0:00

Bitcoin is digital gold. That was actually the way we

0:02

were thinking about it at the time. Deploy

0:05

in Bitcoin and you'll

0:07

do well, kind of thing. We were fully

0:10

open to the opportunity of decentralized

0:12

finance. In a sense, DeFi

0:14

hopefully will become useful by being applied

0:16

as a primitive, as a building block

0:19

to the applications and marketplaces we're trying

0:21

to build. And then in that way,

0:23

we'll see it become much bigger. The

0:25

open movement, the ability to share data,

0:28

the ability to share research, that

0:30

has been the origin, that has been

0:32

the petri dish from which all the notable

0:34

inventions have come. I

0:36

think that the other side of the impact

0:39

of AI on Web3, just a very

0:42

specific one, is that if we

0:45

look at the way in which

0:47

AI copilots are tracking and their

0:49

ability to create applications, that we're

0:51

going to go from a situation

0:53

where there are relatively few solidity

0:55

programmers and it's incredibly hard to

0:58

audit smart contracts to

1:00

a world where... This

1:15

episode is brought to you by Gnosis. Gnosis

1:18

builds decentralized infrastructure for the

1:20

Ethereum ecosystem. With

1:23

a rich history dating back to

1:25

2015 and products like Safe, Cowswap,

1:27

or Gnosis Chain, Gnosis

1:30

combines needs-driven development with

1:32

deep technical expertise. This

1:35

year marks the launch of Gnosis

1:37

Pay, the world's first decentralized payment

1:39

network. With the

1:41

Gnosis card, you can spend self-custody

1:44

crypto at any Visa accepting merchant

1:46

around the world. If

1:48

you're an individual looking to live more

1:50

on-chain or a business looking to

1:52

white-label the stack, visit

1:54

gnosispay.com. There

1:56

are lots of ways you can join the Gnosis

1:59

journey. Drop in the Gnosis

2:01

DAO governance form, become a

2:03

Gnosis validator with a single GNO token

2:05

and low cost hardware, or

2:08

deploy your product on the EVM

2:10

compatible and highly decentralized Gnosis chain.

2:13

Get started today at gnosis.io.

2:17

Cars 1 is one of the biggest

2:20

node operators globally and help you stake your tokens

2:22

on 45 plus networks

2:24

like Ethereum, Cosmos, Celestia

2:26

and DYDX. More

2:29

than 100,000 delegators stake with

2:31

Cars 1 including institutions like

2:33

BitGo and Ledger. Staking

2:35

with Cars 1 not only gets you the

2:37

highest years, but also

2:39

the most robust security practices

2:42

and infrastructure that are usually

2:44

exclusive for institutions. You

2:46

can stake directly to Cars 1's public

2:48

node from your wallet, set up a

2:51

white label node or use the recently

2:53

launched product, Opus, to stake

2:55

up to 8,000 ETH in a

2:57

single transaction. You can

2:59

even offer high yield staking to

3:01

your own customers using their

3:03

API. Your assets always remain

3:05

in your custody, so you can have complete

3:07

peace of mind. Start staking today at

3:10

cars.one. Welcome

3:12

to Epicenter, the show which talks

3:14

about the technologies, projects and people driving

3:17

decentralization and the global blockchain revolution. I'm

3:20

Sebastian Cuccio and I'm here with my co-host Fred

3:22

A. K. Ernst. Today, we're talking with Richard Murad

3:25

who is co-founder of Fabric Ventures and

3:27

also on the NIR Foundation

3:29

Council. So Fabric has been a long

3:31

time VC in the

3:33

Web3 and crypto industry. And it's been

3:35

a long time coming since we should

3:38

have had Richard on the show. But

3:40

10 and a half years later, here he is to

3:43

share Fabric's latest thesis,

3:45

which is about the intersection of

3:47

crypto and AI. And we'll also get in

3:49

some other topics like investing in crypto. And

3:52

as a emerging manager, I'm also

3:54

very interested in asking Richard

3:56

a bunch of questions about how he runs

3:58

his fund. So this is also... So

4:00

a good opportunity here to

4:02

learn about crypto investing. Richard,

4:05

thanks for joining us on the show this week. Fantastic

4:08

to be here. Don't tell anybody that it

4:11

took me 10 and a half years to

4:13

get on the show because they might make

4:15

a jab at my punctuality. And

4:20

also, something you reminded me,

4:22

I said just then, sorry, reminded me of Harry

4:24

Stebbings with his 20-minute BC show where

4:27

he has done an incredible

4:29

job of getting great guests on

4:31

and talking to them about the art of venture

4:33

investing and build that into a great franchise. So

4:35

I wouldn't be shy of that. That seems like

4:37

a fantastic way to

4:39

go. Not that I necessarily have anything to

4:42

add to the picture. And also, I almost

4:44

appeared on his show right at the very

4:46

beginning, but it didn't happen. So

4:48

maybe I'll have to wait until he's on his 10th year as well

4:50

before I get to do that. Well,

4:54

let's hope not. I mean, actually,

4:56

that's sort of not entirely true because

4:58

you were on our 10-year anniversary show.

5:01

You appeared there for a brief moment, but it wasn't

5:03

like a proper interview. But as

5:06

I understood, as I've learned,

5:08

I guess, fairly recently, actually, you've

5:11

been listening to the epicenter since pretty much

5:13

since the beginning, right? Yeah,

5:15

somewhere. I mean, I look

5:17

back at some of the episodes to kind of sort

5:20

of reacquaint myself. But I know

5:22

that I kind of

5:25

got the bug somewhere.

5:27

We got sucked down the rabbit hole or the wormhole,

5:29

as we actually like to call it here at Fabric,

5:34

in the kind of spring of 2013,

5:36

having kind of spent a little while

5:39

building up to that. But I

5:42

know that I then got pulled into kind of running

5:45

one of my sort of portfolio companies in 2014. And

5:48

I don't know, somewhere between the kind of

5:50

the autumn and the spring, epicenter became a

5:52

sort of staple part of my kind of

5:56

stress relieving sort of jogging routine

5:58

and reminding myself that that I

6:00

could give back to the kind

6:03

of romantic glories of decentralization

6:05

and Bitcoin at some point, which

6:08

I finally did. Yeah,

6:11

cool. And I think a lot of people

6:13

were listening to Epicenter while running back then,

6:15

including myself. I'd re-listen to episodes sometimes

6:18

while going

6:20

out for runs when I used to run, which

6:23

is something I haven't done in a long time. But yeah,

6:26

maybe let's feel for folks who don't know you,

6:28

dive into your background a little bit. So I

6:30

mean, Fabric has been around I think since 2015,

6:32

2016. What were you doing before that? And

6:37

how did you get involved in, I mean,

6:40

how did you start a crypto fund basically? The

6:43

moment of self-indulgence to tell about the classic

6:46

cryptic and origin story, but I'll try to

6:48

keep it on point and entertaining. And

6:51

yeah, I've been in it long enough that I've

6:54

heard a lot of crypto origin stories. So

6:57

I did engineering at university and kind

6:59

of always wanted to start a company

7:02

of some description. And

7:04

I kind of fell in love with software, the general

7:06

power of software. I mean, I tried my hand at

7:09

it when I was kind of 9, 10, 11,

7:11

12, I think as many folks do. Discovered, I

7:16

actually got my first computer from Clive Sinclair, if you

7:18

know who he is, his

7:20

parents were neighbors. I got the ZX81 back

7:22

in 1981. Anyway, he's famous in England, at

7:29

least as one of the early major

7:32

protagonist stakeholders in the

7:34

PC revolution. But I

7:36

quickly discovered that I

7:40

didn't have strong aptitude for software development

7:42

itself. And I came

7:44

out of university

7:46

and did some strategy consulting,

7:49

part of which is in the telecom space.

7:51

And essentially, along with my brother

7:53

Charlie had an epiphany

7:56

that there was immense power in

7:58

the openness and

8:01

permissionless nature of the IP

8:04

protocol, internet protocol

8:06

data networks that were rising

8:08

up in the early 90s, but

8:11

that if we were going to use them to

8:13

run the world's economy, that they maybe need to

8:15

be fixed to things like security and quality of

8:18

service and so forth. Actually

8:21

with the advice of someone who turned out to be a

8:23

long-term collaborator, Steven Waterhouse,

8:26

Seben, who was

8:28

an advisor, we

8:31

built a software company to run these very

8:33

large scale IP data networks.

8:37

And critically, I think in this

8:39

question of getting into the

8:42

crypto decentralization space, we

8:45

were quite early, it turns out, with

8:47

using open source software in anger

8:50

at an enterprise or at a carrier scale.

8:52

So we used an

8:55

open source instantiation of Corber

8:57

actually, it's called OmniOrb, it

9:00

was a publish, subscribe kind of middleware.

9:03

And we use that in order to build this

9:06

product and ultimately IPO'd

9:08

London Stock Exchange and then NASDAQ and they

9:11

did a merger and then sold to Oracle

9:13

and that software was still being used

9:15

today actually to manage large IP networks. And

9:18

came out of that, had a brief sort of

9:21

got a taste and this is related again

9:23

to this question of getting into venture. I

9:25

got seduced somewhat to spending

9:27

some time with Axel Parnas, who

9:30

are obviously pretty storied and famous

9:32

and infamous these days, not least

9:34

for Facebook. Back

9:37

now 23 years ago when

9:39

they were setting up their London

9:41

office, actually it came

9:43

about and this is, I

9:46

found interesting, I actually re-energized this

9:48

connection just last week at an event.

9:51

I basically ended up ordering a margarita

9:53

at a bar with Rob

9:56

Glaser who ran Real Networks. So those

9:58

of you who remember Real Networks. from

10:00

last century and it's

10:02

still running today. And

10:04

Jim Briar from Axl Partners, it turns out, and

10:06

he just said, hey, you should meet my partner

10:08

Kevin Camone, he's sitting up in London. And for

10:10

me, interestingly, that's one of

10:12

those fortuitous connections we should always

10:15

be striving to kind of make. I

10:18

incubated a company there at Axl Partners, both

10:20

getting an understanding of what it takes to

10:22

get a new company from scratch, the kind

10:24

of somewhat daunting task of looking into the

10:26

blank sheet of paper, but also learning

10:29

about the Silicon Valley's flavor of

10:32

venture investing, which I

10:34

think is still something that we shouldn't necessarily

10:36

just be looking to ape or copy in the

10:38

rest of the world. We should be plowing

10:40

our own sort of furrow, but of course we can

10:42

be inspired by and, but also

10:45

built a second company that

10:48

used machine learning primitives to

10:50

augment the function of knowledge

10:53

workers operating actually in the IT management

10:56

space. And one of

10:58

the things we encountered there was a

11:00

kind of coordination problem, incentivization and coordination

11:02

problem. So how to get people to

11:05

give up their data, their kind of

11:07

hard one data for

11:09

how to operate a

11:11

particular part of the kind of back office of a

11:14

bank, for example, when actually that

11:16

very same data was the way they were kind

11:18

of clinging onto their job because

11:20

it protected their kind of little fiefdom they had.

11:23

And so through those, I know those

11:25

threads are kind of fully apparent, but early

11:27

work with open source software,

11:29

looking at machine learning, privileges before they

11:32

became famous with AlexNet and ImageNet and

11:34

so forth at the beginning of last

11:36

decade and looking at

11:39

incentive structures and collaboration with people.

11:41

When I stumbled into Bitcoin

11:45

and I genuinely don't know who

11:47

was, but I met somebody who was involved

11:49

in Bitcoin as early as January 2009, but

11:52

I really don't know who that

11:54

was. I can remember the type of person who

11:56

was in Switzerland. And

11:58

then it was a real... energizes a

12:00

conversation by my friend and

12:03

colleague, Steven Waterhouse,

12:05

AKA Seven, in the

12:07

spring of 2013. And

12:10

for me, just very quickly, the shoe

12:13

dropped, not just,

12:15

should we say, into their native

12:17

money, not just

12:20

the extensibility, arguably, of the possibilities

12:22

for the blockchain, but also I

12:24

can remember particularly really

12:27

getting excited by the concept of instantiating,

12:29

if not laws,

12:31

but organizational principles into code

12:34

in what were called

12:36

DAX and then became DAOs in 2013. Then

12:40

in terms of investing, I somewhat

12:42

reluctantly hung up my proverbial

12:45

soccer boots sometime in

12:48

2009, 2010, when I sold my second company

12:50

to BMC and then decided to age-old

12:52

investing and was looking for a thesis

12:56

that felt sufficiently impactful,

12:59

in some senses sufficiently crazy, that

13:02

it really just might work. And

13:06

so that's what Bitcoin, blockchain,

13:09

Crypto at large, Web3, Open

13:11

Web, we may well get into. That's

13:14

what that became, and that's the genesis of Fabric.

13:17

How are you inclined to start another

13:20

company as a founder rather than a

13:22

fund? Oh,

13:25

that gave me severe heartbreak, the

13:27

concept of not starting another

13:29

company again. And in some

13:32

senses, I have successfully scratched

13:34

that itch by

13:36

building Fabric over

13:39

these last years. But

13:41

in some other senses, it probably gets in the

13:43

way of me focusing just on

13:46

the investing. And certainly, we have

13:48

experimented a lot with quite hands-on

13:51

activities, researching what's going

13:53

on, hacking around

13:55

here and there in some of the products,

13:58

mining, state- making, validating,

14:00

nothing that we have scaled, particularly

14:02

within fabric, but we try and

14:04

keep our hand in that regard.

14:08

So, I think, but yeah, but

14:10

it was a bit of a heartache definitely to kind of

14:12

not build another company, because

14:15

whatever you do and look at, I did

14:17

not have household name success, I had some

14:19

okay success, but I felt like I had

14:22

way more kind of

14:24

like tar tracks in my back and scars on

14:26

my scar tissue from things that are not quite

14:28

gone the way I wanted them to, then I

14:30

had some successes. So, you're like, next time I'm

14:32

going to really do it just right and we're

14:35

going to kind of get

14:37

escape velocity. But yeah, it's a good

14:39

question, Frederik. Richard,

14:42

can you talk about your fun thesis for a bit?

14:45

So I mean, you started a

14:47

number of years back and it kind of changed over

14:50

time. So maybe tell us where you

14:52

started and where you're at now. Yeah,

14:55

absolutely. So I mean, inevitably, these things get

14:58

a little blurred over time and you tend

15:00

to have kind of the clarity of hindsight.

15:03

But I'll try to avoid that as

15:05

much as possible. I

15:07

kind of mentioned that we were looking for a thesis

15:09

I found to start a venture firm in Europe was

15:11

quite challenging, if you didn't have tens of millions of

15:13

your money to deploy in it back sort of 10,

15:15

15 years ago. So

15:18

I used my kind of company

15:20

starting kind of instincts that having a focus,

15:23

a distinctive specialization and trying to

15:25

sort of catch a wave would

15:27

be a good way to tackle it. So it was

15:29

kind of on the hunt for something. And

15:33

I had, as I also mentioned in the back

15:35

of my mind, some of

15:38

the principles of digital money,

15:41

should we call it, some of the power of

15:43

getting people to collaborate and share their data. But

15:47

I had not looked too deeply

15:49

at what was going on with

15:51

Bitcoin. And then I

15:53

was exchanging messages

15:55

with seven in the spring

15:57

of 2013. thesis,

18:00

you could have a thesis just around that space.

18:03

I think then of course, and

18:05

this is one of the things I think everybody in this space

18:07

has to wrestle with, when the

18:10

price is up, everybody is massively

18:12

positive and often over

18:14

exuberant and thinks that anything

18:16

is possible. And of

18:18

course, actually, I find ironically, it's at that point

18:20

in time that limited partners or folks who are

18:23

a bit detached from the space tend to get

18:25

jolted into action to look at investing. It

18:28

is ironic because of course, maybe it's when the price

18:30

is not so high that you want to be preparing

18:33

to deploy. And

18:36

we went through that winter

18:40

of 1415 when

18:42

everybody sort of started concocting the kind of

18:45

it's not Bitcoin, it's blockchain kind of narrative

18:47

and going a little bit sort of B2B.

18:51

Let's look at how that can be deployed in

18:54

companies like R3 looking at the kind

18:56

of the back office for banks

18:58

and so forth. I think

19:00

whilst we definitely didn't dismiss that we remained

19:04

believers in the

19:06

openness, permissionlessness and so forth of

19:09

the kind of the public

19:11

blockchain movement. It actually reminded

19:13

me of when I would

19:16

have been like 96, 97 or something. There

19:20

was a chief scientist of UUNet, one

19:23

of the powerful internet service providers at

19:25

the time. I remember spending some time

19:27

with him and he

19:30

was collaborating with IBM and IBM

19:32

was trying to get everybody to

19:34

build sort of private intranets to

19:37

solve supply chain problems and to

19:39

share information in particular with GE

19:41

and Ford and all of the

19:43

kind of clients. It was

19:46

the equivalent of a kind of permissioned blockchain and

19:49

those projects didn't go anywhere. I

19:51

mean, ultimately there was the adoption of the

19:54

underlying technologies but those projects

19:56

went nowhere and so I think for me

19:58

it echoed that. and so

20:00

we kind of remained on the kind of

20:02

public direction. So

20:04

both Max Murch and then Julien Tevena

20:07

of those forthcoming years joined and Julien

20:09

in particular was very early looking at

20:11

the DeFi space and looking

20:13

at the arbitrage of opportunities between

20:16

exchanges, for example, prior

20:18

to that, something that occurred to us

20:20

but other people exploited or took

20:23

advantage or profited from in a

20:26

way that I certainly never did. And

20:29

so we were fully open

20:31

to the opportunity of decentralized finance.

20:34

I think it's an interesting

20:37

question where we reach with the maturity

20:40

of that. I think we

20:42

may be at some kind of local top but

20:44

I think it's really a component of something much

20:46

bigger. I guess the only

20:48

other thing worth mentioning in terms of our

20:51

thesis is that just because

20:53

it had, because of the

20:55

nature where it come from in terms of

20:57

building companies that tried to harness data to

21:01

gather from people in kind

21:03

of networks, we always

21:05

saw ultimately Bitcoin

21:09

blockchain decentralized data structures, the coordinating

21:11

power thereof of tokens that are

21:13

native to them as a way

21:16

of organizing the world's data from

21:18

the bottom up. The

21:21

emergent power of properties of those networks

21:23

are the powerful thing that if we're

21:25

going to have a chance of getting

21:28

the right data to the right algorithm

21:30

at the right time in the right format

21:33

so that we could really all benefit to

21:35

the maximum from what we now see

21:38

accelerating in the world of AI,

21:40

then there's a beautiful marriage between

21:42

those two movements. And

21:44

that was for a long time

21:46

in accord of what we pitched and it's

21:49

gone in and out of favor but it certainly seems to

21:51

be much more in favor today. So

21:53

let's talk about the AI thesis. As

21:57

we were preparing for this, I was reminded that in

21:59

like 20 years, 17,

22:01

18, there was a moment where

22:04

blockchain and AI,

22:07

that narrative had emerged amongst

22:10

all the other narratives at

22:12

the time, DeFi, this enterprise

22:14

blockchain idea that was floating

22:17

around in the 2015s to, I guess, like 17, 18 era. And

22:22

I remember at the time, I remember that

22:24

narrative just feeling very cringe

22:26

to a lot of us folks building

22:28

in DeFi and permissionless networks. And we

22:31

had people pitch us ideas for the

22:33

podcast and maybe sponsors and stuff like

22:35

people reaching out to us to be

22:37

on the show. So

22:40

the pitching this idea of AI and blockchain are

22:42

the perfect technologies, but it just felt so cringe

22:45

at the time. And to me at least, I

22:47

didn't see any sort

22:49

of promise or utility

22:51

in those two technologies merging and somehow

22:54

becoming complimentary for each other. What's

22:57

different now and why is

22:59

this narrative now making a

23:01

resurgence and what specific technological

23:04

advancements perhaps have now made

23:06

it actually interesting and something

23:08

worth looking at? Yeah.

23:11

So I guess

23:13

one of the things I draw from that actually is

23:15

that we all battle, you know,

23:18

market time, market slow with the emotions attached to

23:20

that. We also battle with the fact that trying

23:22

to actually just stick to what

23:25

we believe might be right and trying to ignore the

23:28

opinions of other people, which might be,

23:30

you know, just fighting that instinct of,

23:32

you know, shying

23:35

away from what feels a little bit of cringy,

23:37

should we say. Although

23:40

that being said, of course, you know, and

23:42

maybe we probably wouldn't get into it, but AI has got a long

23:44

history when it, the kind of the

23:46

dawn of the AI age has been heralded and

23:49

it's turned out to be a kind of

23:51

a false horizon, you know, going back decades,

23:53

of course. And

23:56

I think that, you know, it's pretty

23:58

well documented. One of the things that

24:00

is different, has turned out to be

24:02

different now, is that just the computational

24:06

power and the quantity of data

24:09

available to train then

24:12

some innovations in the

24:14

types of AI being generally

24:17

used, obviously notably the transformer

24:20

style LLMs, that is

24:24

really going to be taking things to the next level in

24:26

terms of the wow factor

24:28

for anybody using GPT-3, LLM

24:30

GPT-4. So

24:32

obviously that mainstreaming makes a

24:34

difference on the AI side. I

24:37

think in terms of the connection between the

24:39

two, I think first starting with the web

24:41

3 side of things,

24:44

I think there's often

24:46

in a sense a similar challenge later the

24:48

feat of web 3, which is like when

24:51

are we going to have mainstream apps? When's

24:54

that going to happen? And I

24:56

often get into and go, hold on

24:58

a second, can we just calibrate how

25:01

long it's actually been and how

25:03

far we've come? Just take a

25:05

measure, how many dApps being built,

25:07

how many developers being involved, how

25:10

sophisticated a web 3 stack

25:12

that's being built and say that's part of

25:15

3 trillion in market cap

25:17

that is going to be generated in

25:19

value, that feels pretty impressive. And that

25:21

I think that

25:29

work that has gone in to

25:31

build a web 3 stack that

25:34

truly can scale to support, should

25:36

we say, native on-chain payments, to

25:39

support consumer grade applications

25:41

that has the levels of traction in

25:44

it so that you don't have to

25:46

care about which L2 you're on or

25:48

indeed probably which L1 you're

25:50

on or

25:53

anything about specific to

25:55

the naming service of the aforementioned

25:57

L1. We

26:00

forget about how much has

26:02

happened, and I think it does feel like we're

26:04

at a tipping point there.

26:07

And then I think that all

26:10

of the work that's gone into

26:12

creating the DeFi ecosystem, we always

26:14

thought about those as necessary primitives

26:17

or building blocks in that overall

26:19

stack. This

26:22

is the first time we've had a wave

26:24

of software that has the ability to make

26:26

and keep financial promises, which sounds like a

26:28

pretty powerful tool. And

26:32

I think that's where the

26:35

marriage with AI comes about.

26:40

Because we've seen, and

26:43

I think what it comes about, because AI, we've

26:46

seen how the chokepoint for that

26:48

has been the capital required, not

26:51

so much the electricity consumed, but the sheer capital

26:53

required to buy the GPUs and get them

26:55

in a timely fashion in order to train models.

27:00

There are all sorts of chokepoints around electricity

27:03

distribution and even generation, if you want to

27:05

have that as native geotroglification that are emerging.

27:08

But I think it's going to turn now to data.

27:13

And that a lot of the open

27:15

sourced and commonly available data sets

27:17

that are of the scale required

27:19

for LLMs has already been consumed.

27:22

But there's a huge amount that is

27:24

still trapped in, for example, social media.

27:28

And I think famously, as Zuckerberg said,

27:30

we're not training llama 3 on our

27:33

social data. But

27:35

if we can find a way in which

27:37

we can respect people's privacy and perhaps even

27:39

reward people, incentivize people and

27:42

reward them for sharing that data,

27:46

there's a huge treasure trove of opportunity.

27:48

And I think we can then obviously

27:50

go beyond social network

27:53

data to health data and the

27:55

like and maybe deeper personal financial

27:57

data. And so I

28:00

think. that's the opportunity now, the maturity of

28:02

the stack now that is there, then it's

28:04

the opportunity to use those capabilities

28:06

of that stack to go and use

28:09

token economics to incentivize the sharing of

28:11

that data. I

28:14

think one of the most interesting

28:17

applications for crypto and

28:19

AI, at

28:21

least the one that I find the most useful

28:26

is this idea of user

28:28

owned AI or where

28:31

blockchains effectively act as a

28:34

way for users to

28:36

own their data, to leverage

28:41

their data if they wish to sell

28:43

their data to organizations

28:46

and folks building models. I

28:48

think this notion has also

28:53

emerged previously in crypto

28:55

when talking about social networks

28:59

and creating data networks

29:01

effectively. I think

29:03

that's one of the most interesting intersection

29:05

of crypto and AI aside

29:07

from the obvious marketplace for

29:10

compute and marketplace for GPUs,

29:13

which is an obvious use

29:15

case. In the case of user

29:19

owned AI, given

29:22

how we've

29:24

seen web 2 aggregate to very large players

29:27

and those have become very powerful and specifically

29:30

in the AI vertical have become

29:32

very powerful, do you

29:34

think that it's even possible or conceivable

29:36

that we

29:39

have user generated AI

29:41

or user owned AI

29:44

in 10 years that that is even a significant

29:47

enough portion of the AI market

29:49

to even matter? I

29:52

hope it's possible. It's been

29:55

characterized by Michael Casey

29:57

and his co-author as our greatest

29:59

fund. I think

30:01

it is a recent book. And

30:04

if we go back and look

30:06

at it, it's clear

30:08

that if

30:10

you follow the money, the

30:13

natural concentrating tendency of capitalism,

30:16

we've seen very evident in Web2, and

30:18

that this is already is

30:20

clearly further exaggerated with

30:25

the prevalence of the rise of AI.

30:28

And it's just driven by, we've seen it time and again,

30:32

that there are... Okay,

30:34

step back a bit. The

30:36

open movement, the ability to

30:38

share data, the ability to

30:41

share research, that has been the

30:43

origin, that has been the Petri dish from

30:45

which all the notable inventions have come. But

30:48

what happens, unfortunately, obviously, and we can

30:50

observe this, is that there are a

30:53

few milestones reached in that research and

30:55

the aggregation, using the

30:57

benefit of that data, and then people close off

30:59

their models. I think just recently,

31:02

Francois Chollet from Google went as far as

31:04

to say that along the lines

31:07

of open AI is put back frontier

31:09

research in AI by several

31:11

years, three, four, five years, since

31:14

it closed down through

31:16

its actions, the publishing of that research,

31:19

and also by virtue of the fact

31:21

that it has popularized data so much

31:23

that it's sucking the oxygen away from

31:26

other forms of research. So

31:28

I think we need to remember the power of openness

31:30

that's there, that if we're not careful, this

31:33

concentrating force steals from

31:35

us. And

31:39

I think that at every level of

31:41

the stack, and we would think of

31:43

the stack as, say, compute,

31:48

models, talent, and governance, at

31:51

every layer, there is the potential

31:54

to use a token economic

31:56

driven model, perhaps combined

31:58

with a new form of openness. and

32:00

source license for the models and the

32:02

data that has been generated in

32:04

order to give everybody a share. However,

32:07

that's easy to coin

32:09

a phrase somewhat as user and AI,

32:11

but there are many components to make

32:14

that happen, obviously. I

32:17

think at each layer, you mentioned clearly

32:19

we'll see Airbnb for

32:22

GPUs. We're seeing obviously

32:24

there are great companies like Jensen

32:26

and protagonists like Acash and so

32:29

forth that are out there building things. The

32:32

real challenge is there, of course, of getting,

32:34

and it's always, I always think of this

32:36

a little bit as the martini problem. I

32:38

think it was in the 70s martini out

32:40

as like anytime, anyplace, anywhere or something, presentation

32:43

of a martini. But anyway, it's the

32:46

how do you get the right type

32:48

of compute power available

32:50

at the right price and in the right spot

32:52

in the network at the right time to do

32:54

what it needs to do. That's

32:57

some of the challenges because there's a difference obviously

32:59

between training, for example, high compute

33:01

and inference, a requirement

33:03

for low latency. That

33:06

matching algorithm of supply and

33:08

demand is very challenging and the same applies

33:10

on the dataset. How

33:13

do you get that to work?

33:16

That's what people are wrestling with. But

33:19

I think this is a watershed moment. Either

33:22

we get a bit of a vicious cycle

33:24

where everything gets extremely concentrated, which I

33:27

think there are very strong arguments philosophically

33:29

is a bad thing. Power

33:34

corrupts absolutely. Power corrupts absolutely.

33:37

Without getting into politics

33:40

and the last century or so. And

33:43

then secondly, so we can

33:45

believe in philosophically, that's the right thing to do. I

33:48

think we can actually believe that,

33:50

and I've been thinking of the best

33:52

way to articulate this, but because I

33:54

think we've not seen it yet sort

33:56

of win out. We were an early

33:58

investor, for example. in Ocean Protocol, which

34:01

was looking to get a Maori, you

34:04

know, create a data marketplace. And

34:06

one of the key things is that

34:09

there is kind of no easy marketplace

34:11

for a large

34:13

marketplace for all types of data.

34:16

It's always about kind of very specific

34:18

data in a specific, tiny, you know,

34:20

more applied marketplace where the

34:23

value is. And then you have

34:25

to be able to negotiate the value of

34:27

the data at that point in time. And

34:29

a lot of companies that build software effectively

34:31

are doing that negotiation on behalf of that

34:33

data set. And so we need to create

34:35

something that's similar to that, that allows for

34:38

those collectives that own that data to negotiate

34:40

on the fly or the promise of the

34:42

value of that data in advance. And

34:44

I think that's a, it's an area of

34:46

research and of invention for, I think, kind of

34:48

for founders. And,

34:51

and what that is, in my mind's eye,

34:53

is a rather than, this is

34:55

a crude metaphor, rather than saying,

34:58

we're going to concentrate everything around the Ford Motor Company,

35:00

and there's going to be one type of car. And

35:02

it's, and by the way, you can have any color

35:04

as long as it's black, and it's going to, you

35:07

know, the mass production engine is going to be fed

35:09

and the profits are going to be concentrated like that.

35:12

What we need actually is to ultimately end up

35:14

with, and I said it was a

35:16

metaphor, an analogy, you want

35:18

to end up with having all of the different

35:21

variants of vehicle you could possibly imagine that fills

35:23

out all of the utility curve that

35:26

exists in this space. And

35:28

we need a much more adaptive, organic

35:32

type of system than you're going to get

35:34

from something that's command and control and centralized

35:36

and so forth. And you need, I think,

35:38

what you can get from all of the

35:40

great work that's been done within the Web3

35:43

community. So I think technically, it

35:46

could be much better than the

35:48

centralized alternative. However, if

35:51

we don't find a way in which we

35:53

make it economically better, and not just economically better

35:55

in the long term, but in the short

35:57

term for small groups of people, You

36:00

know, to your point, Sebastian,

36:02

it might be hard to win because

36:05

the kind of follow the money forces

36:07

are very strong. So

36:11

that kind of covers a lot of

36:13

the decentralized ownership aspects. We've

36:16

also seen projects and we ourselves

36:18

at Knossos have not been completely innocent in

36:20

this at actually putting agents on chain, right?

36:22

So in the beginning we did things like

36:24

we said, okay, this is a prediction market.

36:26

I'll give you 10 die. This

36:29

is how it works trade on it sort of thing.

36:31

But actually we've progressed to the point where we can

36:33

just spin up an agent on

36:35

chain and say, here you have a hundred X

36:37

die, make me some money. And the

36:39

bot goes out and does just that. So

36:43

obviously there are very real concerns

36:45

here as to how wise that

36:48

is, right? Do you want to

36:51

put something that

36:54

in principle could compete with

36:56

a human and have different set

36:59

of things it optimizes for

37:02

on a chain that you can't turn off? Do

37:05

you have thoughts as to that, Richard? I

37:09

mean, I think, look, I don't

37:11

fall into the camp of

37:13

there suddenly being an

37:18

acceleration towards the singularity

37:22

when these things start teaching

37:24

themselves and that

37:26

happens overnight. Having said

37:28

the ability to teach

37:31

itself an agent or

37:34

a model to refine itself, that

37:36

kind of second derivative, that

37:39

is something for us to watch out for. I

37:41

think the stuff for

37:43

settlement, for example, would be kind of an

37:45

agreement on that. But

37:48

I will leave those deeper

37:50

elements of the precariousness

37:52

of the AI gift to those who

37:56

are focusing fully on it. But

37:59

I will. I will say, I

38:01

guess where it comes up is, do

38:03

we want closed

38:06

AI, let's call it that, owned by

38:08

corporations and governments with the choices made

38:11

possibly by a very few people who

38:17

has that power and who can harness that

38:19

power? Or do we want something that is

38:21

more openness, open and inspectable? I think I'd

38:23

give for the open and inspectable and widely

38:26

distributed option. Maybe

38:29

we even reach a point where we

38:31

are in a dynamic equilibrium of

38:34

mad, as they call it, of

38:36

mutually assured destruction with

38:38

these things. Maybe that's some kind of

38:40

endpoint we reach. But look,

38:44

knock on words, it seems

38:46

to be in something that has actually in

38:48

some respects at least helped out in

38:50

the context of our transition from

38:53

the Second World War time, Cold

38:55

War and nuclear destruction. Again, I'll

38:57

touch with it a couple more

38:59

times. But

39:02

I guess on the question of, do we

39:04

want to give effectively persistent

39:06

autonomy to agents on chain that

39:09

can do things? It

39:11

feels like it might be fine

39:14

and indeed, unless it

39:16

can teach itself to adapt into something that

39:18

you didn't expect. And maybe

39:20

the problem is we don't know in advance

39:22

whether that's going to happen. I

39:25

think it's going to be quite hard to

39:27

predict what's going to happen with markets. I

39:29

think they're going to become increasingly perfect,

39:33

but maybe they're going to become increasingly volatile

39:36

as well. And

39:38

then I think

39:40

that the other side of the impact

39:42

of AI on Web 3,

39:45

just a very specific one, is

39:47

that if we look

39:49

at the way in which AI copilots

39:51

are tracking and their ability to

39:54

create applications, that we're going to

39:56

go from a situation where there

39:58

are relatively few Solidity programs. and

40:00

it's incredibly hard to audit

40:03

smart contracts to a world where actually,

40:06

and we need to watch out for

40:09

this, that vulnerabilities are possible to spot

40:11

using AI, but obviously, therefore,

40:14

we can use it to audit

40:16

them. And indeed, it's possible to

40:18

use multimodal developer copilots to create

40:21

the complete stack

40:24

of front end, decentralized front ends,

40:26

middleware and smart contracts, poor, whatever

40:31

it might be, whether it's a gig marketplace

40:33

or whether it's a decentralized exchange of some

40:39

description. And therefore,

40:41

the community strength that some

40:44

of these L1s have starts

40:46

to fade into the background. I don't

40:48

know what you've been thinking about on

40:51

that particular topic, Frederica, because you

40:53

seem to be nodding your head

40:55

a little bit that it could

40:57

change the dynamics, the competitive dynamics

40:59

quite quickly. Yeah,

41:02

no, I completely agree. So I totally

41:04

see the dangers. I think stopping is

41:06

not a great option because

41:08

we can't force every participant to

41:11

stop. I'd

41:15

rather we with hopefully

41:17

good intentions are in the

41:20

mix as well, because otherwise,

41:22

if you say, oh, this is a little

41:24

bit too dangerous for me, if someone else

41:26

continues, that does you no good in

41:30

the larger sense. Absolutely. I certainly

41:32

there are people I respect on the AI side,

41:35

including Nathan Beniak at Street

41:38

Capital and folks

41:41

who have commented that

41:43

obviously the human cry

41:45

around let's slow down

41:49

on AI, let's regulate it is

41:52

a classic form of regulatory capture

41:54

by Big Tech. And

41:58

it's self-serving to say slow down. if

42:00

what you're doing is meanwhile, you

42:03

know, having a hearty breakfast and getting ready to

42:05

sprint into the lead. You

42:07

know, so I think, yeah,

42:10

I would agree. I think we need

42:12

to keep going and as many people need to keep

42:14

going as possible. And

42:17

it's ultimately going to be extremely hard to put the

42:19

Kononchini back in the bottle. But I do think a

42:21

lot of the principles around transparent,

42:24

you know, governance,

42:27

decentralized governance, incorruptibility of

42:29

various, you know, points

42:31

of governance, collective ownership,

42:34

modern mutualism, if you will, in

42:36

terms we like, are good

42:39

principles, you know, and in some, from

42:41

Web3 to applied AI, and like in

42:44

some senses, you know, you

42:49

might think that it's hard to

42:51

imagine of a bigger use case

42:53

than in reinventing money and the

42:55

world's financial system, but actually sort

42:58

of reinventing, you know, and

43:01

I think, you know, crypto still has this

43:03

opportunity to reinvent our relationship with our personal

43:05

money. But actually, what's going to

43:07

happen with AI is we're going to reinvent our

43:09

relationship with the use of our personal data. And

43:13

I think the freedom to, you know,

43:15

to have that personal relationship, that direct

43:17

relationship is like, effectively

43:20

a human right. But

43:22

not just a human right in a kind of,

43:27

it's nice to have, but essential for the

43:29

kind of balance we want in

43:31

society. Yeah,

43:33

I agree. I think the

43:36

next hundred years, and maybe

43:38

even less, you know, we'll

43:41

see like tremendous shifts caused by

43:44

AI. And

43:47

we need to really think as

43:50

a society, how the data

43:52

is used and also

43:54

like how the AI is kept safe, sort of

43:57

along the lines of the AI. with

44:01

our objectives as a species

44:03

and doesn't cause chaos, which

44:06

is a tall order. But yeah, maybe

44:08

shifting gears here a little bit, what

44:10

other interesting trends are you seeing

44:13

in this space? And I'd like to ask

44:15

you a little bit about DeFi, because I

44:17

mean, there's been tons of innovation in DeFi

44:19

over the last four,

44:21

five, six years, but

44:23

it appears as though innovation is slowing down

44:25

a little bit. There's not very

44:27

much that's sort of new. It's like a lot of

44:30

rehashed ideas and optimizations.

44:35

Do you see DeFi, have we

44:37

reached like the top or is

44:39

there more innovation to come in by

44:42

way of like new

44:44

types of applications that really

44:47

redefine our

44:49

interaction with finance? So

44:52

I think we're nowhere near the top. I

44:54

think, but I think that realizing that

44:59

the capital that DeFi can handle is

45:01

not just your financial capital, but your

45:03

social capital and

45:06

different, and

45:09

that decentralized

45:11

applications are gonna be intrinsically social,

45:13

and they're gonna have the ability

45:15

to also handle transactions within

45:18

them. And how those

45:20

all get intermingled, I think is gonna be

45:22

the opportunity. I mean, we see people looking

45:24

at building decentralized

45:26

exchanges that can handle

45:29

not just fungible also,

45:31

but various combinations of

45:33

non-fungible token. And

45:37

so I think that there is opportunity

45:39

in that direction. And I think we'll

45:41

see a collision between,

45:46

should we say, AI's

45:49

hunger for data, decentralized

45:51

social media, and the way it

45:54

gives you the ability to mint

45:57

data that you own and

46:00

DeFi. DeFi's ability to create

46:03

marketplaces. And

46:05

I think there's an intersection there

46:08

that's very valuable. And

46:11

also, I guess, if you think about it,

46:13

it's just not just content data, it's also

46:16

things like health data. And

46:21

we've seen how DeFi

46:24

and the World Casino that we once,

46:27

the Web 3 has sort of stumbled into

46:30

can be incredibly good at

46:32

creating these highly liquid marketplaces. Even

46:36

if it is with money that is quite hot,

46:38

as in it's there while the going is good

46:40

and then it sort of disappears. And

46:43

I think one of the next challenges is

46:45

to how to apply that to data sets

46:47

that are actually where

46:49

the kind of value

46:51

is less volatile. For

46:54

example, health data, it's more long lasting,

46:56

but you can use the principles of

46:58

creating liquidity and bootstrapping marketplaces that

47:01

we've learned how to do on the DeFi

47:03

side and have them

47:05

kind of permute across. So I think, in

47:08

a sense, DeFi hopefully will become useful by

47:10

being applied as a primitive, as a building

47:13

block to the applications and marketplaces we're trying

47:15

to build. And then in that way,

47:17

we'll see it become much bigger. And I also think on

47:19

its own legs, I mean, like, I

47:21

guess we could ask ourselves a question, do

47:25

we yet have an app? And

47:27

I think, by the way, as a sidebar,

47:29

we probably feel that apps are disappearing, but

47:32

do we have a

47:35

sidekick set

47:37

of experiences, continuous streaming application experience that

47:39

is sort of tuned into the value

47:42

that we have and all of the

47:44

data that represents our credit score and

47:46

that continuously surfaces for us from

47:49

a financial

47:53

marketplace where there is no

47:56

human intervention and

47:59

presents it to us? us for whatever opportunity

48:01

we're looking at at that particular point in

48:03

time. Well, that's not happened yet. And that

48:05

seems to me the kind of the core,

48:08

still the core open finance or decentralized

48:10

finance opportunity and a lot of the

48:12

ability to do that is only just

48:15

coming to pass. You're

48:17

talking about this

48:20

emerging use case

48:23

of digital

48:26

asset managers that make use

48:28

of AI agents to

48:32

find optimized positions

48:35

for one's portfolio, that kind of use

48:37

case? Yeah, it could be optimizing positions

48:39

in your portfolio, but it could just

48:41

be saying, I'm

48:45

going to buy some clothes for delivery

48:47

tomorrow. I need to use credit for

48:49

that. Maybe I

48:53

have free credit. Maybe it's only two days if I've

48:55

sort of said I'm going to return it. But

48:58

to negotiate the credit for you

49:00

automatically in the background as part

49:02

of that transaction, also

49:05

simultaneously, of course, capture

49:08

a tokenized set of information

49:11

about that interaction that becomes attached. It's

49:14

part of your kind of data resource

49:16

that becomes attached to your profile, which

49:18

then in turn feeds into get a

49:20

future possible office of credit that you'll

49:23

be given sort of instantaneously next time you

49:26

need to do that or indeed not just credit incentives,

49:30

loyalty or whatever. I

49:32

think that the loyalty space,

49:34

which was something I remember looking

49:37

at a company called Ophamatic, best

49:39

part of 15 years ago, which

49:42

was looking to kind of use your transaction

49:44

data across all of your different accounts to

49:46

give you better loyalty.

49:50

Surely we want to get to the point where all of

49:52

the information are already around who you

49:54

are and how your transactions used productively

49:56

for you, hopefully, to give you

49:59

the office. and the loyalty and so

50:01

forth that you want. And then on

50:03

a kind of permissioned basis, you get

50:05

those offers that you really, you care

50:08

about rather than having a ad revenue

50:10

driven model that's sort of inbound. It's

50:12

more kind of permissioned, kind of permissioned

50:14

marketing that takes place. So

50:16

I think all of those are open

50:19

commerce, open loyalty, open payments,

50:22

and the DeFi marketplaces

50:25

behind them, they're all interwoven. And I think

50:27

we have barely got anywhere

50:30

with respect to those so far. And

50:35

within that is the kind

50:37

of real world asset marrying

50:39

to any concept that exists

50:41

today with that

50:43

world is tricky. And

50:45

the reason why obviously it's easier in the

50:47

kind of world casino cases when you have

50:49

a fungible asset and it's all natively on

50:51

chain and you're staying on chain, that's the

50:53

easiest place to get those marketplaces going. But

50:56

I'm sure we will get there. Yeah.

50:59

And so I'm speaking about trends here, maybe

51:01

looking back in your

51:05

journey as an investor in

51:07

the space, which

51:10

notable trends did

51:13

you see over the years that

51:16

you thought would play out or like you thought

51:18

would become a larger

51:21

part of the industry but ended up not?

51:25

What did you learn from those? Yeah.

51:30

I think I just

51:33

mentioned one, which is I think that the loyalty was

51:35

one that we thought would, there

51:38

was a payments

51:40

and loyalty company that we incubated that

51:43

was web two called YoYo. I

51:45

remember trying to pitch my partner, Alaf

51:47

Elise to take that on chain back

51:50

in, I don't know, 2014 or

51:52

15 or something. And

51:55

none of that has taken off

51:57

as quickly as it might. Again,

52:00

I think it's really a question of

52:02

timing rather than anything else. I

52:05

think that on-chain

52:08

games is another one that obviously

52:11

I think has got challenges to take off. A

52:13

lot of people know investors in

52:15

cartridge, a lot of people are doing

52:17

a lot of work, other firms like

52:19

Lattice and so forth to

52:22

build the capabilities to make that work and have

52:24

AAA kind of games

52:26

that can operate scalably

52:29

on-chain and indeed to

52:32

allow people to own, not

52:34

just sort of buy the

52:36

various artifacts in those games and then build

52:39

kind of sustainable token economies rather. I think-

52:41

Yeah, I think the jury's still out on gaming.

52:44

Yeah, exactly. It's not

52:46

happened. So I think that's been on, I

52:50

guess we've been impatient for that stake off. I

52:54

think, so I've always been bullish

52:56

also on just generally the power of

52:58

token economics which I kind of mentioned

53:01

earlier, I think is an essential part

53:03

of how we try and avert sort

53:05

of massive centralization or acute centralization of

53:08

kind of AI systems. And

53:11

I think I do

53:14

think that there's still a lot of work to be

53:16

done to work out how to build

53:20

resilient, sustainable token economic models

53:22

and to value them through

53:24

different inner stages because I

53:27

do think it's fundamentally different

53:29

from discounted cashflow, quite linear

53:31

economics of a centralized company

53:34

versus the self-reinforcing effects

53:36

you get in something that is more akin

53:38

to a city or a forest and it's

53:40

got a lot of always different immersion properties

53:43

and that arguably, or at least we

53:45

would argue can be rather than

53:47

inevitably going from kind of productive

53:50

to extractive as much

53:52

talked about in our kind of sphere can

53:54

be when we get that kind of platform

53:56

power and platform risk can be a

53:59

more positive. some game and

54:01

more sustainable long term. I think that's an

54:03

area that we'd, I guess I expected us

54:06

to advance a little bit further into today,

54:08

but I think it's still to count. I

54:10

mean, we were back as so

54:13

rare back in the day. That was

54:15

one of the ideas that was kind of elegantly

54:17

simple and sort of was almost a kind of

54:20

a trite example of what would

54:22

be powerful on a kind of blockchain and as

54:24

it has gone well- Shout out to the Strata Mafia. There

54:27

you go. And then

54:30

I think one that we missed, I

54:32

guess this is a little bit to do with actually

54:37

trying to apply a more typical

54:39

venture capital approach to the space,

54:41

was that in venture capital in

54:43

general, in the sort

54:46

of famous case study of this when a16z

54:50

backed bourbon that became Instagram and also pick

54:52

a please if I remember what it was

54:54

called, and then they ended

54:56

up on a kind of collision path and obviously

54:58

Instagram was kind of the winner. The

55:02

VCs can only back one player.

55:04

You have to pick up someone

55:06

to partner with and you're really

55:10

kind of full on with that. Whereas I think in the kind

55:12

of obviously L1s, L2s and all of this space,

55:14

a lot of folks have backed multiple

55:16

different players and then there's been enormous, the

55:19

opposite direction, there's been enormous value

55:22

generated and possible returns, distributions for

55:25

venture investors from a whole swathe

55:27

of those L1s and L2s. So that's

55:29

been a bit, I don't think everybody

55:32

predicted that that would happen, that it

55:34

wouldn't all kind of become an Ethereum

55:36

game quickly or some new winner and

55:39

so forth. So I think that's been not,

55:42

I don't think anybody, well, we

55:45

could chat to someone who said that they

55:47

called that, but we didn't play right across

55:52

the field like that. So there was also kind of, should

55:54

we say a missed opportunity in that sense. deals,

56:01

where basically the company pitched you and

56:04

you passed

56:07

and wish you didn't. I

56:11

mean, look, there were definitely, I mean, at the top of my

56:13

head, there were two companies that I,

56:16

by the way, I'm

56:18

very bad at the media recall of these things. But

56:20

the two ones that are on my head is that,

56:22

you know, we have a lot of time for the

56:24

team at Janssen and when

56:28

we first saw them, they were more

56:30

enterprisey and focused, but we became decentralized

56:32

and more, you know, we

56:35

think they're doing great things. And I guess

56:37

the bit I'd like to emphasize that in, you know, in

56:40

talking to them, to the

56:43

founders, Ben and Co, that they're just

56:46

very impressive founders and individuals. And that's

56:48

a huge thing that we can have,

56:51

you know, index on. And you've got

56:53

to be very careful to not

56:56

talk yourself out of some of the challenges in

56:59

these companies. And then another one that

57:01

we should have dug into more, probably

57:05

is lazy ledger that became

57:07

Celestia, where

57:10

Mustafa actually, we

57:13

had a discussion about him joining us in a

57:15

very early when he was still at UCL. So

57:17

we knew him and we knew how the people

57:19

in the space, but that didn't happen. So, and

57:21

look, we're, as we've already

57:24

discussing like kind of, should we say DPN

57:26

and DPN for AI and, you know, we

57:28

think that's an important part of the future

57:31

and interoperable modular open blockchains. We think

57:33

that's also an important part of the

57:36

future. So we definitely kind of do

57:38

his own pieces. Well,

57:41

I'm going to now spend the next

57:44

few minutes that we have here asking

57:46

my own selfish questions as

57:48

an emerging manager of

57:50

a VC fund. So

57:52

yeah, the first thing I wanted to ask

57:54

you is like, crypto

57:59

valuations. obviously have a

58:02

premium and that's because of

58:04

the low time to liquidity. That's one of

58:06

the reasons. At

58:08

the same time, VCs present

58:11

themselves to teams as

58:14

long-term aligned and long-term

58:16

supportive. Do you

58:18

think that there's a contradiction there

58:21

where short-term liquidity and

58:23

the way market cycles

58:25

operate forces you,

58:27

if you're meant to hit is as

58:29

a fund to return cash to your

58:31

investors, return capital, is

58:36

the short-term horizon of

58:40

liquidity and crypto counter

58:44

to what a

58:46

fund should in theory

58:48

be providing to a team is also

58:50

long-term support? How

58:53

do you think about that? Specifically

58:55

when it comes to exiting liquidity?

58:58

Yeah, so it's definitely

59:00

something we've been wrestling with for six,

59:03

seven, eight years. Obviously,

59:05

look, it's a new frontier and

59:07

so therefore there's no reason why

59:09

there should be a well

59:13

understood or a quick

59:15

or easy answer and

59:18

at risk of giving a color, it depends or

59:20

it's somewhere in the middle answer. I

59:24

think that venture firms should be, venture

59:26

backers should be thinking for the long-term

59:28

and backing those projects for the long-term.

59:31

I think there's clearly a segment for which

59:34

that makes no total sense. I

59:37

think that makes sense for

59:40

the projects because they want to

59:42

know that you've got their backs

59:44

through through thick and thin. When

59:47

it comes to the question of when to

59:49

sell, you also

59:52

have a duty obviously to your limited partners

59:54

to try and deliver returns and

59:56

when you come to raise your next fund, ask

1:00:00

where your distributions are so that there's a

1:00:02

pressure on yourself to be

1:00:04

able to achieve that. That's

1:00:06

definitely easier said in a sense than

1:00:09

to do for a few different reasons.

1:00:12

Well, number one, that tension with

1:00:14

the long term interests of the project. So

1:00:16

I think what we have tried to do

1:00:18

is that you need to

1:00:20

have that relationship with the project to understand how they are

1:00:22

hedging. And in some cases, it may well

1:00:25

be the case that they need to hedge. They

1:00:27

should be said to be hedging their own treasury

1:00:30

in order to provide

1:00:32

for a solid basis for

1:00:35

the long term sustainability of the project. And

1:00:37

certainly it's fair if there

1:00:40

is that hedging going on for you

1:00:42

to perhaps be doing similarly hedging your

1:00:44

position at that point in time. But

1:00:46

that obviously involves a dialogue and historically

1:00:48

maybe that dialogue has not been there.

1:00:51

I think it's in the future, but it is

1:00:53

a difficult dynamic to manage because projects may not

1:00:56

want to talk about the fact that they are

1:00:58

hedging because that itself

1:01:00

can have an impact on pricing. I

1:01:03

mean, there's another dynamic for venture

1:01:06

investors, which is that it

1:01:11

can be said and it has been said

1:01:13

that it's actually easier to buy than it

1:01:15

is to sell. So

1:01:19

backing a great project and getting in is one

1:01:22

thing, but knowing when to decide

1:01:24

that it's time to get

1:01:27

off the train, it can be very hard

1:01:29

as well because you think there's more to

1:01:31

go. And then afterwards when the market's

1:01:33

coming down, you can feel like you're trying to catch a

1:01:35

pulling knife and whatever you hesitate. So

1:01:39

I think that muscle needs to be exercised. There

1:01:41

needs to be discipline around that. So it's a

1:01:43

different reason in which it can be difficult to

1:01:45

get it done. So

1:01:48

you need to be sympathetic to the founder of the

1:01:50

project, be integral with it, and then you need to

1:01:53

actually have the discipline internally, the

1:01:56

processes, the people, the numbers, the dashboard,

1:01:58

to think about it. the

1:02:01

structure. And I think it

1:02:03

probably comes back to the fact that expectations

1:02:06

are always key. If

1:02:08

you can be as in it is a backer, you could say,

1:02:10

look, I'm with you through the content. But

1:02:12

if there is a point at which the

1:02:15

market, the liquidity has occurred and

1:02:17

you've done extremely well with

1:02:20

the kind of token price, then it might

1:02:22

make sense for us to hedge our position

1:02:24

and this is how we're going to do

1:02:27

it. And we'll make that collaborative and so

1:02:29

forth. So I think communication and expectations would

1:02:31

be crucial for making sure you remain aligned

1:02:34

with the founding team. Yeah,

1:02:37

that is helpful. I think essentially

1:02:40

requires a measured approach and also

1:02:42

proximity and communication with the founder.

1:02:45

That's helpful. And

1:02:47

one other thing here, you guys have to

1:02:50

look at so many deals over the years,

1:02:52

like possibly thousands of deals invested

1:02:54

in more than 100. What

1:03:01

are the criteria that you look for? So when a deal

1:03:03

comes across your desk, what are

1:03:05

the main three things that you look for

1:03:08

that will

1:03:10

jump out at you as like, this is a potential

1:03:13

100X? Is it the

1:03:15

team or is it way more towards the

1:03:17

tech or like

1:03:19

really true product innovation? What

1:03:21

are the main things

1:03:23

that will jump out and cause

1:03:26

you to say like, okay, this is a priority and we

1:03:28

should pursue this as possibly like

1:03:31

going forward with this? Yeah,

1:03:33

I mean, so clearly

1:03:35

this is a topic that comes up

1:03:38

in many VC

1:03:40

or should

1:03:42

podcasts or conversations and for

1:03:46

me at least, if

1:03:49

you're talking about early stage investment, at

1:03:51

the end of the day, it's just the

1:03:53

team, the team, the team. I mean, it

1:03:56

isn't because you need to see some evidence of

1:04:01

how they're thinking about their initial product

1:04:03

offering. Is this a clean, sharp insertion

1:04:05

point that's going to get immediate traction

1:04:07

and give them learning and

1:04:10

is that elegant in the

1:04:12

way they're constructed it? And you need

1:04:14

to see evidence that they've thought about maybe

1:04:17

not the initial market, but how they

1:04:19

can tack to creating

1:04:22

even a phenomenally sized sort

1:04:24

of market over time. But

1:04:28

in a sense, those are just evidence points that

1:04:30

the team you're talking to are just

1:04:33

incredible. It's

1:04:36

not necessarily the answer, the

1:04:38

specific answer, it's what's gone into generating that

1:04:40

answer and the fact that that indicates that

1:04:42

they'll be able to come up with the

1:04:44

next right answer when new data represents itself

1:04:46

and they work out what to do. And

1:04:50

so for me, that's the

1:04:52

kind of crucial thing. I think

1:04:55

one of the things to get

1:04:58

your head around is Ali Sajam's investor

1:05:01

and my friend Fred Desta, I think

1:05:03

he called his blog post, he runs

1:05:05

Stride, talking about how I learned to

1:05:08

get comfortable with risk and embrace it

1:05:10

or something like that. And

1:05:12

it is about getting comfortable with how

1:05:14

much risk there is. And in fact,

1:05:16

particularly, of course, if you're building a

1:05:18

venture portfolio, we are actually looking for

1:05:20

investments for projects that

1:05:23

are taking more risk than is reasonable

1:05:25

for the founding team to be taking

1:05:28

on a standalone basis, because

1:05:30

we want every single time

1:05:32

that project to be really

1:05:34

swinging to be totally extraordinary

1:05:36

in the outcome that they

1:05:39

produce. But it's

1:05:41

not just risk, it's also a kind of

1:05:43

form of messiness that I

1:05:45

think can exist. Thinking

1:05:47

about it, you both have

1:05:49

these teams that are incredibly capable and

1:05:51

have thought things through

1:05:53

and can evidence that. But

1:05:56

at the same time, getting comfortable. risk,

1:06:01

uncertainty, and even messiness at the

1:06:03

early stage of those

1:06:05

outfits. So if you, I'm

1:06:08

answering the questions, what you're maybe not looking for,

1:06:10

if you're looking for everything to be

1:06:12

buttoned up, and if you're looking

1:06:15

to ensure that you can then feel comfortable

1:06:17

that things at the end of the day

1:06:19

aren't going to go too wrong, that's

1:06:21

the wrong thing to be looking for. If

1:06:23

you're looking for a whole series, if the

1:06:25

sun, the moon, the stars, and the whatever,

1:06:31

some other celestial body kind of line up, then

1:06:34

how if that all happened, how incredible

1:06:36

it could then become, that's what you're

1:06:38

really attuned to. But that

1:06:40

actually makes me think of another point on

1:06:43

actually quickly on this element of

1:06:45

liquidity, which is different and distinct

1:06:47

for the

1:06:49

few capital allocators who

1:06:51

are out there rapping, putting lots of money into venture

1:06:53

at the moment here in the world, let alone Europe,

1:06:56

because we still do have a bit of a dearth

1:06:58

of that, looking forward to

1:07:00

some interest rate costs. But

1:07:03

the other thing is that we've observed is

1:07:06

that because of this kind of organic,

1:07:09

more city-like, positive,

1:07:11

some setup of these networks that

1:07:14

get built where to

1:07:16

take the canonical example, we don't know

1:07:18

who Satoshi is, we don't know who

1:07:20

the founder is, you don't have some

1:07:22

key risks that you might have in

1:07:25

normal startup projects. And so that actually

1:07:27

the fallout rate, the kind of fallout,

1:07:29

the kind of go to zero rate

1:07:31

is lower than you would see statistically

1:07:34

from normal kind of startups. Because

1:07:37

of that kind of open collaborative community-based

1:07:39

nature, I would posit of

1:07:42

these projects. And

1:07:44

so, yeah, so you've got

1:07:46

the earlier liquidity, you've got the

1:07:48

lower fallout rate, but you still

1:07:50

nonetheless, if you really want to

1:07:52

deliver exceptional fund level returns, you

1:07:54

need to be swinging

1:07:57

for the fences as

1:07:59

the proverbial. phrase goes. Well,

1:08:03

thank you so much for that. That's great

1:08:05

insights and this has been a really fantastic

1:08:08

conversation. So Richard, thanks so much for finally

1:08:10

coming on the podcast. Hopefully we can get

1:08:12

you on again in less than 10 years

1:08:15

and yeah, we'll catch you soon.

1:08:17

Well, thank you. I plan to still

1:08:19

be building fabric in 10 years, but

1:08:21

less only in 10 years. Thank

1:08:25

you. Thanks, Richard. Thanks, guys. Really, really

1:08:27

enjoyed it. Thank you very much. If

1:08:57

you want to interact with us, guests or other

1:09:00

podcast listeners, you can follow us on Twitter and

1:09:02

please leave us a review on iTunes. It helps

1:09:04

people find the show and we're always happy to

1:09:06

read them. Well, thanks so much. And we

1:09:08

look forward to being back next week.