00:00:00
J what just happened did you just have a
00:00:01
moment an emotional moment I don't know
00:00:03
you know I've been trying to get over
00:00:05
Sachs and um it's hard to get over your
00:00:07
ex even when they didn't treat you well
00:00:09
we were in a codependent relationship
00:00:12
and we were working on it you have no
00:00:14
one to fight with anymore who are you
00:00:15
gonna fight with jcal that's the
00:00:17
emptiness you feel inside because you're
00:00:18
a fighter you're a street brawler and
00:00:20
you have no one to brawl with anymore so
00:00:21
you feel well you're the first guy to
00:00:23
jump under the table in a bar fight I
00:00:26
mean here you go there you go as I was
00:00:28
saying always want I feel like if if
00:00:31
goes down I feel like sax and I
00:00:33
would have like we would have thrown
00:00:35
down we Bulldogs we would have fought
00:00:37
like tooth and
00:00:39
now I have this I have this I have this
00:00:42
image it's like we're in a bar okay and
00:00:45
all of a sudden you talking to me a
00:00:47
fight is about to break out four thought
00:00:49
bubbles appear okay J's like J's like
00:00:53
let's go get him saak is like this guy
00:00:55
this guy's a dip let's roll I'm like
00:00:58
look at that chick she's so goodlooking
00:01:00
and then freeberg like what will happen
00:01:02
to my invite
00:01:05
duck free BG's
00:01:07
like how do I get out of this
00:01:13
UNSC let your winners
00:01:16
ride Rainman
00:01:20
David and in said we open source it to
00:01:23
the fans and they've just gone crazy
00:01:25
with Love
00:01:28
Queen all right everybody Welcome to the
00:01:31
number one podcast in the world with me
00:01:34
again today for better or worse your
00:01:36
Sultan of science David freedberg is
00:01:39
back how you doing
00:01:41
freeberg I see you're a pawn only a pawn
00:01:44
in their game what's going on with the
00:01:46
pawn
00:01:47
background The Imposter syndrome Pawn or
00:01:51
queen what is that is that from a movie
00:01:54
what movie is that from it's called The
00:01:55
Imposter syndrome uh
00:01:57
by just like some French wave I'm not
00:02:00
aware of how I I've learned that
00:02:02
freeberg backgrounds is some like weird
00:02:05
secret communication language there's
00:02:07
this small but fervent group of people
00:02:09
that are really into these backgrounds
00:02:11
they're always trying to figure
00:02:12
something out absolutely of course with
00:02:14
us again sham paa your chairman dictator
00:02:17
how you doing Shaman nice winner thank
00:02:19
you sir good to see you you ready
00:02:22
excited to see you I know look at this
00:02:24
three of four besties will be hitting
00:02:26
the slopes three of four besties will be
00:02:27
skiing together might be there eron
00:02:29
where are you going for holidays oh
00:02:31
actually I'm I'm uh we'll be in next
00:02:33
week what well there we go new bestie
00:02:35
and Aon are you staying in the town in
00:02:38
just at the hotel okay this is perfect
00:02:41
we're all gonna be there bro just $2,000
00:02:44
a night how many rooms you got with the
00:02:45
kids in the fam you have three boxes at
00:02:47
alltime highs take it easy what did you
00:02:50
do did you start buying Bitcoin with
00:02:51
your treasury how what do you mean boxes
00:02:53
at an alltime high are you uh doing a
00:02:55
Michael salor or something you know it
00:02:57
turns out if uh if you were uh if you
00:03:00
didn't get crushed During the postco
00:03:02
period you can just keep keep cranking
00:03:04
so that's what we've been doing oh
00:03:06
you've been building a real business of
00:03:07
course with us now our new fifth bestie
00:03:09
Aaron Levy he is the CEO of the publicly
00:03:13
traded company box which means he's got
00:03:15
the most to lose by coming here so
00:03:17
welcome back to the
00:03:19
program centrax based on what you guys
00:03:22
were just talking about before this uh
00:03:23
recording started I have no idea what
00:03:25
what uh what we're in for yes listen
00:03:28
we're living in an of meme stocks and
00:03:31
people selling convertible notes to buy
00:03:33
Bitcoin for their Treasury and and and
00:03:35
then becoming a NASDAQ 100 it's that
00:03:37
simple Aaron when you see a company buy
00:03:41
billions of dollars worth of bitcoin and
00:03:42
get added to the NASDAQ 100 what method
00:03:45
of suicide do you think of taking your
00:03:47
own life with um there was a there was a
00:03:50
brief week in uh in 2021 where uh where
00:03:53
the thought kind of crossed my mind so
00:03:56
yes um suu you're just gonna use the
00:03:58
sword the short blade we have a very uh
00:04:00
sophisticated audit committee that that
00:04:03
prevented the action so I uh will you do
00:04:06
this for me yeah just do this for me
00:04:08
take how much cash does box have on the
00:04:11
books ballpark well we just did a
00:04:13
convert so we're probably 6 700 six or
00:04:16
700 million okay here's what I want to
00:04:18
do a little experiment for next week
00:04:19
just buy put 5% of the treasury 30
00:04:22
million in Bitcoin and then you we'll
00:04:24
invite you back in two weeks we'll see
00:04:25
what happens okay just put 5% of the
00:04:27
treasury in Bitcoin hey everybody here's
00:04:30
another announcement a little
00:04:31
housekeeping as you know we successfully
00:04:33
got the allin.com domain that was a big
00:04:36
victory for us and uh we now have an
00:04:39
email list so four years into this
00:04:43
muga uh we now have the ability to take
00:04:45
your email address and spam you with all
00:04:48
kinds of great information like a
00:04:50
notification when the Pod drops wow so
00:04:54
compelling insights from the besties who
00:04:57
you've had enough of and uh first dibs
00:05:00
on overpriced event tickets to come hang
00:05:03
out with us wow you this is the
00:05:04
compelling pitch we have for giving us
00:05:06
your emails so if you'd like to give us
00:05:08
your email and get spammed go to
00:05:10
allin.com and let the spam begin all
00:05:13
right that's out of the way sax is out
00:05:15
again this week I don't know what's
00:05:18
going on we've been trying to figure out
00:05:19
where he is he's Mia if anybody knows
00:05:22
he's in DC this week yeah oh is that
00:05:24
what's going on did you see all the
00:05:27
meetings I know being fous he's he's uh
00:05:31
he's in meetings Mr saxs went to
00:05:34
Washington he creating waves but Mr saxs
00:05:37
is in Washington and have you guys seen
00:05:39
the photos Erin what do you think of
00:05:41
what do you think of saaks in this
00:05:44
role I think it's uh I think it's a
00:05:47
strong pick um so so let the genu
00:05:51
flecting begin go say more Aon so crypto
00:05:55
I'm a little bit indifferent on you know
00:05:57
I I I I haven't you know know so we
00:06:01
haven't SP spent much time there leaned
00:06:03
in much about there but but on
00:06:05
AI I think is a very very strong pick
00:06:08
and um I think you want somebody that is
00:06:13
you know has a general sort of
00:06:16
deregulation anti-regulation bent at
00:06:19
this stage in the evolution of the
00:06:20
technology I think there's risk of of
00:06:23
sort of slowing down too much progress
00:06:25
right now and I think that I think he'll
00:06:28
provide you know some good
00:06:30
kind of parameters and principles around
00:06:31
uh around how to avoid that so I think
00:06:33
very strong and then you know crypto
00:06:35
again don't know don't know too much
00:06:36
about and then we'll we'll see the rest
00:06:37
of the topics as a software CEO yeah of
00:06:40
a public company when the Biden
00:06:43
Administration was putting forward their
00:06:45
proposals on how to regulate Ai and have
00:06:48
definitions on the size of a model and
00:06:50
what the models could or shouldn't do
00:06:52
and the regulatory Authority they would
00:06:53
have over the software that's written
00:06:55
what was your reaction and you were
00:06:57
supporting Harris at the time I believe
00:06:59
or Biden at the time right but like how
00:07:01
did you react to that when you saw those
00:07:03
proposals and just to be clear are you
00:07:05
talking about the EO that that went out
00:07:08
yeah there was the EO but then they were
00:07:10
also drafting they they published a lot
00:07:12
of detailed Drafting and then obviously
00:07:13
California had its bill which you
00:07:15
probably saw as well which specified
00:07:17
like the number of parameters in a model
00:07:19
and the size of a model and yeah all
00:07:21
these kind of constraints in reverse
00:07:22
order was was uh against
00:07:26
s1047 it felt like you know you had two
00:07:28
big issues one you probably don't want
00:07:30
state-by-state legislation on this topic
00:07:32
that that's going to be you're in a
00:07:33
world for of hurt if if every state has
00:07:35
a different approach to this and then
00:07:37
second secondarily if you just look at
00:07:39
how it evolved from the very first
00:07:41
proposal to the final proposal and
00:07:43
unfortunately the kind of underlying
00:07:45
philosophy that was in the the bill it
00:07:47
was it was very clearly a sort of like
00:07:50
like viewing you know basically AI
00:07:52
progress as inherently risky right now
00:07:55
and so it just ratcheted up the the
00:07:57
different you know levels of consequence
00:08:00
for the AI model developers and and the
00:08:03
risk is is is sort of the second or
00:08:05
third order effects of that which is
00:08:06
like does Zuck then want to release the
00:08:07
next version of llama if you're taking
00:08:10
on you know that much risk and and you
00:08:12
know even the incremental updates um the
00:08:14
liability you have yeah in terms of any
00:08:17
of the model advancements and so right
00:08:19
now we're benefiting from just an
00:08:21
incredibly competitive market between
00:08:23
five or six players and you want them
00:08:25
running as fast as possible not having
00:08:27
to have sort of this you know a Council
00:08:30
before every model gets released because
00:08:32
they're they're you know in fear of
00:08:33
getting sued by you know the government
00:08:35
for hundreds of millions of dollars if
00:08:36
one person does something wrong with the
00:08:38
model so that was the problem with sb147
00:08:40
that's been the problem with some of the
00:08:42
the proposals on National legislation I
00:08:44
felt like the first CEO it didn't have a
00:08:46
lot of teeth in it so so it kind of was
00:08:47
more like let's watch you know this
00:08:49
space and continue to study it the
00:08:51
actual current head of s ostp uh Arty
00:08:54
probar a lot of folks in Silicon Valley
00:08:56
know her she's actually very strong very
00:08:58
technical you know understands the
00:09:00
valley well is not a sort of does not
00:09:03
lean into overregulation so I actually
00:09:05
think ostp has had a a pretty good
00:09:07
Steward uh even under
00:09:09
Biden but but I think you know the
00:09:12
efforts that that saaks would you know
00:09:13
clearly be leading I think would would
00:09:14
lean even more toward AI progress and
00:09:17
and and sort of not accidentally
00:09:19
over-regulating too early in uh in this
00:09:23
journey so let me ask you a question
00:09:25
then about crypto you're not into crypto
00:09:28
crypto a little bit harder to regulate
00:09:30
so with saxs being there what do we
00:09:33
think the one two or three things he
00:09:36
could do to actually make
00:09:38
crypto not a scam not have consumers get
00:09:42
left holding the bag obviously
00:09:46
sandboxing projects maybe having people
00:09:48
know your customer you know some basic
00:09:51
regulation there uh the sophisticated
00:09:53
investor test comes to mind shth what do
00:09:55
you think Sac should do in terms of
00:09:58
crypto regul in the short term in the
00:10:00
near term that's a really good question
00:10:03
I think that today there are a lot of
00:10:07
really valuable use cases that can sit
00:10:10
on top
00:10:12
of crypto rails I think the most obvious
00:10:15
one is how you can have real-time
00:10:18
payments using stable coins I think the
00:10:21
United States government is already
00:10:22
using some of these stable coins for a
00:10:25
bunch of purposes the number of
00:10:27
companies that are beginning to adopt
00:10:29
ADT and use stable coins is actually
00:10:31
growing very take a second to to find
00:10:34
stable coins for the audience just to
00:10:35
cat people up let me Define what a
00:10:37
stable coin is which is that you
00:10:40
put a dollar into a wallet somewhere and
00:10:44
in return you get a digital token of
00:10:48
that dollar there are two big purveyors
00:10:50
of stable coins there's tether and then
00:10:53
there's usdc which is this company
00:10:55
called
00:10:56
Circle I think the easiest way to to
00:10:59
disambiguate them is tether is abroad I
00:11:02
think it's run out of somewhere in the
00:11:04
Caribbean and usdc is American it's run
00:11:08
by this company called Circle the other
00:11:10
difference is that there is some
00:11:14
ambiguity around how these assets are
00:11:18
secured so what a stable coin is
00:11:20
supposed to do is when you give them a
00:11:21
dollar they're supposed to go and buy
00:11:23
some short-term treasury security so
00:11:26
that that dollar is always guaranteed
00:11:28
right because if if they if you had a
00:11:30
billion dollars of stable coins but only
00:11:33
$900 million of assets to back them up
00:11:37
there's an insolvency risk there there's
00:11:39
a run on the bank risk right so
00:11:41
theoretically a billion dollars of
00:11:42
stable coins should have a billion
00:11:44
dollars of cash in some short-term
00:11:46
fungible security what's incredible is
00:11:49
at the scale in which these stable coins
00:11:51
operate that has turned out to be an
00:11:53
enormous business why when you give them
00:11:56
a billion dollars and get a billion
00:11:58
dollars of stable coins return they just
00:12:00
go and put it into the bank and you know
00:12:01
when interest rates are 2 3 4 5% they're
00:12:05
making billions and billions of dollars
00:12:08
they get the float they get the float so
00:12:10
these businesses have turned out to be
00:12:12
incredible but that's beyond the point
00:12:14
the point is that a lot of companies
00:12:16
that you would never think so for
00:12:17
example SpaceX uses these stable coins
00:12:20
how how do you think they collect
00:12:22
payments from all the starlink customers
00:12:24
when they aggregate them in all of these
00:12:26
longtail countries they don't want to
00:12:28
necessarily take the FX risk the foreign
00:12:30
exchange risk they don't want to deal
00:12:32
with sending wires so what they do is
00:12:34
they'll swap into stable coin send it
00:12:36
back to the United States and then swap
00:12:37
back to US dollars so it's a very useful
00:12:40
utility so number one is I think we need
00:12:44
to make those rails the standard rails
00:12:47
in the United States and what that does
00:12:50
is it allows us to chip away all of this
00:12:53
decrepit infrastructure that the banks
00:12:56
use to sort of slow down and tax a
00:12:59
process that should never have been
00:13:00
taxed in so this is good competition in
00:13:02
a way Cham now the banks have somebody
00:13:06
to challenge them for money transfer and
00:13:09
money storage and it could be regulated
00:13:11
and stable but I guess question yeah but
00:13:14
that's that's the first thing but then
00:13:15
the second thing it allows you to do is
00:13:16
you you start you can see a world where
00:13:18
now you can have real competition
00:13:20
against the traditional rails
00:13:22
specifically Visa Mastercard American
00:13:24
Express because when you look at Great
00:13:26
companies like stripe I use stripe I pay
00:13:29
them
00:13:30
3% if I use stable coins from stripe I I
00:13:35
don't pay zero but I don't pay 3% it's
00:13:37
kind of somewhere in the middle if I was
00:13:39
technically Adept at implementing stable
00:13:42
coins through the entire stack of of my
00:13:44
product so I I use it for this learn
00:13:47
with me thing where we publish research
00:13:49
I would save a lot of money so yes the
00:13:52
idea that you can take that 300 basis
00:13:55
points you pay to these companies and
00:13:56
crush it to zero would be a boon to
00:14:01
Global GDP because that's 3% on many
00:14:04
tens of trillions of dollars aarin
00:14:06
you're shaking your head this is a
00:14:08
something you're experimenting with at
00:14:10
box or you're aware of or a problem that
00:14:12
you uh recognize no no just the I mean
00:14:14
the credit card rails is I mean the tax
00:14:17
on on transactions is obviously insane
00:14:19
so so the the stable coin being the
00:14:22
intermediary for that in the future
00:14:23
makes total sense if you could get
00:14:25
everybody to kind of coordinate against
00:14:26
that so yeah well and ating it would get
00:14:29
people off of tether hopefully which has
00:14:31
a checkered colorful sorted pass you can
00:14:34
go look it up but they've had many many
00:14:36
legal but I think again I think it looks
00:14:40
leave it at that yeah but In fairness to
00:14:41
them I think again both of these two
00:14:43
companies look as of today again you
00:14:46
have a jurisdictional issue difference
00:14:48
but as of today it looks like they're
00:14:50
both pegged one for one but anyways the
00:14:52
point is if if saaks can really push the
00:14:55
adoption of stable coins number one and
00:14:58
then number two is to
00:15:01
incentivize much much cheaper
00:15:03
transactional rails then I think he can
00:15:06
go to bitcoin
00:15:07
and these other more longtail crypto
00:15:10
projects off of a back of momentum
00:15:13
because these first two things I think
00:15:15
everybody will Embrace and he won't get
00:15:17
caught in a political Quagmire these
00:15:19
other things you have these opponents
00:15:21
always coming into the system and they
00:15:24
have even Bitcoin like when I said this
00:15:26
thing about encryption and and even
00:15:29
though I I thought I was very clear the
00:15:34
crypto community on the internet went
00:15:36
absolutely crazy all last week up in
00:15:38
arms yeah but the thing is like some of
00:15:41
the Maxis piled on then when they
00:15:42
actually took time to understand the
00:15:44
technicalities of what I was saying
00:15:45
other people realized I was right then
00:15:47
they were tweeting it my point
00:15:49
is that world is so religious animated
00:15:53
and energized that I think it's hard to
00:15:56
use them as the first where you find
00:15:59
prog so I would tell go to stable coins
00:16:02
then disrupt the Visa rails and then go
00:16:04
to the other stuff I would go stable
00:16:06
coins and I would go even before that
00:16:08
the accreditation test that I've been
00:16:10
talking about because the SEC has that
00:16:12
mandate people were educated Aaron they
00:16:14
could buy crypto and know they're going
00:16:15
to lose their money or know that
00:16:17
something's a fugazi a fugazi whatever
00:16:20
it
00:16:21
is yeah I mean I uh I I'm sure we want
00:16:24
to move on but I I guess the so there
00:16:26
there's a there's a a parallel universe
00:16:28
where where so no matter what like
00:16:31
obviously gendler did not get his arms
00:16:32
around this whole thing so so that that
00:16:34
was a big big mistake but there's sort
00:16:38
of like like defi Financial crypto where
00:16:43
where almost everything is is
00:16:44
deflationary it improves the rails it
00:16:47
you know if you believe in Bitcoin as
00:16:49
this you know store of value and digital
00:16:52
gold like all of those things can
00:16:53
actually kind of make sense and and be a
00:16:55
bit rational and like improve things and
00:16:57
then you know you know fortunately or
00:16:59
not depending on your views there was
00:17:01
this other sort of you know uh fra
00:17:03
fractal event that happened which was oh
00:17:05
let's also use these things as a a means
00:17:07
of of kind of creating uh a virtual you
00:17:11
know currency and and and Tok and
00:17:13
equities and and tokens for anything
00:17:16
where that then runs into basically the
00:17:19
SEC remit of like are these things
00:17:21
Securities or not and is there insider
00:17:23
trading or not and can anybody issue
00:17:25
them at any time and promote them on
00:17:26
Twitter or not and so so you know I
00:17:29
think to some extent if you could get
00:17:30
back to like crypto 1.0 which was like
00:17:32
this is a this is a financial
00:17:34
infrastructure I I think you would have
00:17:35
avoided a lot of the the sort of noise
00:17:38
and challenges with with crypto now I
00:17:40
don't know you can't put the you you
00:17:41
can't put it back in the bag but but
00:17:44
there's like I don't think you could get
00:17:45
10 crypto people to agree on how you
00:17:47
regulate that second category because
00:17:49
some people some people believe I should
00:17:51
be able to issue an nft on anything and
00:17:53
I should be able to to trade that and
00:17:56
and and at the same time they would
00:17:57
obviously they would sort of claim well
00:17:58
that's just the same as a as an
00:18:00
aftermarket you know uh seat to a
00:18:02
concert and yet another group would
00:18:05
would be treating this as as effectively
00:18:07
you know a security and so you you I
00:18:09
don't know how you're ever going to reel
00:18:10
that in without some people being upset
00:18:13
you know within the crypto Community all
00:18:15
right well more to come and saaks will
00:18:17
be back saaks will be back and we will
00:18:19
be rotating the fifth seat amongst uh
00:18:22
you know Friends of the Pod and
00:18:24
newsmakers sorry did I already did I
00:18:26
already get rotated out uh based on yeah
00:18:29
basically the energy was a little low
00:18:31
but uh you know I mean it's just I think
00:18:34
well what's your you already had to warn
00:18:36
rate what's your resting heart rate
00:18:38
versus we're boys and then we'll just
00:18:40
make a decision got it all right listen
00:18:42
doge is fully
00:18:44
operational Elon and V have helped kill
00:18:48
the last minute Omnibus spending Bill
00:18:50
Wednesday night the bill had been killed
00:18:53
and we were looking at the government
00:18:56
shutting down starting Friday December
00:18:59
20th today when you're listening to this
00:19:01
for some quick background in September
00:19:03
Congress approved a bill that would keep
00:19:04
the government funded through December
00:19:06
20th the day this episode published keep
00:19:09
that December date in mind for a second
00:19:12
this Tuesday December 17th three days
00:19:15
before the deadline leaders in Congress
00:19:17
unveiled what was presented as a
00:19:19
bipartisan stop Gat bill that would keep
00:19:21
the government funded through March 14
00:19:24
that kind of Bill is called a continuing
00:19:26
resolution basically give you more time
00:19:28
for the incoming GOP majority to reach
00:19:31
an agreement on the funding for the
00:19:33
government but there are two major
00:19:34
problems with the bill it's a rush job
00:19:36
it had to pass the house and the Senate
00:19:38
by Friday night after being presented on
00:19:40
Tuesday it's absurdly long 1500 pages
00:19:44
with $340 billion in spending including
00:19:48
pay raises and better health care
00:19:50
benefits for the members of Congress my
00:19:52
Lord read the room gentlemen and ladies
00:19:54
funding for a global engagement Center
00:19:56
for another year that's the
00:19:57
disinformation strug group that was
00:19:59
wrapped up in the Twitter files 130
00:20:01
billion to renew the farm bill for
00:20:03
another year $110 billion do in
00:20:06
hurricane disaster Aid just money being
00:20:09
thrown everywhere a billion three to
00:20:11
replace the Francis Scott Key bridge in
00:20:12
Baltimore V had some spicy comments on
00:20:15
it Congress has known about the deadline
00:20:17
since they created it in late September
00:20:19
he said the urgency is 100% manufacture
00:20:22
designed to avoid serious public debate
00:20:25
but serious public debate is exactly
00:20:27
what's happening on Twitter people are
00:20:28
screenshotting and using chat GPT and
00:20:32
Claude and and Gemini to work their way
00:20:35
through these documents and it looks
00:20:38
like this is not going to happen
00:20:40
freeberg your thoughts so the proposed
00:20:43
bill made no real change to the current
00:20:45
spending level of the federal government
00:20:47
at roughly $6.2 trillion on an
00:20:49
annualized basis which by the way is
00:20:52
roughly call it
00:20:55
23% of GDP just to give you some context
00:20:59
in 1860 nearly 100 years after the
00:21:01
founding of the United States government
00:21:04
federal spending to GDP was less than
00:21:08
1% and it took off during the Civil War
00:21:11
uh for a couple of years but you know
00:21:14
we're at these kind of like
00:21:15
unprecedented levels year after year now
00:21:18
really speaking to how the federal
00:21:20
government has grown as we talked about
00:21:22
many times so much in our life and you
00:21:24
know at our roads were really bad back
00:21:26
then though yeah our were were really
00:21:29
bad back then but but remember the
00:21:32
objective of the Republic was to have
00:21:34
the states make local decisions about
00:21:37
how to take care of their infrastructure
00:21:39
the national highway effort obviously
00:21:41
changed that in the uh the mid 20th
00:21:44
century but this was kind of the
00:21:45
original intention of the Republic it
00:21:47
wasn't to have the federal government
00:21:49
come in and employ people provide uh
00:21:52
Insurance to people provide energy
00:21:55
markets to people own football stadiums
00:21:58
etc etc etc if you go through the list
00:22:00
of things in this bill I think the
00:22:02
hundred billion dollars of natural death
00:22:04
disaster relief everyone says that seems
00:22:06
very reasonable that's something the
00:22:08
federal government should do when we
00:22:09
have a natural disaster we need help we
00:22:11
need support that's a great thing for
00:22:12
the federal government to do but think
00:22:14
about the incentive it creates it
00:22:16
distorts markets so we've talked about
00:22:18
this in the past in areas where you have
00:22:20
a higher probability of natural
00:22:22
disasters and people have paid a lot of
00:22:24
money for their homes pay a lot of money
00:22:25
for infrastructure should the federal
00:22:27
government come in in and rebuild those
00:22:29
homes and Provide Capital to those
00:22:31
individuals and those businesses to help
00:22:33
them rebuild those homes if there's a
00:22:34
high probability of natural disaster
00:22:36
events happening again in the future it
00:22:38
means that the cost of insurance doesn't
00:22:40
matter and the cost of real estate
00:22:42
doesn't matter because the federal
00:22:43
government effectively can come in and
00:22:45
support those prices in the same way the
00:22:47
federal comes in government comes in and
00:22:48
supports the prices in agricultural
00:22:50
products through the work in the farm
00:22:52
bill and through the biofuels mandates
00:22:54
which were also proposed to be extended
00:22:55
in this thing so the federal
00:22:57
government's playing both a Market role
00:23:00
and you know also kind of this role that
00:23:01
I think fills the Gap where people want
00:23:04
to have a customer where there isn't a
00:23:05
customer and an employer where there
00:23:07
isn't one so like how did we get to this
00:23:10
point so from first principal
00:23:11
perspective we've kind of I think lost
00:23:13
the narrative on what the federal
00:23:14
government was meant to do if you think
00:23:16
about the simple rubric in a bill like
00:23:19
just go back some period of time and
00:23:20
someone says hey I I want something I
00:23:22
want to have this in this bill and
00:23:24
you're the representative that's
00:23:25
supposed to vote on that bill and say
00:23:26
yes or no it's very hard to just say no
00:23:30
we are not going to spend that money
00:23:32
what's the incentive to say no the
00:23:35
alternative is you say yes but give me X
00:23:39
as well there is an incentive in that
00:23:41
response and the incentive there is to
00:23:43
get something for your electorate the
00:23:45
people that voted you in as a
00:23:47
representative of the house which is how
00:23:49
we ended up at this point where everyone
00:23:51
says I want something if you're going to
00:23:52
get something and eventually the
00:23:54
government the federal government swells
00:23:56
to spending roughly 24%
00:23:59
of our GDP now the biggest mistake I
00:24:01
think the founding fathers made was that
00:24:03
they didn't create constitutional limits
00:24:05
on spending and enrichment and this was
00:24:06
because they had these deeply held
00:24:08
philosophical beliefs that relied on the
00:24:10
House of Representatives to provide a
00:24:12
check by the people if you read The
00:24:15
Federalist Papers and I went through a
00:24:16
couple of these recently and I use chat
00:24:18
GPT to help me kind of you know bring
00:24:20
out some of I think the key points but
00:24:23
in the Federalist Papers 10 and 51 James
00:24:26
Madison emphasized that the structure of
00:24:28
govern
00:24:29
was meant to ensure that both state and
00:24:30
federal governments would limit each
00:24:32
other's excesses including their
00:24:33
financial ones and then in the
00:24:35
Federalist Paper number 58 he said the
00:24:38
House of Representatives has control
00:24:39
over the quote power of the purse which
00:24:42
gives the people's Representatives
00:24:43
authority over Taxation and spending but
00:24:46
they also warned along with Alexander
00:24:49
Hamilton of the dangers of unchecked
00:24:51
government power through burdensome
00:24:53
Taxation and excess spending which would
00:24:56
ultimately erode individual freedoms so
00:24:59
they recognized that there were going to
00:25:00
be limits but their expectation was that
00:25:02
the house and the individuals that were
00:25:04
electing people to the house that were
00:25:05
members of this Republic would come in
00:25:07
and say we're going to keep that from
00:25:09
happening and clearly something went
00:25:11
wrong along the way that we got to this
00:25:13
point where again spending is 24% of GDP
00:25:16
and I think that the biggest thing they
00:25:19
got wrong was that they didn't create
00:25:20
these constitutional limits on federal
00:25:22
spending or taxation through either a
00:25:24
balanced budget um structure spending as
00:25:26
a percent of GDP no federal debt or term
00:25:29
limits or all of these other mechanisms
00:25:31
that could have been introduced at the
00:25:32
beginning that could have created some
00:25:35
structural limits instead they assumed
00:25:37
that there was this natural limit that
00:25:38
would emerge as a function of the
00:25:40
democratic process because of how they
00:25:42
formed the government but unfortunately
00:25:44
I think they failed to realize that the
00:25:47
electorate would eventually not want the
00:25:50
freedoms of the people of the time back
00:25:54
then in 1776 this was a pioneering
00:25:57
country where everyone wanted to come
00:25:59
here to have freedom to do anything they
00:26:01
wanted everywhere they wanted to build a
00:26:03
business to Homestead to be a rugged
00:26:06
individuals it was entrepreneurial yeah
00:26:08
it was and over the last 250 years we've
00:26:11
gotten used to an increment in lifestyle
00:26:13
every year and discovered that we have a
00:26:16
mechanism to force the increment in
00:26:17
lifestyle through the actions of
00:26:19
government and so the electorate has
00:26:21
stood up and said I want more each year
00:26:24
and I want the government to provide it
00:26:25
for me if the free markets are not doing
00:26:27
it and that's really where we kind of
00:26:29
got to this point where I think we
00:26:30
elected people to the house who
00:26:32
ultimately had this incentive that said
00:26:34
if I give you this I need to get this
00:26:36
and we ended up swelling this so I I
00:26:38
don't know if it cracks with Doge I
00:26:40
really don't know if people step up and
00:26:42
recognize the limits of of government
00:26:44
and what the limits should be of the
00:26:46
federal government on an electorate
00:26:47
basis it's an amazing moment to see that
00:26:50
Elon went on Twitter and said hey guys
00:26:51
this is nuts and everyone said this is
00:26:53
nuts we're not going to elect you if you
00:26:54
do this if that momentum and that
00:26:56
transparency can keep up I I hope that
00:26:58
people start to connect the dots that
00:27:00
this isn't a free lunch that the federal
00:27:01
government spending is not Limitless and
00:27:04
it's not unaccountable Aon I think we
00:27:06
have enough people who are notable now
00:27:11
speaking up about this excess spending
00:27:15
and the outof control debt that it's now
00:27:18
in Vogue to talk about austerity to talk
00:27:21
about inefficiency and that really all
00:27:24
comes back to Doge and dare I say you
00:27:26
know the conversations we've been having
00:27:28
this podcast for 2 years that this is
00:27:29
becoming acute what are your thoughts on
00:27:31
this Vibe shift this complete pivot
00:27:34
where we've gone from my Lord everybody
00:27:36
saying I gotta get mine you got yours
00:27:38
I'm getting mine to name and shame
00:27:42
they're naming and shaming now very
00:27:45
specific pieces of pork in these bills
00:27:48
you know including stadiums for n the
00:27:50
NFL and people are like why is the NFL
00:27:52
getting this if they were 20 3040
00:27:54
billion doll two just quick thoughts one
00:27:57
Patrick Carson had a at a tweet
00:27:58
yesterday that basically said this sort
00:28:00
of this this big
00:28:02
misinformation kind of created by by
00:28:04
people that want to be slow is that you
00:28:06
you you have to sort of choose two of
00:28:08
fast good and cheap and and I think
00:28:12
basically you know elon's companies have
00:28:14
sort of always effectively kind of
00:28:17
proven the opposite which is which is
00:28:19
actually if you just like start to ask
00:28:20
the question like why does that thing
00:28:22
have to cost as much you know if you're
00:28:24
building a rocket or or designing a car
00:28:27
or developing batteries like why you
00:28:28
know if you just do ground up why does
00:28:30
it have to cost as much and so so what's
00:28:31
interesting is is that that probably if
00:28:34
most people looked at what the
00:28:35
government was spending on they wouldn't
00:28:37
even feel like like you know it's not
00:28:39
even helping them in like the disaster
00:28:40
relief sense of of you know I think like
00:28:43
that they're probably actually people
00:28:44
that actually do experience the benefits
00:28:46
of disaster relief it's actually just
00:28:48
all of the all of the overhead that
00:28:51
we've created to getting anything done
00:28:53
in the government that could actually
00:28:54
make the government better serve the all
00:28:56
all the constituents I was talking to
00:28:58
you know sort of a nameless individual
00:29:00
in the government the other uh the other
00:29:02
week where by Congress they have to hire
00:29:05
contractors to do work and the
00:29:07
contractors the Contracting firms charge
00:29:09
them two and a half times the the the
00:29:12
sort of cost of of an individual
00:29:14
employee that they could otherwise hire
00:29:15
and so and so now now they have to
00:29:17
Outsource the work and they don't have U
00:29:20
any accountability mechanism for that
00:29:22
contractor and so I think there's not a
00:29:24
single American that could look at that
00:29:25
and say and say this is like actually
00:29:27
working well like yeah we we are
00:29:30
spending more money to do less and the
00:29:33
ultimate outcome is actually lower
00:29:35
quality and so so you have to at some
00:29:38
point just kind of do a little bit of a
00:29:39
reset moment and that's obviously the
00:29:41
the upside of Doge is like it's like it
00:29:43
breaks every rule of us thinking about
00:29:45
how how you would actually go and attack
00:29:47
this problem we thought you'd attack it
00:29:48
through meetings and and we would do it
00:29:51
through through Congressional you know
00:29:52
sort of processes and and research and
00:29:55
it's actually just it is you know
00:29:57
obviously a much more sort of founder
00:29:59
startup oriented way to approach this
00:30:01
there's going to be lots of things that
00:30:02
are broken glass around the edges there
00:30:05
there's no question but I think what's
00:30:06
interesting about this week's event is I
00:30:08
think that there's been this underlying
00:30:10
kind of notion of like you know Elon and
00:30:13
and and whatnot at all you know don't
00:30:15
understand the the government enough to
00:30:17
be able to change it and it might
00:30:18
actually be the case that the government
00:30:20
doesn't understand Elon in the sense of
00:30:22
of like he will just see this thing
00:30:25
through and the tools at his disposal
00:30:28
and Doge's disposal is is sort of you
00:30:30
know completely unprecedented in terms
00:30:31
of the ability to put any anybody in
00:30:34
congress on notice if you know if
00:30:36
basically they are promoting things that
00:30:38
that are not making the country better
00:30:40
so so the the the you know the thing
00:30:42
that we saw this week was actually that
00:30:44
playing out everybody's been wondering
00:30:45
well what are the actual you know what
00:30:47
are the formal mechanisms Doge has to
00:30:49
accomplish an enact change and it's like
00:30:51
you just saw it like they can just they
00:30:53
can just create enough visibility and
00:30:56
Spotlight on the problem that it it
00:30:58
causes it a level of discomfort in in
00:31:01
supporting moving forward with with
00:31:02
whatever that thing is and so I think
00:31:04
it's interesting I have no opinion on
00:31:06
the actual elements of the bill other
00:31:08
than from a a process standpoint and a
00:31:10
and a new kind of case study of of how
00:31:13
this is going to play out is I think
00:31:14
we're seeing some early indication of
00:31:16
what Doge will will be able to do Aon
00:31:18
how do you feel about this Doge
00:31:21
effort as someone who is a public Harris
00:31:24
supporter so you you've come out I think
00:31:26
you've you've been public about this I
00:31:27
want to understand like why
00:31:30
people wouldn't be supportive of this
00:31:32
effort right like like what is like what
00:31:34
is the motivation for people saying this
00:31:36
isn't a good thing we shouldn't be doing
00:31:38
this like because there's a lot of folks
00:31:40
that have gone on these shows it's like
00:31:42
they they blast Elon they blast the V
00:31:45
but like aren't these principles like
00:31:46
shouldn't they just be Universal that we
00:31:48
should not be wasting money and stuff
00:31:51
sure of course the now I mean so to give
00:31:54
some credit you know you have rokana
00:31:56
supporting it you have fedman supporting
00:31:58
it Bernie Sanders you know everybody
00:32:00
Bernie sand yeah Bernie Sanders had a
00:32:02
good call out for El you have everybody
00:32:04
has their thing in the government budget
00:32:05
that they don't like so assuming that
00:32:07
they see that as something Doge can
00:32:08
contribute to you could probably get
00:32:10
actually broad support you know there's
00:32:12
a classic you know sort of reflex within
00:32:15
within probably Democrat Party On on at
00:32:17
this point just because of elon's
00:32:18
support of trump that that if something
00:32:21
is a is an Elon project they're gonna
00:32:23
they're are gonna instantly respond no
00:32:25
matter what the thing is as a negative
00:32:28
without you know kind of actually saying
00:32:30
is this does this actually support
00:32:31
actually something I do agree with and
00:32:33
so that it's all partisan none of its
00:32:35
first principles right for these people
00:32:38
for this group of people which isn't
00:32:40
everybody that's gonna be true of both
00:32:41
sides like Michelle Obama Michelle Obama
00:32:44
was like let's get kids healthy and all
00:32:45
of a sudden now it's in Vogue to do that
00:32:48
so so I I think I think we're just in an
00:32:50
environment where where any anything
00:32:52
will become partisan what's interesting
00:32:54
is is that because of the you know some
00:32:56
of the the cross party elements of of of
00:32:59
trump and now's cabinet is it might pull
00:33:01
in more more of the Democrat Party than
00:33:04
than would usually happen and and I
00:33:06
think because of Elon and the people
00:33:08
that are s surrounding Trump you
00:33:10
probably have a bit more air cover for
00:33:11
the ranas of the world to also step up
00:33:14
because if it was like Steve Bannon and
00:33:16
Trump doing Doge it would be like ah
00:33:18
okay you know maybe this is not the
00:33:19
thing to to you know lend credibility to
00:33:21
for for Pure political reasons if you
00:33:24
have you know some of the best
00:33:25
entrepreneurs that that are out there
00:33:26
actually like literally in the cabinet
00:33:28
driving this at some point you know it's
00:33:31
it is an IQ test if you're if you're on
00:33:33
board or not shabath this is a c change
00:33:36
that we're seeing and to Aaron's point
00:33:39
this Administration gets to pick their
00:33:41
priorities but everything can't be a
00:33:42
priority or it's not a priority
00:33:44
therefore they've picked the priority of
00:33:46
smaller government more controlled
00:33:47
spending over say Mass deportation or
00:33:51
removing more rights from women to
00:33:53
choose when they have an abortion Etc so
00:33:56
this seems like a distinctly different
00:33:58
Focus for Trump 2.0 the second term what
00:34:01
are your thoughts on what we saw this
00:34:03
past 72 hours relation to this bill it
00:34:06
was the most
00:34:09
incredible change in how politics will
00:34:12
be done going forward I think that
00:34:14
people are probably underestimating what
00:34:18
happened here this was a
00:34:21
multi
00:34:23
billion grift that was stopped on a dime
00:34:27
over over 12 hours of
00:34:29
tweets you would have never thought that
00:34:32
that was possible the point is to put a
00:34:35
dagger in something that big that had so
00:34:39
much broad support just a few hours
00:34:42
earlier I think it's so consequential in
00:34:45
how the United States can run going
00:34:47
forward so building on what you said
00:34:49
shath also very interesting here is the
00:34:52
fact that Trump hasn't taken office yet
00:34:55
and they're having this huge impact this
00:34:57
is a current a month before he's even in
00:35:00
office if they can stop this what will
00:35:02
they do when they actually are in power
00:35:04
I think right now you're seeing the
00:35:06
first order reaction which is about the
00:35:09
bill itself and I think that misses the
00:35:11
point the bigger issue is going forward
00:35:14
you will have the ability to and part of
00:35:18
it is because we have a set of tools now
00:35:20
that allows us to do this to read 1500
00:35:23
pages in a matter of minutes to
00:35:25
summarize it into the key essential
00:35:27
elements to really understand what's
00:35:29
being asked and what's being
00:35:32
offered and then to put it in a
00:35:34
digestible format that normal people can
00:35:37
consume yes then all you'll have to do
00:35:39
is just connect the dots and tell your
00:35:41
Congressman or congresswoman that you'd
00:35:43
like or dislike this thing and what
00:35:45
you're going to see is a much more
00:35:46
active form of government and I think
00:35:48
that's the really big deal the fact that
00:35:51
it's it really becomes the voice of the
00:35:53
people now the alternative can also
00:35:56
happen imagine there's a piece of
00:35:59
legislation that is very controversial
00:36:02
but it turns out that people actually
00:36:04
want it then the opposite will also
00:36:07
happen and can also happen now so I
00:36:09
think that it's a very nuanced and
00:36:11
interesting way that governance can
00:36:12
happen the other thing that I'll say
00:36:14
though which is funny is we should have
00:36:15
a segment called today in hypocrisy and
00:36:19
if I was running the segment what I
00:36:20
would say is today in
00:36:22
hypocrisy you have a group of people I.E
00:36:25
the Democrats who are very upset and who
00:36:29
now point to Elon as sort of like some
00:36:33
Shadow cabinet minister or some Shadow
00:36:35
President elect or some Shadow first
00:36:37
buddy right first buddy I love that
00:36:41
except then I thought to myself well
00:36:42
hold on a second like there was like
00:36:45
some something UNT happening in the
00:36:47
shadows and I thought well actually this
00:36:48
is the exact opposite the guy is
00:36:51
tweeting in real time his stream of
00:36:52
Consciousness you absolutely know
00:36:55
everything that he wants because he just
00:36:57
lays it all bare and at the same time I
00:37:00
thought it was really interesting the
00:37:01
same people who were saying that were
00:37:03
the ones that finally admitted that they
00:37:05
were hiding Biden for the last two years
00:37:08
and I thought did we just miss this like
00:37:10
the same people that are like hold on
00:37:12
Elon I don't like the fact you're
00:37:13
telling us what you actually want on
00:37:15
Twitter transparently while we hide our
00:37:18
president of the United States in a box
00:37:22
well yeah so you're referring to a Wall
00:37:23
Street uh Journal story that broke uh
00:37:26
think this week in hypocrisy Jason take
00:37:27
it away yeah I mean it was just as we
00:37:31
said on this pod we knew that they were
00:37:33
hiding him now the cover up is worse
00:37:36
than the crime and the coverup is a
00:37:37
cover up they were not letting him take
00:37:40
meetings they were limiting access to
00:37:42
him Dean B Phillips came on this program
00:37:45
shout out to him congratulations the
00:37:46
great run you know he he just told the
00:37:49
truth here he's not up for it he's Su
00:37:52
Downing you know terrible situation
00:37:55
people get old and people
00:37:58
have cognitive decline the end question
00:38:00
hold on a second so just to contrast and
00:38:03
compare Trump is not even in office you
00:38:05
know Elon is not a member of the cabinet
00:38:08
per se these
00:38:10
are effectively today still private
00:38:12
citizens when there's all of this noise
00:38:14
about what happened over the last two
00:38:16
days to stop an absolutely ridiculous
00:38:18
pork barrel bill but when are we going
00:38:20
to double click into what decisions have
00:38:22
been made in the last two years that
00:38:23
were actually bidens versus surrogates
00:38:26
that just decided and who gave them the
00:38:28
authority to make those decisions if
00:38:30
they're going to do Aaron and you and I
00:38:32
are left leaning moderates I think that
00:38:34
would be the most accurate way to
00:38:36
describe us I mean if they want to do an
00:38:38
investigation there should be an
00:38:40
investigation to did they know that he
00:38:42
was in massive cognitive client and let
00:38:44
him stay in charge of the nuclear codes
00:38:47
do you agree or
00:38:49
disagree this is uh this is sort of not
00:38:52
not the part of politics I I think about
00:38:54
as much so I'll I'll leave that up to
00:38:55
you as the other left leing moderate so
00:38:58
yeah no I just wonder if like this is a
00:39:00
crime that they like what if it turns
00:39:02
out he actually has Parkinson's or
00:39:04
Alzheimer's like a diagnosis it's not
00:39:07
out of the realm of possibility that
00:39:09
they covered up an Alzheimer diagnosis
00:39:12
chth and if they did is that
00:39:15
criminal I mean it's unethical it's
00:39:18
deeply unethical yeah it's very
00:39:21
dangerous Yes again it just goes back
00:39:24
to the people elected Joe Biden he won
00:39:27
fair and square he ran on a specific
00:39:30
mandate that the people
00:39:32
endorsed I just think it's devious
00:39:34
though that certain figures in that
00:39:37
white house took a level of power and
00:39:40
decision-making Authority that they were
00:39:42
never entitled to if if they wanted that
00:39:45
they should have run and they should
00:39:46
have been elected that's what we all
00:39:48
sign up for and so I think that that
00:39:51
idea that we let that happen or that
00:39:53
that happened to the American voting
00:39:55
public is I think very unfair
00:39:59
that's why I think you have to realize
00:40:01
that and you've said this before we need
00:40:03
these checks and balances going forward
00:40:06
and I think the way that you have these
00:40:07
checks and balances again is Veer
00:40:09
towards transparency the more
00:40:11
transparency there is and again this is
00:40:13
where I'll say you may or you may or may
00:40:16
not like Donald Trump but the one thing
00:40:18
you will never have to be worried about
00:40:21
is whether you will not be able to hear
00:40:23
from him in first person you're GNA hear
00:40:26
from him
00:40:28
that is absolutely true I mean simple
00:40:30
suggestion here chth if you're ever in
00:40:32
doubt you will be able to know very
00:40:34
quickly where he stands on whatever
00:40:37
topic is important to you and I think
00:40:39
that that idea is very important because
00:40:41
then if it's filtered through somebody
00:40:43
else because of some health issue that's
00:40:45
then covered up we're making decisions
00:40:48
that impact the entire world we're
00:40:50
making decisions that impacts the
00:40:52
economy we're making decisions that
00:40:53
touch hundreds of millions of American
00:40:56
Lives who are making those decisions
00:40:59
yeah it's kind of crazy I I have a
00:41:00
simple suggestion here with these bills
00:41:01
by the way when they're 1500 pages how
00:41:03
about for every hundred Pages you
00:41:05
release you have one day of review so if
00:41:07
you want to release 1500 pages 15
00:41:09
business days
00:41:11
review does that sound reasonable maybe
00:41:13
three weeks and then just stop doing
00:41:14
1500 at a time break these things up
00:41:17
into two or 300 Pages at a time I just
00:41:20
love the fact that people are motivated
00:41:22
and they have the will and the desire to
00:41:25
focus on these focus and desire because
00:41:29
you people have to have the will and
00:41:30
people did not have the will to get in
00:41:32
the weeds and examine the spending and
00:41:34
now it's becoming like invogue it's
00:41:37
becoming a Pastime to question the
00:41:39
spending this is a really great moment
00:41:41
in time for America the other thing
00:41:43
Jason that we haven't done is I think
00:41:45
that killing the bill was step one the
00:41:48
thing that America has not yet seen and
00:41:50
I think Aon brought this up with the
00:41:52
Patrick Collison tweet which is just
00:41:54
excellent Nick if you just want to put
00:41:55
it up there it's it it really is true
00:41:58
America still believes that the more you
00:42:01
spend the more you get we do at the core
00:42:05
that's why there are 1500 pages of
00:42:07
spending in here because people want
00:42:09
things because they want things to be
00:42:11
better what we need to train people to
00:42:14
understand is actually that it's the
00:42:16
lies that are told that make you think
00:42:19
that with more money comes a better
00:42:21
outcome and we see every day in Silicon
00:42:24
Valley we all start with next to nothing
00:42:26
as a startup and we outmaneuver and we
00:42:29
out execute companies all day long with
00:42:31
way more resources so it has nothing to
00:42:34
do with the resources contraint leads to
00:42:36
great art Aaron your thoughts no I mean
00:42:38
I I can't add that much more to this I
00:42:40
think it's there there's probably a
00:42:41
little bit of a a disconnected times
00:42:43
from the let's say the the voting public
00:42:46
and and you know broad constituents from
00:42:49
then those that have sort of seen this
00:42:50
in real life being in inside of a
00:42:52
company having to you know do a startup
00:42:54
and scaling up and and just this
00:42:57
um the the perverse incentives to build
00:43:00
bigger teams spend more your project
00:43:02
then is more important the more dollars
00:43:04
it gets we have all of these systems in
00:43:07
place which is the stuff that gets
00:43:08
attention are the things that you spent
00:43:10
more on so you have all these weird
00:43:12
incentives to actually have your thing
00:43:14
literally cost more to have you know
00:43:17
more overhead because you've brought in
00:43:19
more contractors into the project that
00:43:21
then you know you're going to get some
00:43:22
you know kind of future benefit from in
00:43:24
some way so you you have a lot that that
00:43:27
is sort of fully broken in this and um
00:43:29
and there's no there's there's you know
00:43:32
it's hard to imagine any other way to
00:43:34
Veer off from that path other than
00:43:35
something that does shake things up you
00:43:37
know as much as doge is doing all right
00:43:39
listen we can't do any worse than being
00:43:42
massively in debt so just let's have a
00:43:44
culture of yes we can no yes we can you
00:43:46
could have you could have Runway de yeah
00:43:49
I think we already have that I mean we
00:43:50
this is for for people who maybe are
00:43:53
rooting against Trump in the audience or
00:43:55
rooting against this because they're
00:43:56
super part
00:43:57
all I'll say is I know these individuals
00:43:59
around Trump root for them and root for
00:44:02
this process please because this is a
00:44:04
path to fixing the most acute
00:44:07
existential problem we have which is our
00:44:08
debt you don't have to like Trump to
00:44:11
like Elon to like V and to like these
00:44:14
other individuals sacks included there
00:44:16
are great people who are being called to
00:44:18
serve let's judge them based on their
00:44:21
performance that's all I ask for the
00:44:23
people who hate Trump who hate these
00:44:26
indiv ual judge them on their
00:44:28
performance they've come out with a bold
00:44:29
plan let them cook once they've cooked
00:44:33
then judge their results that is what
00:44:35
I'm telling everybody who hates Trump
00:44:38
and who hates this Administration and
00:44:39
who's partisan on the other side let him
00:44:41
cook judge the results oh I'll just
00:44:43
throw out one more one more thing in
00:44:44
this because I think doge is The
00:44:46
Branding of Doge is is often the
00:44:49
efficiency side which people always go
00:44:51
to the the spend side but the corollary
00:44:53
to that is the regulations that that
00:44:56
obviously are expens expensive to
00:44:57
maintain and and that's what creates
00:45:00
layers and layers of overhead on
00:45:01
reviewing everything that's coming in
00:45:03
then to the government and um and you
00:45:06
know unfortunately we have great
00:45:07
examples in California where where
00:45:10
literally we spend more to do less and
00:45:13
and it's because we've ratcheted up
00:45:14
these layers and layers of Regulation
00:45:16
and I have friends literally doing
00:45:18
climate Tech in in like climate Tech you
00:45:22
couldn't imagine something more probably
00:45:23
left leaning Democrat that they can't
00:45:25
actually get things done in California
00:45:27
the state that that you would imagine to
00:45:29
be the most kind of climate first you
00:45:30
know friendly State because of the
00:45:32
amount of Regulation that prevents them
00:45:34
from getting things done so so you know
00:45:37
there's there's actually this like you
00:45:39
know total combination of actually a
00:45:41
fewer regulations you'll spend less
00:45:43
money in government you'll actually grow
00:45:45
the economy faster which will create
00:45:47
more jobs like all of the things get
00:45:49
solved the more efficient you get you
00:45:52
know kind of large on all of the topics
00:45:54
so this is spending great Point Aaron
00:45:57
regulations next let's see what they do
00:45:59
there I was in the room when Antonio
00:46:01
graes was doing zero based budgeting at
00:46:05
Twitter now X and sax and I were looking
00:46:08
at roles for people what the what this
00:46:11
team can do in terms of making things
00:46:13
more efficient it's amazing what can
00:46:16
happen when you do zerob based budgeting
00:46:18
and when you just think from first
00:46:20
principles well do we even need to do
00:46:22
this does this need to exist take all
00:46:24
these regulations put a 20e clock on
00:46:27
half of them a 10year clock on the other
00:46:29
half can can I just give one one more
00:46:30
random example feel free to edit it out
00:46:32
chath you like this because it came out
00:46:33
of meta um uh do you have you followed
00:46:35
the Z standard compression library from
00:46:38
meta so this open source Library kind of
00:46:41
a you NextGen compression on on data and
00:46:44
we finally it took us you know probably
00:46:45
too long but but the team worked
00:46:47
insanely hard implemented this
00:46:48
compression algorithm we literally our
00:46:50
uploads and downloads got faster we
00:46:53
spend less money on on networking and
00:46:55
compute and it you know took some
00:46:57
re-engineering of the system but like it
00:46:59
like that's just like not a concept that
00:47:01
that people go into problem solving with
00:47:02
which is like what if the thing actually
00:47:04
was cheaper to run and it was better and
00:47:07
so think about all the systems of
00:47:08
government that could be that could just
00:47:09
be upgraded and then you would spend
00:47:11
less money actually maintaining them I
00:47:13
mean we spend you know hundreds of
00:47:15
billions of dollars way way too much on
00:47:16
Legacy infrastructure technology we
00:47:18
could automate more you could spend way
00:47:20
less money and then get better outcomes
00:47:22
so this is just happening everywhere and
00:47:24
and I don't think people realize the the
00:47:25
scale of the
00:47:27
so and I'll just give an example when we
00:47:29
went into Twitter nobody was coming to
00:47:31
the office there were people who hadn't
00:47:33
committed who hadn't come to the office
00:47:35
who hadn't committed code in six months
00:47:38
so what were they doing right so you
00:47:39
start looking at the data then they were
00:47:41
spending an enormous amount of money on
00:47:43
SAS software that nobody had logged into
00:47:46
and then they had desk software to Route
00:47:49
people around that was costing $10 per
00:47:51
day per desk per you know location
00:47:54
whatever nobody was coming to the office
00:47:56
but they were paying for software to
00:47:58
Route people to the right desk in office
00:48:00
suting it the waste when I tell you the
00:48:03
waste and the Griff and I'll just call
00:48:05
it what my interpretation is stealing
00:48:08
they were stealing from those
00:48:09
shareholders the government is stealing
00:48:11
from taxpayers it has to be fixed let
00:48:14
our boys cook freeberg you get the final
00:48:17
word before we go on to conspiracy
00:48:19
corner you got nothing all right let's
00:48:21
go straight to conspiracy Corner
00:48:22
everybody put your tin foil hats on
00:48:25
there were drones over New Jersey
00:48:27
Thursday morning the FAA ban drones this
00:48:30
is breaking news in parts of New Jersey
00:48:32
until January 17th due to quote unquote
00:48:36
breaking news special security reasons
00:48:38
they also warn that the government May
00:48:40
respond with deadly force Against drones
00:48:43
posing a threat this thing is getting
00:48:45
Crazier by the day there have been
00:48:47
thousands of reported drone sightings in
00:48:48
New Jersey and bordering states over the
00:48:51
last week here's some examples play the
00:48:53
clip yada yada until Thursday morning
00:48:56
why house officials have been dismissive
00:48:58
saying repeatedly that nothing
00:49:00
significant is going on one of the
00:49:02
theories was that there was a that the
00:49:04
drones were looking for nuclear material
00:49:07
AKA a dirty bomb or lost radioactive
00:49:10
material or the the ultimate a lost
00:49:14
nuclear bomb from like an actual nuclear
00:49:17
warhead from Ukraine on Tuesday Morning
00:49:20
the mayor of Belleville New Jersey
00:49:22
suggested the drones could be looking
00:49:23
for that radioactive material that went
00:49:25
missing on December 2nd that was
00:49:27
radioactive material uh geranium 68
00:49:30
we'll get details on that from freeberg
00:49:32
in a moment and then of course
00:49:35
conspiracy theories are going wild it's
00:49:36
Iran it's UFOs and my favorite Project
00:49:40
Blue beam which is that NASA is using
00:49:42
Holograms and other technology to create
00:49:44
a new world order and religion and
00:49:46
projecting Jesus into the clouds you can
00:49:49
look that stuff up or we'll have Alex
00:49:51
Jones on uh in Sax's seat one week okay
00:49:53
freeberg you're the genius here tell us
00:49:55
what's going on I think there are three
00:49:58
likely explanations the first is that
00:50:00
the government's got some activities
00:50:02
that can't be disclosed so we don't know
00:50:03
and they can't talk about it no one can
00:50:05
neither you know say yes or say no to it
00:50:07
so you know that's kind of one that
00:50:09
that's pretty possible the second is
00:50:11
these are just individuals with a bunch
00:50:13
of drones messing around having fun
00:50:15
trying to wreak havoc could be fun bunch
00:50:17
of kids I think we've all been there the
00:50:20
third is what I would call kind of a bit
00:50:21
of a more nefarious like this is my
00:50:23
conspiracy theory I think it it could be
00:50:27
considered a scop okay
00:50:30
so right now the US has a significant
00:50:35
regulatory burden on on drone
00:50:37
utilization in a commercial setting and
00:50:40
it's very hard to use drones you have to
00:50:42
have lineup sight to the operator these
00:50:43
things aren't supposed to go on their
00:50:44
own there's all these rules and
00:50:45
restrictions and so on and so forth
00:50:47
meanwhile you've got countries like
00:50:49
China rocketing ahead so I don't know if
00:50:51
you guys know the company mwan in China
00:50:53
you know the food delivery company have
00:50:55
you guys did you know that they do a
00:50:57
large amount of drone delivery of their
00:50:59
food now was not aware of that yeah
00:51:00
we're testing that here in the US the
00:51:02
drone delivery business in China is
00:51:04
already $30 billion a year and they're
00:51:07
also launching a pretty significant
00:51:09
Fleet of what we would call kind of the
00:51:11
eeve talls or flying
00:51:13
cars the expectation is that by 2030
00:51:16
there'll be 100,000 flying cars moving
00:51:18
people around in China and these are
00:51:20
huge efficiency gains in fact with mwan
00:51:22
you can now order food while you're on
00:51:24
the Great Wall at one of the ramp parts
00:51:25
and it'll bring the Drone will bring the
00:51:26
food to you while you're walking the
00:51:28
Great Wall as a
00:51:29
tourist and in the US the reason that
00:51:32
these things haven't taken off and the
00:51:34
reason we don't have a large kind of
00:51:35
drone industry which is clearly emerging
00:51:37
is going to be a huge economic driver
00:51:39
for China and others around the world is
00:51:41
simply the regulatory restrictions and
00:51:43
so if you were going to try and mess
00:51:46
with the US's ability to move forward
00:51:49
with the Drone economy you would
00:51:51
probably try and wreak some Havoc Stoke
00:51:53
some fear and get people to say hey this
00:51:57
doesn't seem cool what's going on I
00:51:58
don't like that there's all these drones
00:52:00
in the sky I'm freaking out and try and
00:52:02
get The Regulators to come in and say
00:52:04
hey we're Banning drones and set up
00:52:07
everyone including the people in the
00:52:09
government to say we should take a beat
00:52:11
we should think a little bit before we
00:52:13
deregulate we should who would do this
00:52:14
who's the motivation Uber Eats and
00:52:16
Postmates no no no no it could be the
00:52:20
the other government uh that's my scop
00:52:22
point this could be you're saying this
00:52:24
could be China doing this to try to slow
00:52:25
down our economy
00:52:27
think about it if you're going to pass a
00:52:28
bill in Congress to make drones more
00:52:31
freely rable in the skies you're going
00:52:33
to reference back this crazy story in
00:52:35
New Jersey everyone's freaking out about
00:52:36
it and you're G to say hey wait what
00:52:38
about all that stuff that happened in
00:52:39
New Jersey maybe this doesn't make as
00:52:41
much people are kind of scared about it
00:52:42
we shouldn't rush we shouldn't rush we
00:52:44
shouldn't rush that would be my scop
00:52:46
theory that's my you know kind of
00:52:48
conspiracy theory tinfoil hat I don't
00:52:50
often do them but that's what I would
00:52:52
kind of think about I think the first be
00:52:54
Alex Jones would be proud Aaron uh why
00:52:55
don't you jump fence and tell us your
00:52:57
best conspiracy theory of this no I this
00:53:00
is that was all crazy pills what I just
00:53:02
heard uh there yeah take some I think
00:53:06
there's like there's like 10 higher
00:53:07
scops you would do if you wanted to to
00:53:09
get us to you know have a collapsing
00:53:11
economy than going after drone
00:53:12
deliveries but um no I what would they
00:53:15
be what would they be I mean well first
00:53:17
of all I think I think you'd have ai
00:53:19
like like do a robot I think you'd have
00:53:22
a robot AI thing that goes you know runs
00:53:24
a muck um robot RoboCop yeah yeah I I
00:53:28
think that would be way sooner than
00:53:29
you're worry about food delivery oh you
00:53:30
have 10 self-driving cars hop the you
00:53:33
know hop the curb and and crash into a
00:53:35
storefront self- driving is on ice for
00:53:37
two years that would be uh you know an
00:53:39
example shamat your thoughts you got
00:53:41
some conspiracies here what do you
00:53:43
think's going on
00:53:44
here no but I thought that the most
00:53:48
credible thing was that they were they
00:53:50
were looking for radioactive material
00:53:53
that somehow some got lost and why would
00:53:56
they only look at night actually it's
00:53:59
interesting you say that there was
00:54:01
somebody on Twitter X who claims to be
00:54:04
an expert in this field and there's a
00:54:05
startup actually that I just talked
00:54:07
about which one kakoa the great I think
00:54:09
it might have been King KOA the great
00:54:10
shout out to King KOA the great and his
00:54:11
mom's basement eating why do you hate
00:54:13
Kena he's don't hate Kena I don't
00:54:16
hate Tred to get me bad because he said
00:54:18
I was trying to dox him you always can
00:54:21
kcoa he's great no I just think it's
00:54:23
hilarious that people are retraining K
00:54:25
KOA the great as if he's like this great
00:54:27
journalist and he's in his mom's
00:54:29
basement eating hot pockets or he's
00:54:31
working for Putin he's a really good I
00:54:34
don't know do I want my jcal or kacoa I
00:54:37
don't know I don't know who do you take
00:54:39
I mean your is pretty good likea I
00:54:43
subscribe to kencoa yeah Aaron Ley you
00:54:45
have a favorite who are you are you
00:54:46
autism capital or kcoa what do you oh I
00:54:48
like autism Capital too I think he's
00:54:50
absolutely I'm I'm liking Giger these
00:54:52
days oh you're into Giger capital g
00:54:55
capital
00:54:56
I read all PCS I think the nighttime
00:54:59
thing would be it would be at least
00:55:01
typical of this government to do a
00:55:02
striesand effect of of just like maybe
00:55:06
if we cover it up nobody will see and
00:55:07
then obviously it's the biggest thing
00:55:12
so but if they there is a startup that
00:55:16
makes I we really looked it up and there
00:55:19
is a startup that makes drones to do
00:55:22
this specific task to look for dirty
00:55:24
bombs in ports and ports obviously look
00:55:26
for these things it's a known threat
00:55:29
this is not rocket science sorry and why
00:55:31
and why only at night did free because
00:55:33
that they can read the signatures better
00:55:35
was the theory that at night you can
00:55:37
read the signature which doesn't make
00:55:39
sense to me Science Guy you want to come
00:55:41
in here I don't
00:55:42
see I don't see how nighttime You' get a
00:55:45
better read on radioactive material that
00:55:48
doesn't make any sense so that made no
00:55:49
sense that sound like a crazy thing but
00:55:51
anyway we're living in crazy times
00:55:54
speaking of Crazy by the way that was my
00:55:55
first consp theory in 208 episodes of
00:55:58
the all in pod so very nice I actually
00:56:01
can participate in conspiracy corner now
00:56:04
I'm coming back next week with a better
00:56:06
one okay read us read us some more news
00:56:09
go to Google news and read us some more
00:56:10
news so we can do another to Jesus
00:56:11
Christ man read for us no you do it go
00:56:15
no you got the do in front of you okay
00:56:17
so today the Dow Jones Industrial liage
00:56:19
is up 23,000 to 44,000 TR to guys make
00:56:23
you look kler said he's super happy
00:56:26
sking that's enough I'm going skiing I
00:56:29
thought it was cool how he read the
00:56:30
entire Congressional bill um earlier
00:56:32
yeah all 1500 J loves to filibuster I'm
00:56:36
not filibustering all right listen here
00:56:38
let's do another story let's see if you
00:56:39
can contribute something open a update
00:56:43
M your contribution was amazing I was
00:56:46
actually using chat GPT to go into the
00:56:49
to go into the founding father's papers
00:56:51
and uhal I was reading The Federalist
00:56:54
Papers with G 13 in number 44 I would
00:56:58
like to
00:56:59
quote Hamilton the
00:57:02
musical Hamilton the music I was
00:57:06
comparing the lyrics from Hamilton the
00:57:09
musical oh my God hey Nick can you send
00:57:11
JK another Yahoo news article let's get
00:57:13
going come on let's go
00:57:19
Yahoo let's go let's go can't wait can't
00:57:22
wait can't wait to see you tomorrow
00:57:24
let's go oh my God
00:57:26
we just can we just end it here yeah no
00:57:30
do it do it because I want to talk about
00:57:31
open a you want to talk about open you
00:57:33
do yeah yeah give him his red meat give
00:57:35
chamad his red meat all right here's an
00:57:36
update on closed AI flipping and uh Sam
00:57:41
Alman super villan Sam Allin secure in
00:57:43
the bag meta wrote a letter to
00:57:45
California's attorney open yet on open
00:57:49
for profit conversion two months ago
00:57:53
open aai announced a $6.6 billion fund
00:57:56
around $157 billion valuation they're
00:57:59
producting 3.7 billion in Revenue this
00:58:02
year pretty great Revenue growth but
00:58:05
there's a poison pill in the deal open
00:58:07
AI must convert to a for-profit within
00:58:10
the next two years or investors can ask
00:58:13
for their money back and right before
00:58:16
they announced this round you remember
00:58:18
CTO Mira morat and two other top
00:58:21
researchers resigned many people saw
00:58:23
this as a protest Elon who put the first
00:58:26
50 million in and co-founded open AI is
00:58:28
currently suing the company and seeking
00:58:30
a court order that would stop the
00:58:32
for-profit conversion now Zuck is
00:58:35
joining team Alon Elon and Zuck are in
00:58:39
some weird pairing up they're not in the
00:58:41
ring not in the Octagon no last week
00:58:44
meta sent a letter asking California's
00:58:47
attorney general to stop opening eyes
00:58:49
for profit conversion what do you think
00:58:51
of this we've got them now not in the
00:58:53
Octagon but they're in the political
00:58:55
Arena math your thoughts I think that
00:58:57
this is so interesting so I was looking
00:59:00
at all of these things if you if you put
00:59:02
them all together it paints a really
00:59:03
interesting story so you have Elon
00:59:06
filing an injunction which basically
00:59:08
says you should not allow this
00:59:10
conversion to happen until we sort out
00:59:13
all of these details because if you
00:59:15
allowed it to convert and then I win
00:59:17
it'll be very messy to go backwards I
00:59:20
think that that's a pretty credible
00:59:21
legal argument then you have Zuck basic
00:59:26
Bally say hey elon's right this company
00:59:28
should not convert but the more
00:59:31
interesting thing that really got me
00:59:32
thinking about this was a chart that
00:59:34
Menlo Ventures put out Nick can you just
00:59:37
show this what this shows is just what's
00:59:40
happened in the last year and what do
00:59:43
you notice what you notice is the market
00:59:46
share of open AI has changed pretty
00:59:50
meaningfully from half to about a third
00:59:53
and what you see is anthropic doubl meta
00:59:56
roughly staying the same Google picking
00:59:58
up steam and it started to make me think
01:00:01
this is very similar
01:00:04
to a chart that I would have looked at
01:00:06
when you know
01:00:08
in6 and seven when we were building
01:00:10
Facebook because we had this huge
01:00:12
competitor in Myspace that was the deao
01:00:16
winner and we were this upstart and it
01:00:18
made me think is there a replay of this
01:00:22
same thing all over again where you have
01:00:25
this incumbent that Pioneers an
01:00:28
idea and they start with 100% share and
01:00:32
then all of these upstarts come around
01:00:34
from nowhere and then I thought well
01:00:36
what is better today if you are a
01:00:39
company that's just starting versus if
01:00:42
you were the incumbent and I think that
01:00:43
there's a handful that are worth talking
01:00:45
about so the first is when you look at
01:00:49
what xai is done with respect to invia
01:00:52
gpus the fact that they were able to get
01:00:54
100,000 to work as you know in one
01:00:57
contiguous system and are now rapidly
01:01:00
scaling up to basically a million over
01:01:02
the next year I think what it does Jason
01:01:04
it puts you to the front of the line for
01:01:06
all Hardware so now all of a sudden if
01:01:09
you were third or fourth in line xai is
01:01:11
now first and it pushes everybody else
01:01:13
down and in doing that you either have
01:01:16
to buy it yourself or work with your
01:01:17
partner and I think for the folks like
01:01:20
meta that translates and explains why
01:01:22
they're spending so much more it's sort
01:01:24
of like this arms race if you can't
01:01:26
spend with my competitors I'm just going
01:01:28
to prefer my competitor to you so I I
01:01:31
think that that creates a capital war in
01:01:33
a capital War I think the big companies
01:01:35
like Google Amazon Microsoft meta and
01:01:40
Brands like Elon will always be able to
01:01:43
attract effectively infinite capital and
01:01:45
their cost of capital goes to zero which
01:01:48
means they'll be able to win this
01:01:50
Hardware War okay so put a pin in
01:01:52
that then the second thing is what
01:01:55
happens on the ACT data and experience
01:01:56
side on the training side if you listen
01:01:59
to Ilia suer if you listen to Andre
01:02:01
carpy what they effectively are saying
01:02:03
is there's this terminal ASM toote that
01:02:06
we're seeing right now in model quality
01:02:08
so what happens a lot of these folks are
01:02:11
now experimenting on the things around
01:02:13
the model right the user experience how
01:02:15
you use it can I keep things in memory
01:02:16
can I cut and paste these things from
01:02:18
here and there because what it says is
01:02:21
like the data is kind of static and
01:02:23
brittle but it's actually not that's
01:02:25
what we said before because you have
01:02:26
this Corpus of data on X that's pretty
01:02:29
unique I suppose if Elon fed in all this
01:02:32
kinetic data that he controls through
01:02:34
Tesla that's very unique does that all
01:02:38
of a sudden create this additive pool of
01:02:41
information
01:02:43
possibly and then the third thing is
01:02:45
when you look at that chart what that
01:02:46
chart says is hold on a second why are
01:02:48
corporates moving around and what I can
01:02:51
tell you just through the 8090 lens
01:02:53
is we are complet completely promiscuous
01:02:56
in how we use models and the reason is
01:02:59
because these models offer different
01:03:02
cost quality tradeoffs at different
01:03:05
points in time for different tasks and
01:03:07
so what we are seeing is a world where
01:03:10
instead of having two or three models
01:03:11
you rely on you're going to rely on 30
01:03:13
or 40 or 50 and you're going to trade
01:03:15
them off and you're going to use
01:03:16
effectively like an llm router to dis
01:03:19
you know load balance them yeah and to
01:03:21
Route them or just or just to manage and
01:03:23
Route them and then there's an
01:03:24
intelligence above that that's
01:03:25
constantly tasking and figuring out
01:03:27
prompt optimization across these models
01:03:29
so it's this thing where we were very
01:03:31
reliant on open AI now we're Reliant
01:03:35
across three or four ideally we'll be
01:03:37
reliant on 30 or
01:03:39
40 and so I just see this world where
01:03:41
it's all getting commoditized quite
01:03:44
quickly and so I'm just like sort of
01:03:46
scratching my head like where is the
01:03:48
market value here Aaron your thoughts I
01:03:50
know you're very promiscuous when it
01:03:51
comes to llms as well uh I I mean I we
01:03:54
have a very similar model is what which
01:03:56
M said which is we're agnostic so we
01:03:58
work with multiple AI vendors but the um
01:04:00
uh I think the a a friend deep in AI
01:04:04
land a couple years ago right before
01:04:06
chat BT said there's no secrets in Ai
01:04:08
and I didn't totally understand kind of
01:04:10
at the time it hadn't registered what
01:04:11
that meant but very quickly it kind of
01:04:13
became obvious which is the research
01:04:15
breakthroughs sort
01:04:17
of propagate insanely quickly across the
01:04:20
AI community and so back back to Jas
01:04:23
framework if you just think about it as
01:04:25
as if if the research effectively
01:04:27
becomes open at at some point in time
01:04:29
quickly enough because either the
01:04:30
researchers move or people publish it or
01:04:32
whatnot then it really is a compute game
01:04:35
and then maybe a data access game and
01:04:37
that means that there's four or five ad
01:04:38
scale players that can that can fund
01:04:40
this and and I think as we've seen in
01:04:43
other areas where it's an infrastructure
01:04:45
play you eventually have the underlying
01:04:48
service with with enough competition you
01:04:50
have the underlying service eventually
01:04:51
Trend toward the cost of the
01:04:53
infrastructure so so what we should
01:04:55
expect is that the price of a token in
01:04:57
AI land you know basically will be
01:05:00
whatever the price of running the
01:05:02
computers are and maybe with like a you
01:05:05
know plus 10 20% margin did you see that
01:05:07
same thing happen with storage I
01:05:09
remember in the early days box and
01:05:11
Dropbox and YouTube you all had this
01:05:13
major Innovation with storage how did
01:05:16
that play out let me give you a fun stat
01:05:17
we give our customers unlimited storage
01:05:20
we have 82% gross margin so so the the
01:05:24
what what happened was the of the
01:05:26
underlying storage has gone down by
01:05:28
hundreds of times since we started the
01:05:30
company and then our all our value is in
01:05:32
the software layer on top of the storage
01:05:34
so we've benefited by this incredible
01:05:37
just you know ruthless competition
01:05:38
between Western Digital Seagate other
01:05:40
players that are just trying to pack
01:05:43
more more more you know basically more
01:05:46
storage density into these drives and
01:05:48
every couple years they have a new
01:05:49
breakthrough we're now you know upcoming
01:05:51
we're we're heading toward maybe a 50
01:05:52
terabyte hard drive when we started the
01:05:54
company they were kind of 80 uh 80 gab
01:05:56
how much of your time in the early days
01:05:59
was spent on dealing with this
01:06:01
infrastructure issue and then how much
01:06:03
of your time and your leadership team's
01:06:05
time is spent on this issue now yeah so
01:06:07
so back in the day like if we had you
01:06:09
know 10 10 people in in engineering you
01:06:11
know 80% of them was doing you know pure
01:06:13
infrastructure work now if we have a
01:06:14
thousand people uh it you know be
01:06:17
inverted in terms of the ratio so you
01:06:19
just get you get more leverage uh both
01:06:21
as the the you know as as you get the
01:06:23
advancements in the technology but then
01:06:25
also as you scale but all this is to say
01:06:28
you should basically anticipate a world
01:06:30
where and and I think Zuck is this
01:06:31
interesting counterbalance on all of
01:06:32
this because of Open Source um if at any
01:06:35
moment you know that that Zuck will will
01:06:37
basically provide an open source model
01:06:39
that is at at kind of best-in-class
01:06:41
benchmarks and and at the at the
01:06:43
frontier then there is a limit on how
01:06:46
much you can charge for the tokens of
01:06:47
your hosted model because anybody will
01:06:49
then be able to go host the open model
01:06:51
and and be able to provide
01:06:52
infrastructure around it so so if you
01:06:53
always have that counterbalance and the
01:06:56
tokens eventually kind of look the same
01:06:58
uh the output tokens kind of look the
01:06:59
same isn't it yeah always isn't it also
01:07:02
the truth that major Enterprises Fortune
01:07:05
500s 200s you know 20 years ago weren't
01:07:09
interested in open source and now that's
01:07:10
kind of their default they want to buy
01:07:12
into open source because they don't want
01:07:13
to be locked into a vendor it's actually
01:07:15
not even necessarily the case that the
01:07:17
that the customer has to pick the open
01:07:18
source vendor they might buy it through
01:07:20
an abstraction layer that that is
01:07:21
letting them get the benefit of Open
01:07:23
Source but but still buy through
01:07:25
proprietary so you believe open source
01:07:27
wins I I believe no no no so I I believe
01:07:31
open source causes pricing to always be
01:07:33
extremely low right but okay but in this
01:07:36
case do you think do you think open
01:07:38
source is GNA ultimately win the day in
01:07:40
models no not necessarily because only
01:07:43
only meta has the has the the scale to
01:07:46
be able to provide I think what eron is
01:07:47
saying here let me let me maybe try to
01:07:49
frame it I think what he's saying is
01:07:50
there'll be open source models there'll
01:07:53
be close Source models but the price
01:07:56
that Aaron or me or anybody else pays
01:07:58
these model makers will effectively go
01:08:00
to zero got it and it it'll go to the
01:08:04
cost of the compute to be to be clear
01:08:05
with a little bit of margin for for
01:08:07
theer the physicality of that compute
01:08:09
got it now it's important step back and
01:08:11
say I still think you could probably
01:08:12
have the entirety of the model providers
01:08:14
make 10 times more Revenue than they do
01:08:16
today because we're just literally in
01:08:18
like the first percent of the total Tam
01:08:21
so so it'd be a mistake to think that
01:08:23
that has you know some kind of downward
01:08:24
pressure in in terms of the long-term
01:08:26
economics of these businesses especially
01:08:28
because I think open ai's Revenue stream
01:08:30
is increasingly looking like a SAS
01:08:31
Revenue business as opposed to just you
01:08:33
know just the API kind of token pricing
01:08:37
so so none of this is is to you know
01:08:39
provide any any sort of color on like
01:08:40
what would you bet on today I I agree
01:08:42
with jamat that you're going to have
01:08:43
maybe not 30 providers but let's say you
01:08:45
at least have you know five to 10 good
01:08:46
choices all all competing heavily for
01:08:48
the next breakthrough like literally
01:08:49
this morning Google had a breakthrough
01:08:51
in in sort of this reasoning oriented
01:08:53
model from from their flat from their Je
01:08:55
family 2.0 yeah yeah and uh and and so
01:08:59
what's what's incredibly you know kind
01:09:00
of great is it's it's sort of the best
01:09:02
time ever to be building software
01:09:05
assuming uh you know assuming that that
01:09:07
you have a play in the market that lets
01:09:09
you remain differentiated and the key
01:09:10
there is just do enough on top of the AI
01:09:13
model that there's enough value there
01:09:14
you know how much the world spends on
01:09:16
software and software related things
01:09:18
every year it's about5
01:09:21
trillion so there's like call it like a
01:09:23
trillion and a half of
01:09:26
software licenses a trillion and a half
01:09:28
of Consulting and a trillion and a half
01:09:30
of it folks inside of companies plus or
01:09:33
minus a little bit more you get about
01:09:34
five trillion and that's compounding
01:09:36
133% a year I I'm pretty sure that the
01:09:40
market here shrinks by an order of
01:09:44
magnitude and instead of fighting over
01:09:46
five trillion I think we'll be fighting
01:09:48
over 500 billion what do you think Aaron
01:09:51
you you buy that no not at
01:09:54
all uh I mean I don't know if you want
01:09:56
to make the case more but but the only
01:09:59
reason is that I think that as as we
01:10:02
delever the software development process
01:10:04
from humans I think the unit cost of
01:10:05
creating code effectively becomes so
01:10:07
cheap that it's going to be very hard to
01:10:10
differentially price these products the
01:10:12
way that they are so an example would be
01:10:14
that let's say you use I don't know pick
01:10:18
your favorite piece of
01:10:21
software I I don't want to pick on
01:10:22
anyway that's why I'm not I'm not saying
01:10:24
say office Suite let's pick an office
01:10:26
suite everybody's got one sure actually
01:10:29
let's pick let's pick on Excel because
01:10:30
maybe that's like sure Excel Google
01:10:32
Sheets yeah it's not it's not that Excel
01:10:34
isn't valuable it's incredibly valuable
01:10:36
it's what is the marginal cost of
01:10:38
creating the Excel equivalent that is
01:10:40
good enough that people switch the
01:10:41
marginal cost today is very expensive
01:10:44
and you can see that because it's what
01:10:46
it costs Google to make sheets but
01:10:49
that's humans so the real question is if
01:10:52
you have a legion of B
01:10:55
that works 247 incrementally and
01:10:58
increasingly more accurately every day
01:11:02
the question is what is the marginal
01:11:04
cost and I think the marginal cost of
01:11:05
that is going to be very cheap and when
01:11:08
you do that it's very difficult to price
01:11:11
it anywhere near the same and the reason
01:11:13
is that other companies will then
01:11:15
replicate it and say hold on if if Excel
01:11:18
wants to charge $100 I'll charge 50 and
01:11:20
I'll take a lower margin that's just
01:11:22
Supply demand economics yeah yeah so so
01:11:26
so so I I think I I I think
01:11:28
there's I I just don't agree with the
01:11:30
Tam compression because I think there's
01:11:32
another kind of counter event that's
01:11:33
happening that that AI is is really
01:11:35
going after like services and and so
01:11:37
that then then conversely expands the
01:11:39
the T of software where it budgets
01:11:42
weren't usually applied to those types
01:11:43
of things but we can get in that in a
01:11:44
second but I think we've already seen I
01:11:46
think we've already seen this though and
01:11:48
it hasn't exactly played out as as
01:11:49
you're saying so you already have like a
01:11:51
z like so Zoho is this really
01:11:53
interesting business it's probably a
01:11:54
couple billion Revenue at this point um
01:11:56
and it's basically a suite of of
01:11:58
extremely lowcost affordable software
01:12:01
products by category the cycle time of
01:12:03
Zoho is poor but it's not that's not
01:12:05
been that's not been the reason people
01:12:06
don't switch though they it's all it's
01:12:08
all lever to humans I just think it's
01:12:10
not a good product it's it's decent
01:12:12
enough why do you think people don't
01:12:14
switch
01:12:14
Aaron yeah I do agree with you Aaron by
01:12:17
the way that when you bring C that whole
01:12:20
offline Services category online and you
01:12:22
automate them with AI I agree with you
01:12:24
that that Tam could be ginormous all I'm
01:12:26
saying is right the traditional software
01:12:28
T today what people spend $5.1 trillion
01:12:31
on I think people will spend $500
01:12:33
billion on and barely if
01:12:36
that there there may be other things
01:12:38
that people spend money on that that are
01:12:40
that are wrapped in AI yeah I I guess
01:12:42
the the the the counter and maybe you'd
01:12:45
look at an Erp system or a CRM system or
01:12:47
something else like like it that that is
01:12:49
sort of like those things are totally
01:12:51
screwed no but like but but this is my
01:12:54
point the opposite like the last thing
01:12:56
you want to touch is the system that is
01:12:58
like like powering your supply chain the
01:13:00
companies the companies I talked to
01:13:02
they're consistently like rip it out get
01:13:04
us to a point where we can rip it out
01:13:05
and the reason is because what they've
01:13:07
realized is they'll spend $50 to100
01:13:08
million do a year for five features and
01:13:11
they're like just give me these five
01:13:13
features as workflows yes give me a
01:13:15
simple crud database and just get me get
01:13:17
out of the way and it's like the
01:13:19
tradeoff for that makes a lot of sense
01:13:22
because look let's face it like when you
01:13:24
have to build one piece of software that
01:13:26
has to sell to 50,000 companies the
01:13:28
reality is that that piece of software
01:13:31
is trying to do everything and then some
01:13:33
and it's trying to solve two or three
01:13:35
use cases plus around you know five or
01:13:37
six common use cases that are
01:13:39
generalized yep for 50,000 customers so
01:13:41
you end up with 50,000 features you know
01:13:43
it's really interesting because aarin
01:13:45
you kind of alluded to there's a tam
01:13:48
expansion moment there and I'm seeing
01:13:49
this on the front lines uh we run a
01:13:52
program found University I've talked
01:13:53
about it here before where we see people
01:13:55
pitching us they're year zero startups
01:13:58
two or three person teams and what
01:13:59
they're doing is they're not going after
01:14:01
existing Legacy software they're kind of
01:14:03
going after jobs they're looking at a
01:14:06
position or a job somewhere this is an
01:14:08
accountant we're going to take an
01:14:09
accountant we're going to make of the
01:14:11
number one accountant in the world
01:14:13
that's an AI agentic an agent we're
01:14:16
going to make a podcast producer we're
01:14:17
podcast AI we're going to make a virtual
01:14:19
assistant with Athena we're just seeing
01:14:21
it over and over again that's a whole
01:14:23
another category where you study a
01:14:26
person's Behavior as they work a social
01:14:28
media manager and what they do and then
01:14:31
you replicate it with AI and that's
01:14:34
something that just hasn't previously
01:14:35
been done so there could be two things
01:14:36
occurring here chth and Aaron at the
01:14:39
same time which is a deflation in Legacy
01:14:42
systems and they'll be replaced and then
01:14:45
additionally human capital and jobs that
01:14:48
are easy enough for AI to do as an agent
01:14:51
will also expand the Tam two things at
01:14:54
the same time you are completely right
01:14:57
about the usage all I'm debating is when
01:15:00
you have to price something you have to
01:15:02
look at the total cost of what it took
01:15:05
to make it and then you want to try to
01:15:07
build in a reasonable amount of margin
01:15:09
and some reasonable expansion and you
01:15:11
discount it back and this is what you
01:15:12
think it's worth even though you may not
01:15:15
think you're doing that when you
01:15:16
implicitly price something that way that
01:15:18
is what's happening underneath the hood
01:15:20
to the unemotional buyer of that good
01:15:23
and all I'm saying is if what Aaron says
01:15:26
before which I agree with is true which
01:15:28
is the cost of using a model to get to
01:15:32
an output effectively goes to zero sort
01:15:35
somewhat what I've been saying before
01:15:36
the marginal cost of compute goes to
01:15:39
zero the marginal cost of energy goes to
01:15:41
zero the real question is what does it
01:15:44
take to make a good in the digital world
01:15:46
let me ask freeberg a question here
01:15:48
you're back in the CEO slot at ohal
01:15:50
you're doing it every day Dave you're
01:15:52
making these choices as to what SAS
01:15:54
products and how you're going to solve
01:15:56
problems are you looking at it saying
01:15:58
I'm going to hire developers who can
01:15:59
work 10x because of all these new tools
01:16:01
and I'm going to build my own internal
01:16:03
systems or are you looking and saying
01:16:05
I'm going to buy off-the-shelf SAS
01:16:08
products what are you doing when you
01:16:09
make your own decisions every day David
01:16:12
freeberg well I try and encourage the
01:16:14
teams
01:16:15
to build stuff um that better meets our
01:16:19
needs and then it can actually be a
01:16:22
better solution and it can be built- in
01:16:24
house and I think we did a I think I
01:16:26
mentioned this the last episode but we
01:16:28
did a hackathon where we brought in
01:16:29
people to learn how to use cursor and
01:16:32
chat GPT to build software that had
01:16:34
never done it before to try and create
01:16:36
this as a capability for people broadly
01:16:39
in the organization and they were great
01:16:40
tools that came out of it so I think
01:16:42
that that's really the future as and
01:16:44
this is kind of like you know I'd say
01:16:45
early generation but as we get to
01:16:47
further later generation
01:16:49
capabilities you could see the
01:16:52
instruction to a chat GPT or like
01:16:54
interface hey I'd love to do the
01:16:56
following things with a piece of
01:16:57
software it shows you a couple options
01:16:58
you pick one shows you a couple ux you
01:17:01
pick one it does user testing
01:17:02
automatically for you it does QA for you
01:17:04
and ultimately it puts it into
01:17:06
production for you which is the biggest
01:17:07
challenging step right now you still
01:17:09
need Engineers that can load stuff into
01:17:10
production and do QA but if all of that
01:17:12
gets automated as well now any user in a
01:17:16
company can actually stand up software
01:17:18
to do something for them a non developer
01:17:20
you're saying for n developer stand up
01:17:21
software maybe so jamat just to clarify
01:17:25
you there there's two different ways to
01:17:26
approach the lowering the cost of
01:17:28
developing software one is that it just
01:17:29
creates more competitors in each of
01:17:31
these categories which then lowers my
01:17:33
price because now there's there's some
01:17:34
downward pressure Dave is bringing up an
01:17:35
different example which is I'm going to
01:17:36
build my own software at effectively the
01:17:39
price of
01:17:40
zero the the The Challenge on this
01:17:42
especially the second one is is like
01:17:44
when you like most organizations don't
01:17:46
want to be in the business of having to
01:17:47
think about about building their own
01:17:49
software they just like want it done for
01:17:50
them like it's like not a core part of
01:17:52
of their like so you know you're your
01:17:55
business would be very unique relative
01:17:56
to the broad economy which is like I
01:17:58
just want like like just like I want a
01:17:59
place to put my CRM records I want a
01:18:01
place to like just like have my HR get
01:18:03
managed and I think the downward pricing
01:18:05
pressure due to software cost lowering
01:18:07
makes total sense I don't think it's
01:18:09
it's a 10x Factor um but uh but but you
01:18:12
know I I think that we've always had a
01:18:15
long taale of applications that that
01:18:16
enterprises build uh you did it in
01:18:18
Microsoft Access now you do it in retool
01:18:21
um so the next era of that will be
01:18:22
obviously AI built and there'll be 10
01:18:24
times the amount of that software but
01:18:26
it's not obvious to me why that would go
01:18:28
after the kind of core the the core
01:18:30
systems of running a business because
01:18:32
just most companies are like not looking
01:18:33
to re reinvent the wheel of that we're
01:18:34
working with an aerospace partner won't
01:18:36
say who it is and we sat there and they
01:18:39
walked us
01:18:41
through what they deal with to make the
01:18:43
things that they're making it's
01:18:45
convoluted and this is not a software
01:18:48
problem it's that there is no piece of
01:18:50
software that understands how they want
01:18:52
to run their company and so instead they
01:18:54
have to morph their orc chart yeah to
01:18:56
the tools that are available so I think
01:18:58
what freeberg is saying some is some
01:19:00
version of that as well he's probably
01:19:02
there were probably people there using
01:19:03
the tools that were available and as a
01:19:06
result at some point some HR person said
01:19:08
we need to hire this other person and
01:19:09
this other person and instead what what
01:19:12
it allows companies to do is just
01:19:14
completely reimagine how do you want
01:19:16
your company to work and what is the
01:19:18
business process you actually need to
01:19:21
implement and then let's just get that
01:19:23
built and and you know I I I don't know
01:19:26
if you guys saw this but Saia Nadella
01:19:28
had this clip that's gone pretty viral
01:19:30
in the last couple of days where he
01:19:32
effectively said the same thing and what
01:19:33
he's effectively saying is all these big
01:19:37
software systems are business rules
01:19:39
wrapping a data base plus an incredible
01:19:43
goto Market team you know and this is
01:19:46
what Alex karp points to as sharp knives
01:19:50
steak dinners and basketball game
01:19:53
tickets I don't know if you guys saw
01:19:54
that clip I did see the clip yeah he
01:19:56
likes to win based on product not on
01:19:58
sales effort Nick you should find this
01:20:00
clip we don't play golf what we do do is
01:20:03
we play software we will put if you want
01:20:05
to actually compete compete on your
01:20:08
product and what's very special and and
01:20:11
yes do I enjoy humiliating people who
01:20:13
have better steak dinners and sharper
01:20:14
knives and better golf swings yes I do
01:20:17
you know what I really I really like
01:20:18
that we win in that way it makes me very
01:20:20
happy and it makes our clients happy but
01:20:23
it really points to the heart of how the
01:20:25
software industrial complex that's what
01:20:27
I call it this 5.1 trillion doar this
01:20:29
software industrial complex how has it
01:20:32
evolved I think it's people that build
01:20:35
good point products but when the market
01:20:38
meaning the shareholders and the public
01:20:39
investors demand growth you have this
01:20:42
natural expansion and what do they do
01:20:45
they inflate the feature set they
01:20:47
inflate the cost and then in order to
01:20:50
sell that they inflate the set of
01:20:54
incentives that they give to the buyer
01:20:56
so the CIO inside of these large
01:20:59
organizations they control budgets
01:21:01
they're equivalent to like state level
01:21:04
budgets in some cases you know what do
01:21:06
you think the CIO of a top five bank is
01:21:08
spending it's probably 15 to 20 billion
01:21:10
dollar a year they are getting whed and
01:21:13
DED to a way that you could not even
01:21:16
imagine I don't even think the CEO
01:21:18
probably understands and that filters
01:21:20
down through the org and so they don't
01:21:22
like totally ruin our whole game here
01:21:24
well it's not it's not I'm just pointing
01:21:26
it out like it's I mean I I'll say the
01:21:29
cter but here but here's another part of
01:21:31
the game on the field if these tools
01:21:33
make your team yeah 10 20 6% whatever
01:21:37
you can quantify more effective at their
01:21:40
jobs then it's a dious amount for most
01:21:43
organizations who have created very
01:21:45
profitable businesses I see this where I
01:21:48
was mentioning some companies earlier
01:21:49
and their SAS pricing is broken because
01:21:52
the number of employees at a company has
01:21:54
been going down because people are
01:21:56
getting more efficient so now they're
01:21:57
looking at saying well what value did we
01:21:59
provide with this podcast you know
01:22:01
producer in a box and whatever it does
01:22:04
they say they were starting with like
01:22:05
$99 a month or $49 a seat I said listen
01:22:09
just charge a minimum of $500 to get the
01:22:11
software nobody complained and they got
01:22:14
rid of the bottom tiers so there's a
01:22:16
value being created here that is so
01:22:19
great that I don't think everybody's
01:22:22
going to roll their own like Friedberg
01:22:23
because they're going to be like you
01:22:24
know what what this is so great it's
01:22:26
6,000 a year fck it yolo I'll just buy
01:22:29
it it is today and in five years just
01:22:33
like we're all going to and we're all
01:22:34
using chat like interfaces more
01:22:37
frequently yeah you go to your chat-like
01:22:39
interface and eventually you realize you
01:22:42
can ask it to build some tool for you
01:22:43
that does something and it renders the
01:22:46
tool and it makes it possible and it
01:22:48
stands it up in production and suddenly
01:22:50
you're using it at your company and then
01:22:52
you ask it to do another tool and it
01:22:54
does it exactly to spec and you define
01:22:56
the ux you want you define what you want
01:22:58
it to do and it works and it
01:23:00
interoperates but you and really clearly
01:23:03
and really cleanly with the other tool
01:23:05
and there's a big aspect of this where
01:23:06
you can start to build all of the
01:23:08
software infrastructure that you as an
01:23:10
organization want right but I think you
01:23:13
and Aaron you and Aaron mentioned this
01:23:14
though the extreme difficulty is not
01:23:17
that front end part that's easy it's
01:23:19
like I think it's less than three or 4%
01:23:22
of the work the 96% of the work is how
01:23:26
you actually integrate it in the back
01:23:28
end and how do you provision it how do
01:23:29
you have controls and how do you do
01:23:31
security because those things fail and
01:23:33
those are implemented in a bot in a
01:23:35
highly regulated market as an example
01:23:38
you may not actually be allowed to
01:23:40
operate let's go through that example
01:23:42
and let's say that you're a bank and you
01:23:44
tell the AI figure out all the security
01:23:47
and permissions and and Authority rules
01:23:49
that are necessary for me to operate as
01:23:51
a bank and it can actually render it it
01:23:52
sounds farfetched today but there's no
01:23:54
reason that in seven years that is not
01:23:56
the standard isure that I don't have the
01:23:58
ability to say go look at all the
01:24:00
software that's out there in the world
01:24:01
today help build a tool that meets
01:24:03
compliance standards that meets all of
01:24:05
my security standards and you can
01:24:06
actually instruct the AI to operate like
01:24:08
a large number of software Engineers
01:24:11
need to operate today the Practical
01:24:13
issue where the rubber meets the road
01:24:14
there is when there is a penetration and
01:24:18
a regulator who's a human because it
01:24:19
will not have been a bot comes and
01:24:21
knocks on your door and you're like well
01:24:23
here's this immutable log of the things
01:24:26
that I did and they're like I don't want
01:24:27
an immutable log who the hell tested
01:24:30
this show me the unit test that you
01:24:31
created in signed off on so there's a
01:24:34
there is a critical human in the loop
01:24:36
problem on that other 95% which is where
01:24:39
I agree with Aaron that's in in in reach
01:24:41
in in like stage one like we're in stage
01:24:43
zero stage one that you need no but
01:24:45
you'll I think it's in stage five and
01:24:47
six I think you'll need governments to
01:24:48
take an entirely different risk posture
01:24:50
for highly regulated markets I don't see
01:24:52
that happening in any in any time in the
01:24:54
near future in in life sciences if you
01:24:57
want anything in a in a clinical system
01:24:58
I don't need to tell you guys this but
01:24:59
you have to you have to do a QA test on
01:25:01
every single change that ever happens
01:25:03
and be able to prove that you tested
01:25:05
every single thing so so the idea of a
01:25:07
probabilistic AI system generating the
01:25:10
the kind of code for you you know in a
01:25:11
clinical trial you know workflow like I
01:25:14
just think yeah it's going to take a lot
01:25:15
of change from from Regulators we got
01:25:17
another decade of evolution here to make
01:25:20
these things sustainable and have a high
01:25:23
quality I think it'll happen faster than
01:25:24
everyone thinks yeah and I think
01:25:26
actually that's that's like there's
01:25:28
actually no reason why that wouldn't be
01:25:29
a good thing if that did happen to be
01:25:30
clear I I think though back to to you
01:25:32
know Jason's earlier point though like
01:25:34
then the the counterbalance on the Tam
01:25:36
compression is is just now the the sort
01:25:39
of thing that we think of as software
01:25:41
that market is now is is much larger so
01:25:43
if you're if you're selling if you're
01:25:45
selling yeah so but like but like that
01:25:47
means that if you're selling let's say
01:25:48
$50,000 of security software you know
01:25:51
maybe that goes down a little bit to you
01:25:53
know $225,000 but but you might yeah but
01:25:55
you might sell ,000 of new agent yeah
01:25:58
that I so I agree 100% with that my only
01:26:01
comment is the software industrial
01:26:03
complex today has to shrink because the
01:26:05
strangle hold that it has on how
01:26:07
companies
01:26:08
run is incredibly high for an experience
01:26:12
that's incredibly
01:26:13
poor nobody nobody raises their hand and
01:26:16
says gosh this piece of enterprise
01:26:18
software that I use in my day-to-day
01:26:19
life is as good as Instagram or Tik Tok
01:26:21
nobody says that well I mean except
01:26:24
about box I was reading the reviews
01:26:25
earlier today and I was on G2 or one of
01:26:28
those sites and it just box was just the
01:26:30
ratings were tremendous so I love Bock I
01:26:33
love Bo God it's incredible well I mean
01:26:35
I mean just what's the ticker symbol it
01:26:38
dollar we try to make it very simple for
01:26:39
people so is it dollar box is that what
01:26:42
it is I have to put a I'm putting a j
01:26:44
trade in here because I may have to J
01:26:47
trade this sounds like a really great
01:26:49
opportunity for me all right listen Ain
01:26:51
you just got Kramer what with your lator
01:26:53
yeah yeah you got Kramer FR J trade is
01:26:54
I'm just I'm I'm loading up on uh MST
01:26:58
and Bitcoin I'm doing a I'm shorting
01:27:00
Bitcoin and I'm buying MST I don't know
01:27:02
what that I'm not I'm not trading stocks
01:27:03
right now I'm focused on investing in a
01:27:05
100 startups per year 100 new startups
01:27:09
per year I'm putting the first check
01:27:10
into to them that's what I do all right
01:27:12
listen enough this has been a great
01:27:14
episode Aaron you're amazing thank you
01:27:16
for coming on constantly Everybody
01:27:18
follow what is your before WRA on AI can
01:27:20
I just tell you guys are you at Levy or
01:27:22
at Aaron uh yeah yeahia Leia is that the
01:27:26
French pronunciation l e v i eia have
01:27:29
you guys seen the vo model from Google
01:27:31
yeah I have you guys talked about how
01:27:33
Google has like gotten some incredible
01:27:36
they are they
01:27:37
are I literally paid shout out to Sundar
01:27:40
oh yeah yeah and serge is going to work
01:27:42
every day by the way he's at work what
01:27:44
can Brown do for you jel Sundar back in
01:27:46
the group chat by way you guys see
01:27:48
Genesis yesterday well okay let's let me
01:27:50
just put this up here and then I'll te
01:27:51
it up to freeberg Google has been
01:27:54
putting out new models they have this
01:27:56
new Google
01:27:57
Gemini and they have one with deep
01:27:59
reasoning and their 2.0 model I did a
01:28:01
sidebyside test same prompts with our
01:28:04
friend s deep madra on this week in
01:28:06
startups just this week we did multiple
01:28:08
tests he was he has the 01 Pro the $200
01:28:12
per month Gemini was beating that hands
01:28:15
down for me and I think he conceded that
01:28:18
with their free model and their $20
01:28:20
model if we just gave it decent prompts
01:28:22
I think Google now has reach parody and
01:28:25
they have an app out freeberg what are
01:28:27
your thoughts because people were
01:28:28
counting them out and here we are I
01:28:30
think not only have they caught up I
01:28:33
think they've exceeded what are your
01:28:34
thoughts they were late the compounding
01:28:35
effects are playing out and it's only
01:28:37
going to continue to compound and I will
01:28:38
say the data repository at Google the
01:28:40
engine that they have the infrastructure
01:28:42
the team all everything down to
01:28:43
components is advantaged and so
01:28:45
everyone's been kind of counting them
01:28:46
out but it's clear they're in it to win
01:28:48
it they're here 70% of the usage of open
01:28:52
AI is consumer 70% consumer what does
01:28:55
that mean for them as as Gemini and
01:28:57
Google get get momentum I think you'll
01:28:59
end up having a Gemini that has ads
01:29:04
eventually and you could pay and have no
01:29:06
ads or you could not pay and have ads no
01:29:08
meaning do you provide enough value and
01:29:10
of enough quality where folks don't need
01:29:12
to then folks stay on
01:29:15
Google on google.com I'm asking you
01:29:17
whether it impacts open AI the quality
01:29:19
of Gemini as it increases or do you
01:29:21
think that these usage cohorts are
01:29:23
roughly set I jump back and forth
01:29:26
personally my experience is I don't feel
01:29:28
like I've got embedded data on open AI
01:29:30
or on Gemini that makes me stick with
01:29:32
one I'm just like with Google I'm going
01:29:33
to go to the best interface the best
01:29:35
engine so I do think that there's
01:29:37
fungibility here and people will move
01:29:39
over as they realize better results
01:29:41
better performance but I will say that
01:29:42
what we're seeing now with the data
01:29:44
Advantage at Google so you know the
01:29:46
internet all these llms are are language
01:29:48
models trained on text from the internet
01:29:50
right there's call it 50 billion words
01:29:52
on the internet and I think think if you
01:29:54
estimate the data repository in these
01:29:56
training sets it's like probably a
01:29:58
couple terabytes 1 to 5 terabytes or
01:30:00
something like that but if you look at
01:30:02
the the the the video data that's out
01:30:04
there there's hundreds of billions of
01:30:06
hours or 100 billion plus hours of video
01:30:08
data a large amount of that is sitting
01:30:10
on YouTube and by some estimates there's
01:30:13
a thousand exabytes of video data on the
01:30:15
internet so about a billion times more
01:30:17
video data than there is word or text
01:30:20
data and I think we just saw that play
01:30:22
out with the vo model that laun
01:30:24
yesterday but here's some examples of VO
01:30:26
Google has all of this YouTube data you
01:30:28
know whether or not they're using it to
01:30:29
train I don't know the answer I don't
01:30:31
think they're allowed to is what I heard
01:30:32
from the Insiders they have to redo the
01:30:34
terms of service to get explicit
01:30:37
permission to use it but you're right
01:30:39
these models are tremendous whatever
01:30:40
they're using and and it's basically
01:30:41
rendering physics or it looks like it's
01:30:43
rendering physics now Genesis came out
01:30:45
yesterday and it's open source so you
01:30:48
can actually go play with this Genesis
01:30:49
model which similarly renders the actual
01:30:52
3D objects into a 3D environment so you
01:30:56
so rather than rendering a
01:30:57
two-dimensional set of pixels to look at
01:31:00
a video video visual with Genesis as you
01:31:02
can see here you can type in a prompt
01:31:04
and it renders this extraordinary video
01:31:06
that that also has underlying it the
01:31:09
three dimensional objects that make up
01:31:11
the video Banas which means you could
01:31:14
change the angle right and you could
01:31:16
work with it and also with Google and
01:31:18
this you can start to implement
01:31:20
three-dimensional models based on some
01:31:23
prom that says something like I want to
01:31:25
have the camera angle at this point in
01:31:26
the room I want to have the room look
01:31:28
like this I want to have this color this
01:31:29
wallpaper and suddenly everything starts
01:31:32
to prompt in a way that you can actually
01:31:34
render in real time a video game a movie
01:31:37
a visual experience and it goes to this
01:31:39
point that we're unleashing the capacity
01:31:42
for human imagination and creativity
01:31:44
with these systems because it's no
01:31:45
longer just a lookalike two-dimensional
01:31:48
image it's now an actual
01:31:49
three-dimensional object that then
01:31:50
renders the visuals to make this happen
01:31:52
so we're starting to see the next era of
01:31:55
these models that goes beyond just text
01:31:57
prediction what do you think Aaron yeah
01:31:59
there was another release last week of
01:32:01
experimental project for browser use by
01:32:04
Gemini which is another Advantage
01:32:05
because YouTube has obviously an
01:32:07
incredible amount of screen sharing data
01:32:09
and and videos so so you have you know
01:32:12
like like we think about visual as just
01:32:13
like okay it's it's going to be you know
01:32:15
for for developing you know CG
01:32:18
characters or something but it's it's
01:32:19
actually just like like just all the use
01:32:22
cases of a computer are also on YouTube
01:32:24
Project Mariner is the name of it right
01:32:27
where your Chrome extension AI can
01:32:29
control your browser to do things and
01:32:31
then if you go to the AI Studio from
01:32:33
Gemini they they have a mode where you
01:32:34
can turn on your webcam and and then you
01:32:38
you basically it has full visual uh kind
01:32:41
of access to
01:32:43
anything and and so they're they're
01:32:45
cranking and it's super exciting because
01:32:48
that like just you can tell that
01:32:49
Google's woken up and and they are just
01:32:52
on Full Assault so the I mean just in 10
01:32:54
days the quantum the AI open source
01:32:58
Gemini updates it's like every morning
01:33:00
you're waking up to a Sundar tweet that
01:33:02
is that is some new breakthrough so well
01:33:04
yeah it's amazing when you fire 20,000
01:33:06
people who weren't doing any work and
01:33:07
were involved in di nonsense and then
01:33:10
you
01:33:12
yeah that's not the point the point the
01:33:14
point is they've had this compounding
01:33:16
you know infrastructure Advantage data
01:33:18
Advantage Personnel advantage and now
01:33:21
they were just a little late to the game
01:33:22
but when you they distracted is my point
01:33:25
they were distracted with nonsense you
01:33:26
know that's true come on the first
01:33:28
version of this came out it was making
01:33:30
Abraham afrian American
01:33:33
and Asian you know like they were
01:33:35
distracted okay a big part of Google's
01:33:38
orientation historically has been Don't
01:33:41
Mess the don't mess up the Machine
01:33:42
Classic kind of big company dilemma
01:33:44
where if I do something that's either
01:33:46
disruptive to my core business or if I
01:33:48
do something wrong where I will invite
01:33:50
Regulatory and consumer scrutiny I'm
01:33:51
going to get wrecked and so they avoided
01:33:54
stuff early and what they've done is
01:33:55
they've changed the posture in the last
01:33:57
two years and now the posture is launch
01:33:59
early launch aggressively push hard we
01:34:01
have to win this battle or we're going
01:34:03
to get eaten alive and it's amazing to
01:34:04
see the founders and and Sundar lead
01:34:07
this organization I think there were a
01:34:09
lot of question marks a year ago but I
01:34:11
think that the questions have been
01:34:13
answered I agree I I I'm saying it right
01:34:15
now I think we've hit Peak open AI in
01:34:18
the market I think they're going to be
01:34:19
the number three or four player I think
01:34:22
Gemini meta and xai are going to lead
01:34:26
them if we were sitting here in three
01:34:28
years I think open AI is number three
01:34:31
four five not one or two I don't think
01:34:33
they get the one or two slot what do you
01:34:35
think Aaron open AI who's in the one and
01:34:37
two slot you I don't think you want to
01:34:38
bet against Sam and Greg you don't want
01:34:40
to bet bet against Elon you don't want
01:34:42
to be bet against Sergey now uh you know
01:34:44
in in office with h with Sundar you
01:34:46
don't want to bet bet against zck so I
01:34:48
think I think the rankings are kind of
01:34:50
less interesting as much as just the
01:34:51
fact that that it's like game is on you
01:34:54
do not have you do not have incumbents
01:34:55
that are sleeping at this point that's
01:34:57
just going to push everybody forward so
01:35:00
just an incredible time for everybody
01:35:02
what a time to be alive all right four
01:35:04
the Sultan of science David freeberg the
01:35:06
chairman dictator Jamal payaa and I
01:35:09
don't have a nickname for you Aaron Levy
01:35:11
but we will have one soon this is the
01:35:13
all-in podcast we miss you Saxy poo we
01:35:17
miss you say come back soon Saxy poo
01:35:19
come back soon SX we miss you all right
01:35:21
everybody have a great Christmas break
01:35:23
and we'll see you shortly go sign up for
01:35:25
our spam at allin.com sign up for the
01:35:28
spam newsletter we'll send you some uh
01:35:30
updates on the podcast anything to sign
01:35:32
up for I'm joking I'm joking sign.com
01:35:34
love
01:35:37
you let your winners
01:35:40
ride Rainman
01:35:44
David and instead we open source it to
01:35:47
the fans and they've just gone crazy
01:35:49
with it love you queen of going
01:35:54
[Music]
01:35:58
besties
01:36:00
are my dog taking
01:36:04
driveway man oh man myit will meet me at
01:36:09
we should all just get a room and just
01:36:10
have one big huge orgy cuz they're all
01:36:13
it's like this like sexual tension that
01:36:14
they just need to release
01:36:16
somehow what
01:36:21
you're we need to get merched
01:36:24
I'm do
01:36:27
[Music]
01:36:30
all I'm going all in
