0:00

This podcast is supported by Oracle.

0:02

AI requires a lot of compute

0:04

power and the cost for your

0:07

AI workloads can spiral. That is,

0:09

unless you're running on OCI, Oracle

0:11

Cloud Infrastructure. This was the cloud

0:13

built for AI. A blazing fast

0:15

enterprise-grade platform for your infrastructure database

0:17

apps and all of your AI

0:19

workloads. Right now, Oracle can cut

0:22

your current cloud bill in half

0:24

if you move to OCI. Minimum

0:26

financial commitment and other terms apply.

0:28

Offer ends March 31st. See if

0:30

you qualify at oracle.com/Wall Street. to

2:15

so many emails. I'm Kevin Roos,

2:18

a tech columnist at the New

2:20

York Times. I'm Casey Noon

2:22

from Platformer. And this is

2:24

Hard Fork. This week, Google

2:26

is bringing AI to search in

2:28

a brand new way. How will

2:31

it change the internet? Then, the

2:33

Times, David Yaffe Bellini, joins

2:35

us to discuss the

2:37

idea of a strategic

2:39

crypto reserve, and why

2:42

even some crypto supporters

2:44

think it might be a

2:46

bad idea. And finally, we

2:49

ask you to share your

2:51

experiments in vibe coding. Today,

2:54

we're gonna find out what

2:56

you made. Every once in a while

2:58

we get a preview of a very

3:00

big change to the internet and I

3:02

think we might have gotten one of

3:04

those this week. Tell me about it.

3:06

So on Wednesday, Google said that in

3:08

addition to expanding AI overviews to more

3:10

users and upgrading the underlying model to

3:13

Gemini 2.0, it is also introducing a

3:15

brand new search mode called AI mode.

3:17

And it seems likely to me that

3:19

this may be the future of Google.

3:21

Yeah. This is, I would say, at

3:23

a minimum, a much better name than

3:25

we have seen for any AI feature

3:27

that Google has released over the past

3:30

couple of years. AI mode. It's a

3:32

mode. It does AI. Seems pretty straightforward to

3:34

me. Yeah, and where other companies might

3:36

have called this X-73 hyphen mini, Google

3:38

just picked a normal name for it,

3:40

which is something that Google does a

3:42

lot. That's true. So right now they're

3:45

calling this an early experiment. It is

3:47

rolling out now to paid subscribers to

3:49

Google One's AI premium subscription service. And

3:51

even if you pay, you also have

3:53

to opt into it through something called

3:55

Google Labs, which is a feature that

3:57

Google has to let you opt into.

3:59

new stuff. So it's kind of hidden

4:02

away. We have asked if we could

4:04

use it right away and they said

4:06

no. So we don't have a direct

4:08

report for you today. But what we

4:10

can tell you is that AI mode

4:12

is Google's answer to bringing something like

4:15

chat GPT directly into search. So so

4:17

far it has been dabbling, it's been

4:19

sprinkling these little AI overviews on top

4:21

of the traditional 10 blue links. I'm

4:23

sure you've all seen those. But while

4:26

right now AI mode is separate from

4:28

search, if it's I suspect it will

4:30

gradually merge into the main Google

4:32

search results page. And that matters more

4:34

than you might think, Kevin. Yeah, so let's

4:37

just, before we get into AI mode, let's

4:39

set the scene here. Google, big company,

4:41

most important search engine in the

4:44

world, 90% of search queries run

4:46

through Google, used by billions and

4:48

billions of people every day, indispensable

4:51

part of the architecture of the

4:53

internet for the last several decades.

4:55

A. I. L.M.'s, chat chibitee, comes

4:57

along, and all of a

5:00

sudden, people, including us, think,

5:02

well, why am I going

5:04

to ask Google this

5:06

question when I could ask

5:09

it to an A.I. chatbot

5:11

that might give me a

5:13

better answer without requiring me

5:15

to click on a bunch of

5:17

links and navigate through a bunch

5:20

of ads. I'll just get the

5:22

answer I want to. keep it

5:24

profitable and make sure that people

5:26

aren't using chat GPT or perplexity

5:28

or some other product instead. So

5:30

we want to bring AI into

5:32

our search product, but that is

5:34

complicated because among other reasons, these

5:37

AI overviews that you just described

5:39

got some things famously wrong, including

5:41

telling people to put rocks on

5:43

their pizza. Or was it glue? It

5:45

was telling them to eat rocks and

5:47

to put glue on their pizza. Two

5:49

different answers. Okay. Yeah. So they've spent

5:52

some time retooling this, but they have

5:54

not backed down on their ultimate belief

5:56

that people will use AI for search

5:58

related tasks and that be. because some

6:01

number of people will want to do

6:03

that or are already doing that. They

6:05

need to build that into the core

6:07

Google search engine. That's right. And so

6:09

Kevin, this is really a story about

6:12

Google being caught in between two imperatives.

6:14

Because on one hand, it has to

6:16

change with the times, it has to

6:18

keep pace with the competition. On the

6:20

other hand, it can't break its core

6:23

business, which generated $54 billion last quarter.

6:25

So that is a very tricky balance.

6:27

How is it going to navigate that?

6:29

thinking about it. So let's talk about

6:31

it. Okay, let's talk about AI mode.

6:34

What is it? So AI mode is

6:36

a new mode within Google search, where

6:38

if you ask a query that might

6:40

be a little bit better suited to

6:42

a chat bot, you can click over

6:45

and it is going to give you

6:47

a slightly more chat bot-like experience. But

6:49

it is different from something like the

6:51

Gemini app, which is Google sort of

6:53

straight up chat gPT competitor, and a

6:56

couple of pretty interesting ways. There are

6:58

prominent links to websites to websites with

7:00

in AI mode in what a user

7:02

interface designer would call it carousel. You

7:04

know, those little horizontally scrolling bars of

7:07

links. And so as you move through

7:09

the answer to your query in AI

7:11

mode, Google is going to say, here

7:13

are some actual websites you could visit

7:15

to get more of that information. And

7:18

you just said this is rolling out

7:20

to like a small subset of paying

7:22

Google users who have subscribed to this

7:24

like Google One AI plan. Presumably this

7:26

is something that they're interested in building

7:29

into the Google experience for everyone eventually,

7:31

right? Or is this going to remain

7:33

small? So I talked with Google yesterday

7:35

and tried to get a sense of

7:37

that and the company was loath to

7:40

make too many predictions here. This really

7:42

is an experiment. They told me and

7:44

if it goes super super badly, they

7:46

could presumably pull the plug, but I

7:48

don't think that is going to happen.

7:51

I think Google realizes that this is

7:53

a once in a generation. to reinvent

7:55

the search experience and that is going

7:57

to mean a fundamentally different way of

8:00

presenting it. information. And that's what AI mode

8:02

is. So I haven't tested out AI mode

8:04

on Google either. They have not made

8:06

it available to me. But there are

8:08

some screenshots floating around. I think maybe

8:10

we should just describe what this is

8:12

going to look like for people. They

8:14

have a screenshot of someone asking the

8:16

question, explain how deja vu works and

8:18

how it relates to memory. Now, I

8:20

got a feeling I've heard that question

8:23

before. I thought we talked about that

8:25

on a previous episode. Okay, we made

8:27

the same joke. Good for us. In

8:29

the AI mode, instead of showing the

8:31

user a list of links or an

8:33

AI overview, it sort of writes

8:35

them a little essay. It says,

8:37

Deja vu is that strange and

8:39

fleeting feeling of having experienced something

8:41

before, even though you know you

8:43

haven't. It's like your brain is

8:46

playing tricks, creating a sense of

8:48

familiarity with a situation that's brand

8:50

new, and it continues on, and

8:52

then it has a little thing

8:54

at the top right under the

8:56

query that says four sites, four

8:58

sites, But basically this is. Google's

9:00

version of what tools like ChatGBT

9:02

and Perplexity have been doing for

9:04

a while now, which is sort

9:06

of replacing the 10 blue links

9:09

altogether with this more bespoke AI

9:11

response. Yeah, and you can see

9:13

that after each paragraph, there is

9:15

a little icon of like a

9:17

chain link, which sophisticated Googleers like

9:19

you and I will recognize as

9:21

a hyperlink to a website from

9:23

which Google derives information, and which

9:25

many other people might just think

9:27

is like a fancy period. You

9:29

know what I mean? So, like,

9:31

those particular user interface elements, I

9:33

would say, are not going out of

9:36

their way to entice people to click.

9:38

Right. And we talked last year on

9:40

the show about this idea that Google

9:42

would do the Googling for you, right?

9:44

That they were very invested in trying

9:46

to, in trying to simplify the process

9:48

of searching for information on the internet

9:50

using AI, not require people to click

9:52

through any of these blue links. And

9:55

so I imagined at that they were

9:57

trying to make this available to everyone.

9:59

obstacle. including some of the mistakes

10:01

that we talked about, but also like

10:03

it is expensive to run a search

10:06

engine on a large language model this

10:08

way. These queries, they require inference from

10:10

these large models. It is not. as

10:13

cheap and easy and efficient as just

10:15

running a regular search engine. So did

10:17

you ask them about that, how they've

10:19

dealt with some of the cost concerns?

10:22

So I didn't ask them about that

10:24

this week, but they have said over

10:26

the past several months, effectively just that

10:28

the costs keep coming down here, they're

10:31

figuring out the past several months, effectively

10:33

just that the costs keep coming down

10:35

here. They're figuring out more and more

10:38

efficient. people to eat things that are

10:40

unlike eggs cannot be eaten. Right. Do

10:42

you feel like this is an acknowledgement

10:44

from them that they are losing market

10:47

share to companies like Perplexity, companies that

10:49

are offering a more AI-native search-like experience,

10:51

or that large numbers of consumers are

10:54

already using chat GPT and other AI

10:56

tools for things that they previously would

10:58

have Googled? Absolutely. You know, there was

11:00

some interesting analysis that came out in

11:03

January that showed that over the last

11:05

three months of 2024... for the first

11:07

time since 2015, Google's market share fell

11:09

below 90%. Now, 90% market share is

11:12

still incredible and Google is just barely

11:14

below it. But there is increasing evidence

11:16

that these chat bots are starting to

11:19

eat into Google's audience. Some analysts have

11:21

predicted that chat sheet BT will have

11:23

1% market share in search by the

11:25

end of this year. Bing for what

11:28

it's worth, which has been around forever,

11:30

has something like 4% market share. So

11:32

for chat cheap BT to get from

11:35

zero. to one in a little over

11:37

two years is pretty impressive and speaks

11:39

to why Google feels like it needs

11:41

to do something. Yeah, I mean, I'm

11:44

finding that I'm using Google a lot

11:46

less than I used to. I don't

11:48

know exactly how much less, but I

11:50

basically use Google now for what are

11:53

called navigational queries where you are just

11:55

like I'm looking for this one train

11:57

schedule or this one restaurant menu and

12:00

I know it's a link that's out

12:02

there on the web and so I

12:04

go to Google for that because it's

12:06

right there in my browser bar and

12:09

it's very easy and I can find

12:11

the link. I do not use Google

12:13

anymore for things like product recommendations or

12:16

advice about how to fix an appliance

12:18

in my house or something like that.

12:20

That is the kind of thing that

12:22

I will now go to ChatGPT or

12:25

Clod4 and I imagine that if I'm

12:27

doing that there are probably a lot

12:29

more people out there doing that too.

12:31

There really are and while I don't

12:34

know that this statistic speaks exactly to

12:36

what you said Kevin last month the

12:38

verge published this survey that among other

12:40

things found that 42% of people find

12:43

that search engines like Google are becoming

12:45

less useful. Now, there are a lot

12:47

of reasons. People don't like the ads

12:49

in Google. I think arguably just sort

12:52

of like the quality of websites is

12:54

probably decline. There's a lot of AI

12:56

slop out there, right? But I also

12:58

think part of it is what you

13:01

just named, which is that we now

13:03

actually have a superior technology to a

13:05

problem like, oh, the faucet in my

13:07

sink broke and I want to fix

13:09

it. I bet something like chat cheapity

13:12

can just tell me how to do

13:14

that directly. the Google blog post about

13:16

AI mode that came out this week.

13:18

And it's sort of interesting as an

13:21

artifact of cultural anxiety at Google.

13:23

They clearly know that they have

13:25

to do something big around AI.

13:27

Their competitors are doing it. It's

13:30

eating into their market share. But

13:32

they also seem a little bit

13:34

scared of it. There's this. paragraph

13:36

in there that says, as with

13:38

any early stage AI product, we

13:41

won't always get it right. For

13:43

example, while we aim for AI

13:45

responses in search to present information

13:47

objectively based on what's available on

13:50

the web, it's possible that

13:52

some responses may unintentionally appear

13:54

to take on a persona

13:56

or reflect a particular opinion. What do

13:58

you make of that? of politics.

14:00

Like to this day, if you ask

14:02

questions about politics in Gemini, in my

14:04

experience, it is more likely to refuse

14:06

to answer your request than any other

14:09

AI chatbot that I pay for, right?

14:11

They got in a lot of hot

14:13

water last year that we've talked about

14:15

on the show when, for example, they

14:17

would not appear to depict exclusively white

14:19

founding fathers when you would ask for

14:22

that. And in the wake of that,

14:24

they got a ton of criticism, and

14:26

so they try to strip as much

14:28

politics out of politics out. their products

14:30

as they could. So now they're about

14:32

to put another thing out there where

14:35

people are going to be asking it

14:37

to give them opinions. And depending on

14:39

what opinions are revealed, those people are

14:41

going to go screenshot those and put

14:43

those on social media and Congress people

14:45

are going to see them and it's

14:47

going to like sort of trigger a

14:50

whole cycle. So I understand why they

14:52

are being cautious about this, but I

14:54

think it gets a bit silly because

14:56

ultimately people are turning to these things

14:58

for their opinions. whose opinions you stand

15:00

behind. Yeah. So if you are a

15:03

paying subscriber and you get access to

15:05

AI mode on Google and you turn

15:07

it on in your Google Labs feature.

15:09

Is every query that you type into

15:11

a Google box going to be answered

15:13

by AI from then on? No, so

15:16

you are still going to have to

15:18

opt into it. So think about how

15:20

many hoops Google is making you jump

15:22

through to do this. They are truly

15:24

in that sort of beta stage where

15:26

they are truly in that sort of

15:28

beta stage where they just want to

15:31

gather some data, they want to see,

15:33

are there any obvious horrible problems that

15:35

we can identify and solve before we

15:37

release this to the entire user base?

15:39

kinds of queries at least for a

15:41

while because one I you know I

15:44

want to see how good this AI

15:46

responses are but also I'm so interested

15:48

in this question of do I click

15:50

any of the links that are you

15:52

know in this mode because if I

15:54

don't then it does seem to move

15:57

us closer to that world where the

15:59

entire economic foundation of the internet is

16:01

changing. Okay well you brought this up

16:03

so I let's diving into this issue

16:05

of how this could affect publishers, people

16:07

who make things and put them on

16:09

the internet and rely on Google as

16:12

a source of traffic. What do we

16:14

know about how AI mode is being

16:16

received by the broader internet? Are people

16:18

freaking out about this like they did

16:20

with AI overviews? There is a lot

16:22

of nervousness among publishers in particular, right?

16:25

A lot of publishers in particular, right?

16:27

A lot of people write stories, among

16:29

publishers in particular, right? A lot of

16:31

people write stories about the news and

16:33

then people search for those news queries.

16:35

It seems like that is... now happening

16:38

less and less. But Kevin, the impact

16:40

is going beyond just publishers. There was

16:42

this fascinating lawsuit that got filed last

16:44

week by a company called Cheg. Do

16:46

you know Cheg? The Homework Cheating App.

16:48

Exactly. So if you've been in college

16:51

any time in the past 15 years,

16:53

statistically you use Cheg to cheat on

16:55

your homework. It offers a database of

16:57

more than a hundred million answers to

16:59

test questions and it has been charging

17:01

students about 15 bucks a month to

17:04

use this thing. Well then along comes

17:06

Google and it's AI overviews. and using

17:08

who knows what methods, but I'm gonna

17:10

assume they did some pretty aggressive scraping

17:12

of Czech and other websites. They started

17:14

to put the answers to all these

17:16

homework questions directly in the AI overviews.

17:19

All of a sudden, there is no

17:21

need to visit Czech. And Czech Kevin

17:23

is now on life support. This is

17:25

a company that was riding high during

17:27

the pandemic. It was valued at $12

17:29

billion in 2021. And now its stock

17:31

is basically a penny stock and it

17:33

is exploring strategic alternatives, which is. sort

17:35

of corporate speak for we're going to

17:37

have to unload this thing in a

17:39

fire sale. And in their lawsuit, Cheg

17:41

said that traffic rightfully belonged to us

17:44

and Google destroyed us with AI overviews.

17:46

You really disrupted us in the homework

17:48

cheating business. I realize as I'm saying

17:50

this that Cheg is not a sympathetic

17:52

company and no one is going to

17:54

be sorry to see it go who

17:56

does not work at Cheg, but I'm

17:58

telling you this is a bellwether. check is

18:00

not going to be the last business to be completely

18:02

disrupted by AI. Yeah. So what do we know about

18:04

how people who use AI tools for search like? tasks

18:07

actually interact with links because every AI company that has

18:09

rolled out something like this whether it's perplexity or being

18:11

or Google they've all said like look people are still

18:13

gonna see links don't you worry publishers we're still gonna

18:15

put these little citations on it or maybe we'll put

18:18

some links below it and you might get less traffic

18:20

but it'll be higher quality traffic and you'll still make

18:22

a lot of money from it. So what do we

18:24

know about whether those people will write or not? that

18:27

you can make. One query is an answer where you

18:29

just need something that is good enough, right? You need

18:31

to satisfy your curiosity about something, and if it's a

18:33

little bit wrong in 5% in either direction, you don't

18:35

really care, because you're just looking for the gist of

18:38

something. I think this is the vast majority of all

18:40

queries. There is a second kind of query, which is,

18:42

I need the actual information, because my health, my money,

18:44

my job depends on it. These are the cases where

18:47

I think that people are actually clicking on the links.

18:49

I'm a journalist. I'm constantly Googling for information. I cannot

18:51

rely on an AI overview for one thing, because if

18:53

I'm going to put a statistic, a historical fact, into

18:55

my newsletter, I need to know what the original source

18:58

is. And I need to know that I trust the

19:00

original source, which means that if that link is going

19:02

to the New York Times, the Wall Street Journal, Bloomberg,

19:04

I'm going to the original website. that's a really good

19:07

thing that people should do in a lot of cases,

19:09

but I'm also not naive. That is a minority use

19:11

case for specialists. The vast majority of people, Kevin, do

19:13

not need that level of clarity and they are not

19:15

clicking those dang links. Right. There was a study that

19:18

came out this week that I found totally fascinating and

19:20

quite worrying. It was done by a company called Tolbit.

19:22

They're basically an AI licensing company that sort of works

19:24

as the go between, between public... and companies

19:26

making large language models. And

19:29

they found in their study

19:31

that the AI search interfaces

19:33

they looked at, things that

19:35

are similar to Google's AI

19:38

mode, delivered 91 % less

19:40

click through traffic than standard

19:42

Google searches. And when they

19:44

looked at just pure AI

19:46

chatbots, it was even worse.

19:49

Those generated 96 % fewer clicks

19:51

to links than traditional Google

19:53

searches. Casey, I'm no mathematician,

19:55

but that's pretty bad. It's

19:58

pretty bad. And I should

20:00

say I have some questions

20:02

about the methodology here, you

20:04

know, basically told me it

20:06

took the amount of traffic

20:09

to its publisher sites that

20:11

it could trace from known

20:13

AI scrapers and divided it

20:15

by the total number of

20:18

times that those scrapers hit

20:20

their sites overall. I talked

20:22

to Google about this. Google

20:24

was like, this is a

20:26

really bad way of measuring

20:29

traffic because like Google's crawler

20:31

is constantly scanning websites. Like

20:33

every time the Google crawler

20:35

hits a website, that is

20:37

not an actual Google search.

20:40

So there's some uncertainty here

20:42

about what this really means.

20:44

But at the same time,

20:46

Kevin, these numbers are intuitive

20:49

because you and I both

20:51

use chatbots a lot and

20:53

we know that we're not

20:55

clicking those links a lot

20:57

of the time, particularly in

21:00

cases where we are not

21:02

using the chatbot to do

21:04

our jobs. Yes, yeah. No,

21:06

and I think it's worth

21:09

saying, like even if these

21:11

predictions, these numbers were way

21:13

off, say it was only

21:15

half as bad as Tolbit,

21:17

says it is, and that

21:20

these AI search tools only

21:22

produce about 45 % less

21:24

traffic than the Google searches

21:26

that preceded them, that would

21:29

still be a cataclysmic event

21:31

for much of the media

21:33

industry. Totally. Now, Google has

21:35

said, when I've asked about

21:37

this, that the AI overviews,

21:40

which again is not AI

21:42

mode, but AI overviews are

21:44

sort of like the walkup

21:46

to this AI mode, those

21:48

have led people to search

21:51

more in general, right? So

21:53

Google says basically people like

21:55

AI overviews and they see

21:57

them and they search more

22:00

and that leads to maybe

22:02

more searches in the aggregate,

22:04

but the impact on traffic

22:06

to individual publishers. does not seem to have

22:08

been positive so far. Yeah, I mean, I want to

22:11

play devil's advocate here for a little bit because you

22:13

and I and others have been warning about these AI

22:15

overviews and these AI search products for more than a

22:17

year now, have been predicting that this would sort of

22:20

crater the traffic to publications, would dramatically change the way

22:22

that billions of people interact with the internet. And I

22:24

think it's fair to say so far, that has not

22:26

happened, at least at the scale that. Maybe

22:28

you or I thought it might by now.

22:30

And I think there are some people

22:32

who might be listening to this

22:35

and saying, you guys are panicking

22:37

over nothing. One thing that is

22:39

true that we know is that

22:41

people's habits change very slowly. There

22:44

are still millions of people out

22:46

there using AOL and Yahoo, you

22:48

know, for their email, maybe hot

22:50

mail. These very old, sort of

22:53

we would say antiquated services still

22:55

have. dedicated users because people are

22:57

just used to going to their

22:59

browser and typing in the same

23:02

websites that they've been typing in for

23:04

years. So I think there is a

23:06

case that I can imagine people making

23:08

here that yes these AI features these

23:10

AI search engines they are going to

23:12

appeal to like power users in the

23:14

Bay Area who spend all day looking

23:16

stuff up online, but actually it is

23:18

not an existential threat to the internet

23:20

or to publishers because most people are

23:22

so used to going to Google that

23:24

they are just going to continue to

23:26

do that even if the superior option

23:28

exists. So that is true, which is

23:30

why if one day AI mode is

23:32

no longer a little hidden feature and

23:35

is just actually the front door to

23:37

Google search, then. That's the ball game

23:39

because people will not have to change their

23:41

entire habits in order to have this experience.

23:43

It's just going to be the default. We

23:45

know the power of defaults, right? So that's

23:47

why I want to talk about this today

23:49

is because this could be a preview of

23:52

what Google is in two, three, five years,

23:54

and it's going to be very different from

23:56

the Google yet today. How do you search

23:58

these days? Like, what's your go- if you

24:00

have a question, let's say, let's take

24:02

it out of the realm of work.

24:04

So you're just looking for a good

24:06

restaurant to go. Where do you go?

24:09

So for something where it's like find

24:11

a local business, to me that is

24:13

still an area where Google excels. You

24:15

know, it's like I need to find,

24:17

you know, an eye doctor. I want

24:19

to find, you know, a restaurant that's

24:22

nearby. I'm going to go to Google

24:24

or Google Maps, something like that. set

24:26

of queries now that I am turning

24:28

to chat bots for. I've talked before

24:30

on the show about how I just

24:32

have like a little keyboard shortcut that

24:34

I type in and I can query

24:37

an AI directly. And that's everything from

24:39

how to. It is essentially just random

24:41

trivia. You know, I'm trying to remember,

24:43

you know, when did this album come

24:45

out or like how old is this

24:47

celebrity? And crucially, Kevin, a lot of

24:49

those queries are things that Google does

24:52

still do very well. But there is

24:54

a convenience to just getting a one

24:56

sentence. and not seeing a sprawling web

24:58

page that's full of, you know, six

25:00

different ads and various other widgets that

25:02

I don't care about. So it is

25:04

a mix of things, but chatbot usage

25:07

when it comes to search is on

25:09

the increase for me. How about you?

25:11

Yeah, I'm using chat bots for, I

25:13

would say, at least half of what

25:15

I would previously have used Google for.

25:17

Now, I am not a traditional user.

25:20

I am sort of an early adopter

25:22

of this stuff. I have a whole

25:24

bookmark folder of AI tools that I

25:26

open up every morning and start using

25:28

them for browsing. So I'm not conventional

25:30

in a sense, but I do find

25:32

that for the majority of things that

25:35

I'm looking for, chat bots tend to

25:37

give me better answers than Google. When

25:44

we come back, we have

25:46

reserved some time to talk

25:49

about the strategic crypto reserve.

26:06

Whether you're starting or scaling your

26:08

company's security program, demonstrating top-notch security

26:10

practices and establishing trust is more

26:12

important than ever. Vanta automates compliance

26:14

for SOC 2, ISO 27001, and

26:16

more. With Vanta, you can streamline

26:19

security reviews by automating questionnaires and

26:21

demonstrating your security posture with a

26:23

customer-facing trust center. Over 7,000 global

26:25

companies use Vanta to manage risk

26:27

and prove security in real-time. Get

26:30

$1,000 off Vanta when you go

26:32

to vanta.com/hard fork. That's vanta.com/hard fork

26:34

for $1,000 off. This is Someni

26:36

San Gupta. I'm a reporter for

26:38

the New York Times. I've covered

26:40

nine conflicts written about earthquakes, terror

26:43

attacks, droughts, floods, many humanitarian crises.

26:45

My job is to bear witness.

26:47

Right now, I'm writing about climate

26:49

change. And I'm trying to answer

26:51

some really big and urgent questions

26:54

about life on a hotter planet.

26:56

Like, who is most vulnerable to

26:58

climate change? Should we redesign our

27:00

cities? Should we be eating differently?

27:02

What happens to the millions of

27:04

people who live by the coast

27:07

as the oceans rise? To make

27:09

sense of this, I talk to

27:11

climate scientists and vendors, activists. Mostly,

27:13

I document the impact of global

27:15

warming. And that impact is highly,

27:18

highly unequal. My colleagues and I

27:20

are doing our best to answer

27:22

complicated questions like these, but we

27:24

can't do that without our subscribers.

27:26

If you'd like to subscribe, go

27:28

to nytimes.com/subscribe. And thank you. to

27:31

say about what is happening in

27:33

Washington right now. Elon Musk and

27:35

his Doge team are fanning out

27:37

across the federal bureaucracy, slashing and

27:39

burning wherever they go. But the

27:42

story that I think we can

27:44

really shed some useful light on

27:46

here is about what's been happening

27:48

with crypto and crypto policy. Yeah,

27:50

Kevin. We heard some news over

27:53

the weekend that President Trump was

27:55

interested in something called a crypto

27:57

strategic reserve, and we thought that

27:59

sounds just silly enough that it

28:01

could be a hard-for segment. Yes.

28:03

And then on Thursday night, just

28:06

as we were finishing up this

28:08

episode, news broke that President Trump

28:10

had signed an executive order establishing

28:12

a strategic Bitcoin reserve. And that

28:14

announcement and the executive order attached

28:17

to it clarified a few very

28:19

important things. One is that this

28:21

reserve will be filled with Bitcoin

28:23

that is already owned by the

28:25

federal government that was seized or

28:27

forfeited as part of criminal or

28:30

civil asset forfeiture proceedings. Second thing

28:32

is that the Secretaries of Treasury

28:34

and Commerce, according to this executive

28:36

order, will be authorized to develop

28:38

what they call budget neutral strategies

28:41

for acquiring more Bitcoin. They say

28:43

that this will not have an

28:45

additional cost on the American taxpayer.

28:47

And then the third thing that

28:49

the executive order did is it

28:51

established a separate US digital asset

28:54

stockpile, which consists of cryptocurrencies and

28:56

other digital assets that are not

28:58

Bitcoin. And with respect to this

29:00

stockpile, the executive order said that

29:02

the government won't per... any of

29:05

these other non-bit coin crypto assets,

29:07

but that if it gets some

29:09

in the case of a criminal

29:11

or civil proceeding, it will keep

29:13

it and hold on to it

29:15

as part of the stockpile. So

29:18

we just want to note that

29:20

that happened and that this conversation

29:22

you're about to hear took place

29:24

before the order was official when

29:26

a few more things were still

29:29

up in the air. But I

29:31

think it's fair to say. a

29:33

really big deal. It absolutely is.

29:35

Those currencies, Kevin, are closely tied

29:37

to people in the Trump orbit

29:39

who have investments in those very

29:42

currencies. And so this seems like

29:44

a classic case of self-dealing among

29:46

people in and around the president's

29:48

orbit. Yeah. And so I think

29:50

there's an element of just shedding

29:53

light on what is happening in

29:55

Washington right now with regards to

29:57

crypto policy that is important for

29:59

us as journalists to do. But

30:01

I also am growing more... worried

30:03

about what I see as an

30:06

information and knowledge gap between the

30:08

pro-krypto proponents and the anti-krypto skeptics.

30:10

Would you say that you have

30:12

some reservations? Yes, I have some

30:14

strategic reservations about this. You know,

30:17

back in 2022 I was covering

30:19

crypto much more attentively than I

30:21

am now. This was sort of

30:23

at the height of the last

30:25

crypto boom. This is your pudgy

30:28

penguin's era. I wrote this article,

30:30

a very long article, and I

30:32

basically was arguing that people, whatever

30:34

they thought of crypto, needed to

30:36

understand it, because this is something

30:38

that could become strategically important if,

30:41

for example, a bunch of pro-krypto-partisans

30:43

and activists were to seize control

30:45

of some part of the federal

30:47

government and start making big moves

30:49

to enrich themselves and the people

30:52

around them. And the response I

30:54

got to that article was a

30:56

lot of people saying... Shame on

30:58

you. Shame on you for taking the

31:00

stuff seriously, for trying to explain it,

31:02

for trying to understand it. The mere

31:05

act of wanting to know about crypto

31:07

was sort of coded as being in

31:09

the pocket of the crypto industry. And

31:11

we're in a moment where just ignoring

31:14

it is not going to make crypto

31:16

go away. We know that now, and

31:18

now that we see these really wild

31:20

initiatives being proposed, I think it is

31:23

time once again to see what we

31:25

can learn and to engage with. somebody

31:27

who can hopefully explain what's going on.

31:29

Yes, I think it is time for

31:32

a lot of crypto skeptics, and I

31:34

would include myself in that group, to

31:36

give what's happening in Washington more serious

31:38

attention, because these are big moves with

31:40

big consequences, and they may be hard

31:42

to untangle even after Donald Trump leaves

31:44

office. So today, we're bringing in our

31:46

friend and friend of the pod. David

31:49

Yaffe Bellini. He's a crypto reporter at

31:51

the New York Times, and he has

31:53

been covering all of the twists and

31:55

turns of this latest saga over the

31:57

crypto strategic reserve. Let's bring him

31:59

in. Dave Yaffe Bellini. Welcome back to

32:02

Hard Fork. Thanks so much for

32:04

having me. So let's start with

32:06

some basic questions about this idea

32:08

of a strategic United States crypto

32:10

reserve. When did you start to

32:12

hear about this and what is

32:14

the basic idea? So this was

32:16

an idea that popped up on

32:18

the campaign trail last year. President

32:20

Trump started talking about it, some

32:22

kind of influential figures, and the

32:24

crypto Twitter sphere were talking a

32:26

lot about it. But what it

32:28

actually meant was pretty obscure then

32:30

and remains pretty obscure now, because

32:32

when people say Bitcoin Reserve or

32:34

crypto reserve, they really could be

32:36

talking about 10 different sorts of

32:38

ideas. Right. And I want to

32:40

understand this idea of a strategic

32:42

reserve, because my basic impression is

32:44

that this is not a new

32:46

idea. Governments have... long stockpiled assets

32:48

like gold or oil or other

32:50

foreign currencies to basically give themselves

32:52

some cushion against economic downturns or

32:54

inflation or just maybe running out

32:56

of oil during a foreign conflict.

32:58

So is this similar to those

33:00

kind of strategic reserves? Why do

33:02

people in crypto think that we

33:04

need a US crypto reserve? crypto

33:06

can't power heavy machinery, it can't

33:08

feed people, it's not useful in

33:10

the way that some of these

33:12

other assets are. I think there

33:14

are two key arguments that people

33:16

in the crypto world are making

33:18

and the sort of more reserve-like

33:20

argument for this is that at

33:22

the moment the reserve currency of

33:24

the world is the dollar, the

33:26

world runs on the dollar, but

33:28

maybe not so far in the

33:30

future, the world will run on

33:32

Bitcoin or on cryptocurrencies more broadly,

33:34

and it would be in the

33:36

strategic interests of the United States

33:38

to have a giant. stockpile of

33:40

crypto. What would it mean though?

33:42

Okay, so let's say we're living

33:44

in fantasy land where the world

33:46

now runs on Bitcoin. How does

33:48

it benefit the United States to

33:50

just own a bunch of Bitcoin?

33:52

You know, it's money that you

33:54

could borrow against. It's money that

33:56

you could... spend to fund all

33:58

sorts of projects either foreign or

34:00

domestic, but you're asking the right

34:03

question because it's not totally clear

34:05

how this would work. Okay. The

34:07

other argument that people in the

34:09

crypto world make, I think, aligns

34:11

more with this kind of concept

34:13

of a sovereign wealth fund, which

34:15

Trump has also said that he

34:17

wants to do. And that's the

34:19

idea that crypto is just a

34:21

good investment, and that if we

34:23

bought a bunch of Bitcoin now

34:25

at about $90,000, then 10 years

34:27

from now, when it's worth $20

34:29

trillion, you know, will be set

34:31

up for success. I'm curious how

34:33

this idea of a strategic crypto

34:35

reserve fits with some of the

34:37

Trump administration, which is to... cut

34:39

costs and reduce the size of

34:41

government, not invest a bunch of

34:43

money in some new class of

34:45

asset. The funniest post I saw

34:47

about this said the following, sir,

34:49

we have saved $12 billion that

34:51

were spent on condoms for Gaza

34:53

and transgender comic books in Brazil.

34:55

Excellent, let's buy some Cardano. So

34:57

how are people in the crypto

34:59

policy world reconciling these two views

35:01

of what the government should be

35:03

doing? Well, some of them just

35:05

aren't reconciling it. They're just embracing

35:07

the hypocrisy, and it's not the

35:09

first time they've done something like

35:11

that. But also, there's a huge

35:13

split in the crypto world over

35:15

whether this should be a Bitcoin

35:17

exclusive reserve or whether it should

35:19

be a reserve featuring a bunch

35:21

of different cryptocurrencies. Now, on the

35:23

campaign trail at the Bitcoin Nashville

35:25

conference, Trump said that he was

35:27

going to start a Bitcoin stockpile.

35:29

When he raised this topic again

35:31

a few days ago on truth

35:33

social, he named all these other

35:35

tickers for various cryptocurrencies that the

35:37

people around him have economic interests

35:39

in and said, oh, it's going

35:41

to involve all of these too.

35:43

And a lot of people in

35:45

the crypto world said, no, we

35:47

love Bitcoin. Bitcoin is the most

35:49

established of these assets. It's the

35:51

one that has the kind of

35:53

longest term potential. So, you know,

35:55

don't fill your government coffers with

35:57

Cardano. about a Bitcoin strategic reserve,

35:59

I had actually heard of this

36:01

before because am I right David

36:03

that Al Salvador also built a

36:05

Bitcoin strategic reserve? Yeah, and actually,

36:07

you know, other countries are exploring

36:09

this too. I mean, there are

36:11

various states where there's proposed legislation

36:13

pending that would create Bitcoin reserves,

36:16

but it tends to be focused

36:18

on Bitcoin, which to be clear

36:20

is by far the most valuable.

36:22

cryptocurrency and the one with the

36:24

longest track record. Right. So I

36:26

had at least heard of the

36:28

idea before, but then in this

36:30

post that Trump makes on True Social

36:32

on Sunday, he says, we also want

36:34

to put Solana XRP and Cardano into

36:36

this strategic reserve. David, can you give

36:39

us a little bit of flavor about

36:41

what Solana XRP and Cardano are? So

36:43

these are three examples of what people

36:45

in the crypto world call alt-coins, which

36:47

are, you know, cryptocurrencies that are a

36:50

little bit out of the kind of

36:52

mainstream of Bitcoin. But these are cooler.

36:54

They play smaller clubs. They're not your

36:56

Madison Square Garden crypto currencies. But these

36:59

are not, this isn't like, you know,

37:01

some random coin that got invented a

37:03

day ago. I mean, XRP that's run

37:05

by a company called Ripple, which has

37:07

been a huge player in the crypto

37:10

world for a long time, had this

37:12

really important legal battle with the SEC.

37:14

Solana is one of the top cryptocurrencies.

37:16

It's a competitor to Ethereum. It's the

37:19

platform on which the Trump and Malania

37:21

meme coins were built, perhaps not

37:23

coincidentally. And then Cardano was created by

37:25

a guy who has been a huge

37:28

player in crypto for a long time,

37:30

was involved in the early Ethereum movement.

37:32

So like, these aren't crazy coins, but it's

37:35

a little bit weird to talk about. putting

37:37

them in a government stock pile. It's one

37:39

thing to say these are alt coins that

37:41

have been around a while, they're in fairly

37:44

wide distribution, and another thing to say there

37:46

is strategic value to the United States in

37:48

accumulating a huge stockpile of these things. Yes,

37:50

and I mean like with a lot of

37:53

things that Trump says it's not clear how

37:55

serious he was because he says things sort

37:57

of off the cuff. I mean his first...

37:59

you know, named those kind of three

38:02

tickers and then there was another, another

38:04

treat we was like, oh, also Bitcoin

38:06

and Ethereum will be in it too,

38:08

as if as like an afterthought, you

38:10

know. And so it's tough to read,

38:13

you know, what he's actually saying. The

38:15

other, the other thing that's not clear

38:17

at all. I mean, he said on

38:19

the campaign trail he would create a

38:22

Bitcoin stockpile built on Bitcoin that the

38:24

government already owns, because we've seized a

38:26

lot of Bitcoin from criminals over the

38:28

years. All he said on the campaign

38:30

trail was, we're going to draw a

38:33

box around the Bitcoin we already have

38:35

and not sell it. It's very different

38:37

from saying we're going to hit up

38:39

the open market and buy a ton

38:42

of salana. Yeah, how would that actually

38:44

work in practice? Because I understand that

38:46

the United States does have this Bitcoin

38:48

that it's seized from people who commit

38:50

various crimes and is presumably keeping in

38:53

some secure wallet somewhere. to acquire new

38:55

crypto assets would they just like go

38:57

on to like coin base and buy

38:59

a bunch and like keep it in

39:02

a vault in the treasury building or

39:04

like how would we actually go about

39:06

this as a country? Well when you

39:08

buy a big tranche of cryptocurrency as

39:10

like a huge institution you're usually doing

39:13

it as what's called an over-the-counter trade

39:15

where you're sort of dealing directly with

39:17

another person that you're buying it from

39:19

and the idea is to try to

39:22

like limit the impact. like the Treasury

39:24

Secretary will log on to his coin

39:26

base account and just start hitting the

39:28

buy button. But how would those purchases

39:30

be funded? I mean, none of this

39:33

has been fully explained. And obviously, any

39:35

government spending raises the possibility of tax

39:37

increase or spending some of these perhaps

39:39

imaginary dose savings that we've supposedly been

39:42

accumulating. And so it's all really confusing.

39:44

Yeah. How are people in the crypto

39:46

community responding to this idea of a

39:48

strategic crypto reserve? The response has been,

39:50

I think, fairly negative from a lot

39:53

of people who you might have expected

39:55

to be supportive of it. And that's

39:57

because there's sort of a feeling in

39:59

the crypto world that Wow, this might

40:02

actually be a little bit of a

40:04

conflict of interest. It's like a bunch

40:06

of people are discovering that that's a

40:08

problem in the Trump kind of political

40:10

universe. You know, people are noticing, oh,

40:13

Brad Garlinghouse, the chief executive of Ripple,

40:15

spent a lot of time at Maralago

40:17

before the inauguration and was a huge

40:19

donor to Fairshake, the big crypto pack,

40:22

and suddenly the signature cryptocurrency used by

40:24

Ripple XRP is going to be in

40:26

the crypto reserve. And so people have

40:28

become suspicious of those sorts of behind-the-scenes

40:30

maneuverings and there's also just a lot

40:33

of like loyalty to Bitcoin in the

40:35

crypto world still and in a sense

40:37

that you know we should be treating

40:39

Bitcoin differently from this other stuff. It's

40:42

like it's not all that different than

40:44

if Mark Zuckerberg spent a lot of

40:46

time with Trump at Maralago and then

40:48

President Trump announced that we were going

40:50

to create a strategic Facebook stock reserve

40:53

of just shares of Facebook, right? I

40:55

don't really think there's much difference between

40:57

those things that is likely to happen

40:59

if... And when this strategic crypto reserve

41:02

is created, is that anything that is

41:04

in the basket of stuff that we

41:06

stockpile as a country will become more

41:08

valuable just by virtue of the fact

41:10

that we are stockpiling it as the

41:13

United States government. Yeah, I mean, this

41:15

stuff became more valuable by virtue of

41:17

the fact that Trump tweeted about it,

41:19

right? I mean, the prices of these

41:22

assets immediately went up, though I think

41:24

some of those gains have been paired

41:26

back since then. But yes, I mean,

41:28

there's like a concrete market of effect.

41:30

If someone starts buying a lot of

41:33

an asset, the price will go up,

41:35

so that will happen. But the symbolic

41:37

statement of the US putting its government

41:39

seal on Cardano would be big for

41:42

anyone who currently holds Cardano. Right. And

41:44

I was surprised to see some of

41:46

the people coming out in the crypto

41:48

industry against this proposal. Brian Armstrong, the...

41:50

co-founder and CEO of coin base made

41:53

a post basically saying that just having

41:55

Bitcoin in the strategic reserve quote would

41:57

probably be the best option He didn't

41:59

want all these other coins included Tyler

42:02

Winklevos one of the biggest Bitcoin investors

42:04

in the world also said that he

42:06

doesn't think other Crypto coins should be

42:08

in the strategic reserve is a bad

42:10

idea. And I just want to read

42:13

you a line from this because I

42:15

think it illustrates some of where these

42:17

people are coming from. He writes, a

42:19

crypto reserve would transform Bitcoin from an

42:22

apolitical asset into the placing of the

42:24

government subject to Washington's political cycles. Bitcoiners

42:26

were never ones to hitch their wagon

42:28

to the government, and they shouldn't start

42:30

now. So it seems, David, like there's

42:33

essentially this coalition of more libertarian crypto

42:35

supporters who think... We actually don't want

42:37

to be embraced by the federal government

42:39

in this way because that could undermine

42:42

some of the sort of libertarian ideology

42:44

of the initial crypto wave. Yeah, I

42:46

mean, like you said, Nick was opposed

42:48

even to the idea of a Bitcoin

42:50

reserve before this whole notion of a

42:53

crypto reserve came up. And so it's

42:55

definitely been divisive. I mean, crypto, Bitcoin

42:57

started as this, you know, renegade economic

42:59

movement. And so... to be kind of

43:01

begging the government to like buy up

43:04

a huge stockpile so that the price

43:06

goes up seems sort of antithetical to

43:08

a lot of those principles and you

43:10

know most people in the industry don't

43:13

really care about those principles anymore but

43:15

some still do. Well let's talk about

43:17

this more cynical take that this is

43:19

all just kind of a self-enrichment scheme

43:21

by the Trump administration. I know about

43:24

the Trump meme coins, we've talked about

43:26

those, I know that a lot of

43:28

Trump's supporters and friends are heavily invested

43:30

in the crypto industry, but what do

43:33

we actually know about the Trump family's

43:35

personal interests in crypto beyond this small

43:37

set of meme coins? Like I've seen

43:39

Eric Trump tweeting on and off about

43:41

crypto and various coins that he's interested

43:44

in. Do we know much about... the

43:46

Trump family's overall exposure to the crypto

43:48

markets. So we don't have a window

43:50

into like Eric Trump's personal crypto portfolio?

43:53

But what we do know is that

43:55

World Liberty Financial, the crypto business that

43:57

the Trump family helped start last year,

43:59

in which it profits directly from, has

44:01

over the last few months accumulated a

44:04

huge stash of various cryptocurrencies. It was

44:06

supposed to launch some sort of crypto

44:08

application, defy project. That hasn't come to

44:10

fruition yet, but this firm is sitting

44:13

on a huge stash of tokens. So

44:15

there's arguably like a financial. benefit for

44:17

the Trump family and seeing those tokens

44:19

rise in value. Right. And some of

44:21

the people that President Trump has tapped

44:24

to lead crypto policy efforts in the

44:26

federal government, also have invested at various

44:28

points in crypto stuff. David Sachs, one

44:30

of Trump's advisors, has made some crypto

44:33

investments over the years. He's said that

44:35

he's divested from those to prevent conflicts

44:37

of interest, but there are other people

44:39

in the administration who presumably still do

44:41

have active investments in these categories. So

44:44

what do people in crypto think of

44:46

the argument that this is all just

44:48

a bunch of insiders trying to use

44:50

the powers of the government to enrich

44:53

themselves? People in the crypto world are

44:55

super cynical and love enriching themselves at

44:57

all times, but they're also super paranoid

44:59

and suspicious and are constantly convinced that

45:01

there's some sort of conspiracy to screw

45:04

them over. And so, you know, those

45:06

two instincts have kind of clashed and

45:08

you can sort of see the internal

45:10

battle playing out in a lot of

45:13

crypto people who like Trump where they

45:15

ignore the conflict of interest that benefits

45:17

them when it means SEC suits getting

45:19

dropped, but they pay attention to it

45:21

when it means that like... their favorite

45:24

crypto didn't make it into the reserve,

45:26

but someone else's did. And so I

45:28

think that's sort of the like conflict

45:30

that's playing out in the industry and

45:33

why there's been some public backlash. Yeah.

45:35

So how likely do we think the

45:37

strategic crypto reserve is to actually happen?

45:39

Does Congress need to do anything? Can

45:41

Trump just do this on his own

45:44

by executive action? Like, what are we

45:46

going to see in the next couple

45:48

of weeks and months? Yeah, I mean,

45:50

you know, the way crypto people have

45:53

sort of made this distinction have sort

45:55

of made this distinction is, is, is,

45:57

is, is, is, And if it's a

45:59

stockpile, what they mean is we're just

46:01

keeping the crypto we already have, and

46:04

if it's a reserve, that means we're

46:06

buying more crypto on the market. I

46:08

don't think there's anything inherent in those

46:10

words that means those things, but that's

46:13

how crypto people talk about it. And

46:15

so if the Trump administration goes the

46:17

stockpile route and just keeps the Bitcoin

46:19

it already has, that won't require any

46:21

spending. That just requires a change in

46:24

policy. Let's stop trying to sell this

46:26

stuff. You know, as far as we

46:28

know, the government doesn't actually own any

46:30

cardano right now, so it would actually

46:33

have to buy it to put it

46:35

in the reserve. If the government goes

46:37

that direction, then yeah, I mean, there's

46:39

a question of how you would fund

46:41

that. Congressional appropriation is the most obvious

46:44

route, especially if it's a huge purchase,

46:46

but since the summer, people in the

46:48

crypto world have floated like various kind

46:50

of out there legal ideas about how

46:53

maybe Trump could kind of push this

46:55

through on his own. Okay, so DIY,

46:57

obviously a lot of what's getting attention

46:59

in crypto these days is related to

47:01

meme coins and strategic reserves and whatnot,

47:04

but there's also some quieter, because it's

47:06

not frankly that interesting. One of them

47:08

is this stable coin bill that actually

47:10

looks like it might pass with bipartisan

47:13

support. Tell us about that. So one

47:15

of the big priorities of the crypto

47:17

industry in this administration is to... basically

47:19

convert all the political goodwill it has

47:21

into legislation and there are two key

47:24

pieces of legislation that the industry wants

47:26

to advance. One is this stable coin

47:28

bill which would essentially create rules for

47:30

you know stable coin companies to sort

47:33

of operate in the US. Wouldn't do

47:35

anything super crazy but it would just

47:37

you know essentially create a regulatory framework

47:39

for stable coins that would sort of

47:41

put in some way the kind of

47:44

government seal of approval on this branch

47:46

of the crypto industry which is super

47:48

important. And so that's something that the

47:50

industry has been pushing for. and which

47:53

you know critics say look I mean

47:55

this could be like a kind of

47:57

gateway drug to the US allowing other

47:59

even more dangerous crypto stuff to kind

48:01

of seep into the mainstream economy so

48:04

that's that's the first priority the second

48:06

priority is a market structure bill and

48:08

basically what that means is legislation that

48:10

would strip power away from the SEC,

48:13

which has obviously been super hard on

48:15

crypto, and give it to the CFTC,

48:17

the much kind of weak or less

48:19

aggressive agency. So those are the sort

48:21

of the two big legislative things that

48:24

crypto people want to do. So I

48:26

want to bring this back to the

48:28

question of the information gap that exists

48:30

in crypto right now. experience and maybe

48:33

your experience too is that the people

48:35

who understand crypto the deepest who can

48:37

talk about the L2s and the stable

48:39

coins and the decks is most of

48:41

them are invested in crypto in some

48:44

way which is why they have taken

48:46

the time to understand all the extremely

48:48

complex parts of the crypto ecosystem but

48:50

what worries me about that is that

48:53

there is essentially no principled opposition left

48:55

it seems to me in Washington DC

48:57

the people who are making the policy

48:59

who are having the debates about the

49:01

policy They all come from the same

49:04

universe of people who are bullish by

49:06

and large about crypto. Maybe they have

49:08

some disagreements about which crypto assets should

49:10

be included in a strategic reserve or

49:12

exactly how some regulations would be written.

49:15

But there doesn't seem to me to

49:17

be anyone left in Washington who knows

49:19

what they're talking about and can stand

49:21

up to some of these schemes from

49:24

the industry. Is that an accurate read

49:26

of the situation? Yeah, I mean, partly

49:28

this is a function of the changing

49:30

administration. I mean, partly this is a

49:32

function of the changing administration. I mean,

49:35

there were a huge number of people

49:37

in Washington who knew about crypto and

49:39

were in positions of power. I mean,

49:41

Gary Gemsler himself taught an MIT course

49:44

on MIT course on crypto. Now he's

49:46

out and the crypto people sort of

49:48

have have the run of the town.

49:50

I also think there's an element of

49:52

fatigue here that parallel. kind of like

49:55

the broader fatigue over what's going on

49:57

with Trump. I mean, some of the

49:59

loudest crypto critics, you know, people like

50:01

the actor Ben McKenzie, who was super

50:04

vocal during the last kind of bull

50:06

run and crypto crash, you know, those

50:08

people sort of feel like they made

50:10

their point. I mean, they were screaming

50:12

about how dangerous crypto was and then

50:15

the whole thing crashed and Sam Bankman

50:17

freed went to prison and they were

50:19

proved right in a lot of ways.

50:21

And maybe part of that is just

50:24

the assumption that when that crash happened,

50:26

it seemed like the people who were

50:28

affected were the ones that had decided

50:30

to take the gamble on crypto, whereas

50:32

people who just ignored the whole thing

50:35

were basically fine. And so maybe that's

50:37

why people are sort of prioritizing their

50:39

energy a little bit differently. What I

50:41

think is concerning though is what you

50:44

said, which is, look, if we pass

50:46

a stable coin bill in this country,

50:48

if crypto becomes sort of more connected

50:50

to the main economy, then all of

50:52

a sudden if there's another crash, if

50:55

there's another crash, normal people might start

50:57

to feel a pain. Yeah, absolutely. Crypto

50:59

used to be walled off from the

51:01

real economy, and it still is in

51:04

a lot of important ways, but that's

51:06

beginning to change. The walls are coming

51:08

down. We already have a Bitcoin ETF.

51:10

You know, we could have an XRP

51:12

ETF soon, a Solana ETF. I mean...

51:15

that just creates more avenues for people

51:17

to put their kind of like traditional

51:19

savings into crypto. If your employer tells

51:21

you that they're gonna start paying you

51:24

in Cardano, look for another job. That's

51:26

what I'm gonna say. I will say

51:28

that I think there's also an element

51:30

of complexity here that makes it very

51:32

hard for people who are not spending

51:35

hours a day trying to keep up

51:37

with what's going on on crypto to

51:39

make sense of it all. It reminds

51:41

me a little bit of what happened

51:44

on Wall Street during the financial crisis

51:46

where you suddenly had these assets that

51:48

no regular people had ever heard of,

51:50

mortgage-backed securities, credit default swaps, like synthetic

51:52

bonds, like these things that were very

51:55

esoteric and hard to understand, all of

51:57

a sudden became quite important and the

51:59

people who understood. them had been trading

52:01

them for years trying to make a

52:04

bunch of money but they had not

52:06

been sharing that information because it was

52:08

not in their interest to share that

52:10

information. They didn't want us to know

52:12

what was going on. And so my

52:15

fear is that there's something similar happening

52:17

in crypto right now where the people

52:19

who really understand this stuff are making

52:21

a killing on it possibly at the

52:24

expense of the rest of us. Yeah.

52:26

Yeah and you know the complexity was

52:28

a barrier to regular people getting into

52:30

it for a long time. But you

52:32

know, with these ETFs coming onto the

52:35

market, you know, suddenly there's a way

52:37

for you to buy this from your

52:39

brokerage account, to have it kind of

52:41

factored into your retirement investments, it just

52:44

increases the level of exposure. Yeah. Well,

52:46

DIY. Thanks for coming on. Do you

52:48

have to see you? YVM. Thanks for

52:50

having me. When

52:53

we come back, we'll follow up

52:55

on our episode about vibe coding

52:57

and what you built with your

53:00

vibe coding tools. It's time to

53:02

share with the class. Whether

53:22

you're starting or scaling your company's

53:25

security program, demonstrating top-notch security practices

53:27

and establishing trust is more important

53:29

than ever. Vanta automates compliance for

53:32

SOC 2, ISO 27001, and more.

53:34

With Vanta, you can streamline security

53:36

reviews by automating questionnaires and demonstrating

53:38

your security posture with a customer-facing

53:41

trust center. Over 7,000 global companies

53:43

use Vanta to manage risk and

53:45

prove security in real-time. Get a

53:48

thousand dollars off Santa when you

53:50

go to vanta.com/hard fork. That's vanta.com/hard

53:52

fork for a thousand dollars off.

53:54

New York Times things make me

53:57

feel like I'm amazing. Where it'll

53:59

make me feel things that I

54:01

don't feel from anyone else? I

54:04

absolutely love spelling bee. The Times

54:06

Crosser puzzle is a companion that

54:08

I've had longer than anyone outside

54:10

of my immediate family. When I

54:13

can finish a hard puzzle without pins,

54:15

I feel like the smartest person in

54:17

the world. It gives me joy every

54:19

single day. Join us and play all New

54:21

York Times games games at

54:23

nyimes.com/ games. Subscribe by March 16th

54:25

to get a special offer. Well

54:28

Casey a couple weeks ago on

54:30

the show we talked about AI vibe

54:32

coding This is this idea of using

54:34

AI tools to Build apps and websites

54:37

and other things for yourself even if

54:39

you don't know how to code I

54:41

talked about some of the projects I'd

54:44

been building I built you a hot

54:46

tub time machine for your hot tub

54:48

and we asked listeners to Tell us

54:50

what kinds of projects they were vibe

54:53

coding in case we got a lot

54:55

of responses. We got so many responses.

54:57

I'm told it was more than 60

55:00

in all and it was so fun

55:02

to read through these. Some of the

55:04

people who wrote in had some amount

55:07

of technical expertise and it does seem

55:09

like the more technical expertise they

55:11

had the further they were able

55:13

to get. But we were also

55:15

hearing for people who had never coded

55:17

before and were just trying to. about

55:20

my vibe coding experience and I also

55:22

got just like a ton of feedback

55:24

emails comments text from people who said

55:27

I was pretty skeptical about these tools

55:29

maybe I hadn't used them in a

55:31

while but after I read your article

55:33

or heard your segment I went out

55:36

I tried this stuff I built something

55:38

so I do think that there's something

55:40

about having this first-hand experience encounter with

55:42

this technology that is persuasive in a

55:45

way that like two guys yacking about

55:47

it or not. Yeah, and for what it's

55:49

worth, after we had that conversation, I went back

55:51

to my house and I got out my laptop

55:53

and I attempted a vibe code. Yeah, what did

55:56

you vibe code? I made a little platformer video

55:58

game with a frog in a bathrobe. around various

56:00

platforms collecting coins. I realize that sounds

56:02

absurd. The absurd thing was I was

56:05

able to make that. Yeah. Because I

56:07

just typed a few words into a

56:09

box. It works. Now is it a

56:11

very fun video game? No, but like

56:13

it exists. That's amazing. It is amazing.

56:16

Wait, can I play it? Absolutely. What's

56:18

it called? It's called Frog and Bathrobe.

56:20

Okay, so before we hear about our

56:22

listeners. Vibe coding experiments. I should also

56:24

check on the status of the hot

56:27

tub time machine. How's it doing? Well,

56:29

you know, first of all, I have

56:31

to say thank you. It was very

56:33

kind of you to spend some time

56:35

in your busy life creating a program

56:38

that would attempt to instruct me when

56:40

to instruct me when to put various

56:42

chemicals and do other forms of maintenance,

56:44

like after I'd own my hot tub

56:46

for a week. So. So fortunately the

56:49

emails have slowed down a bit, but

56:51

it did introduce some questions into my

56:53

mind about the reliability of hot time

56:55

machine. Yeah, well I'm sorry about that

56:57

and I have not looked into that

57:00

because frankly I wouldn't even know how

57:02

to start debugging something like this. This

57:04

is one of the like weird things

57:06

about vibe coding is that you can

57:08

create something that works sort of. But

57:10

if it ever breaks, you have no

57:13

idea what you're doing. You just have

57:15

to kind of like poke at it

57:17

and say like fix it. Which is

57:19

the exact approach I take to home

57:21

maintenance. I just, if it breaks, oh

57:24

boy. I'm vibrating in the house all

57:26

day, but then, uh-oh, this stopped working.

57:28

Why can't, why is the ice maker

57:30

making ice? I don't know. I vibe

57:32

assembled an IKEA shelf the other night.

57:35

Let me tell you. Don't stand underneath

57:37

it during an earthquake. That's what I'll

57:39

say about that shelf. Okay, let's get

57:41

to our listeners and let's hope that

57:43

they are doing a little bit better

57:46

than we are with our vibe coding

57:48

experiments. Let's do it. First of all,

57:50

the app I made is a speed

57:52

reader for myself and for my dyslexic...

57:54

ADD friends who can't read without jumbling

57:57

up letters and words and losing focus

57:59

altogether. How it works is you paste

58:01

text in, say, from an article, and

58:03

it displays one word at a time

58:05

at around 400 words per minute versus

58:08

250 words per minute for Normies. I'm

58:10

not a coder, but I did run

58:12

the product team at my last startup,

58:14

and the experience was so similar to

58:16

working with a developer. The better I

58:19

defined the project up front, the better

58:21

the better the better the results. If

58:23

I started with a loose definition, it

58:25

could become really laborious, again, just like

58:27

working with a real developer. Thanks for

58:30

letting me share, keep up the great

58:32

work. So Kevin, how did you understand

58:34

what Mike has built? So this is

58:36

interesting because it's similar to things that

58:38

I think have been around for a

58:41

number of years, which are these speed

58:43

reading apps that basically take blocks of

58:45

text, whether it's an article or a

58:47

book or just a long email, and

58:49

they basically do what is called what

58:52

is called rapid serial visual presentation or

58:54

RSVP, which is basically flashing one word

58:56

at a time on your screen very

58:58

quickly. So instead of like reading left

59:00

to right, you're basically just like getting

59:03

the sort of fire hose directly at

59:05

you. You don't have to move your

59:07

eyes at all. It's just sort of

59:09

one word after another. and people who

59:11

swear by these apps, I've never been

59:14

a big user of them, but people

59:16

swear it helps them read much faster,

59:18

it could also be helpful for something

59:20

like dyslexia where it maybe helps you

59:22

avoid transposing letters and getting confused that

59:25

way. So I love this. You know,

59:27

we hear from listeners sometimes who want

59:29

us to talk more about accessibility issues

59:31

and technology, and maybe one reason why

59:33

we don't talk about it more is

59:35

that it is a topic that platforms

59:38

generally give short shrift to, right? We

59:40

do not see huge investments being made

59:42

in accessibility software. And so this feels

59:44

like a perfect use of vibe coding.

59:46

You are scratching your own itch. You

59:49

are building the thing that you cannot

59:51

trust the company to build for. to

59:53

you and now you have a tool

59:55

that is maybe going to be useful

59:57

to folks who have dyslexia and ADHD.

1:00:00

Yeah, I... I should say I also

1:00:02

think this is a great idea and

1:00:04

I think it also demonstrates how disruptive

1:00:06

vibe coding could be because there are

1:00:08

speed reading apps that work basically as

1:00:11

Mike described that are out there on

1:00:13

the market that you have to pay

1:00:15

for. And so you could go download

1:00:17

one of those and maybe it'll be

1:00:19

a little more polished and have a

1:00:22

few more features or you could just

1:00:24

make one yourself for whatever subscription vibe

1:00:26

coding product you're using and build your

1:00:28

own that you can use over and

1:00:30

over again for free. Yeah, very interesting.

1:00:33

My name is Lauren Buell, and I'm

1:00:36

a full-time practicing anesthesiologist based in Hanover,

1:00:38

New Hampshire, and I started vibe coding

1:00:40

with Claude back in November of 2024

1:00:43

and created this app called Consulcraft that

1:00:45

uses adaptive AI to simulate anesthesiology case

1:00:47

discussions and provide real-time feedback that's based

1:00:50

on case scripts and answers that are

1:00:52

written by me. And I have no

1:00:54

coding experience, and just a very basic

1:00:57

understanding of the command line, so it

1:00:59

took me about six weeks. And now

1:01:01

it's been about three months of adding

1:01:04

new features in cases weekly, including a

1:01:06

database so I can save transcripts for

1:01:08

research. And, you know, I expected to

1:01:11

hear mostly from my residents who were

1:01:13

using it to prepare for their oral

1:01:15

board exams, but I've also gotten a

1:01:18

lot of positive feedback from colleagues who

1:01:20

say things like, you know, I haven't

1:01:22

really done that particular case in a

1:01:25

while in the operating room and this

1:01:27

was a nice refresher. You know, which

1:01:29

has been really great to hear and

1:01:32

really unexpected and really unexpected. Anesthesiologist did

1:01:34

not put me to sleep. No, this

1:01:36

perked me right up. So explain what

1:01:39

Lauren has built here. Okay, so I'm

1:01:41

looking at Lauren's app right now. It's

1:01:43

very pretty. It's called console craft and

1:01:46

it is, it looks like basically a

1:01:48

case simulator where you can say something

1:01:50

that I don't understand. Like say you

1:01:53

have a trans-carotid artery revascularization for carotid

1:01:55

stenosis stenosis. I do have that. Click

1:01:57

on that and you can go into

1:02:00

like. a tutor mode where it explains

1:02:02

the case a 77-year-old man

1:02:04

with history of hypertension, something

1:02:06

else I can't pronounce,

1:02:09

and diabetes malitis presents to

1:02:11

the ED with an episode

1:02:13

of right-sided amorosis. Few jacks.

1:02:15

Wasn't amorosis in the first

1:02:17

Trump administration? That's omorosa. Oh,

1:02:19

that's right. So it basically

1:02:21

says, like, what do you

1:02:23

think of this case? And

1:02:25

then you can type your

1:02:27

response and get some feedback,

1:02:29

or there is a voice

1:02:31

mode. This is very cool.

1:02:33

Yeah. And I get why

1:02:35

this took Lauren six weeks

1:02:37

to build. It is not a

1:02:40

simple tool, but I can see

1:02:42

how this would be quite useful.

1:02:44

Absolutely. Absolutely. in our

1:02:46

professional lives where we feel like

1:02:48

if only we had a tool

1:02:50

like this, it would make my

1:02:52

life so much easier. Up until

1:02:55

now, if you couldn't write software,

1:02:57

you probably couldn't do much about

1:02:59

this, certainly not in the digital

1:03:01

realm. But Lauren figured out, you

1:03:03

know what, in the digital realm,

1:03:05

but Lauren figured out, you know

1:03:08

what, I'm going to apply myself,

1:03:10

but all of my colleagues, this

1:03:12

is just a great example, of

1:03:14

the potential, products that I thought were

1:03:16

only useful for me, like this tool to

1:03:18

help me recommend what to pack for my

1:03:20

kids lunch. And I put it out there

1:03:22

in my story that I built this and

1:03:24

I started hearing from people saying, can

1:03:26

I get access to this tool? So

1:03:28

it turns out that if you build

1:03:30

something just for yourself, there's probably at

1:03:32

least a few other people in the

1:03:34

world who could benefit from it too.

1:03:36

So true. Let's hear from our next

1:03:39

listener. Hey Kevin and Casey, this

1:03:41

is Matt. I'm a graphic designer, I

1:03:43

live in Los Angeles, and I've been

1:03:45

vibe coding for like six months.

1:03:47

I created a project called

1:03:49

Flavor Finder with Claude, and it's

1:03:51

a combination of the data set

1:03:54

of the Flavor Bible, which is

1:03:56

an excellent cooking reference, along with

1:03:58

the UI of a color palette.

1:04:00

generator. So you can kind of go

1:04:02

one by one and you can build

1:04:04

the palate as you go or you

1:04:07

can just push generate and it will

1:04:09

create a five ingredient flavor pairing that

1:04:11

all the ingredients pair well together. You

1:04:13

can lock ingredients and substitute, move them

1:04:16

around. It's actually been a really fun

1:04:18

way to discover a new stuff in

1:04:20

the kitchen. Another exercise that I do

1:04:22

with a lot of the LLLMs is

1:04:25

something I call rosham toe. It's a

1:04:27

combination of the two worst games of

1:04:29

all time. Tic-tecto and rock paper scissors.

1:04:32

And I just say, hey, create a

1:04:34

little playable web app for me combining

1:04:36

these two games. They're so bad, make

1:04:38

something fun. I have to say, unfortunately,

1:04:41

it's struggled to create anything that's very

1:04:43

entertaining, but it is an interesting way

1:04:45

to kind of observe how an LLLM

1:04:47

is thinking. And I should just. Start

1:04:50

by announcing that the New York Times

1:04:52

Company has acquired Rochamto for a billion

1:04:54

dollars. So here's flavor finder and if

1:04:56

you, you see some ingredients that might

1:04:59

work well together here, watercrest, pineapple, pork,

1:05:01

shallot, and olive oil and you can

1:05:03

just have it sort of generate like

1:05:06

different combinations of things that fit the

1:05:08

taste profile that you're looking for that

1:05:10

you're looking for. Look, I am not

1:05:12

a good enough home cook to take

1:05:15

advantage of something like this, but I

1:05:17

know a lot of people probably are.

1:05:19

I might try this. I think you

1:05:21

should. The next step beyond vibe coding

1:05:24

is vibe cooking. And I would like

1:05:26

to challenge him to a game of

1:05:28

rochampto. All right, next one comes to

1:05:30

us from listener Zach. And like Kevin,

1:05:33

I was once a young nerd who

1:05:35

bounced off of coding in exchange for

1:05:37

photography and flesh and dream waiver and

1:05:40

all that. And lately I've been using

1:05:42

like a combination of ChatchebyT and Claude

1:05:44

and Deep Seek. to help me write

1:05:46

these little custom shell or Python scripts

1:05:49

that helped me with some mundane photo

1:05:51

tasks. For example, I had a recent

1:05:53

job where I had almost a thousand

1:05:55

images that needed to be renamed based

1:05:58

on the subject, moved into sub folders,

1:06:00

based on the name, and exported from

1:06:02

their layered Photoshop file into like jepags

1:06:04

and tips, different file sizes. It was

1:06:07

a whole mess. And AI helped me

1:06:09

write a script to automate this organization

1:06:11

process. It even tried to make a

1:06:13

Photoshop script to help the export, but

1:06:16

it was pretty bad that language. But,

1:06:18

you know, of course, with all the

1:06:20

debugging, it probably took twice as long

1:06:23

as it would have taken to do

1:06:25

manually, but it was a lot more

1:06:27

fun. And now I have this cool

1:06:29

script I can tweak in the future.

1:06:32

Thanks. Yeah, so I feel like a

1:06:34

question that we never quite answer enough

1:06:36

is like... What do we actually want

1:06:38

AI for? And to me, an answer

1:06:41

to that question is most people's lives

1:06:43

are filled with unimaginable drudgery and things

1:06:45

that take so long and are so

1:06:47

tedious and require zero creativity, really none

1:06:50

of their human skills whatsoever, and Zach

1:06:52

found himself in exactly this situation, right?

1:06:54

He has a thousand photos, he needs

1:06:57

to manipulate them, it's gonna take forever,

1:06:59

but can he just vibe code a

1:07:01

solution that takes him twice as long

1:07:03

to do as if he addressed on

1:07:06

it himself? And I think he's. being

1:07:08

a little facetious about this because it

1:07:10

does seem like this is a process

1:07:12

that he does frequently. And so this

1:07:15

tool will help him save time in

1:07:17

the future. Yes, there is some setup

1:07:19

cost to building this tool for yourself.

1:07:21

But if this is something that you're

1:07:24

planning to do over and over again,

1:07:26

it may actually make sense to build

1:07:28

a tool rather than doing it manually

1:07:31

every time. Yeah, absolutely. And honestly, like,

1:07:33

to me, the sort of like. BS

1:07:35

I want to deal with is wrangling

1:07:37

with the computer rather than like the

1:07:40

tedium of doing it myself. You know,

1:07:42

I'm happier trying to create the tool

1:07:44

that hopefully has multiple uses in it

1:07:46

than I am just being like, well,

1:07:49

it'll be faster if I just manually

1:07:51

rename a thousand photos. Right. And I

1:07:53

think the part of Zex response that

1:07:55

I've just... resonate with so much is

1:07:58

the joy of it. I mean, this

1:08:00

is something that I got a lot

1:08:02

of blowback on when I published my

1:08:04

call. People said, oh, does any of

1:08:07

this stuff actually work? Does it actually

1:08:09

save you time? How hard is it

1:08:11

to actually, you know, decide what to

1:08:14

put in your kid's lunchbox yourself, you

1:08:16

lazy jerk? And I understand that, like, I

1:08:18

think for me, the point of vibe coding

1:08:20

is not pure efficiency. It is

1:08:22

not pure efficiency. It is also like.

1:08:24

Discovery and exploration and just I

1:08:27

find it very fun to watch the

1:08:29

code fly past as the computer goes

1:08:31

to work building something for me. Yeah,

1:08:33

to make a point that I think

1:08:36

is sometimes underappreciated, it is fun to

1:08:38

learn and it is fun to make

1:08:40

things. Yes. And these tools help people

1:08:42

learn to make things. Yes. Yeah. Thanks Zach.

1:08:45

All right. Let's hear from our last

1:08:47

vibe coding listener of the day. This

1:08:49

one is more of a request. Hi

1:08:51

Kevin and Casey. My name

1:08:53

is Ashley, I'm from Southern

1:08:55

California, and I am a

1:08:57

working mom with four kids,

1:08:59

a nine-year-old, a six-year-old, a

1:09:02

four-year-old, and a two-year-old. And

1:09:04

my vibe code idea is I

1:09:06

just need an app that goes

1:09:08

through all of the social things

1:09:11

that I'm supposed to read and

1:09:13

capture and do something with across

1:09:16

my kids' social calendars. So it's

1:09:18

like email, WhatsApp Chat, Insta Chat,

1:09:20

Insta Chat. text messages, evites

1:09:23

for the birthdays, even Facebook

1:09:25

message, I don't know if

1:09:27

that's still a thing. But

1:09:29

I just need something that's in

1:09:31

summarize, plan, give me a to-do

1:09:33

list, and put it on a

1:09:36

shared calendar, which I guess sounds

1:09:38

like something that a wife would

1:09:40

be able to do, but I'm

1:09:42

not that interested in that. So

1:09:44

if you could just vibe code

1:09:47

me a bit of a wife, that would

1:09:49

be awesome. I love the show, thanks.

1:09:51

Thank you Ashley, truly. I don't

1:09:53

know if I've ever been so

1:09:55

delighted on the show as I

1:09:57

have been listening to Ashley's children

1:09:59

stampede. through her house in the

1:10:01

background of her recording this request,

1:10:03

which I think is a great

1:10:05

request. Yeah. Also, four kids, as

1:10:07

a working mom, I'm tired just

1:10:09

thinking about it. Yeah, God bless

1:10:11

you. God bless Ashley. Now look,

1:10:13

this one is a challenge. First

1:10:15

of all, the idea is great.

1:10:17

Like I want this too, and

1:10:19

I don't even have four kids.

1:10:21

I think the challenge is, Ashley,

1:10:23

it wants many disparate services to

1:10:25

interact with each other. mostly do

1:10:27

not have ways of interacting. But

1:10:29

Kevin, as you, a more seasoned

1:10:31

vibe coder, look at this, do

1:10:33

you feel like you know an

1:10:36

approach that could work? Yes, I

1:10:38

think we can do this. I

1:10:40

don't think it is going to

1:10:42

be very straightforward, because as you

1:10:44

said, it does require the sort

1:10:46

of interoperability of a number of

1:10:48

different apps that don't natively work

1:10:50

all that well together. But actually,

1:10:52

I think we can get this

1:10:54

done for you. And it is

1:10:56

my pledge that we will do

1:10:58

our best. Tell you man, the

1:11:00

toddler's birthday party circuit is out

1:11:02

of control. It never stops. Yeah,

1:11:04

it never stops. Kids are always

1:11:06

having birthdays and there are so

1:11:08

many of them. And some say

1:11:10

no gifts and some say gifts

1:11:12

and some say gifts and some

1:11:14

are at parks and some are

1:11:16

at parks and some are at

1:11:18

zoos. And you got to keep

1:11:20

track of all of it and

1:11:22

my God, how does anyone do

1:11:24

this? How does anyone do it

1:11:26

and can it be vibe coded?

1:11:28

I will. Okay. Well, it makes

1:11:30

sense if the sort of narrative

1:11:32

arc of this show is an

1:11:34

AI trying to get you to

1:11:36

leave your wife to you mastering

1:11:38

AI to vibe code a wife.

1:11:40

That feels like a great sort

1:11:42

of season finale for the Hard

1:11:44

Fork show. I agree. Yeah. Well,

1:11:46

listen, thanks again to everyone who

1:11:48

emailed us. All of the emails

1:11:50

delighted us. We feel like you

1:11:52

guys really like trying new stuff

1:11:54

and showing us what you're trying

1:11:56

and it inspires us. So thank

1:11:58

you. So, Bring

1:12:12

along an American Express card to open

1:12:14

the door to rewards wherever you go.

1:12:16

Morning coffee run with an old friend, earn

1:12:18

cash back. Weekend getaway, earn miles. Dinner at

1:12:21

the hottest restaurant in town, you get the

1:12:23

idea. No matter the place or the plan,

1:12:25

Amex rewards your inner explorer. See

1:12:27

if you pre-qualify for an American

1:12:29

Express card with no impact on

1:12:31

your credit score. Learn more at

1:12:33

American express.com at American express.com/check-4-4-dash offers.

1:12:36

Your credit score may be impacted

1:12:38

if you accept a card, terms

1:12:40

apply, terms apply. Terms Apply. One

1:12:43

more thing before we go this

1:12:45

week, Hard Fork is looking for

1:12:47

an editor, specifically someone who can

1:12:49

take the raw material of what

1:12:51

Kevin says and try to make

1:12:53

and make sense for the rest

1:12:56

of us. So if you think

1:12:58

you might have that skill set,

1:13:00

please go to nytimes.com/careers and look

1:13:02

for the Hard Fork editor job.

1:13:04

We love to talk to you.

1:13:06

Hard Fork is produced by Whitney

1:13:08

Jones and Rachel Cohn. We're edited

1:13:10

by Rachel Dry. We're fact checked

1:13:12

by Ina Alvaradoo. Today's engineeredured by

1:13:14

Chris Wood. Original music by Sophia

1:13:16

Lanman, Diane Wong, and Dan Powell.

1:13:18

Our executive producer is Jen Poiant,

1:13:20

and our audience editor is Nel

1:13:22

Gologli. Video production by Chris Shot,

1:13:24

Sawyer Roquet, and Pat Gunther. You

1:13:26

can watch this full episode on

1:13:29

YouTube at youtube.com/Hard Fork. Special thanks

1:13:31

to Paula Shuman, Kewing Tam, Dahlia

1:13:33

Hadad, and Jeffrey Miranda. You can

1:13:35

email us as always at Hard

1:13:37

Fork at Ny times.com. If you

1:13:39

happen to be in Austin, Texas

1:13:41

for South by Southwest this weekend

1:13:43

or early next week Casey and

1:13:45

I will be there catch us

1:13:47

around town. Say hi. Say hi.

1:13:49

Let's get some tacos. Let's vibe

1:13:51

code