0:00

Is there like a chance that like

0:02

tether which is used for according to

0:04

60 minutes and Congress and all those

0:07

hearings used for really dark stuff

0:09

in the world? Terrorists, human

0:11

traffickers are using it or

0:13

maybe confirmed and they've been banned

0:15

only states. Like is that going

0:17

to become a legit thing? Yeah, I

0:19

think it will. I mean, and just to,

0:22

you know, backup. Yes, tether has been

0:24

used for those. I'm sure has been

0:26

used for crimes. Sure. So is, you

0:28

know, United States dollars in briefcases, right?

0:30

Of course, yeah. By a much larger

0:32

margin. If we were to look at

0:34

it, that historically is correct because tether

0:36

hasn't existed, but today, like

0:38

the average criminal. Cryptocourancy is

0:40

being used for between 40 and

0:43

50 billion dollars of illicit

0:45

transactions per year. That would be

0:47

a magnitude less than US dollars. This

0:49

weekend startups is brought to you by.

0:51

HubSPOT for startups. Smart founders aren't piecing

0:54

together random tools. HubSPOT is the customer

0:56

platform that thousands of startups use to

0:58

scale efficiently. Get up to 75% off

1:01

plus three months of perplexity AI for free.

1:03

Go to hubspot.com/startups. Startups. Squares. Turn your

1:05

idea into a new website. Go to

1:07

squarespace.com/twist for a free trial. When you're

1:10

ready to launch use offer code twist

1:12

to save 10% off your first purchase

1:14

of a website or website or domain.

1:16

And? Oracle. Oracle Cloud Infrastructure, or

1:18

OCI, is a single platform for

1:21

your infrastructure database application of development

1:23

and AI needs. Save up to

1:25

50% on your cloud bill at

1:27

oracle.com slash twist. All right, everybody.

1:29

Welcome back to this week in startups.

1:32

Very exciting show today. We've got an

1:34

office hours at the end with one of

1:36

our founders, but we're very lucky to have

1:38

with us again. Of course, Alex Wilhelm, you

1:40

know him from Tech Crunch and Cautious

1:42

Optimism, his sub stack. And Lon

1:45

Harris is here, original original twist.

1:47

And super lucky to have one

1:49

of my oldest dearest friends

1:51

from web 1.0 in the 90s,

1:53

Mark Jeffrey, when I would go to

1:56

LA and I was broke doing my

1:58

magazine, Mark Lem- sleep quite

2:00

literally on his couch. A very famous

2:02

couch in fact in the history of

2:05

entrepreneurship. Welcome to the program Mark. It

2:07

is the Excalibur of couches. So yeah,

2:09

Travis actually stopped on that couch a

2:12

few times. Okay, Travis from Uber, Travis

2:14

Callanick, myself, Mark had an apartment, you

2:16

know, which when we were in our

2:19

20s was a big deal in LA.

2:21

We would go to LA. Hey, crash

2:23

on Mark's couch. The history of Silicon

2:26

Beach here. We got a little... It's

2:28

Silicon Beach, yes. I remember. I had

2:30

Digital Coast reporter and Silicon Alley reporter.

2:33

I had two different magazines at that

2:35

time. Print magazines and I would do

2:37

digital coast events and Silicon Alley events.

2:40

So I took the two cities that

2:42

were in Silicon Valley and featured the

2:44

startups there. It was an interesting model.

2:47

Mark, I wanted to have you on

2:49

today because you're down the crypto rabbit

2:51

hole, but you are a crypto realist.

2:54

You actually look for crypto projects that

2:56

have some reality to them. And you're

2:58

here in town in Austin because there's

3:01

something going on with a very specific

3:03

crypto project. So maybe you could just

3:05

tell us what that is. Yeah, so

3:08

I'm here in town for a crypto

3:10

project called bit tensor. They had their

3:12

first big event. Now, bit tensor is

3:14

Bitcoin meets AI. And Tao is the

3:17

coin and there's 21 million Tao coins

3:19

only, of which like 8 million have

3:21

been admitted so far. That's the exact

3:24

same as Bitcoin. Yes, that is correct.

3:26

Is it used the same? Open source

3:28

project or do they just thought that

3:31

would be like a clever thing to

3:33

do? They thought that the the Bitcoin

3:35

ethos was the right one to adopt

3:38

So it draws very heavily it's very

3:40

heavily inspired by Bitcoin got it. Okay.

3:42

So they're disciples of Bitcoin. Yes. They're

3:45

doing 21 tokens or coins 21 million

3:47

What do they call them tokens coins

3:49

coins? Got it sure. Okay. What is

3:52

the purpose of the project? So the

3:54

purpose is to build an incentivization network

3:56

Mostly for AI, but not necessarily totally

3:59

for AI. When you say a network,

4:01

you mean a network of computers and

4:03

CPU-G-P-U-S? Yes, I do. I do. So

4:06

what did Bitcoin do really well? Right?

4:08

Like, why did it succeed? Well, Sato...

4:10

she started off saying, I want to

4:13

create kind of this alternative gold and

4:15

I want to be able to move

4:17

it around the world. So I have

4:20

to also create this alternative universe Swift

4:22

system, right? So how do I do

4:24

that? And what he decided to do

4:27

was he decided to incentivize people out

4:29

there to donate electricity and GPU. They

4:31

didn't pay them. There was GPU. It

4:33

was GPU later, CPU in the beginning.

4:36

Got it. So you could mind, you

4:38

know, on your home computer. in the

4:40

early days not right didn't work too

4:43

well so later on no but this

4:45

this whole idea of of you know

4:47

look i'm not gonna pay you we're

4:50

not gonna make a company i'm just

4:52

gonna give you bit coin coins that

4:54

are generated by the network with every

4:57

block this for solving a math equation

4:59

correct well that yeah it's so for

5:01

finding the hash of the previous block

5:04

which i'm not gonna get into the

5:06

technical explanation of that but bottom line

5:08

it boils down to Yes, in the

5:11

number of gumballs in a jar, right?

5:13

That's really, you know, really, in a

5:15

way it was busy work. It's busy

5:18

work, right? It's trying to force your

5:20

CPU to peg up and to prove

5:22

that it has the computing power that

5:25

it says that it has, right? And

5:27

for people who don't know with Bitcoin,

5:29

the reason they created that was so

5:32

that you would have a network of

5:34

computers that essentially act like the swift

5:36

network, like a banking network. So the

5:39

need was to build infrastructure and rails

5:41

to move money around. and to make

5:43

sure you were sending it to the

5:46

right person. Therefore, we'll just have you

5:48

sit here and, you know, solve this,

5:50

how many gumballs in the jar, which

5:52

also forces you to have a computer

5:55

on the network with power. Yes. Correct.

5:57

And I think most people don't understand

5:59

the intent of situation. I haven't. So

6:02

basically, Satoshi proved the point. So what

6:04

was the end result of this, right?

6:06

In aggregate, Satoshi created the world's, right?

6:09

In aggregate, Satoshi created the world's largest

6:11

supercomputer by several orders of magnitude, even

6:13

AIG. you know in aggregate is is

6:16

not going to catch up to the

6:18

Bitcoin networks computing power for at least

6:20

five years if you know if they

6:23

if all the chips are created go

6:25

to AI that still won't catch up

6:27

to Bitcoin for a very long time

6:30

so Bitcoin succeeded at it creating this

6:32

this incredible network through these incentives so

6:34

bit tensor looked at that and said

6:37

that's a really interesting dynamic how can

6:39

we harness that to do two things

6:41

one Let's do something useful with that

6:44

GPU instead of guessing the gumballs on

6:46

a jar. That's dumb. You know, I

6:48

don't want to belittle Bitcoin because it's

6:51

very useful for security, but as an

6:53

activity, it's very dumb. That has been

6:55

the criticism. Hey, we're burning a bunch

6:58

of electricity. We're putting up all these

7:00

GPUs, but conversely, is the open source

7:02

project to replicate this. Tao is the

7:05

coin. Yes. Tao is the coin that

7:07

you essentially earn by putting GPUs on

7:09

the network and then giving them primarily

7:11

to people who are looking for a

7:14

distributed computer network, a distributed computer network

7:16

to run AI jobs on. Sort of,

7:18

yeah. So you're right up to the

7:21

end, you're right. Okay, great. So, yes.

7:23

I'm trying to, I'm recapping this for

7:25

our audience because sometimes. Yes. I know.

7:28

It's not that they have a lot.

7:30

They start, you know, on the, in

7:32

the red zone. They're on like the

7:35

20-yard line and I think people can't

7:37

keep up with it. So I think

7:39

we've set the stage here really nicely.

7:42

Who are the people behind the bit

7:44

tensor project? Are there notable individuals who

7:46

created this? And when was it created?

7:49

Yeah. So it's, I think it's about

7:51

four years old. We do know the

7:53

founders, we do know the people who,

7:56

and it's a team of like five

7:58

or six engineer people all over the

8:00

world. I just met with one of

8:03

them this past, you know. at the

8:05

conference, and then yesterday we spent like

8:07

three or four hours, I was clarifying

8:10

about the stuff that I didn't understand

8:12

about the innards. They are docs, so

8:14

it's not like they're anonymous, like Satoshi.

8:17

Got it. Okay, so bid tensor is

8:19

the project, Tao is the coin, the

8:21

network, instead of doing gumball math, is

8:24

doing, here is a competitor to AWS.

8:26

They're doing many things at once. So

8:28

there's actually under the, under the, it's

8:30

an incentivization network. primarily used to incentivize

8:33

creation of great AI, not only that,

8:35

but mostly that at the moment. And

8:37

beneath that are 100 projects. And each

8:40

of them are mostly AI, but not

8:42

all of them are AI. Got it.

8:44

Right? And we can have a look

8:47

at those projects if you'd like. Absolutely.

8:49

So this platform then allows you in

8:51

a way to plug in your project

8:54

as an individual with server capacity. So

8:56

if I happen to be running, I

8:58

don't know. Squarespace and I have a

9:01

bunch of servers that I stood up

9:03

for whatever reason I wasn't using a

9:05

WS I can say hey, you know

9:08

what I'm going to allocate a couple

9:10

of these H100s whatever to the tau

9:12

to the bit tensor network to earn

9:15

tau Yes, and then other people could

9:17

come in and say yeah, I need

9:19

some compute and the goal would be

9:22

it's cheaper faster or better or just

9:24

cheaper Yeah, so it's using a W3

9:26

so yeah, so let's talk about your

9:29

specific example that you just gave is

9:31

are like two of the of the

9:33

biggest subnets in okay I just guessed

9:36

it yeah so you got it right

9:38

okay so if you go to if

9:40

you show back prop so I said

9:43

there's a there's a hundred subnets in

9:45

the bit tensor universe each one each

9:47

one of the subnets has its own

9:49

coin inside you know that that is

9:52

also part of the the bit tensor

9:54

network sort of like a theorem right

9:56

now the other coins inside of it

9:59

the same so if we're looking at

10:01

this just we'll pause it for a

10:03

second for a second here these are

10:06

the subnets So the network, bit tensor,

10:08

the subnets, the number one one is

10:10

called shoots, as in shoots and ladders.

10:13

It's been around for 120. It has

10:15

a market cap itself of $89 million.

10:17

The price of their coin is $98.

10:20

Their emissions are 16%. What does, say,

10:22

shoots do? Because I see there there's

10:24

a GitHub logo. There's a network logo.

10:27

I don't know what that means. What

10:29

does this all mean? All right. Everyone

10:31

knows that CRM isn't just software. It's

10:34

basically the heartbeat of your business. But...

10:36

It can get ugly quick if your

10:38

data isn't organized and you're dealing with

10:41

a messy tech stack. That's why I

10:43

love Hub Spot for startups. It's the

10:45

all-in-one customer platform. So you don't need

10:48

a frank inside of schools. No. Right

10:50

now, early stage companies are going to

10:52

get 75% off. And with this one

10:55

system, you're going to automate marketing and

10:57

actually converts. Track your sales pipeline without

10:59

spreadsheet chaos. And you're going to manage

11:02

your customers like the Amman Hotels, six

11:04

stars all the way. You're going to

11:06

get investor-ready analytics that tell your story

11:08

perfectly. And man, when you pull up

11:11

Hub Spot and you get those metrics,

11:13

you got those metrics, you got those

11:15

analytics, you got those analytics. Things are

11:18

going to go really faster for you

11:20

as a startup with potential investors. Plus,

11:22

you're plugged into an amazing community of

11:25

founders who've already tackled what's ahead. They've

11:27

been around those sharp turns and they

11:29

can tell you how to navigate them.

11:32

HubSPOT was built by Scrappy founders, I

11:34

know them. And they understand, every dollar

11:36

counts. That's why hundreds and thousands of

11:39

startups trust HubSPOT to scale their businesses.

11:41

Here's an amazing call to action. So

11:43

generous from my friends at HubSPOT, 75.

11:46

That's 7% off, not 5% off, 75%

11:48

off, HubSPOT for startups. You can get

11:50

three months of perplexity AI for free.

11:53

That's a great pot sweetener. Head to

11:55

hubspot.com/startups. So the so shoots does exactly

11:57

what you were just describing. So they're

12:00

they're a decentralized network of of compute

12:02

for AI. So if you want to

12:04

run the you want to run. Deep

12:07

Sea you want to run mistral yeah

12:09

servers here we go I'm watching it

12:11

again I got so all of these

12:14

so basically just come here you normally

12:16

you'd go to AWS right and you've

12:18

rent some instances and then you would

12:21

load up your AI this actually you

12:23

know the the shoots will allow you

12:25

to just go there click a button

12:27

pay a little bit of tow and

12:30

start running your instance and it's about

12:32

85% less than what it costs on

12:34

AWS. Got it. So if you were

12:37

running an AI job and you didn't

12:39

need the stability corporate five-nines of say

12:41

AWS, where you know at a company

12:44

maybe you have to use AWS because

12:46

nobody gets fired for using AWS, Azure,

12:48

Oracle, whatever, you could use this network

12:51

that nobody owns and that is an

12:53

open source distributed project. How stable is

12:55

it? Very stable. I mean, people are

12:58

very happy with it. People think it's

13:00

better, not only cheaper, but there are

13:02

some benchmarks where it's actually quite a

13:05

bit better than what you're seeing out

13:07

of AWS. And typically they have the

13:09

models up and running when Deep Sea

13:12

came out with their latest model, was

13:14

up and running, was up and running

13:16

on shoots before it was up and

13:19

running on anywhere else. And yeah. So

13:21

let's pull up the shoots website. SubNet

13:23

that competes with other cloud computing resources

13:26

out there. You can just base it

13:28

as a developer deploy to it. And

13:30

then I pay them in Tao. Yes.

13:33

So instead of me putting my credit

13:35

card on file with my cloud computing

13:37

company They're going to have that later,

13:40

but right now. Yes, they you pay

13:42

in tow I pay them in tow.

13:44

Targon also Targon is the the other

13:46

one that does the other network that

13:49

does the other subnet that does the

13:51

other subnet that does the same thing

13:53

as shoots great So let's pull up

13:56

that top level of all the projects

13:58

again. Alex if you don't mind we'll

14:00

leave that up? and running. I see

14:03

their number three here with a $44

14:05

million market cap. Correct. And so we

14:07

can load their page and we would

14:10

see it and these just look like

14:12

any other hosting company in the world.

14:14

But instead of using dollars and having

14:17

an office, it's a distributed project, but

14:19

somebody does own Pardon, right? There is

14:21

a subnet owner. Yes. So good question.

14:24

So somebody has defined. the target subnet

14:26

and said, I want people to supply

14:28

GPU to my network to host these

14:31

AI models, right, and load them up.

14:33

And, you know, there are 256 minors

14:35

that are competing to provide, you know,

14:38

this, this, whatever the subnet owner has

14:40

requested. So it is a competition, right?

14:42

Whoever supplies the most, the best, depending

14:45

on how the subnet owner defines the

14:47

competition. Got it. The minors, the people

14:49

who are supplying the computers for this

14:52

decentralized network. They earn Tao. So they're

14:54

earning emissions from the network. So just

14:56

like Bitcoin miners? Sure. Earned Tao? Subnet

14:59

miners are in Tao in this system.

15:01

Who decides who can have a subnet?

15:03

This is always really interesting to me

15:05

is governance. So let's pause for a

15:08

second here. I think we can all

15:10

follow along. You could start essentially a

15:12

project, aka a company, a company. Now,

15:15

would Target also be a company in

15:17

a way? Taragon is a company, they

15:19

put this up, they are making Tao

15:22

by providing this resource to people who

15:24

want it, cloud computing resources, on the

15:26

network. And there are 33 of these

15:29

projects, currently? There's 100. There's 100. 100.

15:31

Is there a limit to the number

15:33

of projects? There is not. So you

15:36

asked how do you do a subnet?

15:38

Yeah, governance wise. You steak Tao. Anybody

15:40

can start a subnet for any reason,

15:43

right? So you steak Tao. You define

15:45

your subnet. Define what staking town means.

15:47

You know, when somebody's not in crypto,

15:50

the term steak means you buy it

15:52

or you put it up for people

15:54

to earn. You basically, you, you. You

15:57

put it in a suspended state. So

15:59

it's like an escrow right so you

16:01

take your tow you have to have

16:04

a certain amount of it You you

16:06

give it to the chain the chain

16:08

is a process for doing this the

16:11

chain locks it up. It says okay.

16:13

I've got it. And you can't have

16:15

it back until you know unless you

16:18

know unless there's some time in the

16:20

future you know unless there's some time

16:22

in the future where you want to

16:24

you know get ready or something that

16:27

I'm not sure the process for on

16:29

the network. And that, and Tao is

16:31

worth about $377 or so dollars per

16:34

coin at the moment. So maybe it's

16:36

100 grand and change to put up

16:38

one on them. Okay, so we're in

16:41

the very early days of this project.

16:43

Yeah. What's like the second most interesting

16:45

use case? You have obviously AI clouds.

16:48

What else is in there? What else

16:50

do you got? Yeah, there's ready AI,

16:52

which is a subnetated. data, you know,

16:55

when you feed data to an AI,

16:57

it's better if it's giving context and

16:59

it's annotated in some way by humans,

17:02

right? But you have to do that

17:04

at scale, it takes a lot of

17:06

humans. So scale that AI had a

17:09

lot of humans around the world, paying

17:11

them to annotate data, which was then

17:13

sold to AI companies to train their

17:16

AIs. Yes, we know this company. I

17:18

think, actually, Alex, is scale AI in

17:20

our TW 500. It absolutely. Anatating data,

17:23

doing reinforcement learning, this is something Google

17:25

was doing a long time ago, you

17:27

may have heard of, what's the Amazon

17:30

project, Mechanical Tour. Mechanical Tour was another

17:32

project where they would say, hey, we're

17:34

going to show you an image, tag

17:37

it with three things, and you'd be

17:39

like, okay, that is a bottle with

17:41

orange liquid in it, with a red

17:43

cap, and it's orange juice, and it's

17:46

190 calories, whatever. And in the background

17:48

is a plastic cup in an iPhone

17:50

and a iPhone. And then they would

17:53

have somebody else do the same task,

17:55

look for which tags they got in

17:57

common, and then that would be... how

18:00

Google would know that there's a smiling

18:02

face in an image. It wasn't that they

18:04

were reading the image for a smiling face,

18:06

it was tagged, then the AI and the machine

18:09

learning learned what a smile is versus a

18:11

frown. Okay, we all know that history.

18:13

So there created a Tao instance

18:15

for people to participate in the

18:17

tagging and learning. So I guess I

18:19

could go in there as an individual with

18:21

no job, but more time on my hand

18:24

or maybe I'm in Manila or a

18:26

developing country frontier market and I

18:28

could just. get jobs on the

18:30

town network. Yes, but there's also, but

18:32

what Brady AI is focused on is

18:35

actually having AI do the tagging, retraining

18:37

the data for AI. Got it, okay.

18:39

So you can better, yeah. And, you

18:41

know, the CEO is in, his CV

18:43

is amazing. These are very serious people

18:46

who are building subnets. Got it. The

18:48

guy who built and sold adwards to

18:50

Google is part of ready AI. Gil.

18:52

So again, yes, Elbaz. Yeah, Gil Elbaz

18:54

has been on the program. I mean,

18:57

we know, Gil, this may or may

18:59

not be the next Bitcoin. You, by

19:01

the way, have your own podcast about

19:03

this topic. You can maybe tell everybody

19:05

about your podcast if they want to

19:08

learn more. Yeah, sure. My podcast is

19:10

called Hash Rate. And I've done about

19:12

110 episodes to date. about crypto.

19:14

It's about crypto, but I added AI

19:17

things in here and there. Got it.

19:19

I've done about 30 episodes on bit

19:21

tensor in Tao, which is spanned over

19:24

the last, you know, year, year and

19:26

a half as I became more and

19:28

more interested in this. You know,

19:30

I've been looking for, you know,

19:33

what is the next, you know, year,

19:35

year and a half as I became

19:37

more and more interested in this. You

19:39

know, I did a lot of deep dives

19:41

that went nowhere, right? Right. where I

19:43

feel like I'm seeing, you know, the

19:45

third great coin, right? Got it. Alongside

19:47

Bitcoin and Ethereum. So I got it.

19:49

Yeah, that's my opinion. And you're an

19:51

investor in crypto projects or early in

19:53

Bitcoin, you first bought Bitcoin when it

19:55

was at what dollar amount? It was

19:57

$2.50 when I bought my first one.

19:59

$200. 50,000 Go

20:58

ahead squarespace.com/twist

21:00

to get a free trial and

21:03

when you're ready to launch go

21:05

to squarespace.com/twist to get 10% off

21:07

your first website or domain purchase

21:09

that squarespace.com/twist Absolutely, well first

21:11

of all Jason Gil Elbaz still

32:06

private, still small-er, get tapped. And

32:08

I do think this should help

32:10

service go public, Jason. Okay,

32:13

this is super interesting. Obviously,

32:15

GROC, GROQ, is not GROC, from

32:18

Elon, it's GROC, from Chemoth,

32:20

and it's tensor chips. These

32:22

are inference chips. So meta,

32:24

does not currently have a

32:26

cloud computing offering for founders,

32:28

right? They... make their own, obviously,

32:31

data centers, but they're not

32:33

in competition with AWS. If what

32:35

I'm hearing is correct here, Met

32:37

is saying we're going to take

32:39

llama, they're open source project,

32:42

and they're going to have it

32:44

hosted with compute from

32:46

Grock and cerebras. Is

32:48

cerebras a data center company

32:50

or a chip company? So

32:52

Rebers is a chip company, they are

32:54

big with G42, and they filed to

32:56

go public, but I remember their IPO

32:59

was so single company revenue that it

33:01

was a little dicey. So now, if that

33:03

does happen, does this mean, is the actual

33:05

story here, that meta is going to

33:07

compete with AWS? To a degree. I mean,

33:09

you can host Lama models around, you know,

33:12

you can run Lama, I think, on AWS

33:14

or on Azure or GCP, but now they're

33:16

going to offer their own kind of homegrown.

33:18

solution if you will with these partners so

33:21

meta is now in a way competing with

33:23

everyone else yes in cloud computing

33:25

is a slice of cloud computing so

33:27

then if they do just that slice

33:30

they're probably going to be

33:32

obligated to offer some storage or

33:34

some transit this could be the

33:36

start of them creating an AWS

33:38

competitor that's actually the real news here

33:40

is this could be their wedge now they have

33:43

a great excuse mark Well, we

33:45

want to have the freshest best

33:47

version of llama available because we

33:49

want the project to win. Maybe they

33:51

offer it a discount. Maybe they offer it as

33:53

a loss. They can price dump this. What do

33:55

you think here, Mark? What? Alex, what?

33:57

Sorry, I don't think... Sorry, I don't

33:59

have some... agreement, but then probably not, Jason,

34:02

because the information reported that meta had actually

34:04

reached out to alphabet and Microsoft. Should I

34:06

get them to subsidize the money they were

34:08

putting into llama? So I doubt they can

34:10

actually take more of a loss here. I

34:12

think this is a way to recoup some

34:15

of their investment, not to further subsidize their

34:17

market share. Well, I mean, if you want

34:19

your model to win and you've got tons

34:21

of cash laying around, you could buy back

34:23

your stock, you could build infrastructure and take

34:26

a loss on it. And they could lose,

34:28

I mean they're losing, what 10 billion a

34:30

year on these? On VR, yeah. On VR,

34:32

I mean they could lose 20 billion a

34:34

year on this. Make it free. The big

34:37

loser here Mark might be Open AI. What

34:39

if this is available for less than Open

34:41

AI charges for their compute? What are your

34:43

thoughts here of Metas, open source, sort of

34:45

strategy, and standing up, hosted compute? Yeah, I

34:47

mean, this is exactly what shoots in Targan

34:50

do, right? So they take these open source

34:52

bottles and stand them up and make them

34:54

available for quite a bit less than you

34:56

can get on AWS. Yeah. So, you know,

34:58

I think the overall friend is that AI

35:01

goes towards zero in terms of cost, right?

35:03

Like, wow. And I think that through various,

35:05

you know, through various mechanisms and various interacting

35:07

market forces. Some of them from bit tensor,

35:09

some of them from, you know, the open

35:11

source community, which is very dangerous for open

35:14

AI, right? Like they're incinerating hundreds of billions

35:16

a year. Yeah, right. So how is that

35:18

sustainable in the face of that area? Single

35:20

digital billions a year. Sorry, single digit billions

35:22

a year. Single digit. Single digit. Single digit.

35:25

Single digit. The company's worth hundreds of billions.

35:27

Yeah. Let's good cleanup and light run. and

35:29

from Sundar and from Satya on exactly how

35:31

much code is being written by AI now.

35:33

This is a trend that I think we

35:36

saw coming but maybe not at the velocity.

35:38

it's coming, I don't know. Q it up

35:40

here. So I'm very impressed by a couple

35:42

of numbers. So during Google's earnings report, Alphabet's

35:44

earnings report, Alphabet CEO, Senator Pachai, said that

35:46

right now over 30% of code that is

35:49

committed from the company, now comes from an

35:51

AI source. Also at Lanakan, Satie Nade, CEO

35:53

of Microsoft, said that right now 20 to

35:55

30% of the code that the company puts

35:57

out, is written by AI. Now this is

36:00

pretty big news. But Cursor is very, very

36:02

proud of how much of its, how much

36:04

of, how much code it's putting out. So

36:06

the CEO said over on X that Cursor

36:08

today writes about a billion lines of accepted

36:10

code per day, and he put that up

36:13

against a global number of several billion lines

36:15

per day. People were a little skeptical of

36:17

that figure, but it just goes to show

36:19

how fast this is moving. Jason, I think

36:21

once we see companies like Light Run, which

36:24

talked about I think on Monday. really kind

36:26

of turn the AI snake back on its

36:28

own tail and begin to have AI improve

36:30

AI generated code, we're going to get to

36:32

80-90% within probably 18 months. Yeah, this is

36:35

pretty amazing and I think it's going to

36:37

be great for humanity because the bottleneck for

36:39

startups has been a moving target over time.

36:41

When startups and PC revolution, server revolution, there

36:43

were hardware constraints. We had an incredible constraint

36:45

of the memory of the computer. We had

36:48

constraints of the storage of the computer. You

36:50

know, 15 floppy disks to run it, that

36:52

everything was too slow, it was too hard

36:54

to even load up a word processor. You

36:56

know, if you saved a large file, it

36:59

was even like the quality of the file

37:01

get corrupted. Like we had really crazy issues.

37:03

Then we went to another phase where the

37:05

bandwidth was the issue. Hey, how do we

37:07

moved this stuff around? Then it became standing

37:10

up server. So if you were starting a

37:12

company in the late 90s and the web

37:14

1.0 era when Mark and I started, you

37:16

had to raise three, four, five million. You

37:18

had to take 18 months, 24 months to

37:20

build your product and stand up your data

37:23

center. Now you can. build your startup in

37:25

three weeks, use somebody else's data center. So

37:27

what has been the blocker the last 10

37:29

years? The blocker has been developers. I can't

37:31

find a developer was what we heard for

37:34

the last 10 years from founders. I think

37:36

what we're going to hear now is because

37:38

you don't need five developers to get your

37:40

project out the door, you need one, and

37:42

that one is getting 30% faster a year,

37:44

maybe your startup ultimately, instead of needing 30

37:47

developers, needs five. Maybe in order to start

37:49

you need one developer not five right you

37:51

know like this is a magnitude change and

37:53

I think it's going to mean we're going

37:55

to make every piece of software that hasn't

37:58

been made yet will get made. What else

38:00

we got on the docket? I want to

38:02

run something by you Jason. There's a venture

38:04

capitalist Charles Hudson from precursor. He's great. Everyone

38:06

knows Charles. He did a post over on

38:09

sub stack, but he says that he's noticed

38:11

that the combination of AI generated cold outreach

38:13

to VCs is pushing people back towards human-driven

38:15

warm intros and referrals. This was very interesting

38:17

to me. I can't imagine handing off my

38:19

VC outreach to AI, but I'm curious how

38:22

founders can take advantage of this to win

38:24

more in 2025. I've always believed it's a

38:26

numbers game, you know, in terms of racing

38:28

capital, especially at the early days. because you

38:30

are selling the promise, right? But even though

38:33

it's a numbers game, because there's so many,

38:35

we just talked about there's 400 funds formed

38:37

a year, so there's 1,200 funds active at

38:39

any one point in time, maybe 1,500, because

38:41

they tend to have a four-year life cycle

38:43

of primary investing. So let's say this 1,500

38:46

firm funds, they probably have six people working

38:48

at each, you know, you're getting to 5,000

38:50

to 10,000 active investors with check writing ability.

38:52

are 10% of them will want to invest

38:54

in your company, 5% of them. It's a

38:57

numbers game in that if it was 10%,

38:59

you might be talking about 500 qualified targets,

39:01

1,000 qualified targets. Then you

39:03

have to look at,

39:05

okay, they do invest

39:08

in my vertical, I'm

39:10

in marketplaces, I'm in

39:12

military. Okay, of those,

39:14

which ones invest at

39:16

my stage? Seed rounds,

39:18

pre -seed, series A, series

39:21

B. Now you have

39:23

to parse that list.

39:25

And then you have

39:27

to start a real

39:29

sales process of going

39:32

to them. The problem

39:34

is, because of databases

39:36

like CrunchBase and other

39:38

ones that exist, sometimes

39:40

the founder will get

39:43

overzealous and they will

39:45

send too many emails.

39:47

And that upsets people.

39:49

And that's where you

39:51

get VCs complaining. Like,

39:53

I am not doing

39:56

medical devices, I don't

39:58

invest in pizzerias. So

40:00

what you want to

40:02

do is make a

40:04

nice big list. And

40:07

then you want to

40:09

really understand on that

40:11

list, have they invested

40:13

in companies adjacent to

40:15

yours? Do they only

40:17

invest in certain regions?

40:20

So it is a

40:22

numbers game, but you

40:24

have to also curate

40:26

that number. So it's

40:28

very easy to get

40:31

to a list of,

40:33

I would say you

40:35

should have, in your

40:37

seed round, there should

40:39

be probably at the

40:42

top level 200 firms

40:44

that you've identified would

40:46

actually invest in your

40:48

company. And then you

40:50

should see of those

40:52

200, how many can

40:55

you get a warm

40:57

intro to? And then

40:59

how many do you

41:01

need to do a

41:03

cold intro to? And

41:06

then you should be

41:08

very thoughtful when you

41:10

send that email, instead

41:12

of trying to do

41:14

it as quick as

41:16

possible, go slow. Which

41:19

is, if you were to meet somebody at

41:21

a party, you know, and you were a real

41:23

estate broker, you wouldn't be like, are you

41:25

selling your home or buying a new home anytime

41:27

soon? You'd be like, oh, where do your

41:29

kids go to school? Oh yeah, no, I'm over

41:31

here and you kind of warm up the

41:33

lead. So you want

41:36

to warm up the leads. We have a great

41:38

video from Alexis Ohanian about the perfect cold email,

41:40

if you want to just throw to that real

41:42

quick. It's on the from the feeds section. This

41:44

was shared a few weeks ago, I've had it

41:46

in the docket in case we ever got to

41:48

talk about it. This is such a great example.

41:50

It's only one minute of video, and I feel

41:52

like he lays out exactly

41:54

how to write the best cold email

41:56

you've ever heard. I

42:00

love a good cold email that

42:02

is to the point. It is

42:04

no longer than like three or

42:06

four sentences. It very clearly up

42:08

front states who you are and

42:11

why you're real. Basically what value

42:13

you have to provide to the

42:15

person that you're reaching out to

42:17

and then makes a very specific

42:19

request. And says thanks. That's it.

42:21

That's it. That's it. Up front

42:24

you want to demonstrate the value.

42:26

Why am I going to spend

42:28

another 30 seconds reading this email

42:30

and then immediately follow up with

42:32

the request and ideally the offering.

42:34

Old emails are deeply, deeply empathetic

42:36

for the person you're emailing. You've

42:38

taken the time to understand who

42:40

they are, what they're about, what

42:42

they like. One of the easiest

42:45

ways to mess it up is by not

42:47

doing that work, getting their name wrong. Putting

42:49

20 paragraphs into an email, you know,

42:51

you can tell you can't show your

42:53

value up front, make your clear ask.

42:55

That's it, good exercise, to ease for

42:57

the rejection too. Yeah. He's nailing it

42:59

there. You gotta be concise, you gotta

43:01

be concise, you gotta be concise, like.

43:04

Gosh, if you want, it's so easy with VC

43:06

who is publicly active to just say,

43:08

I saw you on this week in startups

43:10

or you had this tweet, it

43:13

really resonated with me because I'm

43:15

building something inspired by that tweet.

43:17

Boom. And then you're, all of

43:19

a sudden, you've created some commonalities,

43:21

some common ground, and then you

43:23

get to the ask. We're raising

43:26

our seed round. I also think a chart,

43:28

if you have, I like leading with

43:30

what's strong. So a chart is the strongest

43:32

thing in the world for VCs because

43:34

we like up and to the right,

43:36

we like things that are going to grow.

43:38

So we've had our product in

43:41

market for seven weeks. We've grown

43:43

on average 18% week over week.

43:45

We're doubling every three to four weeks.

43:47

And here's a link to our app and

43:49

here's a link to our deck. We'd love

43:51

to, if you're interested, we'd

43:53

love to do a quick follow-up

43:56

meeting. I always added something extra,

43:58

which was happy to meet. Anytime,

44:00

I know you're in Palo Alto, any time,

44:02

you know, Saturday, Sunday, 7am to midnight, anytime

44:05

I can meet you for 15 minutes, happy

44:07

to go where you are. So you're actually

44:09

even putting out there, like, you're a dog

44:11

and a rabid person who will meet any

44:14

time anywhere if you want to do an

44:16

introductory call. And yeah, I also love sometimes

44:18

people ask me a question. What do you

44:20

think of this design? That kind of like,

44:23

oh, okay, yeah, maybe I'll. give you actually

44:25

some feedback. Go ahead Mark. There's something that

44:27

you used to say at Mahalo a lot

44:29

that I that really made me focus a

44:32

lot more on this and you know the

44:34

video talks about clarity right getting getting to

44:36

the point yeah used to say to people

44:38

you know answer the question you'd ask someone

44:41

a question they wouldn't answer the question they'd

44:43

like give you 50 paragraphs literally that's my

44:45

line. Like all the context. We've always want

44:47

to give you context. I never really thought

44:49

about how often people don't answer the question

44:52

and how rare clarity is. And you kind

44:54

of, you know, tuned my brain to that

44:56

a lot more than it had been. All

44:58

right. We have an office hours. We're going

45:01

to get to our office hours now? Let's

45:03

do it. Yes, we are. Okay. All right.

45:05

So the company in question is layer next.

45:07

The co-founder and CEO is bootica madam. Now,

45:10

layer next, if you don't know, is taking

45:12

the world of business intelligence to the next

45:14

level, using AI to get all that structured

45:16

and unstructured and unstructured corporate data. Let's talk

45:19

to later next. All right, how are you,

45:21

sir? Nice to see you. How are you

45:23

going to do it? Tell us, what's going

45:25

on with Lara next? What's challenging? What are

45:28

the wins? What are the fails? Layer next

45:30

is the strategic business intelligence platform. And we

45:32

are helping CFO to generate strategies to grow

45:34

their business or increase the efficiencies. The challenge

45:37

is right now with business on board. Because...

45:39

Every customer has a different data set. We

45:41

are especially going for these mid-market companies in

45:43

the manufacturing and transportation. The problem is that

45:46

data is not AI ready. So we have

45:48

to do a lot of upfront work in

45:50

order to make our system work with their

45:52

data. That's the challenge. So you have AI

45:55

that goes in, looks at a CFO's data

45:57

from their company, and then gives them some

45:59

strategic intelligence, an example of intelligence or strategy

46:01

that you've given. to a CFO, what would

46:04

you say is the best example of a

46:06

wow moment a CFO had when you deployed

46:08

layer next at their company and against their

46:10

data sets? What was the biggest wow moment?

46:13

So we had one customer, he wanted to

46:15

understand whether we want to hire more sales

46:17

people or not. Okay? Yeah. So then we

46:19

analyzed how much sales people they have today,

46:22

how much? sales they make in today and

46:24

also the cash flow. So then AI generating

46:26

strategies, if you add the one sales agent,

46:28

this could be the revenue. Then how much

46:31

is your margin would be? Those are well,

46:33

yeah. So this is a great idea, but

46:35

it's hard. And so some ideas are hard.

46:37

What's hard about this idea? Well, what one

46:40

CFO wants to solve might be very different

46:42

than another CFO. If you don't have sales

46:44

people, well. and you have retailers, you have

46:46

a different task to do here. And as

46:49

you mentioned, there's different stages. Some companies don't

46:51

have a CFO, then companies have an outsourced

46:53

CFO, then you hire your first CFO, then

46:55

you have a public market CFO, you know,

46:58

you have a real range of different stages

47:00

of companies, you have different goals, and then

47:02

also you're trying to tell them things, actionable

47:04

items, that maybe they're not even aware. So

47:07

that would be like doing a blood test

47:09

with superpower and it comes back to you

47:11

and says, hey, and you're doing superpower, get

47:13

your blood drawn soon, you just signed up.

47:16

Maybe they say to you, oh, you're.

47:18

vitamin D. Now you

47:20

would never say to

47:22

them, I need to

47:25

take a vitamin D

47:27

test. So they take

47:29

all the tests. So

47:31

this business, you probably

47:33

do need vitamin D

47:36

because you're not outside

47:38

enough. This is your

47:40

problem. You have disparate

47:42

systems and you have

47:45

insights you can give

47:47

them that they may

47:49

not even know they

47:51

need. So the value

47:54

of this product is

47:56

hard for them to

47:58

know. And so what

48:00

you probably have to

48:03

do is figure out

48:05

what is the most,

48:07

which group of CFOs

48:09

are gonna have the

48:12

most wow moments and

48:14

get the most value

48:16

from your product. If

48:18

it's people with sales

48:21

teams, you'll know that

48:23

because Salesforce exists, HubSpot

48:25

exists. If it's people

48:27

with retailers, maybe they

48:30

use SAP, who knows?

48:32

Maybe they use NetSuite.

48:34

So I think you

48:36

have to plan to

48:39

flag early on to

48:41

find an ideal customer

48:43

profile, narrow the focus

48:45

down to, hey, you

48:48

know, this product, Snowflake,

48:50

NetSuite is the industry

48:52

standard. These people have

48:54

money to spend and

48:57

we can help them.

48:59

So Ikigai, do you

49:01

know that? Ikigai. Ikigai?

49:03

Yeah. Pull up the

49:06

Ikigai chart. I'm gonna

49:08

show you something that

49:10

might blow your mind. Ikigai,

49:13

have you heard of this before? Oh

49:15

no, I don't know. Okay, Ikigai is a

49:17

Japanese philosophy. What are you good at? What

49:21

does the world need? What are people

49:23

willing to pay for it? And there's other

49:25

circles. People have made all kinds of

49:28

different spins on it, but somebody's gonna pull

49:30

up the Ikigai chart. Okay, here we

49:32

go. So we'll look at this for a

49:34

second. And I just want you to

49:36

slow down. We're not talking about your startup

49:38

here. We're talking about life. So we

49:40

have what you love. Okay, you love data,

49:42

don't you? Okay, what the world needs.

49:45

Analysis of that data to get insights. What

49:47

are you good at? You're good at

49:49

making that software and will people pay for

49:51

it, right? Somewhere in here is your

49:53

Ikigai of your startup. What

49:57

software the world uses.

50:00

for data. It's going to be hub

50:02

spot, it's going to be net suite,

50:04

etc. What do they love? They love saving

50:06

money, they love making money. What are you

50:08

good at? You're good at telling them how

50:10

to save money, how to make money, how

50:12

to avoid maybe one of your value

50:15

propositions, how to avoid tax issues

50:17

in the future, or how to

50:19

save money on taxes. It could

50:21

be all of those things. And

50:23

would people pay for it? Well.

50:25

They'll be overlapping circles here. So

50:27

IKEG-I-K-I-G-A-I is a way for a

50:29

human being to look at the

50:31

world and say, what should I do

50:33

with my life? What's my purpose?

50:35

IKEI for startups is a

50:37

new concept that I'm just debuting

50:39

here right now for the first

50:42

time, but it came to my

50:44

mind, which is, IKEI for

50:46

startups is, who are these

50:48

customers? What do they covet? And

50:50

what can you do with them, right?

50:52

So what would you say for that for

50:55

that company that got the

50:57

wow moment and it provided

50:59

great value for them? Do you

51:01

think they'd be willing to pay for

51:04

that or is that like a

51:06

one-time insight? Or is that

51:08

a reoccurring inside? I'm curious. Is

51:10

the reoccurring? Yeah, because perfect.

51:12

Yeah, because you have to

51:15

monitor the sales people. Yeah. So

51:17

you found something. Where was data

51:19

held? They have the sales force and

51:21

they pump in the data to the

51:24

data warehouse every night. Okay, so

51:26

they have sales forces where the

51:28

data resides. Did you cross-reference

51:31

it with any other data? They have the

51:33

county system, it's the legacy accounting

51:35

system. Oh, so you had the

51:37

legacy accounting system. Do you have

51:40

the legacy accounting system? Do you

51:42

know what name of that is?

51:45

It's not QuickBooks or something? So

51:47

while you're figuring out... your ideal

51:49

customer profile, you're going to have

51:51

to figure out how many people have

51:54

sales force and this accounting thing

51:56

and maybe start with that group. Now

51:58

you have identified a subset. Maybe in

52:00

the future you want to go in

52:02

and take every piece of data from

52:04

every system and give this magical, you

52:07

know, here's how to run your business.

52:09

Like it's almost like a shadow CEO,

52:11

a shadow CFO advising people. It's like

52:13

a clone, right? You've got this like

52:15

perfect clone that's out there working as

52:17

an agent 24 hours a day trying

52:20

to figure this stuff out. But maybe

52:22

we start and say, you know what?

52:24

There's enough people with sales teams and

52:26

sales data and customer engagement data. that

52:28

we can just go in and tackle

52:30

that first. So you have a feature

52:32

you can say to people, hey, you

52:35

got sales force, you have over 50

52:37

sales people, we can really help you

52:39

figure out how to make decisions in

52:41

your sales group. Then you say to

52:43

the marketing group, hey, we know your

52:45

CAC, we know where you're spending money,

52:47

we can help you spend money more

52:50

efficiently to then get it into the

52:52

sales group. Then you say, okay, now

52:54

we're gonna work with our accounts and

52:56

our tax people. We can tell you

52:58

how to save money internationally figuring out

53:00

your tax status and where to put

53:03

these sales etc. But you start with

53:05

one, then you build the adjacencies. Does

53:07

that make sense? I think makes sense.

53:09

Yeah. So we get a lot of

53:11

custom. Some people is the early stage

53:13

of the data journey, data maturity journey.

53:15

So we are in the waiting list

53:18

still. So makes sense to me. Yeah.

53:20

I mean, the good news also is

53:22

and most VCs will not say this

53:24

because they really want you to scale

53:26

and not build custom software right because

53:28

custom software is custom and it does

53:30

it's not repeatable but if you did

53:33

some client engagements where they needed your

53:35

help with some servicey kind of stuff

53:37

and it was custom and it was

53:39

bespoke if that bespoke work gets you

53:41

a lighthouse customer I'm gonna say go

53:43

ahead and do it if that bespoke

53:46

custom work for them and consulting if

53:48

they're paying for it and it makes

53:50

your product better and more scalable for

53:52

the next customer. on the IP for

53:54

that stuff, I'm going to say, go

53:56

ahead and do it because this is

53:58

going to be years of you grinding

54:01

it out to get this data and

54:03

normalize it. And if you can make

54:05

a little bit of money along the

54:07

way to keep the lights on and

54:09

have to raise less money from VCs,

54:11

that can be good too because you

54:13

keep more of your equity. Now, VC

54:16

would tell you don't do that. You

54:18

know, build the platform, we'll give you

54:20

the money, we get your equity, but

54:22

it's a way for you to not

54:24

do it. Mark, you have any thoughts

54:26

on here and advice? But I think

54:28

you've got enough to go on here.

54:31

You know, let's try to define that

54:33

ideal customer profile and then I want

54:35

you to also bear hug them. Did

54:37

I ever talk to you about the

54:39

bear hug strategy? when you're trying to

54:41

find these lighthouse customers, the one who

54:44

shine this beacon of light that other

54:46

customers follow. Oh, you know, we got

54:48

this company, that's got a lot of

54:50

sales people in it. It's or it's

54:52

IBM and IBM uses sales force and

54:54

IBM's got this glow or it's KPMG.

54:56

KPMG's got all these sales people all

54:59

over the world selling, you know, audits

55:01

or whatever. And, you know, we now

55:03

have them as our lighthouse customers to

55:05

Ernst and Young and, you know, other

55:07

groups might follow their follow their lead.

55:09

If you can get one of those,

55:11

and then you can embed yourself at

55:14

their office, so they have this problem,

55:16

they've got data problems. You say, hey,

55:18

you know what? We want you to

55:20

be our lighthouse customer, you're super up

55:22

front with them. Would it be possible

55:24

for us to get like a war

55:27

room, a conference room, or are you

55:29

super up front with them? Would it

55:31

be possible for us to get like

55:33

a war room, a conference room at

55:35

your office? Those other opportunities could inform

55:37

you like these could be circles you

55:39

didn't in your eke guy didn't anticipate

55:42

emerging so I like the idea of

55:44

like getting really close to a couple

55:46

of people who Love your product and

55:48

learning from them and you'd learn so

55:50

much being embedded in that way. There's

55:52

things you can't pick up just on

55:54

a phone call or resume. People don't

55:57

know about themselves and their own business

55:59

that you'd pick up being in their

56:01

face all day. If you were Panavision

56:03

or you were the red camera company

56:05

or one of these, it would be

56:07

the equivalent of saying like, I make

56:10

these incredible cameras, you're making a movie

56:12

or a television show. Can we send

56:14

a couple of technicians to hang out

56:16

on set with you and answer any

56:18

questions you have? work well and you

56:20

know they keep slipping out of people's

56:22

hands we're gonna make a grip that

56:25

doesn't fall out and you know you

56:27

get you gain some incredible knowledge so

56:29

great job and we wish you great

56:31

success. I mean maybe we'll do one.

56:33

credit rapid response. Oh, okay. We got

56:35

we got a few of those. We

56:37

got a few of those. You were

56:40

hanging out. So Jason was hanging out

56:42

on the R slash anti-work sub-read. Yes,

56:44

capitalist in the anti-work Sub-ready. Yeah. Just

56:46

to get mad. Jason. They do show

56:48

up. But just like I get them

56:50

out. You know what? I have been

56:53

we've been working on return to office

56:55

for our company, you know, and have

56:57

a certain philosophy. Extremely high performers can

56:59

be remote. But some jobs need to

57:01

be in person so we've been slowly

57:03

working on this getting back to in

57:05

office and I guess because I was

57:08

researching active track which is like productivity

57:10

software and we have to lock all

57:12

computers down I started I think the

57:14

way I stumbled upon anti work was.

57:16

People were talking about active track in

57:18

there, which is tracking software that you

57:20

put on your corporate laptop and it

57:23

watches everything you do watches everything you

57:25

do But it's really for a finance

57:27

company also to secure your laptop. So

57:29

if somebody were downloading the database sharing

57:31

it with people who shouldn't be faced

57:33

Basically, like so, you know, anyway, people

57:36

were talking about active track on there

57:38

and I saw this Thread and it

57:40

just resonated with me so maybe you

57:42

could cue up sure the person line

57:44

from user electric horsepower They're asking why

57:46

the big push to return to office

57:48

I get a sense that the majority

57:51

of domestic employers want everyone to return

57:53

to office. I understand that leases on

57:55

buildings need to be maximized, but is

57:57

there anything other than money that would

57:59

make a company have all of its

58:01

employees come back to the office? Yeah.

58:03

So I just thought I would explain

58:06

to them what are some of the

58:08

other reasons that people are actually doing

58:10

this and the why. And a little

58:12

bit of my philosophy, so maybe you

58:14

could just read a little bit of

58:16

my response. If there's typos or things

58:19

in there, please feel free. I usually

58:21

leave my typos in now, like Grammarly

58:23

tries to correct my typos. I'll clean

58:25

up egregious ones, but I leave a

58:27

couple of typos in so people know

58:29

it's real. People though it's not AI,

58:31

it's you. So Jason wrote, I wrote

58:34

a couple of podcasts, all in, this

58:36

week at venture capital firm, launch you've

58:38

out of university. We're in a competitive

58:40

space and being in person makes us

58:42

faster at everything. Two, energy level, intensity,

58:44

and pace is different. Three, after four

58:46

years of working for home, people had

58:49

lost their intensity, the culture had disappeared.

58:51

Four, after starting to beat with founders

58:53

in person again, it was a huge

58:55

advantage. Five, on the margin, we had

58:57

10 to 20% of folks abusing work

58:59

from home, which you figured out via

59:02

active track. Great software for teams of

59:04

high performers because it exposes folks who

59:06

are phoning it or abusingusing it or

59:08

abusing the system. However, that's just one

59:10

data point, important to keep in mind.

59:12

And finally, creativity, when folks are in

59:14

creative meetings in person, they bring their

59:17

A game, but when they're on Zoom,

59:19

they can get distracted, they could disappear

59:21

or whatever, and then you filled it

59:23

out with some other important, you know,

59:25

personal sort of information, said happy to

59:27

do an AMA. Yeah. So thoughts on

59:29

my response. I think you're correct. I

59:32

actually chibed and responded to this comment

59:34

myself. And I was, I was very

59:36

skeptical. I was a, I loved working

59:38

from home. I thought that was like

59:40

the dream. Yes. Like I can hang

59:42

out in my PJs all day just

59:45

at my computer. I don't have to

59:47

commute. Yes. And I enjoyed it for

59:49

a while, but after several years. I

59:51

think exactly what you said about being

59:53

distracted losing some of the intensity and

59:55

focus and I'm I like to work

59:57

I know you're a worker bay for

1:00:00

sure hard worker yeah I'm not somebody

1:00:02

who's like naturally like I don't want

1:00:04

to do that I'm going to hang

1:00:06

out for an hour but even I

1:00:08

found that after a while it's very

1:00:10

easy to get distracted when you're in

1:00:12

your house when you're never face to

1:00:15

face with your coworkers when everybody is

1:00:17

just like a little line of text

1:00:19

on your chat app instead of being

1:00:21

in your face You just lose that

1:00:23

personal connection. And when you're in an

1:00:25

office with other people, they're your peers

1:00:28

and your coworkers. You raise your own

1:00:30

game to keep up with all of

1:00:32

them. And you don't do that when

1:00:34

you're at home. You set your own

1:00:36

energy level and everybody kind of has

1:00:38

to come to you. And I think

1:00:40

it just slowly at your fees. I

1:00:43

know it did for me. If you

1:00:45

are a young person, I think, and

1:00:47

you've been doing this for a while

1:00:49

and you're convinced like, it's some crazy

1:00:51

capitalist and they're forcingcing you to come

1:00:53

to come to come to work. It's

1:00:55

actually in your benefit. I think not

1:00:58

socializing with people is making a generation

1:01:00

of very weird people. I've been talking

1:01:02

to some parents who have kids older

1:01:04

than us who lost their entire college

1:01:06

years or lost their high school years

1:01:08

to the COVID lot then. So they

1:01:11

literally lost graduation of college or graduation

1:01:13

in high school. You should demand as

1:01:15

a young person to be in office

1:01:17

and to be near the locus of

1:01:19

power and to be mentored and to

1:01:21

be professionally developed. You should be demanding

1:01:23

that. You're getting ripped off. Alex and

1:01:26

I came up at a time, you

1:01:28

know, me a little bit earlier, where

1:01:30

we were in rooms with editors, reading

1:01:32

our work out loud, telling us we

1:01:34

sucked, telling us how to be better.

1:01:36

We got to watch other people do

1:01:38

the job. And we had people model

1:01:41

at first and the professional development that's

1:01:43

happening on when we have our Monday

1:01:45

editorial meeting for founding university, which I've

1:01:47

been at two or three of them.

1:01:49

I mean, the learning, the feedback I'm

1:01:51

getting back from people is like, whoa,

1:01:54

that is like the best part of

1:01:56

the job. I mean, you're... first couple

1:01:58

of jobs, you don't know how to

1:02:00

be a good employee, yet you haven't

1:02:02

done it. It's your first couple of

1:02:04

jobs. And yeah, I can't imagine starting

1:02:06

my career in a work from home

1:02:09

like I was doing it after a

1:02:11

decade of being in an office every day,

1:02:13

being told exactly what to do. I can't

1:02:15

imagine just starting your career from

1:02:17

that. I want to double clearly what someone

1:02:19

Jason said though, being near the locus of

1:02:22

power is the real hack here. People talk

1:02:24

about mentorship and culture and all of that

1:02:26

to some degree. But if you want your

1:02:28

career to accelerate, what you want is time

1:02:31

with the SVP or the CEO or whatever,

1:02:33

and they're never going to have time for

1:02:35

you on slack, but you can find them

1:02:37

in the office, you can make you guys

1:02:39

collide and then get to know them, lever

1:02:42

that, I mean, if you're ambitious, I don't

1:02:44

think promotes for you, if you are an

1:02:46

individual contributor who has a defined role and

1:02:48

you're very good at it, sure. But I

1:02:50

mean for everyone else who's not bad. I

1:02:53

mean, it's kind of for everyone else who's

1:02:55

not bad. extremely high performer in a

1:02:57

specific vertical with a very tight skill set

1:02:59

with a tight arrangement. You're going to do

1:03:01

X, Y, and Z. Perfect. I'm trying to

1:03:03

be the return to office armistice person.

1:03:05

Right. Which is somebody asked me kindly here a

1:03:07

young person. I'm going to New York. I'm going

1:03:10

to a wedding. I don't want to use one

1:03:12

of my vacation days. I'll work really hard remote,

1:03:14

remote, but I don't want to go see New

1:03:16

York. I've been to New York a million times.

1:03:18

I don't like the city. I don't like the

1:03:20

city. I was like, well, that's a bit of

1:03:22

an insult. I'm from there, but I get it.

1:03:24

I like to kind of get it. One shout

1:03:26

out to Maddie and Maddie was like, can

1:03:29

I work remote one day? This one spent

1:03:31

any time in New York. I was devastated.

1:03:33

I was like, I'll give you 10 things

1:03:35

to do. It doesn't even sleep. But I

1:03:37

think she wants to save the day for

1:03:40

a proper vacation. You know what I said

1:03:42

to her? You're a high performer. Certainly

1:03:44

fine, just let the team know if you

1:03:46

need to take a remote day at least

1:03:48

go to Castile or something like our town.

1:03:51

There's some Chinatown, get some peaking

1:03:53

dog, that's what I have. Go for a

1:03:55

walk and such a food. I haven't had

1:03:58

lunch yet, and it's hilly meat. Four. Alex

1:04:00

Wilhelm, the Wilhelm. He's at Alex

1:04:02

on Twitter, at Alex optimism. Go

1:04:04

give him the Hyundai. Get

1:04:06

his go give day on his

1:04:08

newsletter. get his insights every day on his newsletter,

1:04:10

Mark Search for hash rate

1:04:12

right now. the pod, pod. Go

1:04:14

to hash rate now, pause sign

1:04:16

up to hash rate, and sign up and be

1:04:18

a little out of your

1:04:21

depth, but you'll catch up

1:04:23

pretty quickly. but And Lon

1:04:25

Harris, he's quickly. And You're Harris, on

1:04:27

at Lon, You're at Mark You're a

1:04:29

Trump supporter. active, you're a You

1:04:31

have Trump dedication syndrome. TDS. Ron and

1:04:33

Alex have Trump flip it around syndrome.

1:04:35

You have Trump syndrome. You have

1:04:37

Trump dedication syndrome. And I

1:04:39

call strike. See you all next

1:04:41

time. Bye bye. time. Bye bye.