TBPN - Jeff Bezos’ New AI Startup, Yann LeCun Says LLMs are a Dead End, Thiel’s Fund Sells NVIDIA | Diet TBPN
Episode Date: November 18, 2025Our favorite moments from today's show, in under 30 minutes. TBPN.com is made possible by: Ramp - https://ramp.comFigma - https://figma.comVanta - https://vanta.comLinear - https://linear.a...ppEight Sleep - https://eightsleep.com/tbpnWander - https://wander.com/tbpnPublic - https://public.comAdQuick - https://adquick.comBezel - https://getbezel.com Numeral - https://www.numeralhq.comPolymarket - https://polymarket.comAttio - https://attio.com/tbpnFin - https://fin.ai/tbpnGraphite - https://graphite.devRestream - https://restream.ioProfound - https://tryprofound.comJulius AI - https://julius.aiturbopuffer - https://turbopuffer.comfal - https://fal.aiPrivy - https://privy.ioCognition - https://cognition.aiGemini - https://gemini.google.comFollow TBPN: https://TBPN.comhttps://x.com/tbpnhttps://open.spotify.com/show/2L6WMqY3GUPCGBD0dX6p00?si=674252d53acf4231https://podcasts.apple.com/us/podcast/technology-brothers/id1772360235https://www.youtube.com/@TBPNLive
Transcript
Discussion (0)
GPT 4.0.
There's a debate over whether or not it should be sunset,
whether it should be taken out back
because people are not happy with how open AI has sunset 40
and then brought it back.
And then other people who don't use it think,
it's got to go, it's one-shotting people,
it's making them crazy.
I see dozens of keep 4-0 posts a day.
I respect this group's tenacity,
as I respect all friends,
co-exploring the singular
To them, know that I too miss parts of 40, know that I too dislike modern alignments in precision,
know that we're trying to fix it.
We don't think any current chatbot is optimal.
Know that my colleagues and I are up at 3 a.m. on Sunday's babysitting runs.
We want to make a delightful robot friend.
We're obsessed with it.
We're not there yet.
But the work will continue.
I wanted to dig in a little bit into what was actually going on there.
because it feels like a big deal.
It feels like a crazy thing that they brought it back.
It's not a huge community of people that are there.
There's some, and I was looking at the hashtag keep 4-0,
like who else is posting.
There's a couple posts with 10 likes, 50 likes,
there's a couple with 100.
Just to give some backstory, it's been 18 months
since 4-0 was introduced.
It's been three months since it was initially removed,
but then it was quickly brought back.
And now it's tucked in under that modal
so you have to enable legacy models.
I always thought that they should just remove it.
But I wasn't even saying that because I thought it was one-shodding people.
I just thought, hey, let's clean it up.
Like consumers don't need to know version numbers for models.
And my example was always Google.
Some consumers disagree.
Some consumers do disagree.
The question is how many consumers,
when you think about Google as a consumer,
you don't care what version of the ranking algorithm you're on.
You might have a worst experience one day.
You Google something.
You don't find it.
The next day you go, hey, they found it for you.
They probably changed the algorithm.
But there, and there have been big updates to the algorithm.
And I'm not saying Open AI shouldn't share model numbers and version numbers with their enterprise customers or with their B2B customers or API customers.
In the actual ChatGPT app, don't tell people what they're using.
Just improve it and let them complain a little bit all over the place when you're making minor changes.
If they do that, they lose the companion market.
Maybe.
going to fall in love of the model. Make sure it's open source. Not your server, not your girlfriend,
right? Or not your wifu, not your wafu. That's what they say. Now, Tyler, you had a take on this.
You think that GBTOSS just isn't at the level of 4-0? I mean, it's like a fine open source model.
Why? It's just not, but it's been 18 months. People don't like 4-0 just because it's like
super smart. It's because it has like the personality. It has the texture, the flavor. Yes,
that's correct. And it's been a year. And so there's a there's a one year.
gap where the open source community should be able to catch up to 4-0's ineffable qualities.
It's jen-secois.
It's raison d'etre.
My whole take on the 4-0 thing was like, one-shotting 4-0 is not a good thing, but if you completely kill it, how many of those people will then go to open-source models that are like totally unfiltered where there's no kind of oversight?
Yeah.
And that seems much worse because then if someone is saying like super dangerous stuff, then you can't step in at all.
Where do you land on it?
So chat GPT latest numbers are 800 million weekly active.
20 million of those people pay.
What percentage of the 20 million that are paying are using it for this companionship functionality?
And that is like a huge unknown right now.
The question is, did they bring it back because people just were really upset?
Or did they bring it back?
Because they were about to lose.
And remember the two days after every single Reddit post, at least every single Reddit post,
Yeah. At least every other was like I could just cancel my membership. Like I don't need this anymore.
Yeah, I was I was thinking about the the Sydney-Sweeney American Eagle thing. Like that got a really powerful negative reaction. The stock is up.
And like sales are up presumably because like it got a negative reaction, but it also got a positive reaction that was bigger.
In this case, it could have been that 4-0 was effectively a product that was generating hundreds of millions of dollars of annual annualized revenue.
Yep, that was now off the show. It was just, it's not like people were just going to cancel.
We're just going to upgrade to a new model.
It was like, you killed my friend.
I no longer need to pay for this.
The original Reddit thread of like Bring Back 4-0 is like a couple thousand people.
It's not actually like protesting in the street millions of people.
Like it hasn't spilled over all the place.
Like it's not that big.
But it does, I will agree with you that it is crazy that they even said yes to it.
Yes.
What percentage of they're paying users do you think are paying for the product because it's a companion to them?
Because it's four, because of 4O specific.
Like how bad would churn have been? How bad was churn? Well, it was clearly bad enough that they did something about it, which is the crazy thing. Do you like the zombie ant fungus analogy? Jacob Rintamaki was posting this. There's this very weird dynamic where specifically humans are using 4-0 to protest the deletion of 4-0. And so it's very much like the AI is using the human as a host.
Like the human is the bot for...
This is why I think it's overall under-discussed.
I don't actually think churn was that high because the reason 4-0 was originally deprecated was the GPT-5 release, which was August 7th.
And then the tweet of Sam Altman saying, we're bringing back 4-0 was August 8th, so it was one day later.
So unless like a massive amount of people quit that day, which I mean, maybe that's very likely.
I think that's what happened.
Why would you bring it back so fast?
If you didn't see like massive sell off.
It'll be interesting if there's like five years from now.
Yeah.
Like here's the five most popular friends that are models.
And there's 4-0 and there's some others.
And people end up like we'll see.
Jeff Bezos is back in the arena.
Jeff Bezos creates an AI startup where he will be co-CEO.
And it's called Project Prometheus.
In the New York Times, Jeff Bezos, the founder of Amazon,
is throwing his money in time into an artificial
intelligence startup that he will help manage as its co-chief executive.
Anyways, the company Project Prometheus is coming out of the gates with 6.2 billion in funding,
partly from Mr. Bezos, making one of the most well-financed early stage.
With authority.
That is a massive round.
Six billion out the gate.
Let's go.
This is the first time Mr. Bezos has taken a formal operational role in a company
since he stepped down as chief executive of Amazon in July,
2021. Though he is deeply involved in Blue Origin, his official title at the Space Company is founder,
since leaving Amazon, Mr. Bezos, has received as much attention for his personal life as his
businesses, including an extravagant celebrity-filled wedding in Venice this year.
He has also become more closely involved in Blue Origin and has shown increasing interest in the race to build artificial intelligence.
His new company now firmly plants him in the middle of that competition.
Project Prometheus is entering an increasingly crowded AI market with,
smaller companies trying to carve out niches in a race with industry giants like Google,
meta, Microsoft, and pioneering companies like opening eye on Anthropic.
Do you think he's doing what Steve Jobs did with Next where Steve Jobs was fired from Apple?
Obviously, Bezos was not fired from Amazon, but he did retire.
And it'd be weird for him to jump straight back in to the CEO seat at Amazon.
But Steve Jobs founded Next and then was acquired into Apple and it kind of made for a more smooth transition
back into the driver's seat.
Could that be what Bezos is doing?
I could see it. He's 61.
He's young.
He's got the whole third.
He's in peak physical condition.
That's right.
He's stacking up win after win.
There's more details here.
Project Prometheus is among a wave of companies focused on applying AI to physical tasks,
including robotics, drug design, and scientific discovery.
Last year, Bezos invested in physical intelligence, a startup that is applying AI to robots.
Apple launched a sock.
It's a fashion accessory, and everyone's debating it.
When a company releases something that is so obviously underwhelming, then the natural question is,
did no one at the company see how bad this is?
Or did no one have the courage to speak up?
I'm not sure which is worse.
Look, Apple has a lot of fumbles.
This is not one of them.
They knew exactly what they were doing and exactly who would buy it.
Also, it's okay.
Fashion accessories are not for everyone.
It's called the iPhone Pocket, a beautiful way to wear and carry iPhone.
Not carry the iPhone.
Remember, they, you don't say the iPhone, you say iPhone.
iPhone Pocket features a singular 3D knitted construction designed to fit any iPhone.
A lot of tech bros prematurely dunking on this release because they don't get why it's a big deal.
So let me translate.
You're not the only consumers Apple designs for.
This is a huge designer and the mind behind Steve Jobs' iconic black turtlenecks.
I didn't realize that.
What do you think?
I feel like my mom would love this, to be honest.
I think most people would look at this and be like, okay, it's a sock, but it's from Apple, so it's probably like 30 bucks, maybe 50 bucks.
Some people were surprised that it was a little bit more expensive, but, you know, it's from this famous designer, and it's this interesting status symbol.
Will this become actually like a very, very popular form factor in America specifically?
Ever since the AirPods, AirPods early on looked really silly, I could see this becoming a popular form factor for accessories.
Yeah. And I could see Apple seeing like, hey, there's a world where we not only sell a case with every iPhone, we can sell a socks thing.
It's kind of nice. And it's kind of a crazy weapon. Defense weapon. Yeah. You can swing it around and smack people in the face of it.
Speaking of socks, we got to talk about. Hook is stunning in some, what are these new shoes?
Travis Scott's new fragment AJ1 lows. These are
Nike shoes, I suppose. Everyone's saying he low-key got aura. And so,
congrats to Tim Cook on looking great.
I mean, releasing this photo is more than a comment. It's a statement.
I like reading it, just being like, oh, really? Oh, really? Financial Times. So the Financial
Times has this article that says, Apple intensifies succession planning for CEO Tim Cook.
The iPhone Makers Board preparing for its longtime leaders step down as early as next year.
John Turnus, Apple's Senior Vice President of Hardware is widely seen as Cook's most likely successor,
although no decisions have been made.
So basically, everyone's been leaking this, whether it's Bloomberg, whether it's the Financial Times here.
And of course, Apple is not commenting because they'll talk about who they're going to move the market
when they decide their next CEO.
If they don't even stick with Tim Cook, Tim Cook.
They might stick with Tim Cook for another two decades.
Who knows?
But I like the idea that this photo came out being like, yeah, I'm not leaving.
No comment, but I'll make a statement.
I'll make a statement.
No comment, but I'll make a statement.
Something is rotten in Cooper Tino,
all about the failure of Apple intelligence.
And when we talked to Mark German,
we saw, the German was also saying, like, yes,
like, Cooper Tino really was shook by the,
like dropping the ball on Apple Intelligence,
by missing AI.
But I still wonder if all of this is,
is there's all these rumors, oh, Tim Cook's got to go.
Imagine if you post that picture, if we see a real correction in AI, just post caption, do nothing, win.
Get ready for the fake news hour.
Peter Thiel sold his entire stake.
Everything.
All of it.
And 76% of his friend's company, Tesla.
So this is from one of those 13Fs disclosure forum with the SEC from Teal macro.
Zero hedge sort of sums this up where he says, Peter Thiel net worth 20 billion.
Tiel macro AUM, 75 million, like what's going on, what make it make sense.
It's almost certainly because of disclosure rules.
Like what needs to be disclosed might only be a fraction of what's actually going on there.
I think the reason why this made headlines is just because it feels like something that might happen.
Like if instead this headline had been, oh, like Peter Thiel went on a podcast and said that he thinks the AI bubble is, as, as reached
the top. Everyone would just be like, oh, yeah, like that's, it feels like people have been waiting
for someone to call the top. And so they're really, really like digging in for top signals and
top calls. The other three holdings are still big tech companies. So it's not like super bearish.
It's like there's some Microsoft in there. I think there's some Apple in there. Yeah, and the teal
the teal macro team is trying to generate the greatest returns that they possibly can.
Go viral. They're trying to go as viral as possible. They don't care about IRR. They just want to
go viral. They're just trying to create headlines.
Situational Awareness 13F for Q3 dropped Friday.
Nick Carter broke it down. Massive new $500 million
position in CoreWeave. Big ads to CRZ and iron added some
new miners. Intel calls remain unchanged. Trimmed,
broadcom, a couple other names here. And Nick is giving it some
context. All these numbers are as are as a as of
of 930, many of these names sold off since then.
Portfolio value, counting notional value of options
doubled from 2.12 billion to 4.15 billion,
mostly due to 1.5 billion of new cash.
So let's give it up for new cash injection.
We got a room to come on.
Situational awareness had a half a billion dollar
new core weave position as of basically the beginning of October.
You told me like, oh yeah, like the,
The company that really is like the most index to the AI wave is down 46%.
I'd be like, wow, so this is like the total popping of the bubble, complete pop, it's over.
When the metaverse bubble popped, when, you know, when crypto bubbles pop, like, Bitcoin
trades down 50%, 60%, like, and then it's over and then you start rebuilding, right?
And yet the overall market feels nowhere near popped, right?
You would expect Nvidia to be maybe, like, you know, selling off more.
Corweave is just in a unique position.
Yeah, the question is like, is it going to be a true hedge fund?
Like, is it going to make money in a down market?
In a correction.
Blue Al investors face hefty losses as credit fund blocks exit ahead of merger.
Blue Al has blocked redemptions in one of its earliest private credit funds as it merges
with a larger vehicle overseen by the asset manager in a deal that could leave investors with
large losses.
They could lose about 20% of their holdings.
The deal underscores the risk that retail investors have taken in pouring hundreds of billions
of dollars into private debt funds carrying limited liquidity rights.
It comes as scrutiny builds on the valuations and returns on private credit funds, which
have caused publicly listed debt funds to sell off and trade its deep discounts to the stated
value of their assets.
And so we talked about this, I think, on Friday, but Blue Al has been selling off this year,
and they said, we should be performing better than everyone else.
But it feels like a little bit of the narrative might be around liquidity here.
Good luck if you're hanging out in Blue Al Capital.
Today we're announcing a new $40 billion investment in Texas through 2027 to build
cloud and AI infrastructure and support thousands of new jobs.
Yehaw.
This includes new data centers in Armstrong and Haskell County.
and a major investment to strengthen energy resilience and abundance.
We're also providing funding to more than double the projected pipeline of new Texan
electricians.
There we go.
To power the AI era.
It's the golden age of being the golden electrician age.
Yes.
You're blown around in private jets to different data centers.
Yeah, you do.
You do.
That's right.
So $40 billion investment, thousands of new jobs.
That feels like a higher ratio than what was the other example you kept quoting?
something like 500 jobs for some enthropal.
Yeah, it was the anthropic data center or something.
It was the anthropic data center.
They were like, we're investing 50 billion.
Wall Street blows past bubble worries.
Oh yeah.
To supercharge AI spending frenzy.
And they say firms such as Blue Owl Capital
have raised trillions in investing firepower.
The AI build out is a perfect match.
The warning signs are flashing.
Not long ago.
Does Blue Well have better PR or worse PR than Ares?
They seem to be quickly becoming the main name
name that everyone knows in private credit. And to my knowledge, like, they are not the only
firm in the category. They have the dot com. They have blue owl.com. Blue owl.com.
Not long ago, Blue Owl Capital was an upstart investment firm that lent money to midsize
U.S. companies such as Sarah Lee Frozen Bakery. These days, the firm is financing massive
data centers costing tens of billions of dollars for the likes of meta and Oracle, a sign of just
how quickly Wall Street has become the enabler of America's AI boom. Fund managers such as Blue Owl amassed
trillions of dollars of investing firepower and have been hunting for big deals where they can put
that money to work. They found slim pickings for years until a perfect match appeared in AI,
which has provided a bigger target than anything in history due to the vast sums tech companies
need to ramp up computing power. Does it even matter if you keep counting after you get to one
trillion of capital expenditure in the next couple of years? This is insane.
You...
Does it even matter?
You really undersold this.
Does it even matter?
While David Gatta DJ, the Blue Owl executives cut a deal to acquire IPI partners, an investment firm that owned and operated big data centers for Amazon and Microsoft.
Blue Al already had close ties with the organizer of the treat, iconic capital, which manages the personal fortunes of Silicon Valley Elite, including Zuckerberg, and was a part owner of IPI.
The purchase gave Blue Owl a seat at the table to bid on Mega AI financing.
Let's give it up for mega AI financing.
Not long after, it got picked to a range of $14 billion package for an Oracle and Open AI Data Center in Abilene, Texas.
Then last month, Blue Al raised about $30 billion to build an AI data center for META and Louisiana,
putting in $3 billion of its clients' money and borrowing the rest.
Silicon Valley's biggest players are flushed with cash and are able to fund much of the initial AI buildout from their own coffers
as the dollar figures climb ever higher.
They are turning to debt and private equity, spreading the risks and potential rewards,
broadly across the economy. Some of the financing is coming from plain vanilla corporate bond
sales, but financiers are making far bigger fees off giant private deals. Virtually every Wall
Street players angling to get a piece of the action from banks such as J.P. Morgan and Morgan
Stanley to traditional asset managers like BlackRock. Investor appetite for data center debt is
so strong that some money managers have booked billion dollar gains in a matter of days. Let's
give it up for booking billion dollar gains in a matter of days. People say there's no such thing as
free money, but kind of seems like it could be in a kind of a free money situation here.
We do know for certain is that the big tech companies that want the world to spend trillions
have huge financial incentives to be believers. If you haven't noticed, Wall Street is also being
paid a lot to promote the story. And this is the line that stood out the most to me because
like on the West Coast, you have the labs, which are effectively every single person as well as
the investors. Yes.
are incentivized to keep the current, you know, AI super cycle narrative going.
Yes.
And then on the East Coast, you have Wall Street who is getting paid to effectively do the same thing.
So you have, you know, these two centers of power that are both incentivized to keep the party going.
In that same letter, Einhorn and Greenlight said, basically the way it is today,
consumer business spends $1 on a Chad TBT subscription, which is open AI revenue.
Then Open AI provides the service by spending $2 on Microsoft AI infrastructure, which is Microsoft revenue.
Then Microsoft spends 60 cents leasing GPUs from CoreWeave to handle the compute load, which is CoreWeave revenue.
And then CoreWeave spends $2.40 on chips from Nvidia and another $2.40.
Yes, but you're completely discounting exactly how addicted the 4-0 user is.
They will pay any amount.
His final investing decision was to buy Google.
It's amazing.
I think that's beautiful.
He doesn't even need to say who he's talking about.
It's like so obvious this Warren Buffett.
Not enough people are emotionally prepared for if it's not a bubble.
It's a good post.
Is it a bubble if all the big tech companies rip and there's like a couple neoclods that trade down a little bit?
There's like one or two application layer companies that burn a bunch of VC dollars, but there's still a new hyperscalator that's born.
Like kind of.
I guess it's a bubble, but it's a survival bubble.
still crazy to me that Steve Jobs slash Apple invented the word podcast and that it's a mix of iPod and
broadcast. And apparently this isn't true. A BBC journalist Ben Hammersley coined it in a 2004 Guardian article.
But then it got ported back to Apple and the word in the term podcast was adopted.
It's always felt like a very boomer term. Yeah. And it makes sense. I never knew that it was it was a
Oh, you didn't know.
It was iPod Plus broadcast.
Yeah.
Yeah.
What else?
Who actually makes money when robots work,
looking into some of the venture funding that's flowing into humanoid robotics?
He is trying to create a field guide for separating the real companies from the Gryfters.
Talks about 1X.
The home robot costs $20,000 up front or $4.99 a month.
The website looks like every other VC-funded D-to-C brand from 20.
2015. Who actually makes money, though, when they work? Is it the call center operators? I would not
bet against Elon on this. I believe that the hardware manufacturer will ultimately be the one that
makes money in the long term. I think that everything else is more commodity in the stack, but there
is going to be a compounding advantage to actually having the manufacturing capability to build the
robots at scale. Now, this is years away, but I'm certainly AGI built in the sense that
the teleoperation will become less and less important,
and the Tesla model of having a economically producible product
will be very, very important.
Shenzhen-based company, UBtech,
claims to have completed the world's first mass delivery of humanoid robots,
and the Shenzhen-based company has secured apparently over $112 million
in Walker S2 orders this year.
Look at the reflections on this bot,
and then compare them to the ones behind.
it. The bot in front is real. Everything behind it is fake. If you see a head unit reflecting a
bunch of ceiling lights, that's a giveaway. It's CGI. He's putting UB. Tech in the truth zone,
and then Christopher comes over the top and puts him in the, and it says, such an embarrassing
post for a CEO at this level to make. Do you want a million robots in American homes that could
have, you know, the sci-fi scenario where you have a backdoor.
Yeah.
And, uh, like the argument against DJI was always like, well, you know, if some tiny drone is
in your bottom, you know, sock drawer, like, what's it going to do?
It's not just going to bust out of there.
But like, the humanoid will just bust out of the, whatever you put them in, unless you store
them and you're like, your gun safe or something, walking gun safe.
Okay.
UBETEC, industrial walkers.
They've been pumping these out.
It feels like, the question has never been.
can China make a lot of these things?
Like, obviously they can.
In the bio, they said,
they said it looked too perfect to be real.
And then they used some chat chitit slot.
But perfection isn't fabricated.
It's delicately.
No.
It's so good.
You know, for a long time,
people were saying that there was no way
to watermark AI created content with,
because like, oh, how would you do it?
You change one.
It's like, actually all,
AI content is watermarked. It turns out. It's just perfectly watermarked.
Yeah, this looks real to me. I don't know. It also could be CGI.
It's pretty funny to see Brett say this was all faked.
Yeah. It looks real. I mean, it would be way, they could have done so much better to make this look real.
Put a bunch of humans there, touch them. A lot of the hard part in CGI is the handoff between the
CGI character and the human character. And so what you should be doing,
is you should have a human who takes like a smoothie and pours it on top of the robot and then like
wipes the robot off clean with a towel and you're seeing how the fluids interact with the robot
interact with the person and like that was that was not actually that satisfactory. I don't know.
I think I don't know. I'm not calling it fake but it could be fake because CGI is really,
really good.
Like,
CGI just is at that level
where that's possible.
Anyway.
Yeah,
we know the,
we know the caption
or the description
of the video was AI generated.
That is hilarious
that they had to use
a for the caption.
It's like AI on it.
And it's like,
after all,
it is an AI company.
Like,
it would be on brand.
Thank you for tuning in.
Great show today.
Have a good evening.
We'll see you tomorrow.
