The War on ChatGPT | Ep 007 Lemonade Stand 🍋
We launched a Patreon! - Patreon.com/LemonadeStand for bonus episodes, discord access, and many more ways to interact with the show. First Ep. out now!
This week... Atrioc goes panning for gold, DougDoug is back on the AI train, and Aiden goes house hunting in Vienna
Recorded on: April 16th. 2025
Clips Channel: https://www.youtube.com/channel/UCurXaZAZPKtl8EgH1ymuZgg
Follow us
TikTok - https://www.tiktok.com/@thelemonadecast
Instagram - https://www.instagram.com/thelemonadecast/
Twitter - https://x.com/LemonadeCast
The C-suite
Aiden - https://x.com/aidencalvin
Atrioc - https://x.com/Atrioc
DougDoug - https://x.com/DougDougFood
Edited by Aedish - https://x.com/aedishedits
New takes on Business, Tech, and Politics. Squeezed fresh every Thursday.
#lemonadestand #dougdoug #atrioc #aiden
Listen and follow along
Transcript
I had my first lemonade stand fan interaction this morning.
The first, the first ever person to come up to me and say, I love you on lemonade stand.
I love what you guys are doing.
It happened this morning at the coffee shop.
Is that your mother?
No, it wasn't my mother.
It was a guy who was granny-old male who looks like around college age or post-college graduate.
Yeah, you've got it,
you did it right.
I think we kind of know most of the audience.
Yeah, we got the audience.
That's cool.
it was kind of a nice, it was a nice little thing.
He was like, You're saving.
He's like, Your show's hitting a nice niche.
I think you're saving a lot of mid-20-year-olds from their brain rot.
Saving America, if you will.
We've been doing it.
Can you stop him and say we're doing more than that?
We're so much more than the saviors of we've done so much more.
We've accomplished one big thing every episode.
And I reminded him of that before he walked away from me.
That's good.
Literally, we talk about tariffs.
He repeals the tariffs.
Okay.
It's just, it has to be us.
We, we, for our fans, like this one I met this morning, are launching a Patreon.
Oh, finally happening.
Uh,
child that you met, we need to extract money from his wallet and put it into our pockets.
I kind of shook it
like dollars from the laugh.
Did he pay you like a donation on stream?
Like, did he give you three dollars to read out the donation?
I said, if you're a real fan, you could buy my coffee.
And
thanks for the call, man.
It doesn't pay the bill.
Okay.
This coffee's come on, get involved, dude.
Get involved, get involved.
You said I saved you?
You said I saved your mind?
Did at the end, after he talked to you, did you leave a cool little character at the end of his sentence?
Oh, yeah, like a little apostrophe.
And then he tried to make with his hands, he tried to do two carats.
Oh, that's adorable, man.
He definitely watched the full apostrophe.
I heard he said he left a negative comment, and then you shivved him and put him in the yeah.
I did hear that.
There's a lot of ambulances.
You put him in the alley, dude.
Because I just can't stand for that.
I'm not.
Finally, a commenter can shake down in person.
I said, dude, it's like the slime slime thing where he asks people who ask for a photo.
Let me see your Twitter.
That is crazy.
And he sees what sort of profile they have.
Oh, my God.
We should ask for a person's stock portfolio.
Yeah, they're like, let's see.
Let's see.
Open your Robin Hood.
Yeah.
Politics quiz.
Yeah, it's like a caption.
We show them four crypto coins.
We have to pick one that we approve of.
You can take a picture with me if you make one options play of my choosing.
That's right.
We're launching a Patreon.
The biggest benefit is that every week we're recording an extra hour that is going to be basically a bonus episode that you get to watch.
Just to be clear, this is not like we're recording two and a half hours and then we chop off the last hour and hoard it for money.
It's a separate episode, separate recording, all this stuff.
It's going to be a little more like probably conversational, looser, and responding to what people are saying, other things we didn't pick up on.
We already have this one recorded.
If you go sign up for Patreon right now, there is another hour episode that you will get and a new one coming out every single week.
Plus, on Patreon, you get
access
i was just saying give it a yeah you get discord access you get to talk to us in a thing you can suggest topic ideas for the next episode you can give your responses that aiden will crash out on and then also in the discord uh in the second tier we're gonna do a book club every month so every month
yeah get that thing on you this book was suggested by aiden our first book i'm like a little into it so far it's almost entirely about orgies yeah and i am not kidding believe it or not this book about canes the The Economist is mostly about swinging
so far.
Insane.
I walked in this room.
You guys are talking about how it's like a suck and fuck off between.
It's a sucking fuckathon.
It's crazy.
It looks like the most boring book about
in the first, I think, paragraph.
It's like.
John Maynard Keynes fell in love with a woman and it was shocking because she was a woman.
And I was like, what?
What?
And then it goes into like 100 pages of his gay orgies.
And I would love to get to the
economy at some point in the book.
It's a slow burn.
It is a slow burn.
It is a slow burn.
Anyway,
that is the first book for the first month that we're going to be reading.
We're going to choose a new book every month.
And at the end, when we've all read or listened to the book, we'll record another episode where we just
talk about the book and get people's feedback on the book.
So we're also going to be doing that with movies once a month.
We're going to pick like famous, like
Wolf of Walls, like business finance movies
and just do watch-alongs and also talk about them afterwards.
And I think the last thing, I kind of skipped over this in the first tier.
You mentioned that people could like submit topics and articles, basically
engage more directly with the show and with us.
Those are the only two tiers at the moment.
We haven't added a third one or anything yet.
We're going to add Lemonade Stand Plus.
That's going to chop 10 minutes off of this episode every week.
And slowly but surely chop more and more of it as you go higher in tiers until the entire thing is a payment.
It's like you only get a few words and then they say you're going to pay more.
And then you really just have to ask
pops up.
But that's live right now at patreon.com/slash lemonade stand.
Wait, do we have the big goal, the long-term goal that we're seeing for?
Is that are we launching?
Have we agreed on that?
We said, whatever, we just committed to it.
Okay.
At 10,000 paid Patreon members, we are going to plan a trip to China together and film, I think, a couple episodes in China from our trip and our experience there.
I think the idea is we're going to go with a translator, someone to show us around
and try to, and maybe, maybe just freeball it.
Maybe let me take the willie now.
Dude, I heard your Chinese.
And
I'm really, really looking forward to that because we have all, I think we all are really, really interested in China individually.
And we've seen a bunch of people recently, like from blog posts or videos we've watched, you know, anything like, you know, all the way to like iShow Speeds Clips to
who was the article that we all read.
The blog post.
Article My Life.
The blog post.
Oh, Dwarfesh Mattel.
Yeah.
Yeah, yeah.
About his experience fitting there, visiting there.
I think we just talk about China.
We're talking about structure cities that we just aren't aware of in the West, really, at all.
That are these massive cities.
Like, yeah, there's so much interesting stuff to do there.
So, it wouldn't just be like hitting tourist traps.
It would be like we're really trying to see some of the stuff that people talk about and really compare this.
Yeah, I truly think China and the U.S.'s relationship is like the defining story of the next decade.
Yes.
And, like, to not understand China on a deeper level is like a dereliction of duty.
And what's great about this is if you hate us, you can sign up for the Patreon and get us to china faster and i think there's a real chance that we get kidnapped because we have said not nice things about you for sure right like this so i mentioned this to my family
like going through chinese immigration but only showing the episodes where we've gassed them up it's like but then unlisting the episodes where there's any criticism did i get like four
main channel videos about pro-taiwan on dick down
no literally i was like talking to the family and like is that smart to do i was like oh yeah we're not that nice necessarily.
I think it's chilly.
I think we're gonna prove that it's chill and it's gonna be
can you say in Mandarin, uh, what do you really think about Xi Jinping?
Oh,
uh,
Shema, uh, nope, nope, can't get there.
That's gonna go really well.
I think you want us, I only can say what in that sentence.
This is us in downtown Shenzhen in six months, completely lost, can't get home, no taxi.
Woman should
Take them, leave me, at least.
But I think this was something we had talked about before this episode is, you know, we've talked about a few things recently on the show that are pretty heavy, dramatic,
not very, not, not always the most like positive topics, right?
And I think we want to keep a positive outlook on this show.
Doug mentioned this and like, oh, what big breakthroughs that we can be excited about?
Could we find those for like an episode in the future?
And I kind of went searching for some articles with that in in mind the first article i found uh titled as fentanyl deaths slow meth comes for main
oh let's go
just amazing meth's chance to be on time you know what i'm saying a breakthrough it's felt like fentanyl's been dominating the headlines for so long it's got to be the worst like if you're you know if you're like one of the best soccer players in the world then a messy comes in right that's how meth probably felt that is right fentanyl comes in and just they're the new star right they're like some 17 year old wonderkin they can score a goal with like an ounce of
fentanyl.
Like, God damn.
And now it's out.
He's back number one.
No, meth is back on top.
That's big for meth.
That aside, the actual first, so I found that article and I was like, you know what?
That one seems a little maybe not positive enough.
And I went and I found a different article about the Trump gold card that he has been talking about.
Neither of these are that positive.
The best two things you can follow in the world is that meth deaths are climbing and Trump is selling a gold card to pay our national debt.
Both of big, I can't imagine being anything but excited about both of these two things.
I think this topic has always been,
he's been circulating the idea of this gold card for, I don't know, a few months at least.
And apparently, in about two weeks,
Elon Musk's team or whoever is at Doge, including the guy that pseudonym is Big Balls,
is
helping launch the buying process for the gold card.
For those who don't know, the idea is that much like a green card, you have a residency card, a permanent residency card that you can just flat out buy.
You can find your way to become an American permanent resident.
You can pay $5 million for the gold card,
which is apparently going to pay off our national debt.
That's just like the price of two dozen eggs.
It's just a few eggs.
Dude, how long until rappers are wearing the gold card on like a chain?
Like the way
when Charizards are worth a lot, didn't Logan Paul have like like a Charizard?
Yeah, that's good.
If it's $5 million, the gold card.
It's every PSA 10 Charizard.
That's rare.
That's scary.
When I was in high school, that's what I did with my green card.
I wore it around my neck.
You've been straight painted at gold.
Have you?
No.
Oh, my God.
That's such a bad idea.
No, you put out a bad image if you believed that too easily.
Was that the good, the positive story?
Was you wearing the gold card?
Yeah, that I could have been wearing a gold chain in high school instead of my so they're launching it.
It's happening.
Yeah, it's supposed to.
I think much uh much like elon musk's other time estimates for his projects we'll see but uh i think as funny as this is the idea that trump with everything about him comes into power and makes a gold card that you can buy for five million dollars to become a part of the united states at least your permanent residency uh it seems kind of silly i think this is i think in the grand scheme of things this is pretty unimportant but i actually don't think it's that bad and i was wondering if you guys could point out any big flaws flaws in this because a lot of countries offer permanent residencies or citizenships by investment.
A lot of places do this.
For instance, in Ireland, you can just pay a million dollars to become a citizen.
It's like a million flat.
In Switzerland, you can pay $200.
And in a way, American citizenship is five times more valuable than Irish citizenship.
Not even citizenship, just the people.
Just human people are better.
Five zero people better.
It's 5x better than Irish.
The market has valued it.
Okay.
Yeah, exactly.
Exactly.
And the free market
And we are, and at 20,000 patrons, we're going to go to Ireland and tell people in the streets that,
but the examples are, there's so many examples of this.
I think
somewhat famously, me and Atriok have joked about how Gabe Newell during COVID apparently bought his permanent residency in New Zealand.
He just purchased it.
And other countries like Switzerland, too.
Just to be clear, we talked about if there was an apocalypse, we would sneak into Gabe Newell's bunker in New Zealand, kill him, and then live in it.
Yeah, that was the context.
That was the context.
We might have to bleed that out.
But I was wondering what you guys think.
Because I think by no means is this going to pay off the national debt.
But if anything,
what does this do?
It gets rich people who want American citizenship to be a part of our tax base.
Like, that feels not that bad.
Because if you're a foreign investor in the U.S., you can already buy assets and like influence things within American society and politics.
I feel like if you really want to, like, if you're just a foreign investor in a company or in housing, you can pretty much do that already.
So it's like, okay, well, welcome to being taxed by the United States to some degree.
How many zeros are in a trillion?
Oh, yeah.
I mean,
isn't it 12?
One, two, three, one, two, three, one, two, three, one, two, three divided by five million, which is what, one, two, three, one, two, three, right?
Yeah.
So
if we sell 200,000.
That's so awesome.
If we sell 200,000 gold cards, we make a trillion dollars.
And so I actually think that's not unrealistic.
200,000 is a lot.
That's not going to happen.
200,000 people with $5 million to buy it.
To be taxed and chased by the United States.
Not like in a month, not right away.
But if you think over the course of like a decade, absolutely.
If you think of many major countries, countries, so it's like, there's a number of reasons why it's smart.
It's like an exit strategy.
Let's say you're in China, right?
And you're concerned, although I don't know if Chinese citizens would be able to buy this, but in any sort of any country that has a more, let's say, authoritarian or government that might rapidly crack down on certain industries or things like that, you'd be incentivized to have a backup strategy, basically of being able to be in the U.S.
For most people who have $5 million, they're probably thinking a lot about how to make money.
One of the best places to do that in the world by far is America.
So having a permanent residency here would be great.
And then I think part of the thinking is then it incentivizes people who do that and buy this to come to America, set up a little office.
They're renting an apartment.
They're doing whatever else.
So they're also then dumping more money into the country.
I don't see a lot of downside with it, to be honest.
Like I'm sure it will be abused by some people, but like.
If we as America, we're not 35 trillion, whatever it is in debt, then like, yeah, this feels less necessary.
But if we're literally just trying to figure out how to get out of potentially country ruining debt and you can just have like 200,000 people give you a ton of money at basically no cost and no problems to you, like that seems pretty good.
I mean, I'll villain share a little bit in that I agree that there's no real downsides.
I don't think it's going to get abused that much.
I think if you made the number 500 million or something for the cost, everyone could see, oh, okay, there's no downside, but no one's going to buy it.
That's essentially what's true for the 5 million in that there's not enough people people in the world who have 100 million plus assets, which is what you really need to afford this kind of 5 million thing.
There's not 200,000 of them in the world.
There's not, I mean, there's the number of billionaires in the world is relatively minuscule.
It's just not enough people to like make a serious dent.
So I'm frustrated with it in that it seems like another Doge style thing where it's almost like a look over here at how much we're doing to these are all another thing.
Like, we're going to get the debt, but the debt is going up.
The debt is going up as we speak.
The debt continues to go up.
Continue to not talk about anything other than the real problems.
So while this is not, not, I think a few people will buy this and it will give us $5 million for each one, you know, for the tax base and they'll be citizens.
But the idea there's so many people that really want to pay American taxes and give us $5 million.
It's not, it's not a ton of people.
It's not a huge market for this.
Yeah, I think if you look at the places that offer this, it's often a place that has some sort of tax or financial benefit to you,
like a place like Switzerland or like a place like Ireland.
They come with some sort of
system, amazing passport that lets you take advantage.
Whereas that's what I, when I've looked at moving and
going somewhere else, one of the things that when you're an American citizen is that the tax taxation follows you.
You're not taxed by your residency.
You're taxed by your citizenship.
So when you're,
I guess because you're not a citizen,
this not wouldn't necessarily apply, but I can't imagine a lot of people getting residency and not then trying to become citizens.
And then not trying to become citizens after.
If you're spending $5 million on it, I assume they have a plan to become a citizen.
I would need to see a further breakdown of
the benefits.
The people who are rich enough to spend $5 million to get what is essentially a green card
maybe don't want to deal with the taxes of America.
So they're incentivized to have this as a sort of safe haven for them and for their business.
And so almost like treating it as a foreign office or something and then not actually going for citizenship.
So they dodge taxes.
I don't know.
Cause this thing does not give you citizenship.
You would then have to go through the whole process of whatever it is, like years.
Yeah.
It takes a while.
Yeah.
What is like 10 years?
It's not super quick.
I think it took me from when I got my green card until I became a citizen.
I think it took me seven years.
Would you say that was worth $5 million?
If you, okay, if you right then were given the option, we'll pay you $5 million, but you have to leave and never come back.
And you're looking like, who is this?
Nine-year-old Aiden in the eyes.
And you're like they're like five
game stop you're like looking at all the games you could buy they're like five million dollars we have to leave the country is it is it worth it it's i really
you give up but i could buy a lot of copies of like super monkey ball instead but it's i mean the counterpoint is it's an investment in your future you've easily made that back since being a citizen
yeah Yeah, I've come out on top, if anything.
I think in all seriousness, the 5 million part is tough.
I actually think a better use for this, because
the framing of this being used to pay or make an impact on the debt because of the reasons we've talked about, I feel like is the misleading part.
But the idea of this existing is something I'm not opposed to.
I actually think offering this at a lower price point kind of makes sense.
I don't think our immigration system should be pay to win in the first place, but I do think you think it should be a loot box.
A loot box.
You think you put in a small amount of money?
Wait, wait, this is just the immigration lottery.
This actually is the immigration lottery lottery.
We do currently have this.
This is how it works.
Oh, but it's got shiny.
It's not like I really like it.
Explaining the immigration lottery through the context of CSK since it's finally, it's clicking.
So you get an immigration key in an immigration box.
Yeah.
And then sometimes it's good, sometimes it's bad.
I think that having a simpler version of this or a cheaper version of this might have some type of utility, but it feels bad that the solution to bad immigration policy is, okay, okay, to get through all the paperwork of the bureaucracy, just pay enough money.
I don't like that part of it.
It seems very transparently, it's just trying to make short-term revenue.
I can't see it being like this, because if you're super rich, you can find a way to live in America, right?
Like, I think this is a way to make that process just straightforward and just like, look, you can just buy your way in if you're like mega rich, just as short-term cash inflow, right?
Like, it's like you said, 200,000 is probably extremely over-optimistic, but I bet that's for 1 trillion.
That's for a trillion.
Yeah.
I bet they can get tens of thousands thousands of people.
That was for trillion.
Oh my God.
I didn't even do them anymore.
That's for 1 trillion.
It's not like it's going to pay off the debt, but it's going to take a lot of things to pay off the debt.
A lot of revenue.
If you get 500 billion from this, that's incredible.
I'm not opposed to it.
I just think it's over talked about.
For the place it takes up in the news and the idea of this being significant, I feel like that's the misleading part.
It's like, oh,
it's fine that this exists, but it's not meaningful.
It's like not a meaningful thing.
You know why that is?
It's because of the name Trump card.
That's so well marketed.
It's truly unbelievable.
They're calling it the Trump card.
And that is, you have to admit it, incredible, right?
You're going to talk about it.
Because then he literally says our Trump card is going to get us back into green this year.
You rolling up to border security and flipping it over like you're playing Yu-Gi-Oh!
They speak Japanese at border like the
Blade Blade Runner feature.
Wait, wait, wait.
Where we're all Japanese.
Speaking of gold, there's a little gold rush.
Is this another positive story?
This is a positive.
Well, debatable.
Okay.
Debatable.
But I'm going to give you the positive side.
Happy episode.
Yeah.
So far, I'm just
a bit of a gold rush going on underseas.
I don't know if our producer can pull this up, but
we have a little stock market game we've been playing that honestly we've all been doing pretty badly in because the stock market's been tanking since January.
But one person, Aiden, over here, made a possibly brilliant call where you invested in a really small company that was like almost unheard of called the Metals Company that you saw in like a CNBC video or something.
Yeah.
Anyway,
two days ago,
Trump announces that as a way to get back from China, because they're blocking the export of rare earth metals.
China is blocking the world, actually.
They own the entire supply, like 98% supply chain for like
lithium and a lot of these rare earth metals.
Okay.
So they're blocking it off as part of this trade war tit-for-tat.
So one of the things Trump approved is like, we're going to do more deep sea mining to get our
rare metals.
The point is, the one company that was mentioned, the one company that's benefited from this is TMC, which had like a
hundred percent spike in the stock.
Aiden is now leading.
I mean, that's the number one stock in our whole little game of the 50 we drafted.
And
I don't know.
I want to know what Warren Buffett thinks about.
The only one I did actual research on.
Wait,
did they like find new stuff and that made their stock go up?
Or is it just specifically Trump is like, we are now going to invest in deep sea mining?
The spike is because of that.
Okay.
Interesting.
And they don't even know that they're going to go with them yet.
So it's still speculative.
But the idea is that they're the leading.
Politicians just pick companies now.
That's so sick.
So it's interesting because
actually, just quick context.
The leading stock by far before this was an AI company, I believe Tempest AI
that has no real impact in the AI market at all, but Nancy Pelosi invested in it.
And so that single-handedly bumped it way above anything else.
And now this one is crushing it, not because of what they're doing, because Donald Trump just kind of shouted them out.
Anyway, yeah.
Sort of.
Yes.
And this is funny because we have talked about deep sea mining on this show.
in one of our test episodes.
So we filmed two test episodes before we started working on this show.
The loose idea is that there's companies out there that have discovered that you can pick up these nodules of rock on the ocean floor and you can mine the like rare earth process them into the rare earth metals that we need for a bunch of different stuff uh things like electric car batteries for instance uh and the metals company was the most prominently featured company exploring this type of mining over the past few years and it was the company mentioned in the cnbc piece that i watched on youtube yeah i think uh literally I checked the date almost two years ago.
And I just kept this in my head because I thought this was interesting.
This dynamic of is this a new source for these things that we would be mining with a terrible environmental impact otherwise?
Or what are the environmental consequences of doing this?
And then I started getting into the regulation and this new frontier of mining
in general, like how it actually works.
It's all administrated by this like small office in the Bahamas administered by the UN.
So it's never supposed to be in charge of all of this.
There's a bunch of really interesting stuff here
about the industry in general, but I just bought it because of that.
I was like, I need to make one actual pick in this competition.
And they had been doing pretty well.
I mean, their stock when I bought it was at like $1.50 and now it's at $280.
Yeah, you can see here.
I mean, you bought it at this very, you know,
the very bottom dip of 2025, right above the zero.
And now it's massively spiked.
And that's just pretty crazy.
That's That's what you call massively spiked?
Well, it's more than double.
The chart's pretty skewed because it started at 10 way over here in a weird era.
But, you know, if you skewed it down to just this part, it is a massive spike.
It's like a 200% increase or something.
Damn.
I think I want to let kind of the explanation of the industry and the company itself speak for itself.
We've talked about, we want to edit those test episodes and put those up on the Patreon as well, which we talked about at the beginning of the episode.
But for now, I think it's cool.
The main reason I bought it was: oh, upcoming energy industry, basically, something that ties into the energy of the future.
Oh, let's just buy that stock and see how it works.
We'll see.
The Canadian company, so maybe Trump won't give them federal funding or whatever.
But the idea is that maybe they're going to be a Canadian company run by an Australian.
But they're the leader in this weird space.
And there's this whole thing where they like
pay
small Pacific island nations to mine it on their behalf.
It's a crazy system of
companies
abusing the privilege of the small Pacific nations to try and start mining.
Another good story in this
story.
Unfortunately, it is a dark explanation behind it.
But I think if you have time, looking into it is really cool.
But
I kind of want to get into your topic for this week, which was primarily about open source AI models, which is something we've briefly talked about on the show.
AI bros, rise up.
It's our time again.
It's been two whole episodes.
It has been two episodes.
We have to tie him down.
I was literally shaking.
We shane him in the basement between episodes.
Like every 10 minutes I go without talking about AI.
I start like having a shaking.
I start
veins popping.
So you've been good.
I mean, you've been a real noble global citizen to avoid every episode.
All right, now I'm going to do a huge line of AI discussion.
Okay.
We're going to talk about open source.
So I think this is a very interesting
element that is happening in the broader AI space.
Even if you are tired of AI and think it's all obnoxious, I think the open source versus closed source thing is super fascinating.
And it sort of come, it questions the entire industry and what's going on and the way a lot of the companies like ChatGPT are basically running.
And it, I think, in many ways threatens their ability to keep making a product, which you've talked about in the past.
This really kicked off with DeepSeek releasing a few months ago now, I want to say.
So we're going to touch on that.
But broadly, there's a lot of questions being asked and various companies are responding to it in interesting ways.
So first primer, open source versus closed source.
Open source code means that when you write code for something, the code is available publicly.
So anybody can find it and they can download it on the internet
and run it and edit it.
So there's a lot of famous, you know, OBS, which we use for
streaming.
That's used.
PyTorch.
I believe is the one, not rather than TensorFlow.
A lot of the underpinnings of
what runs AI models themselves are open source.
Many of the most important software and programs like Linux are open source.
So basically the community broadly can all download these things, use them, modify them.
And it's really cool.
That means they're free too, right?
Or how does that work?
Yes.
You know, asterisks.
So there are some companies that do like a hybrid model, which we'll get into, where they make an open source product, but then they have a kind of paid version on top of that.
Premium tier.
Yeah.
Like what we're doing right now with Patreon.
Yeah, we're kind of like the open source.
Yeah, so for example, WordPress.
WordPress is like a blogging software that is powers like most of the internet, or at least some giant, giant percentage of it.
WordPress as a software is open source and free, but then WordPress, the Matt Mullenweg, who like kind of runs the open source element of it, then has his own company on top of that where you pay for it.
So that's how they justify running a company.
They have their own extra features they added.
So you basically go to them and you say, hey, we are going to use your version.
So a lot of companies that manage open source software run it like this, where they are both the shepherd of this big, giant community project that anybody can use, but there's this weird balance where some portion of their work is not accessible to the public and they're like, you have to pay us for it.
And it actually, it works really well in a lot of different cases.
Closed source is the opposite of that.
It's what you'd think.
A company makes a product and they have it, right?
Fortnite is not open source.
Nobody has the code to Fortnite.
You can't go take Fortnite.
You can't do anything with Fortnite.
You can only play Fortnite on Epic's servers because it is closed source.
They are the only ones who have access to the code.
Just another corporate win.
I hate that.
Exactly.
So, you know, you can kind of think about it in these different categories.
And then there's this new one that's kind of coming out called open weight.
So in the context of AI, you have kind of three pieces going on.
When you make an AI model, there's the training phase where you actually make the thing.
Okay.
So this is where you give it a whole bunch of data.
You write the code that does the machine learning specifics that actually has this thing learn.
I give it a million chess games.
You get a million chess games.
And it's not only the million chess games you give it, it's also the mechanics of how you've designed it to process those chess games.
So the actual, let's say, ingredients that go into this thing being made.
And then the thing is made at the end.
And now it's an AI model that is done, it's trained and can be used to answer questions like a chat GBT type thing.
So what some companies have started doing is rather than open sourcing everything, which would mean that you give all the tools to making the AI out publicly, they call, they do it open weight, where they don't give you the code that they used and they don't give you the training data that they used, but they give you the model at the end to run yourself.
So the equivalent of this would be like Fortnite gives you a copy of Fortnite that you can run on your computer, but you can't modify it, right?
You can just run it exactly like it is.
Isn't that what I already have when I play Fortnite?
No, because you can't really play the game unless you're connected to Epic's servers.
I guess it's not a perfect knowledge.
Offline, I could run the game.
Here's a better example.
I don't need to be connected to the internet to play Sabrina Carpenter and Fortnite.
I could just do it without
Rick Sanchez, and we can like.
This is not a great analogy, and it's getting increasingly confusing.
And the storm is like society.
No, closing in on AI.
And the gold card is when you get a gold AI.
That's the battle bus.
Yeah.
Trump is the storm.
Are all the characters in Fortnite open source?
Sabrina is open source.
So LeBron Carpenter.
She's open weight.
Have you seen her?
Talk about Sabrina Carpenter, actually.
Sabrina's weight, I guess.
Yeah, Sabrina Carpenter.
Is that public information what her weight is?
Well, you can't change it, but it is open.
You can't talk about her.
Yeah, that is.
Okay, so a better example, Facebook.
Okay, Facebook, you cannot run Facebook locally on your computer.
You have to go use Facebook's servers that are running it.
You're accessing their servers they control entirely.
I could download, in this example, I could download Facebook.
I could run my own copy on my own website.
I could run Facebook too on a website.
Yes, if it was open source.
No, if it was open open weights.
So open weight is specifically for AI.
So me trying to convert this into an analogy for something that doesn't fit doesn't make a lot of sense and it's probably confusing.
You're telling me you can't synthesize all the big ideas you talk about on the show into Fortnite.
Imagine Sabrina Carpenter.
Okay.
Okay.
Right now you have to go to a theater to watch Sabrina Carpenter.
All right.
But there's a copy of Sabrina Carpenter that you can get and hang out with at home.
That's awesome.
But you can't change her personality at all.
That's
perfect the way that she's.
an open weight model.
And a lot of people are happy with that.
Open source is that you can then go and change Sabrina's code and how she acts.
I see.
It's like a Blade Runner where you're doing the sliders on.
Okay, I see.
Which
is actually based on Sabrina Carpenter.
That's, I did not know that.
So this is an
interesting mix of how this works.
And so
there's different basically strategical elements of what making a certain AI in a certain category does.
So most of the ones that people think about and talk about, ChatGPT is closed source.
You can't run that on your own computer.
You have to go to their servers.
So that for them is what we call in the business world, a remote.
Yeah.
All right.
So it means that competitors can't go like just copy everything ChatGPT is doing because they don't have the code.
They don't know how to make their own, right?
They can guess.
They can try to emulate it in the same way you could try to make your own game that's like Fortnite, but you don't have Fortnite's code.
You can't go recreate it somewhere.
So the dominant AI models over the past couple of years of this whole AI craze have been closed source.
And that's like most software that companies make, where it's their own product.
They're, you know, they're keeping it to themselves and they basically offer it, you know, in their own terms and they control it entirely.
And that's, that's super valuable for ChatGPT.
It means like they can charge 20 bucks to use it every month because they can't, you know.
Right.
Only if their product is better.
Only significantly.
If their product is significantly better.
And this is true of Claude or Gemini or Grok or all the main ones.
Up until DeepSeek a few months ago, there were some other notable ones that were open weight/slash open source.
Llama, which is Mark Zuckerberg's one, that is a major open weight slash well, open weight model.
And then Mistral is another one, which is out of France.
And these were you know the French, guess what the French version of uh Chat GPT is called?
I just said it Mistral.
No, that's that's their product name.
Guess what it's called?
Oh, actually, it's called LeChat
France.
That's like going
when you go to Paris.
Scholars don't know, Doug.
Scholars have tried to figure it out, but it's
undecipherable.
When you go to Paris and every restaurant is just La Something,
can we, come on, guys.
Can we get some culture in France, please?
Can we
pick it up?
Pick it up.
Put a McDonald's.
Put a Walmart in there.
But
I imagine you might be.
So Llama as an example of open weight, but DeepSeek is open source.
No, it's also open way.
So we'll get to that.
So this was how the world worked.
Most of the big AI companies that you've heard about were all closed source, which means those companies are investing.
billions of dollars into making these crazy ass models that they put out and they're like, this is the best model out there.
You should pay us 20 bucks a month to use it.
Right.
And it's like any kind of product that you would think of, where you then have to pick between what people are selling, right?
Llama and Mistral were kind of fighting against that.
Llama would come out and they'd be like, ooh, maybe I'll use Llama instead.
But what really, really changed things is DeepSeek coming out a few months ago.
So this model was from China.
Very unexpected because nobody thought China was doing any sort of like really high-end AI stuff because America actually has sanctions to prevent them from buying the best GPUs, to prevent them from being able to develop the best AIs.
So not only were we trying to stop them from selling it, also it just felt like America tech companies and a little bit in Europe are the dominant ones that are leading this.
It's unbelievably complex stuff that requires really smart people and tons of resources to make critically.
And we're totally fine.
China's not.
And then just boom, this thing drops and it's open weight.
And what that critically means is that anybody could download it and run it on their computer for free.
Anybody can do that.
So if you are a company that before was paying ChatGPT and you're like, look, I'm going to pay a couple million a month to use your service for my program.
We're doing AI customer service or we're doing AI analysis of our business or whatever, agents, you know, all the things that are growing.
Suddenly this new competitor comes onto the market and it's not just that they're like hey we're cheaper it's like you can download this and use it for free in whatever way you want we don't even have the ability to stop it from running once you've downloaded it it's yours forever we can't stop it and this completely blew up the sort of ecosystem that was happening i mean there's two things i want to say that one was about the the sanctions because i think there's an interesting story today that's like a little Yeah, diversion.
So we'll get into like how it was made as well.
But yeah, you can go.
Okay, so you know, as he was saying, NVIDIA high-end GPUs are banned from sale in China because they don't want them to be catching up in the AI race.
So, NVIDIA, in response, and I worked there when they were doing this, made a nerfed version of their H100 that is like slightly worse.
You know, it's worse on every aspect so they can sell it in China.
And then just two days ago, that version, or maybe even yesterday,
very recent by the time you see this.
uh got banned that version after they made it before they got to sell it but after they spent billions making it got banned and so nvidia stock yesterday huge drop huge drop because that was very expensive to make and now they can't sell it again and so there's becoming this problem where even the nerfed ones are getting banned but because they're getting banned china is now finding ways of doing it either with smuggled h1 gpus h100 gpus which is happening but also without them so like the um
chinese smit connector companies are getting their own they're catching not they're not there yet but they're catching up in a way so it's becoming this weird war over that but that was that was high point i just wanted to bring it up yeah So it gets into the, so it's relevant.
It's about the development of this thing.
So imagine that you are these AI companies the past couple of years.
You're riding high.
You have ChatGPT.
It's this crazy ass product.
You have whatever it is, like 100 million people paying for it or using it every, I forget the exact numbers, using it every month.
It's this smash success.
And the entire stock market has massively inflated over the past few years, mostly from tech companies investing in AI.
That's like most of the stock gains the past couple of years.
And we're all talking about how great the stock market is, largely based on all of these companies like Google, Microsoft, facebook everybody is going massive massive massive all in on ai yep um and then imagine you're one of those companies you're you know you're sundar pachai at google and you've told your entire company to focus on ai and then this model drops from china and again it's not that a chinese company came out and said we're going to come out and offer it for five dollars a month they said it's free you can download it and use it on any pc on any computer we have we don't even have control over it no it is open weight and what that means is we don't know how it was made.
So, again, they're giving us the finished product that you can run on your own computer.
ChatGPT, you can't run on your own computer.
There's no way of like having it yourself.
If you are a company that wants to start using AI, you are basically dependent on ChatGPT to give you access, to set the pricing, to do everything.
But now, if you want to do it, you can set up your own server farm, run DeepSeek on it, and just use that entirely.
And the only thing you pay is energy costs.
Essentially, it's free.
And even if you use their API, which is basically you are using DeepSeek's servers, that's 1 20th of the price of ChatGPT.
So either it's free if you run it yourself, or it's ridiculously less expensive, 95% less expensive.
I want to jump in.
Yeah.
Because that's where the real, because I've used both now.
I tried, you know, I would use it to summarize articles or to quiz me or whatever.
And I tried DeepSeek for a while because on the benchmarks, it did as good or in some cases better than ChatGPT.
I was like, and I found...
qualitatively pretty similar but i found myself drifting back to chat gpt i like the interface there were some things that were cleaner The voice thing, they have some new features, you know, they're updating.
But I'm kind of price insensitive in that it's a business expense for me.
I can spend a little bit more on it.
Here's where I think, that's the area that I think this really shakes up the business model is that like for an average consumer, they probably, they trust Chat BG GPT.
It's the new Google for them.
They might keep using it.
But if you're a company that's a middleman company built on top of the AI layer.
So an example I use is like, there's like character AI where you can like chat with Genghis Khan or you can chat with, but basically what that's doing is it's a fancy fancy portal, but it just calls back to ChatGPT.
But they have to pay ChatGPT a fuck ton of money.
So if I spend, you know, there's a lot of people out there that are like trying to get AI boyfriends.
There's girls that are like chatting with their AI boyfriends.
They spend $150 a month on character AI to talk to their AI boyfriend.
They have to then give $100 of that to ChatGPT.
So they have to make a small profit.
If they build the exact same thing, but on the deep seek layer, they get to keep all of the profit.
And only a little bit goes to deep seek if they use their servers or not.
So that's the big change.
And that's where all of of the money that these AI companies were hoping to make might evaporate.
Because like as a direct surge engine replacement, yeah, you might go to ChatGPT, but for all the things that are going to get built, like for example, you know, we might got a world where the next Grand Theft Auto 7 has AI voiced characters that's built on top of something else.
If it's an open source model, then Rockstar keeps all the money.
But if it's ChatGPT,
and that's...
That's what they were hoping.
That's where all this inflated stock market came from was these future profit expectations that are not possibly showing.
That's my, that's it.
Yeah, and it's worth reiterating.
So the AI companies that have been making all this stuff that's been going on, where every week you're like, there's this new model from whatever company, all of them are operating at an absolutely massive loss.
They are basically spending billions of dollars every year in order to train these AIs because it takes, you know, they're taking tens or hundreds of thousands of GPUs that are running these massive training cycles with massive quantities of data.
They're paying like the smartest people in the world.
So this costs unbelievable amounts of money.
And the whole industry is predicated on the idea and the hope that, yes, we're blowing tens of billions right now, but it'll pay off in the future, right?
Like this is the future.
Everybody's going to be using AI for everything.
But if then China drops this model, which is free,
again, all of your customers, like, why would they use your product, right?
If it's, if even if it's
better, it has to be so much better.
It has to be like a little better.
And so, that is, I think, the really interesting question with open source.
And this has really been going on the entire time AI has been happening.
It was a big deal last year when Mark Zuckerberg came out and randomly dropped like Meta 3 is out.
And sorry, excuse me, Llama 3 is out, which was his open weight model.
And that alone was whole was like, oh shit, Mark Zuckerberg's strategy is basically to just set fire to the other companies by releasing their models for free.
Because the instant you do that, everybody else has a harder time competing in the market.
DeepSeek was that, but like on crack.
So it was just this insane level of competency and insane level of being cheap.
And so that is the question.
So now I want to open up.
And there's a lot of interesting angles of this going forward, which is like, one,
how if you are OpenAI, which is a company that, again, is spending tens of billions of dollars a month or not a month, a year, but you're starting to make more money,
but open source models are going to keep releasing.
How can you ever stay competitive and recoup all this investment, right?
Microsoft is giving them tens of billions of dollars.
How are they ever going to make that money back?
And if they can't, and if it doesn't seem like they're going to, are they going to stop investing in all of this, right?
And so there's things that you mentioned.
Like right now, OpenAI is still doing well because they're trying to basically make features and models that are so much better or so unique compared to what DeepSeek does.
So they have like this deep research feature, incredible.
Yeah.
Their advanced voice mode is incredible.
I've been using it like every day nonstop to like learn things, practice a learn a new language, whatever else.
Operator models where it's going to be controlling your computer or remembering your conversations is a new one where Sam Altman has talked about how he wants the AI to really like get to understand who you are and then it tailors its answers to you.
And the idea, like, while that sounds maybe creepy, that totally makes sense, right?
You would want it to kind of understand your way of learning.
It would want to know like your history.
It would want to make fun of you in the way that Slime does, right?
Like, it would, it would really, like, learn.
Is that what I want?
Yeah.
AI designed to treat you like Slime.
Sam wants for me.
Yeah.
And he said this and he said, Aiden has gotten too complacent, and Chad GBT is now going to.
So they've been training it on thousands of Slime tweets.
I didn't know Sam was talking about me so much.
Yeah.
It's wild when the Yard fans come over to the comments.
You're not bullying Aiden enough.
I know, it's
like this was a happy episode.
We should have the mean to Aiden episode.
That's every episode of the yard.
That's every episode.
Would it be funny if one week could come in?
He and I aren't here.
It's just slime and nick, and we're like, hey, we couldn't make it today.
We found replacements.
That would be the, I would just launch into berating.
That would be the worst episode.
I would be so mad the whole time.
Okay, I, I, my first question, because the introductory, let me, let me do one interesting thing.
So, so, OpenAI and all of the AI companies are kind of in a, what the fuck do we do kind of mode the past few months?
Some people are like, well, DeepSeek is an indication that it's going to be even more popular.
Other people are saying we need to do more to stop China from doing this.
Other people, so like they're taking the sort of national security concern.
Other people are saying, oh, yeah, we need to like make crazier products that can't just be copied by a deep seek type of model.
And then very interestingly, OpenAI, who makes ChatGPT, and I mentioned this, I think two weeks ago, they announced they're making an open source model or open weight model.
So
seemingly the pressure from DeepSeek has made them start to push towards an open source model, which then in turn makes it be like, how are you guys going to make any money?
So, and Sam Altman, the CEO is seeming, the way he phrased it is like, we're going to release this for the community.
And that's all great.
But it really does make you wonder, like, what is your business plan?
Like, I continue to be very optimistic about AI in general, but I do wonder, like, dude, how do you justify this?
So now I'm curious.
I, one thing I wanted to clarify was the terminology with open weight and open source again.
Yes.
Because in this example,
this situation, you said open weight is something that is specifically applicable to AI.
Yes.
Is there a step further than this where an AI company or project could be open source?
Like, what is the additional layer of benefit if, say, some random company came out and said, here's our model.
This is open source, not open weight.
Is there any further benefit there?
Yes.
So, and a few people have done that.
For example, Mistral has a fully open source model, but it's much less powerful.
They're kind of doing similar to what I mentioned earlier, where they have models that are like open source.
And then they're like, but if you pay us a lot, you know, for a subscription, then we'll give you the premium one that only we have.
Functionally, what would be.
I thought your example of customer service at scale was really good.
Like the difference between what I could do with DeepSeek right now versus paying ChatGPT a bunch to run the customer service for maybe the giant clothing business.
So, here let me give you an example.
Let's say you are the customer service company, okay?
And you're like, we want to integrate AI in some way.
This is going to be really helpful for our customers or whatever.
And you were using ChatGPT and now DeepSeek comes out and you want to use DeepSeek.
And you start using DeepSeek, and it's much cheaper and it seems to be about as good.
But you realize what you really want to do is make changes to it.
It still doesn't quite talk to customers the way you want.
You want to sort of fine-tune it in a way that really can
be tied into your your business specifically rather than just how it was trained in China.
You can't do that because all you're given is basically a model that you can run on your computer.
Imagine just like you have ChatGBT, but it's on your computer locally.
You can run it anywhere for free.
But you right now, when you go to chatgbt.com cannot change chat GBT.
You can't go into the code and say, I want to train you on different data.
I want you to focus on different things.
I want you to emphasize your personality in a certain way.
You can't actually affect how the AI functions in any way.
So open weight means that they've made the thing and you can use it, but you cannot change it at all.
Open source means you also have the code and could actually rebuild it however you want.
And that would mean that companies could take, if DeepSeek was fully open source, they could tape DeepSeek and then customize it to whatever their needs are.
Or
a new company could come in and they're like, we want to be what DeepSeek was.
And two years from now, they take the starting point of DeepSeek's model and iterate on that and then make it some crazy new thing that they release.
But they can't do that because there's no code available.
We don't know how DeepSeek was made.
Yeah, that makes sense to me.
I have maybe a more basic question about how AI works to begin with.
Yes.
When my impression, because you're talking about downloading something like DeepSeek, whatever the open weight model is on my computer.
And then I turn the internet off on my laptop and then I can still use that model.
Like I can still interact with it.
My
understanding of AI up until this conversation right now was I interact with an interface and then I type to the AI and then it's communicating with like the data center that trained it in order to output the response that it gives to me.
But how can I do that if the device is not connected to the internet?
How can it, I guess, what my question is.
With something like DeepSeek, and if I've downloaded it and I'm not accessing their servers anymore, how does it answer what I need to be answered without the pool of data that it was trained on yeah i mean basically imagine it is a video game so you know if if you know cloud gaming which is the big new thing so with cloud gaming the whole idea is that the game is being run on a external server somebody's data server so you're playing the game or you know if you go to facebook right most of the stuff most of the calculations that are happening are on facebook servers somewhere so when you access facebook.com you're sending a request to their servers they do all the work it gets sent back to you with whatever the result is um with uh with an open weight model, you can download it and the entire thing is running on your computer.
So instead of it being running on the data servers and you're talking to it, it's now running on your computer.
The way that actually works is with GPUs.
So the reason.
This might be a silly example, but say I'm offline on my computer and I type a really complex math question into the offline version of the open weight model I'm interacting with.
Will my computer
take a while and it'll heat up because it's putting a lot of work into answering that question?
Do you see like in the same way when I'm like running a video game or running maybe like an editing software?
Yeah.
So the way the way the model is actually run, it's not.
So there's the training phase of AI and then what's called the inference phase.
Really, what that means is the time where they're making it and training the thing.
And then once it's done and made, and then it is just, to put it simply, a program that you can download and run on a computer.
So once a company like OpenAI finishes making their new ChatGPT model, they have the program just copied a whole bunch of times on their servers.
But if you theoretically were to to take a copy of that, put it on your own computer and run it, it's now running on your computer.
But when you say run on a computer, what that means is it is using a GPU to do a shit ton of math.
And so the reason GPUs are so valuable and NVIDIA is so valuable is because fundamentally it's just a lot of linear algebra.
It's machine learning.
It's a giant machine learning system.
And it is just many, many, many millions of
simple linear algebra equations that are being calculated.
GPUs are basically designed for lots of simple math equations.
That's why they're so good.
So if you have a crappy GPU in your computer, which you probably do in that Mac.
Yeah, relative to what?
It either wouldn't run well or it'd run like it wouldn't run at all or it'd run extremely.
It might take a long time for the question to get answered.
Because it just literally has to do millions of tiny math equations.
But if you then run it on like my CPU at home, which has like a 4080 or whatever, like it'll be able to run that.
It'll be able to do it fairly well.
It's still putting a lot of load on your computer.
It's needing the CPU and particularly GPU to just do a lot of stuff, but it'll do it and it'll calculate it.
And that model, that thing that was made, is basically a giant math equation that will calculate the answer.
It just needs a GPU to run on.
But if you give it the GPU, then yeah, you're good.
When I used to work at NVIDIA, I used to have to do all that by hand.
All the math equations.
All the linear algebra.
It was a pain in the ass.
I think technology has really advanced quite a bit in the past years.
Well, it was honest work.
It was honest work.
Don't get me wrong.
They're already automating those jobs away they did
no i i i mean there's a lot of there's a lot of ways we can take this conversation this is like a pretty because there's a there's a really interesting angle about this which is a country's security and how open source gets into that it's that like there's a lot more going on to china releasing this model than just like they made a cool thing and wanted to release it i want to give the business thousand yard view just real quick just yeah just of what i've because i've been talking about this on a broader scale for a while which is the idea that especially flashback three months ago because the tariffs have changed a lot of things but three months ago
the entire world not just America but the entire world has been slowly but surely piling into one trade which is that American stocks have an exceptionalism they're just doing better so everybody not just our pension funds but Bwedish pension everyone has been putting their money in American stocks of the American stocks pretty much all that's been the SB 500.
Of that, pretty much all that has been in seven tech stocks at the top.
It's been the magnificent seven of Apple and you got NVIDIA.
You got Apple.
You got the Metals Company.
You got.
Pretty soon.
And everybody's favorite company, Tesla.
Tesla should not be a next seven million.
You know what?
Elon Musk actually has some really fucking good ideas.
There we go.
I feel it taking me over, Aiden.
So that is the context we found ourselves in three months ago, which is just like the entire world is like, yeah, this is smart.
And we're all agreeing.
We're all that's, it's one of the weird things with everybody like freaking out about the stock market ups and downs over the past few weeks, which again, totally makes sense.
But like, there's a real argument that everything over the past two or three years of all these stock market games is just a giant AI inflationary thing.
I don't know whether it's a bubble, but it is certainly probably overvalued.
And the idea of like, America's doing so great
because these six companies are investing in AI a shitload is like and haven't made money yet.
Like they, right, nobody's making money.
They've thrown billions and billions and billions at this, and they're all promising future returns.
And, you know, I think we all agree that AI is a real technology.
It's getting better every year.
It's going to happen, but they have to make money off it to even justify these insane $3 trillion, $3 trillion, $3 trillion.
They're insane valuations that have to be justified with profits at some point.
I have a question for both of you.
Lay it on me.
Have any of these companies, because we just talked about the clearest example of how this helps these companies make money.
Basically, selling the AI as a service that is the underlying software for something like a customer support service, right?
And they can charge a big licensing fee for that to be used.
And that's revenue in finally from the work that they've done.
Have any of these companies laid out their long-term plans for like how this is supposed to make revenue?
Is it just that?
Like, was that the plan all along?
Or has there been more of an approach of move fast, make this stuff, because it will be monetizable down the road in some way.
And the plan with most of these companies is to just figure it out.
Like, do you see what I mean?
There's probably multiple answers to that question.
I'm interested in your.
I have so, I mean, obviously AI is so vast at this point.
I, I have not heard of that a single time.
Basically, everybody's just talking about how they're going to integrate AI and everything and AI is going to improve everything.
Yeah.
I haven't heard a specific concrete plan, but even if somebody were to give that, it's probably bullshit because we have no idea what's going to happen.
A lot of the AI, let's say research labs like ChatGPT, OpenAI and Anthropic, their stated goal is to make AGI, which is artificial general intelligence.
So their stated goal, and who knows if this is real, is like they want to make an intelligence that far surpasses a human's intelligence and can like solve all the world's problems.
In theory, that's what they're going for.
They're not, they don't worry about profit.
Presumably they are actually worried about profit, but that's what they're saying.
And then you have companies like Google, which have more been in the line of like, we're just going to put AI into everything.
I haven't heard it vocalized.
What I've seen is them basically saying like across all of our services, AI is going to make it better for the customer.
So with Google, they're like, okay, our search, when you search something on Google, as people have probably seen, it pops up that AI summary, right?
And they're like, we are going to make sure that all of our features have all of these things.
So for them, it's making their existing product and infrastructure and company and business and products all just like even better and retain what they have and then get even more people.
I have not heard anybody be like, here's what it's going to look like in five or 10 years because nobody knows.
Yeah, you're right.
Nobody knows.
It's all changing so insanely rapidly.
I think the way you describe it is kind of like, you know, Microsoft has Copilot that they integrated into PowerPoint and all these things.
But the idea is that
people have to buy more than they were buying before.
Or like Apple's putting AI stuff into the iPhone.
But the idea is that that makes your phone enough better that you're selling more iPhones.
There's more increase in profit or revenue.
And that has not materialized yet.
In that people are throwing it into Google search, they're throwing it into Copilot, they're throwing it into phones.
But so far, the iPhone sales are actually down a little bit this year it's not nothing is going up
in a way they want they want Clippy to be so smart that I pay more for Microsoft Word they want you to have a romantic relationship with Clippy where you actually that's probably what they want the most now that I think about it in terms of getting money out of me they absolutely want me to fall
much like canes I turned 38 years old and I fell in love with Clippy yeah you become robose because like that's why Microsoft like imagine character AI and all the women who have like AI boyfriends now yeah but it is it is then the AI boyfriend is telling them, wow, Copilot could really help you schedule your appointment today, you know?
And so it's like once you fall, once they fall in love, then Microsoft sells them their products.
Yeah.
Have you guys seen her?
Yeah.
Yeah.
Great movie.
More prescient than ever.
I mean, it feels very applicable.
Yeah.
One thing that's interesting is, is, and it's probably not the best episode to discuss it, but like,
people are using these services more and more.
Like, I think one of the that I think there
has been some profit growth is direct subscriptions to ChatGPT.
Like, they're not profitable because they spend so much money.
But, can I give you guys an example?
I hadn't used AI
at all up until about a month ago.
And then I downloaded DeepSeek and I downloaded ChatGPT and I interacted with, you know, just the free version of ChatGPT and
mostly using it to try language stuff.
I was, you know, getting Swedish sentence help and things like that.
And then I tried to do research.
And then I heard that, you know, the paid model of ChatGPT is a lot better for that type of stuff.
So I upgraded like a week ago and started asking questions to help research
topics I wanted to do for this show.
Because the nice thing about ChatGPT was it would give me like a sentence summary of like some point I had asked it.
And then it would link all the supplemental material so that I could go click into it and then look further.
Basically, aggregating things in a way that was much better than just Googling stuff
and upgrading to the $20 a month plan, not an ad.
I noticed it.
It was, it was, it, it was interesting to me to see the difference in quality and the way I, the way it produced responses.
Right.
Uh, so, just from a day-to-day helping perspective, very cool for me to use.
Like, it has a, it has a material use in my life.
One angle I wanted to come at this from was
we're talking about the benefits and the consequences of open weight models versus something that's potentially open source versus the closed models that you pay for.
If I'm the person who maybe I work at a call center right now, maybe I'm an artist and I'm worried about the future of AI and how it affects my ability to be employed in a very broad and general sense.
Does this type of difference in model play out very differently for me?
Because I'm worried about my job being taken away and my job, not only my job being taken away, but my job being taken away and then not being
not being able to step into anything new in any reasonable amount of time.
So what are, how do you think the difference in these models affects all of the people that are worried about the consequences of this technology?
So I've talked about this on stream and I'm going to repeat that idea.
One of the reasons I think a frequent criticism and concern, which is certainly valid, of all this AI stuff, is even if AI is as powerful as everybody says and it becomes this incredible thing, it's just going to concentrate wealth in the hands of a few, right?
Because it's going to allow certain companies and certain leaders and certain people to just become even more powerful with AI, and they're just going to reap all the benefits, and everybody else is going to lose their jobs and suffer.
I think one of the counterpoints and the reason why I think open source is so incredibly valuable and important is because when a company like DeepSeek, when a model like DeepSeek is released,
the accessibility of AI models just skyrocketed, right?
Because before ChatGPT was the leader, they had the best products and they could put up this paywall right here, according to this on Google.
And this is Google's Gemini AI.
So who knows if it's correct?
But it looks like they have 20 million paid subscribers right now on ChatGPT.
It's $400 million a year on people paying $20 a month to use ChatGPT.
That's losing money on that.
Wait, that's 400 million a month 400 million dollars a month then there's like 20 million pay 20 million paying the minimum 20 tier yeah that's what this is 20 the minimum it's not a five no it's 20 is the minimum yeah so you know again it's hard to know chat gpt is not open ai is not super forthcoming with all their data but this is the estimate but the problem is if they're spending a billion dollars uh or multiple billions of dollars every year the same amount of time right and then and then they're and that's just what they're currently doing there they're trying to ramp it up constantly because the next thing is going to be more expensive the next thing is going to be more expensive i so I think there's a real world where if you just had three AI companies that were making these things, they were growing bigger and bigger and bigger and bigger, that you might see essentially all the value of AI consolidate with them.
And you would need to trust them to continue to make it available to everybody for a reasonable price so that the average person in the world can benefit from it.
But in that situation, it's kind of like if
anything becomes utility-esque in society, but private companies still control it, the degree that they can gouge you on it because it's essentially required for day-to-day life.
I would, in a way,
how we basically all need a phone and the internet to interact with society, right?
If you're applying for jobs, we can't, you can't just show up and hand them a resume anymore.
The interaction is very online.
So, in this circumstance, by making something open source, you potentially relieve that.
Yeah, imagine.
So, imagine that right now you can go buy an Android phone or an iPhone, right?
Yeah.
Both about $1,000.
And then you got to pay for what, $100 a month internet subscription, whatever it is, like whether that's a home or part of your mobile device or whatever.
And that's probably considered the bare minimum to exist in society and be a, you know, and can like work with businesses and whatever else, right?
Imagine that a company that a Chinese company released a phone that costs 5% of that.
So that'd be $50, right?
They release a $50 phone that's just as good and a $50 internet subscription.
And that's why this episode is brought to you by the Huawei P60s.
Yes.
Huawei P60 p60 pro it doesn't work in america
so if you have these models that are released and they're open the price that another company who's closed source like open ai like chat gpt can charge goes down dramatically right competition right it's just it's just competition right and so open source is like it's not just a matter of you come in with a better product that's cheaper again if deep seek had come in and said we're way cheaper you can come use our closed source thing that's just a competitor to ChatGPT.
But every time they release an open weight model, they've basically set this is the standard that's free now.
So you have to be way above that in order to possibly justify your expenses, which to me means that every year or few months, you're going to have the private companies pushing the boundaries and the quality of AIs way up.
But then every few months, the open source competitors are going to catch up almost to them, release a model that suddenly becomes the floor.
The other other private companies can't charge that much anymore because it doesn't make any sense.
And now they have to leapfrog again.
So eventually extremely
go back to their investors and say, hey, we need another $100 billion to train the new model.
And eventually
came out with their new one.
Eventually the investors say, fuck no.
That's the point we're getting at where people are starting to question.
We keep shoveling money into this furnace for you to buy 10 billion more GPUs and make the next model that's going to make us money.
But then it doesn't.
Then you make it and it's cool and people like it.
But then there's an open source one three months later.
That's this, that's the scary part for the corporate
retirement to everybody.
So and this is why it's not directly answering the job thing, but I'm not seeing a world right now.
And there's other folks who agree with this, like Mark Andreessen has voiced similar things about the merits of open source, which is that because companies are incentivized to open source things, and this isn't even getting into the fact that other countries are incentivized to do this.
China is massively incentivized to drop DeepSeek open weight because they just fucked every American AI company, right?
By releasing DeepSeek, they're just like all of your attempts to just have like total control over the AI, nope, it's all gone.
And they arguably cheated to make DeepSeek, which is also, it gets into a whole set of questions.
They probably made DeepSeek by copying ChatGPT, which you're not supposed to do legally.
So interesting questions there.
But in a world where companies and countries are incentivized to drop very, very powerful open source models, it will basically make it completely infeasible for any private company to get too too large of a market share because the open source models will come in.
And to the average person who is trying to start their own business or they're, you know, in rural Nigeria or whatever, and they're trying to
get onto the internet or learn a new language or have one of these tools assist them to make a business or whatever it's going to be.
Like there's going to be many, many, many different applications of AI.
All that is going to be incredibly accessible and incredibly cheap because open source will force that to happen.
If I were to draw a line to something like the internet, I think something that's very cool about what you're you're saying is the open weight model that keeps getting better is so easily deployable to so many people.
Because the equivalent of this, let's use the internet as an example.
There isn't a free because it's a physical thing that needs to be built and a service that needs to be provided in that way.
Nobody can just keep deploying the free internet infrastructure.
Somebody needs to pay and lay like the fiber optic cable for the internet to be like brought to that new place.
But anywhere with an internet connection is somewhere that this can be deployed to.
So like any equivalent or similarly impactful technology prior to this, because it has physical infrastructure behind it, can't be, the free version can't be deployed as quickly as this.
And I think that's so interesting.
That's, this makes this so different from any sort of larger technology before it.
Yeah, it's like, it's like if a company came in and they said, hey, we just put up a billion satellites into space, and here's free satellite internet for anybody who wants it anytime.
Exactly.
All of the ISPs right now who have these like physical connections that run our internet would be like, What the fuck?
What?
And they would not be able to justify their monopoly.
It's kind of it.
That's where I was in my head.
I was like, if China just announced, Hey, America, see mobile, get on it.
It's zero dollars a month on limited data.
Nobody can connect it.
That'd be wild, man.
It's like what they did.
It's crazy.
It's so cool.
I think, I think
I guess the pushback would be it has yet to prove the full utility of something like the internet.
We haven't got to the point where it's fully understood of like how it's going to impact and improve
any aspect of every aspect of life.
We can use it in like very concrete ways now, but the scale at which it will affect everything is not really known yet.
I mean, I think that's a fair thing to say.
We don't know if it's going to be as big as the internet or whatever, but I'd say if you look around, I mean, what I'm seeing just in not even in big business headlines, but in regular people's lives, people are using ChatGPT instead of Google in a lot of situations.
Like it's, it's beginning to seep its way into the regular, you know, I mentioned before the show, I was watching a podcast where one of the hosts of the podcast pulled up a large language model and was like using it to ask a question and get a response to a guest he was having.
I mean, it's this weird changing of of
and by the way, I'm not in supportive of a lot of, I mean, a lot of this ways are kind of scary to me, but it's happening.
happening.
Like it's happening.
And I think it is impacting society in a way.
The real question on my mind is not whether it will do these things.
It's whether,
you know, everyone's retirement is going to plummet to 50% less because there's no, the way they thought it would
crash.
You mentioned, it was funny, you mentioned cables being laid.
You know, in 2001, when we had the internet bubble at the beginning, the big company that blew up then and had a huge stock rise was Cisco, who everyone thought would make, you know, they're like the picks and shovels of this gold rush.
They sell the fiber optic cable, they lay the systems, they help set up servers.
It turns out that's not where the money was.
So right now, you know, people are looking at NVIDIA and trying to see, is that the same thing going to happen?
Because NVIDIA sells all these GPUs.
Everyone thinks that's where the money is in AI.
But if we get to a spot where, you know, most people are running these low GPU inference models open source.
And we don't need to keep buying 200 billion, 300,000 I mean, NVIDIA's money comes from five customers, basically, and it's the five big tech companies that buy it.
Then NVIDIA could be in a less,
they could be the Cisco of the that's what we don't know, but that's where we could be headed.
I have a counterpoint.
I think even NVIDIA, who makes the GPUs that run the AI, I think they're more safe because I don't think that something like DeepSeek changes the demand for AI, but it does change the demand for the AI being 10% better than last year, right?
So, for OpenAI, can they justify $10 billion to make GPT-7?
Maybe not.
But, regardless of whether people are using GPT-7 or DeepSeek 5, they need a GPU to run AI.
So for me, NVIDIA is still like, I don't know if their valuation is good.
It's probably overinflated, but they're still going to be highly sought out because the demand for AI is increasing.
I think it's just the demand for the model that is slightly better and costs 10 billion more.
Right.
Well, that's, so I ask them, it's a valuation question because Cisco didn't disappear.
They just got cut to 10% and then have grown slowly back to where they are.
Because that's where NVIDIA's big profit comes from is like them needing to get 10% better.
That's true.
So they got to buy out.
Those companies stop.
Like, if that stops and it's like everyone's running on a normal RPU setup, then it changes.
We've been going for a while, so I want to ask one final thing.
I'm just curious, and I'll give my pitch on this.
So let's say you are open AI, right?
For the past three, whatever it is, four years, you release ChatGPT.
You're the big hot AI, you know, hottest person at the dance.
You are dance queen up until Deep Sea comes in and takes your crown and kicks you over and rips your dress, and they're the hot new thing.
Okay.
That's your setup?
No, it's fine.
Tragedy is Kane's at the orgy.
Everyone wants a piece of his eponymous.
So my question is, and then Deep Sea comes in and they start like, I don't want to say eat your lunch, but they basically are saying like, hey, we are offering a product that is 95% as good as yours for free or at most
5% of it.
Right.
How do you fight back?
What do you do?
This is, I would argue, an existential threat to a company like OpenAI.
I have a theory about what they need to do in order to survive what is basically an increasing, like increasingly likely reality that they will never have much of an advantage in the quality of their big models because DeepSeek or whoever else will just copy it and make something almost as good a few months later.
That's a great question.
All right.
So
I want to counter like a core axiom of the question, which is that I think actually the way I'm seeing it is that ChatGPT is,
well, they're hurt by this.
Yeah, they're the one that kind of makes it out okay.
The real loser of this deep seek ChatGPT problem is almost everybody else.
Everybody who's not the number one brand.
I feel like that's kind of what I was thinking: is how many people are using llama, which is something I hadn't even heard of until you brought it up this week.
Quick counterpoint: a lot of people, but it's not that's
not B to see.
Is it like Bing, where it's like secretly the search engine?
No, no, no, no, no.
No, it's not Bing.
It's more.
So what we're talking about, where you guys are, let's say, more like casual AI users.
Yeah.
Right.
You guys are talking about chatbots.
Chatbots are a small percentage of what AI is being used for.
If you're talking about a company, for example, Case Text, who made co-counsel AI for law firms and sold it to Roters for $650 million.
So this is a company who's like very successful that made a law firm specific AI that's being used right now by lawyers, right?
This is a hugely successful case.
I imagine you and almost nobody listening has ever heard of this, right?
But they, as the company who made a product on top of the core kind of foundational AI concepts, were able to make this incredibly successful specialized business.
And they are incentivized to go use Llama because then they keep all the money rather than having to pay ChatGPT for it.
So Llama is actually used very broadly, but you're not going to hear about it because it's not a chat bot that people go use.
Oh, this is, this is like when you tell people about Oracle and they're so highly valued and the guy who owns Oracle is one of the richest people in the world.
The average person is
no idea
at all.
Yeah, but I, okay, so first of all, I think chatbots are becoming more and more of a part of regular people's
information diet or discourse or discovery.
And I think ChatGPT, if you ask the regular person, they don't know anything other than ChatGPT.
Yes.
It is like the brand.
Yes.
So, you know, it's almost like an Apple versus Android situation where the deep seat can be the Android where it's more mass adopted.
The phones can be jailbroken.
There's a lot of variety.
You can make your own on Android.
And then there's this one premium charging more brand that has, you know, whatever, a little bit more trust, a little bit more, people prefer that.
I think ChatGPT is isolating from the pack in that it's the only one that has a standout brand that can survive this.
And once they cut down on some of their insane expenditures, they're still making a
crap ton of money.
I mean, this is billions in subscription revenue direct, and they have other ways to make money on top of that.
So my world based on this would be like, oh, yeah,
I totally get that.
Yeah.
Yeah.
Well, yeah, no, keep going.
And then I want to pull up a very quick graph and then we'll, we'll move on to new things.
But
just that in that I think Google is a very valuable business.
90% of their profit comes from search.
Yes.
Even if they become the new Google or take a big chunk of that market share, that's an incredibly valuable business.
And I don't think ChatGPT is so much in trouble as I think the other competitors who thought they could make their own LLMs and stuff like that.
I think they're more in trouble versus this open source area.
That's my feedback.
The quick theory I have is that, so this is a chart.
This is from a Substack article that is,
I would recommend for people interested in the environmental impact of AI.
I think this is a very interesting deep dive into it called using ChatGPT is not bad for the environment.
But one of the things that he talks about in this long Substack article is if you look at the total AI usage in all data centers, chatbots are a tiny percentage of it.
So what we as common,
as the plebs think about with AI is that AI is chat GBT.
It's a site where you go and ask it questions.
But the reality is where it's really, I think, going to have the most value is not that.
It's where AI is integrated into all these other services across society to make everything much more efficient.
And again, some examples, there's like finance models that companies are using that helps them like make vastly improved decisions.
The quants are going to be in the streets.
Yeah.
And we'll send up for the quants.
Yeah.
Yeah.
Quants are going to be.
No, but now quants are going to have these ammo.
And these things exist right now.
AlphaSense and Bloomberg GPT are two examples of like very, very, very successful companies that are using these things in manufacturing technology.
Like John Deere has C and Spray technology.
There's the medical industry is starting to use AI.
Dude, I imagine John Deere, and I fear what they create with AI.
I was going to say, John Deere, we could talk about John Deere's predatory business model in another episode.
We'll bring that up.
You know, like AI, again, AI AI is used for all of the internet, for search, for algorithms on social media, for self-driving cars.
Like AI is sort of crypto bots in my YouTube comments.
Like all this needs to be powerful.
So it's worth acknowledging chatbots are a lot less sort of a part of the pie than maybe it feels like as a casual user.
My quick thesis is that specialized models like that, where you are saying, look, we're going to make a really focused business that is helping one industry or sub-industry really, really, really accentuate what they're doing and make it better and easier for customers.
There are examples of that right now that are incredibly successful.
And I think that's going to really grow.
And I think it is actually, you're in a weaker position if you are open AI with ChatGPT, where you're trying to be everything to everyone.
They have what's called a foundational model, which is that it's meant to answer any question.
And I think what is more likely to be a sustainable business model is if you go make a really, really, really strong, focused, specialized AI model that can be based off of ChatGPT stuff initially, but you build on top of it.
And then you make an incredible tool for doctors or for the government to do tax remittance or whatever, right?
And there's so many industries that can benefit from these types of things.
So you want to be the specialized company rather than the company that is ChatGPT trying to make the tool for everybody.
You don't want to be the internet provider.
You want to be Facebook.
You have a product on top of the core technology.
A government provided tax GPT that just does your taxes for you.
Or a private one.
That is going to happen soon.
It's going to happen really soon within a few years.
Awesome.
I know everybody loves to.
There are so many scary things about AI, but there's so many shitty things we deal with in the world on a day-to-day basis that AI can and will fix.
Get this.
We don't even need to use AI to fix that.
Yeah.
It's just a city of it.
I don't know if you know this.
That is a fact.
But other places have made it better pre-AI.
Nope.
Nope.
This is a good idea.
I don't understand this.
Okay.
I liked the part of the conversation earlier where we were talking about how free models basically help rein in pricing and competition in the marketplace.
And that brings me to our last topic in this episode because I wanted to talk about housing in Vienna, which is often lauded as this pinnacle of housing policy,
especially by
people on the left, I would say, where the
socialist party in Vienna a long time ago started offering public housing at a scale that most other cities just do not have available and public housing without the intention of it being for low-income people specifically.
Public housing for everybody.
Not the only city in the world to do this,
but now their rental market is a little over 50% public housing.
And that splits into two groups, basically.
The way the rental market breaks down is there's normal private rentals, and then there is public, fully socialized housing that the government offers.
And then there's this thing called semi-public housing, which is basically housing that was previously owned by some sort of private company or person that was bought out by the government and operates a little differently from normal public housing.
What this has done, or a success of this policy, is that Vienna, compared to all the other big Western cities in Europe, in Western Europe,
has a very low average rental price for a one-bedroom apartment.
It's a lot cheaper than places like London, like Amsterdam, like Madrid.
And Vienna is used as an example of really successful housing policy because of this.
And I, in preparation for talking about this, I spoke to a couple friends that live in Vienna and got their breakdown and like perspective of how this actually functions.
And then also, you know, generally just
did
generally did research on this topic.
But the idea is that because so much of the housing offered is publicly owned and not run with the profit incentive in mind, and because it takes up so much of the market, the remaining private sector of the market, which is allowed to exist,
has to compete with the
public housing available.
And they continue to build public housing now.
And the share of public housing over time has remained pretty consistent.
It's shifted a little more private in recent years, but the split between the public, the two public options and the private options is pretty, has remained pretty even for a long time.
And when you look at other cities with cost of living or housing crises, a good example is Vancouver.
They've introduced public housing projects like this before.
But the issue is that
when you're starting from square one and very little of your housing is public to begin with, any nice building that you offer up at a way lower price than market that is equal quality to the rest of the housing being built, well, it's going to be snatched up right away.
A bunch of people are going to apply and get those slots immediately, right?
And you don't really solve the problem, a problem in the rest of the market with the rest of the market's housing being expensive because you need enough control over the market and enough available units to compete against the private ones to be able to suppress the price.
And they have managed to do this super successfully for a long time.
The socialist party that introduced this policy is under fire more recently for something related to this, but has been in power in Vienna for a long time.
And they were who introduced this policy.
I
also want to add that talking to one of my friends, He, you know, how it branches down into three different types of housing, basically.
Public, semi-public, private.
He said there's a secret fourth option that a lot of people go with.
Like an in-and-out secret item menu.
I was intrigued.
And I said, you get the house animal style.
They cover it in chili.
And then they knock $200 off your rent.
But he said that
the other common way a lot of people get cheap rentals in Vienna is just
by knowing people.
And he said,
your homie way into private renting is actually the cheapest option of them all.
And he listed off himself, his girlfriend, and a bunch of other people he knew as examples of I just met someone through a friend or through a family member and then got my rental through them.
And then because we could dodge all the agency and paperwork aspect of all of the other options, but specifically the private option,
you could get the place at a way lower price and you're way less likely to be to be gouged.
So there was this secret like fourth private option where you avoid all the paperwork.
And also interesting is prices have actually started to climb at a way higher rate in the last couple of years.
And the reasoning, as explained to me, was they have these agencies that sell or introduce people to homes.
So these private agencies, if you're looking for private housing, will
they'll be the guiding light to like finding you an apartment in Vienna, give you like private options that you could potentially rent, right?
And that process apparently takes a long time, which is one of the reasons reasons why you might try to find your homie option instead, because it's just more straightforward.
But the prices of that, you paid the agency a fee as the person looking for an apartment.
You pay them a fee to like help you find a place to stay.
That fee, two, three years ago, a policy was passed that passed that fee on to the landlords.
And this is the policy that has backfired on the socialist party that has
the largest share of seats in Vienna,
where the agencies and the landlords have basically started colluding because they can privately negotiate what this price is and then tack it onto the rent after with no interaction from the renters,
has started raising average rental prices at a way higher rate than they ever have in past years.
There's a couple other reasons why rents have gone up as well, but this policy has totally backfired because of the ways the agencies and the landlords have been able to gouge that specific fee out of of people looking for private rentals.
And
this just gives you a little like landscape of what housing is like.
I can answer like more specific questions if you have them, but because of the time we have left, I know you had specific pushback to
you were saying you don't want people to have affordable housing.
I did say that I was saying that off pod and I want to say it on pod because I'm in the villain chair.
I don't think people should be able to live.
Can I pull this up?
Yeah, yeah.
No, listen.
But if I were here, I'll give a synthesized kind of
because like I said, there's a lot of details that I dug into with both the people I interviewed in the research into this topic.
But the synthesized point I want to make here is because we could say that because of this city's approach to housing and the amount of public housing that competes against the private market and importantly, important to note, they've kept building public housing along the way.
So you can't just like have a static.
you know, quantity and then not do anything about it after.
That competition from the public option that has no profit incentive is a huge part of what has kept rents low and in control in VNF for a long, long time.
So I'm curious what you guys have to say.
So what I want to say is,
I'm going to go slightly free market, bro, on this one.
Yeah.
Village.
Yes.
Village.
And this is one area.
Like, you know, I've talked about this before.
Like, I'll watch a Bernie Sanders speech.
I fucking love the guy.
I'm cheering for him.
And then he says something about rent control.
And I'm like, because I've lived through,
I've had a rent control apartment.
I've seen the experience firsthand.
I've seen apartments without it.
I've looked into it a lot.
And over what I've found,
to me, it feels like
a kink in the hose of supply and demand that causes pain to a lot of people outside.
A few people benefit, but a lot of people are hurt.
And this is sort of somewhat in that.
So I want to try to explain.
I hear Vienna brought up a lot.
And just to overview what happened, I think you said this pretty well, but the idea is: you know,
Vienna was the capital of Austria-Hungary, a big populous city had 2.3 million.
yes, sir.
Yes, sir.
Corner of the world had 2.3 million people in 1916.
Obviously, Austria-Hungary had a problem in two different wars.
Pretty bad prime.
Things went to shit.
Country got
devastated.
Vienna turned into Vienna sausages.
Yeah, exactly.
And so we flash forward 100 years later, and they are at
about the same level of population, 2 million.
So again, a housing problem is much easier to solve when your population barely grows or barely changes in 100 years.
When it's mostly flat, the problems that you might have in like a SF Bay area where a lot of people are trying to get in at once
aren't as apparent.
What they did was, that I think deserves commendation, is not whether the house was socialized, whether the house was rent controlled, whether that, it's that they built fucking houses.
They built a lot of them.
That was the, when you really look into this policy, the thing that differentiates it from other cities is that they said, every time we add a new block or a new new area, we're going to set aside some area that has to be used for building housing.
And it has to be dense housing, dense, socialized housing.
So whether that housing is socialized or not, it is zoned for dense, affordable housing.
So it gets built in actuality.
So I did some research for this and I wanted to see what actual,
you know, to be to your credit, like I didn't expect you to bring up so many points from people that live there, the good and the bad.
So I don't need to bring all this up, but I wanted to look at what actual people that live in Vienna are saying.
And I tried to find comments that people are very positive on the city, except for a certain thing about rent.
So I brought a couple of them up.
This guy has a ton list of pros.
He loves the city, but he mentions that cheap housing is kind of a lie.
There's a lot of public housing, but to qualify at all, you need to have lived two years in Vienna.
I think you know this as well, right?
Yeah.
You can't get it if you haven't been there two years.
So if you're a new student that moves to Vienna, your options are only private housing.
And actually, the private housing is rather expensive.
It's actually gotten to the level of like Berlin and other major cities.
So if you're there for two years, you can apply to get on a list to wait in a waiting line to get one in socialized housing.
But there's a a difference between the good socialized housing and some of the more
not as good.
And generally what happens is to get the good stuff, you have to know somebody.
You have to have an in.
You have to know a politician or somebody that already has one and wants to rent subletted out to you.
So it becomes a game of who you know
rather than anybody of you to buy it yourself.
So for example, he says the good ones never go vacant.
If you don't have these connections, you get a cheap flat with cardboard walls.
You have to rely on private market, which is on par with other big European cities.
Rent for apartments are pretty steep, especially for newcomers.
They don't qualify for social housing.
Basically,
what this ends up being is a subsidy for a few people that punishes anyone who is new or moving to the city.
And it creates these weird externalities that I've seen with rent-controlled apartments, even the ones that I've lived in, where people that have, maybe they get a better job, maybe they have an opportunity somewhere that they cannot leave for because the...
the subsidy they have for the rent control now makes it so that they they can't the deal is too good to leave to go anywhere else so it kind of of clamps down on normal economic movement it clamps down on the ability for new people to flood into the city and enjoy it it sort of stagnates it kind of locks you in time now i think we all know as you can say from your open source model or from um this example for anything is when you have a lot of competition the price goes down and i fully agree with that and i think if this is the way you have to do it it's better than what we're doing in the bay area where there's just nothing happening totally but i think um the negative externalities like for a good example zen is because the government is the largest landlord, they own almost all the houses in the city,
they've done a pretty bad, outside of the rent being low, they've done a terrible job at it.
In terms of like, there's many loud, vociferous complaints about trying to get anything fixed.
In fact, you generally have to factor in the cost of doing the repairs yourself because the time it takes is so long, the lag time of getting anything fixed is so long that you have to pay for yourself.
And when you add that into your rent cost,
it makes it more even with other places.
So my idea is like simply, whatever way you got to do it, if we're making areas zoned for the demand is there.
People want to live in these parts of the city or they want to live close to the city center at an affordable rate.
People want it.
If you make it so that's allowed to be built, someone will build it to take advantage of that opportunity and they will sell it.
So
I think their big win, the big takeaway is just the great zoning and not really so much that the government has to make and build and decide what the houses look like, where they are.
We're going to run your landlord.
We're going to do, I think that's the area where I feel like is it's overstated.
And the idea they have to set the rent price with rent control.
Again, all that is is kind of a well, they have a so they do, I don't know if I would call it rent control, but they have an additional policy that applies across everything, as far as I understand.
I forget the German phrase, but it's basically called the index adjustment, which means that landlords cannot raise prices
unless they scale with inflation for the year.
And then even then, there's a cap on that.
So it prevents huge spikes in rent from happening over short periods of time.
So that's also something that keeps prices low.
I think the idea there being that it prevents somebody like a landlord from gouging you on the fact that you need to pay for this place to live in.
They can't just keep scaling rents at a rate that outpaces inflation, basically.
Yeah.
Like, okay, let's imagine that you are going to a
this is Aiden's desert rave.
okay?
You're going to, you're going to go to a big rave party in the desert.
It's going to be awesome.
And there's going to be a concert and there's going to be light lights and everyone's going to have a good time.
And everyone out there is having a good time at the rave in the desert.
Okay.
And someone brings
a bottle of, or a case of Aquafina water to sell because everyone's going to be thirsty.
It's going to be a, it's a sweaty ass rave.
And they know that no one else there is selling.
So they can offer it for $10 a bottle or $20 a bottle and it sells out.
If the government comes in and says you have to only only charge $2 a bottle, that's it.
There's a price cap on that.
You just can't, you cannot raise the price.
But the demand is there for $20 a bottle.
Then what usually happens in real life is that they'll say, okay, those, but then, you know, mysteriously, a crate of water falls off the truck and now it's being sold in the black market because the demand is there.
Like even if you followed the law to the letter, some people get the $2 water in the front, then it's out and nobody.
It doesn't, you know, you're not saving more people.
The same amount of people are getting water.
There's a better price.
So the only real way to fix it is to get more water.
The only way is to have more people selling water that they're driving the price down.
So when you cap the rents like that, you end up with situations like your friends talk about where the real way to get it is you have to know somebody.
It ends up with these weird back channel.
I said maybe I
maybe somebody from Austria or Vienna can chip in and clarify here.
But the way it was explained to me is that the wait they're talking about avoiding is
specific.
They were talking about the wait time for private housing, finding private housing the normal, the normal way.
They also said it can take a while.
The way it broke down is with normal public housing, which doesn't, sounds like it doesn't have the same restrictions that you were talking about.
The public seems to be available to any
EU citizen
or refugee.
Well, after two years, right?
No, no, public.
Sorry, the purely public housing is any EU citizen can apply to live in it, right?
But the semi-public housing has a Vienna residency requirement requirement to it.
So there's a distinction between those two things.
If they broke down the semi-public option is you basically apply for like a ticket.
It sounds like something you apply for like way in advance if you're a Vienna resident.
And then
after,
you know, eventually you, your ticket comes up in the queue and then you're fed options up to three options of flats that you could move into.
And then if you veto three times in a row, you reset your position to to the back of the queue or you reset your position back like a year.
Yeah.
And that was an example of how the semi-public stuff broke down.
So it seems like there's different wait times and different systems for like how to fight through these things.
I think another piece of a context here is when you're talking about knowing somebody to like get past the system.
He talked about the broader problem of Austrian bureaucracy in general.
He said, getting things done in Austria in general is a miserable experience.
Yeah, I got that.
Getting your marriage acknowledged, for instance.
He was talking about how his sister just got married and she's getting married to someone not from Austria.
And they don't even live in Austria right now, but they want their marriage to be recognized by the Austrian government.
And it's a huge pain in the ass.
It's a very complicated process for basically no reason.
And I saw a lot of complaints about that in general from just strangers online, like Austrian bureaucracy specifically.
I think what you're saying is
like,
I think what you're saying broadly,
I kind of agree with.
Like there's, there's, I think building is important.
And if they weren't building more public housing units along the way, then it wouldn't make, it wouldn't make sense.
But they're building a supply of public housing that is competitive with the private market over time.
And those places are available to get into or apply to.
That's something that does affect the market regardless.
No, it does.
More housing is the biggest solution to the problem.
Yeah.
I just think there's a lot of negative externality, a lot of annoyances that crop up when it's done this way that are still better than not building at all.
Okay, so this is, I actually think this synthesizes nicely down into a point that my friend basically signed the conversation off of.
And because I'm not trying to get him to go one way or the other, I want him to tell me what his experience and his friends and his families are like.
How do you guys talk about this?
Basically, what are the vibes with housing in Vienna?
And he's someone who's thinking about moving to London pretty soon for a master's degree.
And he's like, housing in Vienna is something that me, if I were to talk about with a close friend, we would both be like, ugh, like we'd, we'd complain about it, but then any reference to cities or places outside of it, we agree that we are in a way better situation.
Totally.
And I think that is important to recognize is like he looks at other cities in Europe.
Still views it as as a positive situation and views it as much better the examples he gave is like Amsterdam is a nightmare right now and London is awful Toronto or Sydney or yeah 100% and I also want to make a point that uh I think I arguably I'll I'll be honest with you, I probably need to spend more time looking into this.
But another famous city for its public housing policies is Berlin, like historically.
Berlin has had a lot of its housing eaten up by the private market over time.
So like the percentage split that I was talking about with Vienna's market has degraded in Berlin, where private
people or companies have a much larger share of the housing now, and the prices have skyrocketed in that same period of time.
So as private
interests gain that majority share, the prices increase at a much higher rate than they did before.
Did you look into seeing if the rate of housing building also went up?
Because I think I, correct me if I'm wrong, I think that's kind of the point you're making.
What if Fianna, if none, if they had just upzoned and none of those buildings that they made were social, but it was just they just made a shape like that.
But they just built anyway?
Because that would be interesting for us to do is compare like a similar amount of building that happened in another city around the world where it was all private, but it was just a large quantity of housing and compare those.
I can break down my response into
three things here.
One, we could go back to the Tokyo slash Japan episode.
You guys are absolutely right.
I think people in the, some people in the comments made fun of me for like, oh, Tokyo cheap.
No, no, no.
No, that's not the point.
Tokyo is Tokyo housing is much more affordable relative to the income of people there than a lot of other large cities in the world.
Like compared to the LAs and the San Francisco's and the Londons, Tokyo is weirdly, considering the scale that it exists at, cheaper than proportionally than those cities are.
That was the point I was making.
And they don't have an enormous amount of public housing in Tokyo, from my understanding.
It comes from the speed and amount that they build.
So it's not that I disagree with that part, but the Berlin thing, if we were to go back to the population, for example,
a lot of the big cities in Europe have had pretty stable populations for a long time.
If you look at population differences between like early 1900s, mid-1900s, and now, most of these cities are pretty similar.
Like I think London back then was,
or in 1950, is like 7 million people and now it's nine.
Like that is two million more people London's grown a lot specifically but I think Berlin is almost exactly the same I want to say I looked it up like right before so the population thing is
I if the population is about the same during that time period it's not that you can't be right it's not that there isn't a version of this argument that I think has value but I do want to point that out is like population stable
You know, other things like maybe family size and occupants per unit could have changed and like cultural trends in that time.
But
the point I'm making is it scales suspiciously with when private investment gets a larger hold of the housing market.
And I think that not
when I say private investment, I don't mean like firms or private equity necessarily.
I mean just private buyers, private buyers owning parts of the rental.
Yeah, I can't comment till I know more about how many houses they built in Orlando.
This is the last thing I would do.
According to Chad GBT, which, you know, asterisk,
4 million in 1925 and about 4 million in 2025.
So similar to Vienna.
So about the same.
Okay, about the same last hundred years.
It's like when we keep these factors consistent across cities that are relatively similar, I don't have an in-depth understanding of Berlin zoning.
So there's probably a deeper conversation there.
The last part of what I wanted to say with this.
Well, I just want to say on that because like if I'm a student and I move to Vienna or I move to Berlin,
based on the wages of each area, the cost of the apartment is the same.
It's essentially the same.
Because
I can't get the two-year special deal.
Okay.
If I were to take that argument the other way, right?
Yeah.
If I were to say, oh, only the private or
semi-public options are, sorry, all the public housing is basically gone because it's cheaper.
I only have the private, more expensive options to go to.
The private process is like pretty arduous because of all the bureaucracy behind it.
It's like, couldn't the solution to that to make build more public housing?
Like, I guess you're not denying that either.
You're just saying to build more in general.
And the argument I would give to that is: I think a big part of this and why housing gets to be priced lower.
And I think the
sorry, this was a while ago when I watched this, but it was breaking down that the Vancouver is increasing the amount of public housing projects they have, but it's a very small percentage of the overall market, right?
These places aren't being built at a loss.
They are pricing them at something that covers the cost going into the buildings.
They do pay themselves off.
So removal of the profit incentive does keep the price lower for these specific units.
You can price them at lower margins and have them remain affordable for more people.
Do you get what I'm saying?
If you have a market that is 100% private, then every one of those actors has the profit motive, even if it's a market competing against one another.
And the government option in this case, if run correctly, is something that will never have that motivation because it's a public works project that's like funded by revenue or taxes or whatever.
I agree.
Yeah.
I mean, I agree.
I think, again, I think it's better than where most of these cities are.
And if I was my button, I could only press, I would pass it 100%.
I think, I think this is, this is, if I were to, sorry, let me add one more thing.
I know I've been talking a lot.
I apologize.
It's interesting.
I think the common argument I see against when we were talking about abundance, that the number one leftist critique of abundance, often from people that have no idea, that didn't read the book book, clearly.
I'm just going to be honest with you.
Not a single person I've talked to that has a heavy-handed criticism about the book has read the book.
But
what the
criticism of the book is that they claim it is asking for more things to be handed off to private interests when that's not what the book says
at all.
It's saying, arguably the opposite.
It's saying,
I would say it even argues for more public autonomy, like government autonomy to take action.
But regardless of, in the case of housing, regardless, building needs to happen, get it done by any means, kind of.
That's basically what it's saying.
And not even building housing, it's talking about way more.
I think people get wrapped up in like the first 50 pages where housing gets talked about a lot.
But the whole book is like make and build and innovate.
through whatever means necessary because a lot of the blockers that affect these things are the same thing blocking government agencies or government people from making them as well as like private businesses.
Yeah.
And then you have the government create incentives that make things happen rather than have the government.
Yeah, they don't necessarily need to do everything, which is what he was talking about in the invention portion.
I thought it was like an interesting comparison to this, which I like in concept, although nothing actually gets built, is San Francisco has an affordable unit policy where if you make a residential building with 10 or more units, there has to be a certain percentage of them that are affordable.
So most residential buildings that go up in San Francisco have some portion.
It's usually 12 to 16 percent.
It's like 15 percent of them are like designated designated as these are affordable permanently.
And I think that's a really interesting concept where you still are incentivizing and saying, hey, developers, you can still make a shitload of money, but some portion, you have to allocate some portion of that profit and basically drop that so that affordable housing can get built.
Unfortunately, it seems like that then gets used a lot to stop anything from being built at all.
So they're like, it's not enough affordable units.
And that's, that's the criticism I've heard is then that in turn is basically just used by NIMBYs to make sure nothing gets built.
And then that is, again, the worst scenario of any of them, which is you don't build anything.
Yeah, that would be the worst outcome of basically this version of the conversation, right?
It's like, because I'm saying, oh, this way is good.
And you're saying, maybe like this way isn't perfect.
And then the two people resolve to build nothing.
And then you've
then you're definitely not sorry.
Yeah, which I don't think is what we're saying, but I think is a version of that, that conversation.
It's how that is happening in practice in many places in the country, unfortunately.
If you know that if the first step was just changing the zoning laws, we we would be a huge improvement it would be it would change so much so quickly things would get built already and then you'd go okay maybe we need some more right socialization there's so many for-profit developers that would just are salivating at the idea of building in san francisco because the prices are so high
even if you're like 20 of it has to be affordable they can still be like well we'll still make money because the prices in sf are so high and then enough of those developers come in and the prices drop because you've just added 50 000 100 000 units you know it's like and i just think back on my life i lived in a lot of different apartments and some of them were ranked.
Actually, one of them was rank control in NSF.
Yeah.
And I, I remember my experience with that landlord and how much he fucking hated me and probably wanted to burn the place down so he could get more money.
But the only time I've ever had a landlord
that was even remotely at all trying to improve things, fix things, help me out, be responsive was when there was competition in that area.
They had just built this thing.
They needed to fill the units.
They needed, you know what I'm saying?
They had, they had a financial incentive early to fill these units because otherwise they're going to take a loss on it.
So they needed me to be happy.
That's the only only time I've ever seen it actually like work.
And I think in our real lives, you know,
I can think of only so many things are monopolized and have been fucking destroyed that we hate it.
But the few industries I think regular people like are the ones that have a lot like indie games.
People fucking love indie games because it's an open marketplace.
A ton of people can do it.
Everyone can make it.
There's a hundred competition.
We're always getting new good stuff all the time at a low price.
TVs have gotten cheaper over the past few decades.
In the few areas where we have real, sustained competition from a lot of actors,
there's a consumer benefits dramatically.
We see that, and so I think government's role should be to try the best they can, like real strong antitrust, real strong competitive enforcement, trying what they can do to make it so, you know,
things are as competitive as possible.
Things are as competitive as possible.
I think my argument would be this is a way of doing that.
I agree.
Yeah, I'm not even that opposed.
Honestly, in this example, the government is basically one giant competitor, which is to say, in this market, if you are a private developer, you have to compete against the government who is setting it at a certain price.
So it's somewhat forced, but it's still again, it's better than just don't build anything.
There's more nuance.
This is a there's oh, surprise, more nuance.
Uh,
no, there isn't.
One thing I literally
read
some about Singapore's healthcare system recently as a refresher, and there's similar components at play there.
Basically, they this idea that by allowing private uh industry to exist and giving it like incentives and restrictions, and then also having public options that compete against it with the same rules.
And without the need to seek profit,
you create a market that is more healthy.
So it's very, it's very interesting.
This thing is also crazy.
We could do a whole thing on Singapore housing.
It's like a totally different housing is way different, but also good in a lot of ways.
So I, I, you know, lots of, lots to talk about both on the Patreon
and in future episodes.
So check that Patreon out, huh?
Yeah.
I know we've plugged a bunch, but this is the, you know, we're just kicking it off.
And I think at the beginning is like when there's the biggest push.
So patreon.com/slash lemonade stand.
Obviously, if you don't want to subscribe, you just want to keep listening to this show.
This will always be available for free as normal.
This is just a bunch of, if you want an extra hour of the show that we record on a separate day,
and a few other bonus shows.
Oh, I forgot to mention one of the main things also is
either individually in groups of twos, we're going to be taking more comments from the show and replying to them directly, doing direct follow-ups as another separate hour that we put in tier two of the Patreon.
It should be about our physical looks.
Yeah.
Just comments about how we're going to be able to do that.
How old I am.
Where's it at this week?
Yeah.
We'll keep a marker and go up.
Yeah, I mean, I want to say, you know, direct to Cam here, we recorded our first Patreon episode and I really, just me individually, I really loved it.
I thought you guys both had really
authentic, heartfelt things to say about the way you engage with,
I don't know, the world and thoughts on the way you've been evolving and learning about this stuff.
I thought it was really enjoyable.
So I highly recommend at least checking that one episode out and seeing if it's for you.
But
yeah, I'm excited.
I've been enjoying this podcast.
I think I've been learning stuff and it's cool.
This is great.
I've had such a good time.
Yeah, it's fucking dead.
Great time went long today, but I thought it was a good discussion.
But Trump didn't do anything cool today.
There was no more problems.
That was the one problem.
Today's episode.
Kind of a flaw.
We're becoming snoozer on tariffs, dude.
It's like
what if we split a gold card?
What if we just
20,000 patron gold?
We buy the gold card.
Yes, sir.
Who gets it, though?
Because you already have perfect.
You know what we should do?
We should crowdfund it and then we lottery it to a fan.
Oh, we get one fan for one fan gets the gold card.
We could make a whole market.
If we get enough of these and we're buying and selling, let's get into the gold card market.
Like bonds, dude.
Dude, wait.
Couldn't you theoretically run a gold card lottery at a profit?
You so could.
This is evil.
If enough people put in $5, one of them gets the gold card and we still make a profit.
We scrape the...
Oh.
Okay.
This is actually horrendous.
And we've done it again.
Officially launching
every $6 subscription to Patreon does get you a lottery entry into the gold card.
Assuming that Trump allows us to lottery it off to some random person, right?
Because we don't end up in genius.
Let me know.
A couple asterisks that you need to check out.
In fact, so many asterisks that you could put at the end of your comment if you got to the end of this episode.
Prove that you're a real fan.
All right.
Dude, if you made it to 153, put the ending paragraph of the book abundance into the comments so we know you read it.
Yeah.
Don't do that.
All right.
Well, thank you for watching this Limited Stand.
We'll be back
next Thursday.
Farewell.
Bye.