetic causes right if you look at babies that are in the NICU that unfortunately don't make it they've sequenced them and they found that at least 20% of infant deaths are due to genetic disease. Right? So, I think it's Right?
1:46:11
So, I think it's insane to tell parents who are doing IVF that they shouldn't scan for, you know, lethal disorders when, you know, we're already doing um, you know, elective terminations for totally non-medical reasons, right?
1:46:24
Just because people, you know, don't want to have a pregnancy at a specific time or for, you know, a specific medical reason, right?
1:46:31
Much much later than, you know, an embryo that that's five days old.
1:46:34
So, I don't I think that's something that I found kind of interesting is that um >> back to like Ross's like like Ross's like core point uh I the way I took it was like like it's pretty easy to make like a utilitarian argument that that this is a net good and this trend not even orchid specifically but like this this general trend of IVF is is a net good because it results in like less suffering.
1:46:59
Um, but his point was that there is a downside and and uh and I'm wondering if you debate that point or if you or if you say yes there is a downside.
1:47:11
We are losing something that poets write about but it's worth it because there I could make this argument about basically everything like space travel it's going to like SpaceX is going to wind up killing >> Earth.
1:47:24
Another way to look at it is like if >> the water used in data centers I'm okay.
1:47:28
I think that's a good use of water.
1:47:30
There's a million reasons why where I'm okay with trade-offs. >> Yeah.
1:47:32
I mean, I think if somebody just watched that short clip and and clip was so brutal to fullness of the interview.
1:47:39
>> They they might not pick up that.
1:47:39
I think it'd be insane if somebody's doing IVF and they're not using orchid, right?
1:47:45
Because if you're already going through this process and then you're saying, "Well, I don't Yeah, I'll just I'll just let the doctor kind of like do a vibes-based analysis of the embryos and he he'll just pick >> um you know, a beauty contest as you described it.
1:47:58
uh the you know seems like an obvious decision if you're going through that process to to leverage the max amount of data that you can have. Totally.
1:48:07
But then I think a lot of people were being triggered about this broader idea of like what what would we lose if if every if every baby >> Yeah. Yeah.
1:48:15
That's Ross's question is like is like if there's a future where where being able to do more genetic testing pushes more people towards IVF, do we lose something?
1:48:24
Do you do you agree with that? Do you debate with that?
1:48:27
Do you think that uh it's worth it? >> Yeah. Yeah. Yeah.
1:48:31
I think it's sort of uh basically fundamentally a you know personal decision like a private decision, right?
1:48:37
I think for each individual couple, each individual person, you know, they'll decide, right?
1:48:42
Do I lose more or do I gain more from that decision?
1:48:44
That's why I thought I think I was kind of surprised that he was asking me sort of like, hey, it's it's not really my decision, right?
1:48:48
Each parent is going to decide if they're losing more or gaining more.
1:48:52
And I think that yeah, if you think about any technology, right, like when we swapped um, you know, candles for electricity, when we swapped horses for cars, right?
1:49:03
Um, you know, there's always something lost, right?
1:49:05
The question isn't, you know, is something lost?
1:49:08
Every with every technology, something is lost.
1:49:09
The question is, you know, what's gained, right?
1:49:12
So, you know, when you have the decision of you, do I want to use an epidural during my pregnancy? Is something lost?
1:49:18
I mean certainly some women who choose not to use epidural because you know they think that you know what they gain from you know natural child birth is better but it's just it's just fundamentally a personal decision and it seems strange to um you know dictate to people or stigmatize people who choose you know epidural or not or you know to to screw their embryos or not. >> Yeah.
1:49:39
it kind of like uh I mean that plays out to all of technology.
1:49:40
plays out to all of technology. like we recently found out that the the Amish population is doubling every 20 years and that uh by the year 2030 uh uh 2300 there will be more than 7 billion Amish people and so like even though the people I mean people people can make a choice they they can choose not to
1:50:01
choose not to use IVL choose not to use anything >> there's a potential future where where groups like the Amish continue to just operate without technology and then there's a large amount of people that decide I it's worth it for me to go through this process and leverage technology to avoid >> a child with a a a permanent heart defect or or something of that sort. But
1:50:22
But I think that's a scary >> Yeah, >> that's a scary it's it's a scary world I think for a lot of people to imagine.
1:50:30
>> Uh which one's the scary world?
1:50:30
Uh, I think a lot of people that have I mean, it's easy for parents that have gone and and had a healthy child through the natural process. >> Yep.
1:50:45
>> To think about a scenario where nobody humanity no longer experiences that at scale.
1:50:51
>> And then >> you you the the other side of that is parents that have had a child with uh some type of of um compromised health to some degree.
1:51:00
uh that that would probably in some cases do anything to go back and and avoid having a child that you know died shortly after childirth or didn't make it through the full >> or or the opposite.
1:51:16
>> or or the opposite. I mean there I'm sure if you talk to parents who have had health complications and gotten through it in many ways they would say they wouldn't trade that for the world like like they they see that as like something that was a crucible that they
1:51:28
needed to cross and that was something that even though it feels very bad they >> they feel like it made them stronger in the long term but yeah I don't know it is tough it's a very hard situation >> what yeah sorry go continue >> oh yeah I mean I just think that it's just a super private decision, right? I
1:51:45
I just think it it seems super strange for people to be um you know trying to push their ideology on other people.
1:51:52
I think it's just like, you know, we there's so much um in the way right now of, you know, people having babies and fundamentally one of the biggest anxieties people have is, you know, is the disease that I'm affected by, is the disease that, you know, um affected my my sibling or my parent going to affect my child?
1:52:10
That's sort of like front and center, I think, for a lot of people is, you know, most basic moral um desire I think people have is, you know, I want my child to suffer less than me, right?
1:52:20
So any tool that's uh available I think should be you know just um you know should be available to parents so that they can make that decision for what's right for their family and you know if they think you know more is lost by you know not doing it the old fashioned way then that's what they should do and no
1:52:36
one should um you know stigmatize that but I think you know the same is true in the opposite direction and I think that unfortunately there's um yeah there's still like a huge amount of of stigma around uh IVF and embryo screening that I think shouldn't exist because it's such a massive force for good, right? Like there's sort of this huge category
1:52:51
Like there's sort of this huge category um you know of illnesses that previously we just had absolutely no control over.
1:52:58
We just had to roll the dice, right?
1:52:58
And I think for a lot of parents, they're thinking that hey, this is going to be the most important parenting decision that they'll ever make.
1:53:05
I mean, that's, you know, fundamentally what what I think, right?
1:53:08
I mean, there's no parenting decision that I'm going to make that I think is going to be as significant as is my child going to be affected by pediatric cancer?
1:53:13
Are they going to be affected by developmental delay? Right?
1:53:16
delay? Right? If you look at children today um you know 60% of kids with moderate to severe intellectual disability there's a definitive molecular cause there's a genetic cause for that right and a lot of those are denovo mutations that means they happen spontaneously in the embryos so that so
1:53:33
that means that even if you scan the parents ahead of time you you don't you can't mitigate that risk so um you have to screen the embryos you have to look at the earliest possible stage and I think that it's more compassionate to do that than to put women in this position
1:53:47
where you know they're already pregnant and then now they have to make this decision you know during pregnancy as opposed to let me actually have a pregnancy that's successful right so 50% of miscarriages actually are you know due to genetic causes as well so it's like >> you know as a woman you're going through
1:54:03
sort of the most intense physical experience right a pregnancy is like a marathon every single day um it would be it would be nice if you know you didn't have to have as many miscarriages right so >> what very obvious like it's just information that people should should have access to. >> Yeah. What's actually going on on the >> Yeah.
1:54:20
What's actually going on on the science side?
1:54:22
I feel like we sequenced the genome like 20 years ago.
1:54:25
People were getting genome sequenced in the mail like 15 years ago.
1:54:30
Uh but just now we're able to sequence embryos before they get implanted for IVF.
1:54:36
Like has there been some sort of fundamental scientific advancement? Is it a cost thing?
1:54:42
Like did the FDA approve something?
1:54:44
Like why is this now instead of like 15 years ago. >> Yeah. Yeah. Yeah.
1:54:50
There's a couple of things.
1:54:51
So, one, um, you know, specifically in embryo, there's a really tiny number of cells, right?
1:54:55
So, you have about 125 cells on day five and then five cells get sent to orchid for analysis and in those five cells, you only have about 30 pigs of DNA.
1:55:03
So, you have a really, really tiny amount of DNA. Got it?
1:55:07
>> And you have to amplify that DNA.
1:55:07
So, Orchid had to invent a new protocol, a new amplification technology in order to get really high quality whole genome data off of embryos.
1:55:17
>> Um, so when you have blood or saliva, you can get a whole genome sequence through a commodity process because you don't have to amplify it.
1:55:21
You have enough DNA to just throw it on a sequencer.
1:55:24
So, that's one piece of the puzzle.
1:55:26
>> And that's like the shotgun sequencing that uh 23 andme does where they're where they're just kind of like taking the average DNA that they're seeing in a bunch of different saliva and they have so much volume that they can like piece it together. >> Yeah.
1:55:37
So, so, so basically 23 and me is a saliva based test, right?
1:55:39
So, when you have saliva or blood, so where's the DNA?
1:55:42
The DNA is in the nucleus, right?
1:55:44
So, if you have saliva or blood, you have a ton of cells, right? >> Yeah. >> Millions of cells.
1:55:48
So, you don't have to amplify the DNA, right?
1:55:49
You just have enough DNA in the in the blood to just do it straight up.
1:55:53
And the other thing is that 23 and does not do whole genome.
1:55:57
They do something called an array.
1:55:57
So, an array is just a subset of the whole genome.
1:56:00
So, instead of looking at three billion letters, you're looking at like maybe 500,000 letters.
1:56:03
So it's again that sort of less far less than 1% number.
1:56:07
So if you look at a tiny fraction of the whole genome, you just can't scan for all the diseases.
1:56:11
So basically one piece of it is just can you read the data off of the off of an embryo, right?
1:56:16
So orchid the first to be able to do that.
1:56:18
And the second is what can you actually tell when you have the whole genome.
1:56:21
So that's um you know sort of the collective genetics and scientific community has worked on that over the last two decades to be able to catalog these thousands of different genetic diseases. Right?
1:56:30
We didn't used to know what is the genetic basis of um you know lots of different heart defects.
1:56:36
We didn't know before the genetic basis of of u you know syndroic forms of you know some of the most severe forms of autism right we had to actually catalog that by sequencing you know millions of people over the last decade or two.
1:56:47
So part of that is okay >> you know it's the first time you can actually read the data at that super early embryo stage and the second is okay what is the actual things that are going to be clinically meaningful to parents uh during that stage.
1:56:57
So and and then sort of the third part of that is you know what models can you use to be able to um predict and quantify risk for not the just those binary yes or no is that embryo affected or unaffected by a specific genetic disease but can you actually quantify genetic susceptibility for conditions where it's not just a binary yes or no that's kind of like the polygenic side of it.
1:57:17
So there sort of three different things that are kind of coming together to make um embryo screening. >> Yeah.
1:57:23
So the science kind of advanced.
1:57:25
Um what what what's going on in the FDA side?
1:57:28
Have you do you have approval? Do you need approval?
1:57:29
Like uh can you is it a different pathway than kind of the traditional uh like pharma drug pathway that we're familiar with from you know cancer drugs phase one phase two phase three.
1:57:41
>> The way that the way the FDA uh regulates all testing not just embryo testing is through something called LDTS.
1:57:47
So laboratory developed tests. Sure.
1:57:49
>> So um yeah there's basically two different agencies.
1:57:51
One is called CLEA, one is called CAP.
1:57:52
They operate at the federal and state level and they come and you know inspect our lab and make sure that um you know uh you know all the machines work and all of the you know all the analysis that we're doing is correct.
1:58:03
So sort of these independent um audits that happen at the um yeah basically at the state and federal level annually. >> Interesting.
1:58:10
So it's more like an ongoing process than like you get approved and then you have a patent and then 10 years later there's like a copycat knockoff product that's like the the the generic version.
1:58:19
It's a very different pathway.
1:58:23
>> Uh well, so sort of I guess there's a couple of different things, right?
1:58:25
the the patent on the amplification technology is sort of separate from the separate FDA.
1:58:29
separate FDA. But the regulatory environment is that laboratory developed tests are regulated via LDTS which means CLA and CAP these two different agencies actually come and examine labs at the at a physical level and at an analytical level right they come and you know verify that um there's something called like proficiency testing right so they
1:58:47
basically send you DNA sequences and they test you and they say hey did you you know call this correctly or incorrectly right so they sort of uh you know blind you to the results and that's kind of the the process for actually validating genetic tests right so orchid is genetic test on embryos and a sort of larger framework which is how do you validate genetic testing on you know blood or saliv? No, I think I think uh it's such a I
1:59:11
No, I think I think uh it's such a I think it's like such a personal technology.
1:59:17
It's like a like the >> most businesses do not have uh you know as they're growing and marketing their product service technology they're not there there's I think few things few products on on earth that would be potentially more controversial where one person might see Orchid and be massively relieved and uh you know immediately reach out and and want to learn more and then another would probably send you a very nasty DM.
1:59:51
Uh but uh but I think it's >> um so yeah, I think, you know, it's it's not even I don't even feel like somebody that traditionally comments on business, you know, having having had two kids going through, you know, that that process as a family.
2:00:04
I think there's so many decisions along the way that are just deeply personal that should be made within the family unit.
2:00:10
And I think, you know, whether somebody uh decides to use something like this or not should be again like >> everyone should just make their own.
2:00:20
>> You shouldn't listen to a podcast to get two, three, four kids the old way.
2:00:22
Have two, three, four kids with Orchid. See how each plays out.
2:00:27
Then you can judge the company. >> AB test. >> Just AB test it.
2:00:30
Or or run a run a Gatka experiment.
2:00:32
Have one the oldfashioned way. Have one superhuman.
2:00:34
And then see who can swim the farthest.
2:00:37
I know you haven't seen Gatka. It's a great movie.
2:00:40
Anyway, thank you so much for stopping by, Nor.
2:00:43
>> Yeah, thanks for having me guys. Have a good night. Cheers. >> Bye. >> Bye.
2:00:46
>> Um, you you really haven't seen Gatka? That's so insane.
2:00:49
Have you seen Gatka, Tyler? >> I've not.
2:00:52
>> You haven't seen Gatka?
2:00:52
Has anyone here seen Gatka?
2:00:54
>> John, you need to create a John's movie club. >> I do. I do.
2:00:56
I need to get more people watching. It's a fantastic movie.
2:00:58
The the plot is basically that that in the future you can design the perfect embryo and so that there are two brothers that are born.
2:01:06
are born. one like the oldfashioned way the the family leaves it up to chance and with the other brother they do like the perfect like jeans like everything and so one of them has a ton of like uh no like crazy disabilities but just like
2:01:22
isn't gifted athletically doesn't have the right like you know like musculature all those different things and and like the superhuman one is like totally ready to go uh to space and join like the most elite group and they're getting he's getting genetic genetically screened. Um
2:01:36
getting genetic genetically screened. Um but uh but like the brother who is I think played by Ethan Hawk uh who who doesn't have the advantages like uses the sheer abominable indomitable uh human will to like overcome everything like fake his way into this elite group and uh and ultimately like succeed and
2:01:57
uh and there's this famous quote you want to know how I did it this is how I did it Anton I never saved anything for the swim back because they would go out and swim in the ocean and he would always beat his brother who was like more athletic and it purely because like he had the drive and he was just like I was willing. >> He's on a mission. >> He's on a mission. >> Yeah, that's great. >> He's on a mission.
2:02:12
Well, Emos, back to the timeline.
2:02:14
Emos says, "Rengineering Reddit r/engineering.
2:02:17
I've applied to 10,000 jobs and haven't heard back from one engineering X.
2:02:21
I built a transducing comb with $20 >> combobulator."
2:02:28
>> Combulator combobulator with $20 of sheet metal parts.
2:02:31
Here are the instructions.
2:02:32
Please stop offering me jobs. Incredibly real.
2:02:34
If you uh build almost anything on X, you'll probably start getting job offers. So, >> yep. Yeah. >> Real alpha there. >> Yeah. Yeah.
2:02:44
The the the new the new job application is like vibe code something.
2:02:48
I mean, that's how we met Tyler.
2:02:48
That's how we met like uh just pretty much everyone has like has like kind of done something to showcase their Yeah, same thing too.
2:02:57
Um like what we've done and uh and and what they want to do and uh and I think I think people are really receptive to that.
2:03:03
You probably need to go kind of up the chain.
2:03:04
Um, but that's where people like the the founders are hanging out on Ax and you will get you'll get flooded with DMs.
2:03:13
>> Well, Gothamist is saying, "Call it a comeback.
2:03:14
Manhattan foot traffic finally tops preandemic times."
2:03:19
>> Let's hear it for Manhattan foot traffic.
2:03:21
>> The first time since the city. Uh, so yeah.
2:03:24
>> Did you notice a lot of foot traffic when we were in New York?
2:03:25
Seemed like a moderate amount. I don't know. >> I don't know.
2:03:30
It's hard to tell in summer and it just depends on the neighborhood. >> Yeah, I guess.
2:03:33
Well, if you're walking around Manhattan, you need a hitter on your wrist. Go to getbzzle. com.
2:03:38
Your bezel concier is available now to source you any watch on the planet. Seriously, any watch.
2:03:43
>> There's been enough robberies that maybe you want to leave the hitters. >> Oh, yeah.
2:03:47
Maybe you don't want it or keep it in the watch box.
2:03:49
Keep it in the watch winder. Keep it keep it wound.
2:03:53
>> And then save this for the Hamptons. >> Brother-in-law. >> Yeah. Rough.
2:03:57
>> Had a had a tragic incident.
2:03:57
Uh, Rune says, "George hots tried to fix Twitter search and decided it would be easier to take down Nvidia.
2:04:05
>> Search is the hardest thing."
2:04:07
>> Did you I didn't know that George George hot was at >> joined X post Elon buyout whose intern for days came in and was like, "Oh, what do you guys want to be a work on search?"
2:04:19
search?" kind of realized that like it is a truly massive system and there are some that like it is a big company and I think that uh X has obviously evolved a ton there's been a ton of change of of leadership but uh even under Elon didn't happen overnight um why Tyler do you
2:04:38
know why search is so difficult why is it so intractable is it just a consumer >> preference issue like there's just so much stuff >> yeah you'd think like yeah just indexes or something would be better, but it's just so hard to I use a lot of like the filter colon follows to search like people that I follow for a topic. So, if
2:04:54
So, if I want to if I if I want to understand what people are saying about Nvidia today, I'll search Nvidia space filter colon follows and it's just who I follow in the feed of results. That works pretty well.
2:05:08
But, um, Twitter search doesn't seem that bad these days. It seems actually okay.
2:05:12
I think it's probably a function of the fact that like there are so many different subcommunities on X like teapot is its own like thing and then separately there will be like basketball Twitter and the these two areas will almost never collide and they're like these separate loose clusters that that have very little overlap.
2:05:33
>> They're absolutely beautiful. >> It's beautiful. It's beautiful. every once in a while.
2:05:36
I I mean I think we literally did collide when we put up the traded post uh from Meta Super Intelligence hiring an AI researcher because that was that definitely hit with the sports fans as well. >> Um got something.
2:05:48
>> Yeah, I still see that post on Instagram sometimes.
2:05:52
>> Sometimes >> somebody put it on Instagram and got like like five million views or something like it it it rips no matter where it goes.
2:05:57
There's some combination of like that guy his name meta super intelligence poached is just so good.
2:06:05
Uh, it was very, very funny. Anyway, book a wander. Find your happy place. Find your happy place.
2:06:12
>> Book a wander with inspiring views, hotel grade amenities, dreamy beds, top tier cleaning, 247 concier service.
2:06:15
It's a vacation home, but better folks. >> Better.
2:06:20
>> Um, Defense Analyses and Research Corporation says we need a Teal Fellow style program that works to identify the next generation of talented private military contractor operators.
2:06:30
Who's going to be the next Zoomer? Eric Prince.
2:06:34
Um, >> the oldest members of Gen Z are 28 years old. >> Yep.
2:06:42
>> Should they be operating PMC's yet?
2:06:42
Are they >> Well, when we talked to Eric Prince, we found out that he was in a very unique situation where he couldn't necessarily get pipelined into it.
2:06:51
He had uh I think he had a family business that he was taking over.
2:06:55
He was doing training with the military.
2:06:58
So the Zoomer Eric Prince might be out there, but they aren't necessarily capital constrained.
2:07:03
It feels like starting the next generation of PMC is not something that you can buy with a $40 million series A or Mango Seed round.
2:07:12
Like it needs to be based on connections, experience, all these like loose piecing things together.
2:07:18
Um it doesn't seem like it's something that is just like a more elegant design or like a cracked team. What about you, Tyler?
2:07:26
You want to start one of these?
2:07:26
I would like to see like a company like 1X >> get into it to PMC.
2:07:32
>> Uh I think they're not quite there yet.
2:07:34
Did you see the the the the Chinese robotics fight that happened over the weekend?
2:07:40
>> Oh yeah, they had the world the humanoid robot games.
2:07:43
>> Yes, in uh in Beijing.
2:07:43
Uh before they can rule the world, robots need to master basic chores.
2:07:49
This is Hannah over Journal.
2:07:53
Human-like robots are great for entertainment.
2:07:55
Mean menial labor, not so much.
2:07:57
Now, you got to put this in the truth zone a little bit because did you see the back and forth between Gnome Brown at OpenAI and uh uh who's that humanoid robotic Brett Adcock at Figure?
2:08:09
So Brad, so figure put out a video of a robot folding clothes, folding towels, and Gnome Brown came in, spicy reply, says, "What happens if you raise the the table six inches?"
2:08:22
Uh, and so then Brett Adcock, uh, raised the table six inches and and did it again and was like, "You are you not are you not entertained? Are you not satisfied?"
2:08:32
Um, >> what if he just put the robot on a platform, though? >> Well, no. No.
2:08:36
So, so that the the the video that Brett I shared was uh was I I feel like it it passed the test that Gnome Brown uh was throwing down, which was that while the robot was doing the chores, the the folding, they came up and adjusted the standing desk and raised it 6 in and be and so it it did adapt to it kind of in real time.
2:08:57
Could have been scripted, could have been teleoperated. We don't know.
2:09:00
But uh it seems like uh it seems like it at least satisfied no Brown.
2:09:05
said, "Congrats, like good job.
2:09:07
Excited to see where this goes."
2:09:09
Um, so at the humanoid robot games in Beijing this weekend, more than 500 humanoid robots, that's a lot of contenders, uh, competed in both sporting events and real world tasks such as moving boxes, delivering luggage, and cleaning rooms.
2:09:24
Some were remarkably fast and agile, but most were clumsy and inconsistent.
2:09:28
Throwing away nine pieces of trash in a mock hotel room took more than 17 minutes for one robot.
2:09:35
I thought you were going to say 17 hours. >> 17 minutes isn't bad.
2:09:38
I mean, really, like the these are all of these things are on exponential growth.
2:09:42
>> We're so far from the Robo X games.
2:09:45
Humanoid X the humanoid robot X games. >> Seven. Yeah, we are. But may maybe not. I mean, 17 hours.
2:09:49
That was probably like last year, right?
2:09:51
And then the year before it was like infinite hours.
2:09:52
So, we really are coming down at an escalating rate. It is exciting.
2:09:56
Uh, in a pharmacy simulation, a robot spent nearly five minutes grabbing three boxes of medicine.
2:10:05
In a factory scenario, a robot spent about 2 minutes placing two containers on design designated shelves.
2:10:09
Everything that is easy for humans is a challenge for robots, says a PhD student from Germany who helped train robot soccer players. >> Yeah.
2:10:18
Who was talking about the challenge of just trying to get something out of your pocket?
2:10:21
>> Oh yeah, that's analysis.
2:10:21
Level five is are are tasks that are force dependent.
2:10:27
So, if you want to pick if you want to pick up and fold this newspaper without ripping it or you want to pull a phone out of your pocket without ripping your pants, like it's very different than like if you're grabbing a hunk of steel, like you can kind of apply like the same level of force.
2:10:38
Um, but something that you don't want to crush.
2:10:40
Uh, so it's like, you know, assembling assembling food on a plate in a restaurant.
2:10:43
Like you don't want to squish the you don't want to crack all the chicken bones if you're putting wings down, I guess. I don't know.
2:10:50
Uh, China has said it wants to be a world leader in humanoid robots by 2027.
2:10:55
Is it >> is it not >> a world leader?
2:10:58
It's got to be in the top two.
2:10:59
Like, who else is in the game?
2:11:02
It's it's America and China.
2:11:02
Like, mission accomplished, China. Congratulations.
2:11:06
You are world leader as far as I'm concerned.
2:11:08
If not the world leader, what they're doing is very, very impressive.
2:11:11
Yeah, we're we're over here in the US already talking about how do we make sure that China can't do for humanoids what they did for drones and flood the market, make it impossible for American companies to >> like the default is that they will win.
2:11:23
But, uh, you know, they're setting the bar low so they can beat expectations.
2:11:26
Respect to Beijing on this one.
2:11:26
Uh, the Olympic style event, however, uh, suggested that truly useful humanoids are still years away.
2:11:34
a robot trained by the Beijing Institute for General Artificial Intelligence in collaboration with Unitry Robots.
2:11:39
Um, a leading Chinese robot maker comp complete competed in a hotel reception scenario using a three-fingered hand to drag a suitcase to a designated door.
2:11:50
It moved in small stomping motions, freezing at times.
2:11:55
Many of our humanoid robot algorithms are still in the lab demo stage, says a researcher.
2:11:59
another competition simulating a hotel environment involve involved.
2:12:05
>> Figure can figure Figure can really >> like uh rehab their brand in the valley by putting one of their their guys in in >> Oh, if they go to China.
2:12:15
>> If they go to China with the figure Chad humanoid just walks out and just dominates.
2:12:21
>> Yeah, we need uh uh what what are those famous >> If they can do that, I've seen enough.
2:12:25
give them give him the 40 billion post or 38 billion whatever the whatever the the the potentially uh not real fund raise uh was >> yeah we didn't really get an update on that.
2:12:36
Uh the whole idea was like it was it was kicking around it was going to happen but then nothing nothing really came came out of it. Um I don't know.
2:12:42
I mean company's still cooking.
2:12:45
They're putting out videos so you know still around.
2:12:48
Um, yeah, we need uh we need an Usain Bolt versus Johan Blake at the 2012 Olympics.
2:12:54
Bolt after losing to Blake at Jamaica's Olympic trials.
2:12:58
Silence doubts by >> I just want to see I want to see I've said it before.
2:13:02
I want to see cliff jumping. I want to see skydiving.
2:13:03
I want to see big wave surfing.
2:13:06
Like these are the important ch I don't care that the humanoid robot can fold my laundry.
2:13:10
I want to see it jump off a cliff.
2:13:15
>> Oh, this is a good one.
2:13:15
1972 Munich Olympics, USA versus USSR in the men's basketball final.
2:13:21
In a controversial finish, the Soviets beat the United States 51 to 50 after officials gave them multiple chances to score in the final seconds.
2:13:30
The US team famously refused to accept their silver medals.
2:13:32
I don't know if that's hallucinated from TP.
2:13:35
>> Wait, the Soviet Union beat the United States at basketball in the >> This was the This is the 70s.
2:13:41
>> Still, >> we hadn't Michael Jordan yet. have the the technology.
2:13:45
Uh >> no no you know what we need uh do you know the story of Nancy Carrian versus Tanya Harding >> 1994 Olympics.
2:13:52
So Carrian was attacked weeks before the games and a plot linked to Harding's ex-husband.
2:13:59
Uh the showdown captivated the world. Carrian won silver.
2:14:04
Harding p placed eighth and the scandal became one of the most infamous Olympic stories.
2:14:08
Uh, it I I think Tanya Harding got her exhusband to whack Na Nancy Carrian in the knee with like a pole to injure her so that she could beat her because they had a fierce rivalry.
2:14:21
So that's what we that's that's Brett Adcock's real like play.
2:14:23
He goes over to the unit and just or like electrical interference.
2:14:29
>> I mean, just have one of the humanoids do it and be like, "It was a bug."
2:14:32
>> Yeah, maybe >> it was a bug. >> Oh, yeah. Yeah. Yeah.
2:14:33
Just it just spazzed out. No worries. >> Yeah.
2:14:35
The they do these things all the time.
2:14:37
We everybody's seen the videos at these points of of >> what's going on.
2:14:41
>> I have to issue a correction on on the on the kid testing.
2:14:42
So the testing that happens uh in the first stage of the tri uh in the first trimester uh you don't actually touch the baby.
2:14:49
You test on the blood and the placenta. So it's much lower risk.
2:14:55
>> Uh it's called CVS testing.
2:14:55
CVS is typically performed earlier in the >> I thought that sounded a bit >> then amnocentesis.
2:15:00
CVS involves sampling placental tissue while amnneoccentesis involves sampling amniotic fluid.
2:15:06
Both tests carry small risk of miscarriage.
2:15:10
>> The real test for humanoid.
2:15:11
>> So there is there is still a risk of of miscarriage for both of those.
2:15:13
Very very small though like like so so small that million. >> Yeah. Like one in a million. Exactly.
2:15:20
And so so everyone basically everyone does this these tests and no one has problems.
2:15:24
>> Well, you know what the real test for humanoids is? >> What?
2:15:27
>> If they can beat a 70-year-old an average 70-year-old at pickle ball. and thoughts.
2:15:32
>> I didn't realize that Joe was on your team. >> He's on my team.
2:15:34
Of course he's on my team.
2:15:35
>> Of course he's on your team.
2:15:36
>> Joe says, "Look, I'm done hating on pickle ball.
2:15:38
Live and let live, but hard for me to take seriously any sport where 70 year olds can beat 20 year olds."
2:15:42
And Trace Cohen says is responding earlier saying, "It's a lot of fun for all ages.
2:15:50
So, you can see 70 year olds beating 20-year-olds, which almost no other sport can do."
2:15:53
I got a side with uh I gota side with Joe here.
2:15:57
>> I'll take the other side of this. I'll take this.
2:15:59
I'll I'll take the other side of this. Hard to take seriously.
2:16:00
Take it seriously from a financial perspective.
2:16:05
One pickle ball court is one quarter the size of a tennis court.
2:16:08
So if you own a tennis court, you can you can put four pickle ball courts on the tennis court. >> 4x your earnings.
2:16:17
>> Put your money where where your mouth is.
2:16:18
>> If Soho House has one tennis court, they can >> take take 20% of your liquid assets and rotate them into various pickle ball bets.
2:16:27
If you're so if you're so bullish, John, if you're so bullish, why don't you why don't you invest in a pickle ball team? >> Okay.
2:16:33
Well, I I will consult the pickle ball expert, Tyler Cosgrove over there. How you doing? >> All right.
2:16:37
I wouldn't say pickle ball expert, but but okay, Jordan, do you think golf is a sport?
2:16:40
Because in golf, you could easily see a 70-year-old beat a 20-year-old.
2:16:44
>> Yeah, Donald Trump, he's gonna >> I'm not saying I mean, I just mostly don't like I'm just saying I don't like I don't like watching it. Don't put it on the TV.
2:16:53
Don't put it on my neighborhood's court. >> Okay.
2:16:58
And uh don't put uh Nick over there. I see you smiling.
2:17:01
Nick was trying to put the TVPN logo on a pickle ball paddle. I said absolutely not.
2:17:06
>> Yeah, >> absolutely not.
2:17:08
>> Uh Tyler, when you play pickle ball, uh do you wear flip flops?
2:17:14
>> I I I was I was road testing.
2:17:14
I didn't just fully send it, but I I'm thinking of uh of a nickname for Tyler.
2:17:20
of uh of a nickname for Tyler. just call them flip flops because if you have a following why not just flip flops flip-flops over here has been doing the full murf every week in flip flops >> which I think >> I have to make it harder for myself >> hilarious and iconic >> just running in flipflops is >> he said the reason he gave me was like I
2:17:39
don't have another pair of shoes and I'm like you have a MetaQuest 3S Xbox edition sitting in a box that you could easily put on eBay and trade for a pair of shoes you have no excuses I should have traded up to a house >> except the fact that it does give you insane aura because doing a full hero wad weekly in flipflops is is a it's a good move. >> Yeah, you must have an insane callus
2:18:02
>> Yeah, you must have an insane callus where we're like in between the >> totally. Yeah, rainbows. They're worn in.
2:18:08
>> Want to think about that.
2:18:08
Let's change the subject.
2:18:09
Uh Dylan Field says Benjamin Franklin's routine is goals >> and >> yes, >> the good morning questions.
2:18:17
What good shall I do this day?
2:18:19
Rise, wash, and address powerful goodness.
2:18:21
Contrive day's business and take the resolution of the day. >> What does that say? Prosecute.
2:18:29
>> Prosecute the president and breakfast.
2:18:31
And then from 8 to 12, >> 11 to noon. Yeah.
2:18:36
>> At noon, read and overlook my accounts and dine.
2:18:39
>> It's a 2hour twohour lunch. I love it. >> Insane. And then >> 2 to 5. >> Two to five.
2:18:44
He's putting up hours working again.
2:18:46
And then of course around six or seven foot >> four hours of putting things in their places, supper, music or diversion or conversation >> exclamation of the day.
2:18:57
>> You know, uh I was thinking about it and >> there's been this debate around alcohol recently.
2:19:04
Is it because people are getting more health consscious?
2:19:05
Is it because people are being less social? >> Sure.
2:19:10
>> And I think it's because nobody gets bored anymore.
2:19:14
>> This is my my working theory, right?
2:19:14
It used to be somebody >> maybe there wasn't any good television on >> and so they're end of the day they're bored >> and they just start drinking.
2:19:24
Now >> they can numb the mind with uh with a scroll >> or a thousand >> or an ancient scroll if you're not Freriedman. >> That's true. That's true.
2:19:34
But I think there's I think there's something there.
2:19:37
It's um if you look at social media as effectively you know its effect on the brain is similar to uh to to a drug in in many ways that uh you know I think I think when you when you look back in history these different eras where people were just drinking for you know the entire day.
2:19:54
I think part of that had to have been >> just like general boredom right like wanting to like change change their mental state.
2:20:03
Yeah, I I'm very rarely bored. I'm very rarely bored.
2:20:07
It happened this weekend.
2:20:09
I tried to watch the new Superman movie and I was not I was not very into it.
2:20:12
I couldn't get into it and I was thinking like >> I might crack open a beer, but I didn't have any.
2:20:17
So, I didn't crack open a beer.
2:20:20
>> Um, back more timeline.
2:20:20
Molly Contilian says, "Nothing screams I am the main character more than America having the plus one phone code viral."
2:20:31
And then somebody quotes and dunks on her.
2:20:33
It goes even more viral saying, "We invented the telephone.
2:20:38
>> We did invent the telephone. We deserve it. We earned it."
2:20:40
>> Still pretty main character activity to the telephone.
2:20:44
>> Is is that you would you would assume that I feel like the uh the phone code should be like just a power ranking like like in Dubai, you know, where the license plates >> it is a power ranking. >> No, it's not.
2:20:57
>> I mean, >> is China number two? >> No.
2:20:59
>> Is Japan number three?
2:20:59
But I really only care about number four.
2:21:02
>> If you're not first, you're last. >> I What?
2:21:04
What are the other like what what are the other phone codes? Like uh top 10.
2:21:08
>> I think China >> countries plus 86 >> codes. 86.
2:21:14
>> They're not the 86th than anything.
2:21:15
Certainly not in humanoid robots.
2:21:17
They're easily in the top 10.
2:21:18
>> They're not 86 in spam.
2:21:18
Like getting a a plus 86.
2:21:22
>> Oh, and Canada is piggybacking. Canada also gets a one. See? Yeah. UK 44. >> Stolen valor. Stolen valor.
2:21:28
United Kingdom really the 44th country to adopt the phone? Like what happened? What happened? >> What are they doing? >> Japan.
2:21:36
I feel like they invented telephones. Like so many phones.
2:21:37
They like adopted technology very early. They're 81.
2:21:40
Germany, Russia got on the got on the program pretty quickly. They're number seven. >> Um France 33. >> This is funny.
2:21:48
So Lewis Hamilton's racing number is 44. >> Oh, okay.
2:21:53
>> Could that be because he's British?
2:21:55
>> Oh, maybe >> he modeled it after the phone code. >> Maybe. Maybe.
2:21:57
Uh, I love this post from uh uh friend friend and uh and repeat guest via his post geot tickets.
2:22:04
Me when I see an AI generated video that looks two standard deviations below the current quality benchmark for Frontier Model Labs. >> Wow. >> Wow.
2:22:17
>> Which uh been seeing quite a lot.
2:22:20
>> Um >> there's a lot of slop on the timeline right now.
2:22:22
But Nikita Beer's working on it.
2:22:24
You saw that he had the he's adjusting the dial human post. AI slot posts.
2:22:28
Every every once in a while there's an AI slot post that's so bad it's good. You want to see it. >> That's true. That's true.
2:22:32
Uh, another post I'm slowing in here. We don't have it. >> Also underrated.
2:22:35
It is possible to create AI slop without the use of AI technology. You can one.
2:22:40
And I was actually thinking about this.
2:22:42
So, uh, obviously this was somewhat inspired by Superman.
2:22:46
Incredible ver visual effects team on that.
2:22:48
No hate to them, but the cinematography just doesn't feel grounded.
2:22:52
And I was hoping for this next reboot of a DC character to be like the Nolan Dark Knight.
2:22:57
Like give me something a little more gritty, a little bit more grounded, bring in a serious filmmaker and reboot Superman properly.
2:23:04
And instead, it's like this very like jokey, crazy.
2:23:06
There's like a whole bunch of funny scenes.
2:23:09
Like it's funny, but the CGI is like the camera's flying all over the place.
2:23:13
It's not like grounded, and there's a real way to do this.
2:23:15
So, uh, I know you haven't seen Pacific Rim, but, uh, it's a fantastic film by an award-winning filmmaker, uh, who did Pan's Labyrinth. What's his name? Pans's Labyrinth.
2:23:28
Uh, who who did Pan's Labyrinth? >> GMO Del Toro. That's right.
2:23:33
So, GMO Del Toro does does um does Pacific Rim.
2:23:35
And in Pacific Rim, it's this crazy huge robots, these joers that fight these massive monsters.
2:23:45
could easily be completely CGI, over-the-top, crazy Marvel movie, camera all over the place.
2:23:52
But he decided to ground the cinematography in shots that could be filmed practically, but obviously weren't because you can't go film a massive Godzilla fighting a massive robot.
2:24:02
But so he like when you're watching the movie, it feels like, okay, I'm filming from a helicopter.
2:24:07
Now I'm filming from the ground.
2:24:09
Now I'm filming from the top of a building.
2:24:11
like as if someone were looking off the top of a building watching this fight happen.
2:24:17
It's not just like the camera flying all over the place like it's impossible.
2:24:19
And uh and then for this the sequel, he's not on the he's he's not on the film.
2:24:25
He's not attached to the film anymore.
2:24:27
Uh and for uh uh the for the sequel, the camera's kind of flying all over the place, and it got a lot worse reviews.
2:24:33
Um and I was hoping that Superman would do that, but they didn't.
2:24:36
It was kind of a crazy film.
2:24:37
I didn't really I didn't really enjoy it that much, but um it felt like even though they didn't use AI, it felt very downstream of AI stylistically and kind of like hallucinatory even though they clearly used a uh like a traditional CGI pipeline.
2:24:55
pipeline. And I was kind of thinking it'd be funny to to try and uh to try and we were talking about another another video that we saw that that you you your initial reaction to the video was this is AI slop and I told you no it's actually all filmed it's just a lot of VFX but you're like VFX is typically in the unclop this was this was handmade
2:25:17
slop but I was thinking what if you went a step further and tried to make an AI slop video entirely practically so you don't do any VFX But it's like we're wearing prosthetic sixth finger prosthetic sixth finger and like really like like really like you know like grease on the lens to like make it more hallucinatory. Like the
2:25:35
Like the actors are constantly changing so you're swapping people in and out so the face doesn't always look exactly the same.
2:25:43
You get like seven different brothers to play the same actor and they're always slightly different and in every scene it's slightly different.
2:25:48
like you could create AI slop using traditional uh methods.
2:25:53
And I think I think the takeaway is like like even AI slop is merely one form of slop.
2:26:00
You need to be in the business of not making slop at all no matter what your tool of choice. >> Great.
2:26:04
Well, Super Superman could have used AI uh potent it would have been potentially helpful for one of the co-stars in that one scene. Did you see this?
2:26:14
This video is going viral.
2:26:16
>> Uh one of the kissing Yeah.
2:26:16
Rachel >> Barznahan the director calls cut.
2:26:19
She keeps >> she keeps going in.
2:26:23
Maybe maybe she was just in character but uh >> it's a funny outake.
2:26:27
How did that outtake leak?
2:26:29
That's a crazy behindthe-scenes video to like put out on the internet.
2:26:34
Do you think that's viral marketing?
2:26:35
Do you think it's intentional?
2:26:36
Do you think everyone's bought in?
2:26:37
Because sometimes like the the >> Didn't seem like Rachel's husband was that bought in.
2:26:43
>> Sometimes the sometimes the production team will be like that was amazing.
2:26:45
I'm like offloading this footage.
2:26:46
Uh, but you can get in a lot of trouble for that.
2:26:49
Uh, and then sometimes they do it deliberately.
2:26:51
I think during the promotion of the uh of the latest Spider-Man movie, they intentionally leaked a VFX shot of the Andrew Garfield who played a previous Spider-Man in the VFX scene and they made it look like, oh, they like they like this is like fake CGI or something, but it was really like a teaser for him being in the film.
2:27:13
Spoiler alert, the the most recent Spider-Man features like multiple Spider-Man from previous previous eras. They come back.
2:27:19
So, it's not just it's not just is it Tom Holland who is the main one?
2:27:23
Tom Holland's the main one.
2:27:25
Andrew Garfield is the previous one.
2:27:25
And then there's someone else who's the pre the one before that. And they all appear.
2:27:30
Anyway, uh Kate has assembled a list of blogs based on TBPN's Metis list. Go check this out.
2:27:37
She says it's the fastest way to get smarter is to read smarter people.
2:27:40
Um, and she lists a whole bunch of interesting blogs here.
2:27:45
Carpathy, Li, Lillian Wang, Sergey Lavine, Jeff Dean has a blog. Go check it out.
2:27:50
Uh, Daario of course is blogging. Great stuff.
2:27:52
Lots of lots of good stuff to and fun to see people riffing off the medicine list.
2:27:56
Uh, also Gabby Goldberg gassing you up says, "Quote from Jordi Hayes that unironically changed my brain chemistry.
2:28:04
The best neutropic is being on a mission." >> Really is true. I've tried.
2:28:07
I've tried every neutropic under the sun and uh nothing hits like loving your your work and having a clear vision for what you want to accomplish. >> Yeah.
2:28:19
This is kind of a >> there's like a lot of people that are that are out there and I've been this person before >> where you're searching for this like hack >> totally >> to get yourself to be focused have that drive dayto day. >> Totally.
2:28:34
And you can't, it's almost impossible to um that you can you can briefly hack it, right?
2:28:42
You know, people that that take Adderall when they have an exam coming up, right?
2:28:46
You can kind of force it, but it's still forced.
2:28:47
And the only thing that is that is durable is is like truly being on a mission. >> Yeah.
2:28:55
It's kind of the this is kind of the flip side of uh of your take about being commercial.
2:28:59
Like it was like you're you're you're early to the magnesium thing, but like would you be happy if you were like could you see your life's work being running a magnesium company or do you feel you were a better fit for this?
2:29:11
Would you be more on a mission if you had if you had stayed there versus coming here and doing this?
2:29:15
Um >> yeah, it's it's interesting.
2:29:16
I I I certainly don't wish I could go back in time and start that company, right?
2:29:20
I think that I think that uh one thing that I do think is is addicting to people and it's a positive loop is that >> somebody can be working on something that they don't feel like is their life's work but they can get addicted to the process of winning and making the number go up. Yep.
2:29:37
>> So make number go up >> and being extremely fixated on that >> is also effectively a powerful neutropic because you wake up every day >> and you want to make the number go up. Yep. >> Right.
2:29:47
And that's that's most of uh >> most of our modern lives as humans >> indeed. >> For better or worse.
2:29:53
>> Uh this Sean Franco post was funny because I thought he was going to be talking about uh rapper companies and and AI companies.
2:30:00
Uh he says unpopular opinion.
2:30:03
Gross margin is not product costs and I was like, "Oh, he's going to talk about application layer companies paying foundation model companies through the nose."
2:30:10
Uh, but he says, "Gross margin must include product costs, freight cost from factory to warehouse."
2:30:17
I was like, "From the AI token factory to the Amazon AWS data warehouse." Okay, got it. Okay.
2:30:21
Payment processing rates, of course, like you have to pay Stripe or whatever you're using.
2:30:26
Uh, returns and exchanges.
2:30:26
And that's where I got confused because I was like, I I didn't know that you could return tokens generated from from the claw.
2:30:35
>> This one wasn't that great. Take it back. >> Take it back.
2:30:37
Sales channel fees, you do got to pay those.
2:30:39
um fulfillment to customer.
2:30:42
Uh that's the that's the bandwidth, baby.
2:30:44
Uh this is the gap definition of gross margin cost to get the goods in the hands of the customer.
2:30:51
Obviously, this is a great takeaway for folks who run physical goods businesses.
2:30:54
Uh also relevant to the AI labs, but >> physical product founders would say accounting rules aside, >> yeah, >> I have 80% gross margins. >> Yeah.
2:31:06
Uh, cope or not cope, buco capital bloke says, "Chat GPT adoption did not reduce Google search usage.
2:31:11
People did not substitute their typical Googling with chatgptting.
2:31:15
In fact, there was a slight increase in average Google search usage after chatgpt uh chat gpt adoption.
2:31:20
Now, oddly this is from SEM Rush.
2:31:23
Their business is in Google SEO.
2:31:26
So, they might not be the most reliable narrator. What do you think? Do you believe this?"
2:31:32
So the data says that uh before using Google before using chat GPT users were doing 10.
2:31:38
5 Google search sessions per week that feels extremely low.
2:31:42
Um but after using chat GPT that increased to 12.
2:31:45
6 I could see it being additive.
2:31:48
My my joke on this was like it turns out you can spend more time on the computer.
2:31:53
Um but uh and and I do find I do find that I I go to chat I kick off like something I want to learn about and then I'll also be googling like that happened with our story about uh Rishard Mill.
2:32:05
I found it in deep research the history of Rishard Mill.
2:32:10
There was that interesting story about them getting the price wrong.
2:32:12
I went to Google to fact check it.
2:32:14
I went to Google to find images of Rishard Mills.
2:32:17
I wound up doing a number of traditional Google search queries.
2:32:19
So, um I I wonder maybe I should uh like track my screen time or something like that and actually or I bet I could just pull my actual Google results, my Google data.
2:32:30
Maybe if I'm logged in, I could see how many searches I've been doing, whether it's really fallen off a cliff or actually it's it's additive. I don't know.
2:32:37
Do does it feel like you're using Google less?
2:32:39
Everyone, everyone affiliated with OpenAI says like I never use Google anymore. Um but it's unclear.
2:32:45
I mean, they're clearly aligned the other direction.
2:32:46
You know, they they want to tell the story that they've replaced Google entirely.
2:32:49
I definitely use Google a lot less on my phone.
2:32:54
>> Yeah, >> I still >> I find I'm using it a ton for Google images.
2:32:58
Uh because if I want a real image to cut out, like I'm not going to go to chat GPT for that because like it's either going to trigger generate an image, which is not what I want.
2:33:07
Maybe it will be able to find it for me, but then I still have to click out to find the actual image.
2:33:11
It's certainly not going to like give me a grid of images like Google images >> yet. >> Yet. Maybe soon.
2:33:17
Anyway, Andy two cents money says, "A marvel uh a marvelous testament to the tax benefits of the CC corporation that tens of thousands of Americans can raise standing armies off a whim and yet there have been so few uprisings in our history.
2:33:33
People just stop overthrowing the government once it lets them get rich. This is a good take.
2:33:38
Uh startups, there's there's so much intel in this one image yet so often ignored.
2:33:41
Number of firms and employment. Um interesting.
2:33:46
So there are uh yeah there are like >> 20 20,000 firms.
2:33:50
This is 2019 data from the Census Bureau. >> Yeah.
2:33:56
>> Uh >> 20,000 over 500 employees.
2:34:01
>> So if you have 500 people that are willing to go to war for you potentially, overthrow the government.
2:34:06
>> Sean Frank is in the chat pumping pumping the TBPN Substack. TV tbp. substack. com.
2:34:13
>> Go subscribe if you haven't already.
2:34:13
We send out >> Great to see you, Sean.
2:34:17
>> We send out our show notes every morning.
2:34:18
>> Sean, when are you coming on the show in person? >> Yeah, come on.
2:34:21
>> You're not that far away. Let's make it happen.
2:34:24
>> Um >> uh and before you give away the the the the Lambo, come drive it onto the set.
2:34:30
That should be how how >> Yeah, definitely bring the Dorado for sure. Um we'll hang out. We'll go out. We'll hang out.
2:34:35
Uh and then when we reflect, we'll say, "Yo, last night was a team's meeting."
2:34:39
This is Atlas creatine cycle. >> Banger. >> Banger. It's a great one.
2:34:44
Uh, people are enjoying the merch.
2:34:45
Erica says, "Posting a p a pick in the TBPN jacket is the 2025 version of a thirst trap."
2:34:50
I don't know if that bodess well for thirst traps or us.
2:34:55
I think it's more just people showing their support.
2:34:57
But, uh, this is a funny post, sir. >> It's a good jacket.
2:35:01
Uh, this is the same story with AI.
2:35:03
The killer apps will be the ones that collapse latency, friction, and context switching into something instant.
2:35:08
So, this is a quote that signal posts.
2:35:11
I heard a story years ago about Steve Jobs after the release of the original iPad.
2:35:16
Jobs had been on medical leave in 2009 when he returned to Apple.
2:35:21
He was focused almost entirely on the iPad.
2:35:23
In 2010, after the iPad was introduced, he had a meeting scheduled with engineers on the MacBook team.
2:35:28
That meeting was big picture.
2:35:31
What's the future of the MacBook? That sort of thing.
2:35:34
These engineers had prepared a ton of material to present to Jobs.
2:35:36
Jobs comes into the meeting carrying an iPad.
2:35:40
He goes to a then shipping MacBook on a table and wakes it up. It takes a few seconds.
2:35:45
He says something like, "Look at how long this takes." He puts it to sleep. He wakes it up.
2:35:49
It takes a few moments each time.
2:35:51
Then he puts the iPad on the table and hits the power button. On, off, on, off. Instantly.
2:35:56
Job said something like, "I want you to make this."
2:36:00
He pointed to the MacBook like this. He pointed to the iPad.
2:36:01
And then he walked out of the room.
2:36:04
And that was >> people I I almost forgot that it used used to go and open up your laptop computer and there was a real delay before you could actually get value.
2:36:13
And now aside from you know having to authenticate it's pretty much instant. >> Yeah. Yeah.
2:36:18
Uh I mean I guess some of the delay of like I I think it might be good that there's like fingerprint reader and type a password every once in a while.
2:36:27
It feels like I if I'm doing something more secure I could do that on the on the on the MacBook instead of the iPad.
2:36:32
Never been into the iPad, but I I I think I'm gonna try for the next one.
2:36:37
I've seen too many memes about the hierarchy of like the CEO does all his work on the on the phone, doesn't need the MacBook, and then the more monitors you have, the lower you are on the totem pole.
2:36:47
Shout out to our incredible production team that has >> high ontem of our hearts.
2:36:54
>> They have they have the most monitors. Oh, yeah.
2:36:56
This is this is the money shot.
2:36:57
Thank you for putting this up.
2:36:57
And and I and I lost a serious bet with uh producer Ben over this because I told him that mounting would not hold more than 24 hours. It's been up for a week.
2:37:09
And so I I owe him uh I owe I owe him a diet of Coke, I suppose.
2:37:12
Uh anyway, uh thank you, Techno Chief.
2:37:14
Techno Chief says the Substack is elite. We appreciate that.
2:37:18
We're working hard on it every single day to make it better, making it 1% better. I'm learning from it.
2:37:24
We have a team that's focused on it as well.
2:37:25
I write uh I write the run of show.
2:37:27
I uh collaborate with everyone to put together uh most of the stories, but we have a whole bunch more folks that are contributing stories there.
2:37:33
And uh we're going to experiment with Substack.
2:37:36
We're we we're talking to Chris Best, figuring out what else we can do there. We're streaming there.
2:37:38
Um but we're going to be asking people questions, surfacing those on Substack, collaborating, crosspollination, all >> very long. Good stuff.
2:37:46
And the reason we're doing it is because math nerds are not having a good time.
2:37:52
According to Peter Teal, this is uh a a little over a year ago.
2:37:56
Peter Teal says AI will be worse for math nerds than for writers.
2:38:01
So, time to fire up a Substack.
2:38:01
Uh >> it is interesting that that the that many of the things that LLMs and and models do well today, generating designs, generating marketing copy. Yep.
2:38:15
I still feel like the number of truly elite creatives like that there's an overwhelming demand for truly elite creative talent. >> Yeah.
2:38:24
The power laws get >> whether you're a copywriter, whether you're a designer, whether you're even, you know, people that are doing Photoshop, they haven't been >> Yeah.
2:38:33
It's It really is a bullcase for finding like an odd an odd life's work.
2:38:38
I I I I was talking to you about this earlier, but like uh you know when when kids come to you and tell you what they want to be when they grow up, like the previous generation was very attuned to, oh, you want to be a firefighter?
2:38:50
Like, haha, yeah, like that'll be fun for a couple years, but like you're probably going to be a lawyer, bro.
2:38:55
Like that's really the only way you're going to like pay the bills in the future is like doctor, lawyer, merchant chief.
2:39:00
I really like merchant chief. It's a good one.
2:39:02
Uh but uh but increasingly I find that uh if you can be top 1% in any of like the kid careers, the the pilot, the police officer, the the firefighter, >> the international businessman, >> the businessman.
2:39:20
Uh really anything anything that uh is in a Richard Scary book, anything that can be done by a cartoon pig.
2:39:26
Uh if you can be top 1% of the like pottery, if you can be the top 1% potter, like you're probably going to have a job forever.
2:39:32
And AI will never displace you.
2:39:34
There will never be the the the reinforcement learning training set.
2:39:38
They won't be able to pull it out of you.
2:39:40
But Tyler, you got the you got the counter take.
2:39:41
You think you think you're gonna wait.
2:39:43
Me and Tyler went to the Hollywood Bowl the other day and there's there's someone at the Hollywood Bowl that manually switches the lights in the street during hightra uh times.
2:39:53
So because there's so much crazy traffic when the Hollywood Bowl is funneling people in and out, they have a they have a physical bowl that someone goes and sits up in and then they control the lights manually.
2:40:02
And Tyler's just like, I'm going to automate that.
2:40:03
I could I could vibe code better.
2:40:06
>> It feels like a a ripe target because there there's there's there's a human can do sort of a vibes based analysis like, ah, we got a lot of people coming this way, cars.
2:40:14
Let's let them let's let them through.
2:40:17
Uh but but a machine I think could potentially take on 10 times the amount of data and be like I have a camera down here and there's actually yes there's people coming from this direction but there's actually a bigger backlog here and if we let them through.
2:40:33
So I'm on team Tyler on this. >> Yeah.
2:40:35
Give me the full bullcase for uh AI stoplights. Is it good business?
2:40:40
>> Well I mean I don't know how normal stop lights work.
2:40:42
Like I always have thought like do they work with cameras? I don't know. Are they just on timers?
2:40:45
I don't know anything about this space.
2:40:48
>> They're all on timers at best.
2:40:51
>> Most of them are just completely on timers.
2:40:54
>> So, it seems like >> So, bit of lore.
2:40:55
Blake Scho, founder of Boom Supersonic, before he was going to start Boom, the other idea that he was going to work on was AI stoplights, >> put a camera on it, see if there's a car there.
2:41:07
If there's no car here and there's a car here, >> too easy. >> I don't know.
2:41:11
It might It might be equally hard. It might be hard.
2:41:13
We don't have we there are multiple like there are multiple uh supersonic plane companies now.
2:41:21
I' I have yet to talk to an AI stoplight company. It seems obvious.
2:41:25
It seems like something flock safety could expand into.
2:41:26
But uh I even looked up like what are all the stoplight companies?
2:41:31
Could you buy one, roll it up?
2:41:32
Like how do these things get worked out?
2:41:34
But you're just doing things in the physical world with the government and budgets are tight and there's just not that much value and people are fine with the current status quo.
2:41:42
you'd get to work with the Department of Motor Vehicles. >> Could be fun. >> Could be fun. DMV enthusiasts. >> Very agile. Very agile.
2:41:48
And so, yeah, I think even I think I think it's a it's a it's a week-long hack project to develop something that's, you know, an order of magnitude better than the current status quo and then uh probably a decade to actually get it into the hands of the American populace.
2:42:04
>> Well, here's a post from Aaron Bali uh from Carbon Health.
2:42:08
He says uh it's not in the stack, but he says with AI, everybody will make their own software just like how we all 3D print our own furniture.
2:42:18
[Laughter] >> You 3D printed that, right, John?
2:42:22
Because 3D printing is is so advanced that, >> you know, I mean, the future's here.
2:42:28
It's just not evenly distributed yet.
2:42:31
People underestimate what can happen in a decade.
2:42:33
They overestimate what can happen in a year.
2:42:35
Um, I saw an incredible 3D print.
2:42:38
It was somebody who 3D printed chain mail for like uh uh what do they call it?
2:42:43
Like a live action role playing LAR.
2:42:46
Uh what what what was that festival that people go to where they dress up like Game of Thrones?
2:42:51
You know what I'm talking about? I don't know.
2:42:54
Why am I blanking on this? Uh Renaissance fair. >> Renaissance fair. >> Renaissance fair. Yeah. >> As a kid. >> Yeah.
2:43:00
Do you remember stumbling upon like you're just going to the park stumble upon a restaurant talking about laring or >> Yeah.
2:43:06
The people that are the people Yeah.
2:43:08
I I just remember I I was uh grew up in in Berkeley and uh the local park every there was there was >> there was a big LAR community really in in just be like going >> I'd be like go using the swings and I'd look over and a bunch of adults would come over in medieval costume and and start uh battling. It was it was cool. Yeah, >> it was cool.
2:43:34
I I respect uh I respect their uh imaginations.
2:43:40
>> Yeah, respect the larers.
2:43:41
>> Luke Metro says, "How to get rich as a founding engineer in 2025.
2:43:42
One, find a startup from a famous founder or a VC incubation. Two, get hired.
2:43:47
Watch it get marked up to billions.
2:43:49
Three, lots of growth is priced in, so silently sell a few million in sketchy forward contracts to cash out. Four, don't tell anyone.
2:43:55
I wonder, >> isn't this illegal?
2:44:00
This sounds like security fraud.
2:44:03
>> I don't know that it's uh That's a good question.
2:44:05
We we should get a true expert like a venture capitalist on but um >> uh because as you know they >> Silicon Valley is an iterative game.
2:44:13
>> No, but I think this was this is more like >> uh you know against uh uh against the startup's wishes.
2:44:20
They've probably told employees do not do this >> but I don't necessarily think that it is a doesn't seem like a uh seems like it'd be done in a way that is not uh broadly illegal.
2:44:33
But you could probably get your equity clawed back.
2:44:36
>> Can you imagine if So Periq was doing this.
2:44:41
>> You'd be like, >> everybody was trying to clock So's run rate.
2:44:45
It's like is it is it a million, two million?
2:44:48
>> He was just selling >> 20 million in in forward contracts.
2:44:52
Yeah, no one came forward on the timeline.
2:44:54
>> We have another post from Aaron Bali two in one show. Nice work.
2:44:57
He says, uh, showing the chat with AI characters.
2:44:59
So people are uh so I guess people can make their own >> um they can make their own characters in AI studio.
2:45:09
>> So you kind of go into AI studio you want to chat with Russian chat with it >> or chat with stepmom >> stepmom >> and this this spawned a lot of jokes like imagine leaving X.
2:45:20
>> No but here here's the thing.
2:45:20
So >> Elon is competing with Mark Zuckerberg on AI companionship. >> Yes.
2:45:29
It's just that one of them has tens of billions of free cash flow. >> Yes.
2:45:34
>> And the other one has has to raise a combination of debt and equity. >> Yep.
2:45:40
And also just a more aligned business model where time Zuck feeds right into it.
2:45:47
>> Does Zuck have 10 times the user B?
2:45:47
Like >> probably three billion daily active users I think something like that. It's in the billions.
2:45:54
Um, and also it's just like all of the product features from chatting and DMing and and groups WhatsApp or Meta or >> you're really just taking like social networking and then bolting on like a companion could be like I don't think I would ever ever have like a one-on-one conversation with an AI companion, but I could imagine adding an AI to a group chat.
2:46:16
What if I make the the chats and like >> roughly a day >> roughly eight hours a day I'm sleeping usually stay up later than me because you don't have as long of a commute. >> Yep.
2:46:28
>> And so what if I made a meta AI of me and I trained it on >> all the data that I have and then you could chat with that at night because sometimes >> sometimes you'll text me it's like at like 10 I'm sleeping and you could get a faster response. >> True. True.
2:46:42
>> So >> uh I don't know. I haven't tried these. I don't know.
2:46:45
So, I don't know how good it would be. It really is.
2:46:46
It really, but the numbers are staggering.
2:46:49
I mean, 5 million messages to Russian girl.
2:46:53
Maybe that's just the scale of of of FB, but the blue app is is undefeated.
2:46:58
>> He was quoting He was quoting, "Imagine you're one of the smartest AI engineers in the world.
2:47:01
You just joined a company that has access to a GPU cluster worth tens of billions of dollars.
2:47:05
Your work can one day help solve all diseases, create unlimited abundance for humanity.
2:47:10
But first, you have to build >> spicy mode."
2:47:12
So >> it doesn't and and and who it was will um >> Will Brown was posting you know that this uh hypothetical engineer that that quits their job at one lab >> at XAI >> at Xi Valentine >> spicy mode with Valentine >> and then goes over to meta and has to work on step >> or stepmom.
2:47:34
>> Yeah I know I don't >> of course these are these are human uh these are created by >> other uh >> yeah you have to think about them like like meme pages.
2:47:41
I mean I I do I do wonder if people will will like the fact that this tool will be out there.
2:47:48
Someone will probably create something interesting like I always give that example of like character AI.
2:47:51
Um I went on character AI and was trying to debate Vladimir Putin about the benefits of capitalism.
2:47:59
>> It folded pretty quickly, right? >> Yeah. Exactly.
2:48:00
No, it was actually Stalin. Yeah.
2:48:02
>> I was debating with Stalin.
2:48:02
I was debating communism and capitalism with Stalin and he kept admitting like he was clearly RLHF broadly on Western ideals and capitalism and then like fine-tuned in the prompt level to be Stalin.
2:48:15
And so I was able to very quickly get him to be like, "Yeah, I did some bad stuff. I wouldn't recommend it.
2:48:24
It's not really that good." I'm like, "Victory. Flawless victory."
2:48:29
Uh which of course >> another win for capitalism.
2:48:31
But I but I think my takeaway is like I do think if you put this in the hands of like of like billions of creators, you might get someone that comes up with an interesting chatbot that is funny in the same way that you get a Harry Potter Balenciaga every once in a while or you get like a like a 4chan green text, right?
2:48:49
And so something like that, I don't know what it would look like, but some sort of thing that you can I mean we've seen like chat with cow like already that's like funnier in the sense that like I might actually click on that.
2:48:59
I don't know exactly what that would be like, but there's probably >> has his own horos AI now. >> Yeah.
2:49:06
Did we ever solve how much he made? I saw one slide.
2:49:09
>> Tyler, figure out how much made if you're not familiar, Alexi.
2:49:13
>> He did a big book launch, course launch, >> business influencer.
2:49:17
>> He said that he spent $30 million on consulting. >> What?
2:49:21
>> And then use that to train the model. >> Wow, that's a lot.
2:49:24
So now instead >> the numbers are all over the place like the brain >> which I thought was an interesting selling point because obviously like a lot of the models are trained on Harvard, HBS like case studies and and just general like think about how much like uh work that like the big four consulting firms had produced that eventually ended up >> Yeah.
2:49:46
Deis would like a word actually.
2:49:48
Dennis would like a word.
2:49:48
Alex Dennis says, "Yeah, yeah, we actually trained Gemini on on five billion dollars of of training data." >> Yeah.
2:49:56
I think I think if anything, uh, so going back to this debate of like XAI going heavy into AI companions and it's like every time I open X now, it's like an Elon post promoting Yep. >> Valentine or Annie. Yep.
2:50:11
>> And >> I don't enjoy it.
2:50:12
I don't want to, you know, a bunch of stuff that Elon >> married, so you're not looking for a real girlfriend or an AI girlfriend or Yeah. boyfriend. >> Yeah.
2:50:22
It's just like I don't want like like normally if somebody was posting that stuff, I would just mute them or or unfollow them, but it's Elon.
2:50:28
He posts a lot of stuff that I am interested in.
2:50:30
Uh but S for Satoshi here says Elon is in real danger of falling off culturally.
2:50:37
He must remember it was cultural re relevance due to SpaceX and Tesla meme stocks that catapulted him.
2:50:41
X is cool but XAI is not.
2:50:44
He can fix it but he is currently not on the path to do so.
2:50:47
This may tick him off but it's true.
2:50:50
>> And and S for Satoshi is like a an Elon Bull, right? Famously. >> Yeah. >> Yeah.
2:50:55
So, >> and I think this is how I think this is how like people are feeling broadly. >> Yeah. Yeah. Yeah.
2:51:00
I I I think the it goes back to those like filter bubbles can be a feature, not a bug.
2:51:04
And so like there is a world there there are groups of users out on X that are interested in chatting with Valentine and Annie, but you should probably never show that post to anyone in Teapot. >> Full stop.
2:51:20
Like you should just say like we're going to show teapot >> papers on what we're doing at XAI on the reinforcement learning side and surface that instead it's like because Elon's account's so big and it gets so much promotion in the feed like everyone saw that and everyone's like why am I seeing this?
2:51:39
this? this is not relevant to me and and it's just like kind of a it's a it's a bad it's just a bad like signal like to land that post in this particular community because >> yeah especially when people see the outputs from Grock imagine and it's just
2:51:54
not on the level of many other the outputs are not on the level of many of the other >> totally like totally >> do I need to see two marble statues you know kissing >> and that's not and that's and what's interesting is that like that like you
2:52:09
the not the same the same cannot be said for the cluster that XAI built that was super impressive like when that >> and the results >> yeah I mean the the the the biggest like XAI bullc case was just laid out by Casey Hammer on Doresh he was saying
2:52:25
like Elon understands he gave this amazing analogy this amazing historical story about um I forget who it was someone who was building boats in World War II and he ran out of like he was like I'm trying to buy uh materials to build the boats. I need steel. There's I I need steel.
2:52:39
There's I bought all the steel and so now I'm going to buy a mine and start mining for steel.
2:52:47
Um anyway, do you want to give a shout out?
2:52:51
>> I got to give a shout out to Isaac Mini Katana. Mini Katana.
2:52:55
>> Isaac is a legendary entrepreneur.
2:52:55
I would say that he is the the Willy Wonka of our time. >> Really? >> Yeah. >> Okay.
2:53:03
>> He's an he's an absolute legend. >> Please subscribe.
2:53:05
Mini Katana, thank you for being in the chat.
2:53:07
Uh they will shout you out. They love business. We do love business. >> We do love business.
2:53:15
He has a company called Campai Foods which is >> probably you know he is just absolutely crushing.
2:53:21
It's uh his facility is or was in LA I've visited.
2:53:24
Uh but he's just >> he's an an algorithm animal.
2:53:27
I will just say that both his both his companies he's able to get billions of views totally organically and uh he's an absolute uh >> Campai freeze-dried candy says they subbed. >> There we go. >> Fantastic. Well, shout out.
2:53:42
>> Um >> let's go to sucks.
2:53:42
Texting and driving is such a funny crime.
2:53:44
You already have six hours of screen time, but you need another 30 minutes.
2:53:47
Even if it's life or death, you know, you can literally die, but it's worth it to you to scroll for 30 more perfect minutes in high-speed traffic.
2:53:56
>> It's a good reminder to everybody. Get a comma. >> Get a Whimo. >> Get a Whimo. >> Get a Whimo. Get a driver.
2:54:02
>> You actually can't text in uh in a comma AI.
2:54:04
It has driver monitoring.
2:54:04
It watches you like a hawk. And it's very very good.
2:54:08
And it actually it's it's such a good experience.
2:54:10
I can't shelf for it enough because you it it it watches you.
2:54:14
And so if you're on your phone, it will disengage.
2:54:16
But if you're not on your phone, it's the chillst driving experience ever because you're just like sitting there watching the road, just having a conversation, listening to music.
2:54:23
like it puts you in a much it's much less active than actually having to drive and steer >> praying for exits.
2:54:31
>> Things are about to go absolutely nuts in America because there's an image here of the first nicotine energy drink.
2:54:34
I couldn't find this online.
2:54:38
I don't know if this is just like a like a mockup or something, but this seems to have come out of nowhere.
2:54:43
>> I think this might have been something that literally was just a mockup. Yep.
2:54:48
>> Because you're you're one of the foremost experts in the world on nicotine and >> nicotine from what you've told me because I asked you this.
2:54:57
Why has nobody put nicotine in an energy drink?
2:55:00
Caffeine and nicotine together in one drink would be amazing.
2:55:02
And there's like regulatory reasons for it and then there's like actually um biological reasons that nicotine doesn't >> process in the same way if it's in your stomach. >> Yeah.
2:55:14
it it needs an alkaline environment.
2:55:16
Your your stomach is too acidic and so it doesn't absorb.
2:55:19
So it like it can kind of give you an upset stomach, but mostly it just like doesn't absorb.
2:55:25
So people will always say like, "Oh, what if I swallowed a nicotine pouch?"
2:55:28
Like is am I over is it over for me?
2:55:31
It's like uh no, it'll probably just pass through you and it's just not like you didn't get any value out of it. Uh it's fine.
2:55:38
Um, you probably shouldn't do it a lot, but uh, also like it's it people would be eating them if they could.
2:55:44
Uh, obviously people would would would try all sorts of things.
2:55:46
Um, but, uh, there are theories on how to actually get it to work.
2:55:51
You could you can't make your stomach not acidic because then you'll throw up.
2:55:56
Um, but you can potentially wrap the nicotine in some sort of molecule that diffuses through the stomach lining and then enters the bloodstream and gives you the nicotine buzz.
2:56:06
So, it is like theoretically possible, but it's like this huge science challenge that no one's worked on.
2:56:11
There was a company that raised money um in the 80s, I believe, or the 90s to to do this.
2:56:15
They worked on it for years, couldn't get it across the finish line.
2:56:19
I I we we looked into it a ton, couldn't figure it out.
2:56:23
Um never really got any any way.
2:56:23
It's possible that this company solved it.
2:56:26
Um I think what's more interesting is this scenario.
2:56:31
Oh, and also like you can't mix active ingredients, so the FDA would probably not like appreciate this.
2:56:35
And even if you did solve it, the FDA would have to approve it and that would take like a decade potentially.
2:56:40
Um, but we're in this weird regime where uh maybe we're in the beg for forgiveness era uh with a lot of government stuff and uh and there's a lot of there's a lot of nicotine companies, most from overseas, that do not care what the FDA has to say about their legality.
2:56:58
Uh, and there's a lot of value to just going viral.
2:57:00
And if this is something that uh, goes viral, the real hack here, I think, would be to brand this as nicotina energy.
2:57:08
Don't put any nicotine in it because that would trigger FDA and then just sell an energy drink and people will be like, oh wow, like it has nicotine in it.
2:57:18
And really just put aine in it and just don't have nicotine on the ingredients. No one checks.
2:57:23
It goes viral and then eventually you just end up with just a energy drink brand.
2:57:26
But that's a little bit disingenuous.
2:57:29
Probably uh not the best thing to do. Not the best. >> Yeah.
2:57:33
I think somebody made a mockup.
2:57:35
They hadn't thought through how you'd actually take this to market. Yeah.
2:57:37
And now everyone thinks it's on the way, but when you Google search it, it doesn't exist. >> Yeah. >> Last post of the day.
2:57:43
Luke Metro says, "My pet theory is that most people uh so so any goes back to >> Gold Rock calls it five loco.
2:57:49
That's actually really good."
2:57:51
Uh, so, uh, Cold Healing said, "Bro, you make $500,000 at OpenAI.
2:57:57
You can go to the art fair and buy a little $10,000 painting to hang up in your SF apartment's living room."
2:58:03
Luke Metro says, "My pet theory is that most people here with big money had to spend their most their formative years with most of it being illquid.
2:58:09
So, you're used to modest spending habits for a while, unlike bankers who can blow their whole bonus on a weekend in Miami starting at age 22."
2:58:16
I think that's a good take.
2:58:18
good take. the uh the the nature of having most of your net worth be illquid is uh as Felpix says besting is good for the soul it seems or ill liquidity is good for the soul >> but it's it's easy to get started with
2:58:33
art if you have if you have a bunch of money you you're good >> no Dylan Dylan Dylan Abscatoto close friend of the show he he routinely buys art from artists >> for like a thousand dollars and then three years later they're worth >> tens of thousands dollars. I mean, he's
2:58:48
I mean, he's got a great great taste and a great eye uh for for emerging artists.
2:58:52
But uh but yeah, you don't need to be a a beanire to indulge in in fine art.
2:59:01
>> Yeah, we need a we need a fine art.
2:59:03
>> But on that note, we got to get on with Taipei.
2:59:05
Uh thank you for tuning in today.
2:59:08
Uh give us uh five stars on Apple Podcast and Spotify.
2:59:11
And of course, subscribe to our Substack, which we are working uh very hard on. TBPN. substack. substack.
2:59:18
com and we made a bunch of upgrades to the studio today. Yep.
2:59:22
Uh some of which are working well, some of which aren't.
2:59:24
Um so, uh >> we will continue to iterate.
2:59:27
Thank you for your patience and your support. See you tomorrow.