0:00

Ladies and gentlemen, thanks so much for

0:02

tuning into the scoop on your first

0:04

Frank Shaparov. And joining us on the

0:06

other side of the mic is

0:08

not a stranger to the program,

0:10

but it is his first time

0:12

appearing in his new seat as

0:14

the co -founder of Double Zero, Austin

0:17

Federa. Thanks so much

0:19

for taking the time during your

0:21

travels and the midst of your travels

0:23

to take a moment to

0:25

Walk the audience through what you've been

0:27

working on so diligently these past few quarters. Hi

0:31

Frank, thanks for having me. Appreciate it. Of

0:33

course. For crypto and DeFi professionals looking

0:36

to contribute to a team

0:38

committed to innovation and digital assets,

0:40

Fidelity Investments is now hiring.

0:42

Why Fidelity? For starters, they've been

0:44

exploring Bitcoin and blockchain since

0:46

2014. They were the first traditional

0:48

asset manager to launch a

0:50

native crypto business and among the

0:53

first to launch a Bitcoin

0:55

exchange traded product, innovation is in

0:57

their DNA. Fidelity invests in

0:59

their talents by supporting and developing

1:01

associates through continued education opportunities,

1:03

casual crypto -focused discussion groups and on

1:05

-the -job experience of working with

1:07

new technologies. If you want to

1:09

work on cutting -edge products and

1:11

explore the new frontier of

1:13

DeFi and blockchain, Fidelity is for

1:16

you. Their products continue to

1:18

improve and evolve, staying ahead of

1:20

this rapidly changing digital asset

1:22

landscape. So go ahead, explore Fidelity

1:24

Crypto Careers today by going

1:26

to crypto .fidelitycareers .com to learn more.

1:28

That's crypto .fidelitycareers .com. Fidelity is

1:30

an equal opportunity employer. So

1:34

I guess for the unbaptized,

1:36

right? And I would even

1:38

maybe include myself in that

1:40

category, given that we haven't

1:43

talked too much about the project. You know,

1:45

we're always too busy joking about

1:47

the height of my television

1:49

and our own respective travels around

1:51

the world and airplane configurations,

1:53

just the sort of nerdy stuff

1:55

that we often go back and

1:57

forth on in the DMs. But

2:00

walk us through, we'll just dive right into it. What's

2:02

the core thesis behind the double

2:04

zero project? And

2:07

just, you know, lean

2:09

into sort of this gap,

2:11

right? This gap in

2:13

internet infrastructure that you're hoping to

2:15

solve. Obviously, I think, and forgive

2:17

me, audience, if you

2:19

don't know Austin's background, I mean, he

2:21

kind of has been through many, he's

2:24

been in the trenches as it were,

2:26

building out Solana. and launch

2:28

double zero last year. But

2:31

anyway, with that all

2:33

being said, let's get into

2:35

that that core thesis. Yeah.

2:37

So, you know, to start out with Frank's TV

2:39

is like at least five and a half feet

2:41

up the wall. It's it's way too high, though.

2:44

Someone needs to make an intervention in New

2:46

Jersey. I sent him a picture of

2:48

me watching it in the POV is like

2:50

this. I have all these

2:53

neck problems. Yeah,

2:56

00 is basically an attempt

2:58

to rebuild the internet

3:00

infrastructure for high performance distributed

3:03

systems. And we do

3:05

this by using physical fiber cables.

3:07

It's kind of the unsexiest project

3:09

in crypto, because what we're talking

3:11

about is physical fiber infrastructure in the

3:13

ground. People in like crypto like to

3:15

talk about like, oh, we're building like

3:17

infrastructure, and there's a lot of awesome software infrastructure

3:19

that's been built. But this

3:21

is like a physically interconnected

3:23

network of fiber where we

3:26

have multiple contributors all around the world. And

3:29

the goal of the project really

3:31

is to take the best of private

3:33

networking technology, which is several decades more

3:35

advanced than the technology that you

3:37

get on the public internet today. And

3:39

it's been used for decades by

3:41

high frequency trading firms and, you know,

3:43

the cloud providers and all of

3:45

kind of the big tech companies you

3:47

think of today to enable their

3:49

systems to run high performance and run distributed

3:51

around the world. We don't think of

3:53

Google as a good distributed systems company because

3:56

at the end of the day, there

3:58

is one company that holds all of the

4:00

keys, for lack of a better term, to

4:02

Google. But they have many points of presence

4:04

all over the world to make sure that

4:06

when you send a Google request in New

4:08

York or Buenos Aires or anywhere in the

4:10

world, you get a response back really quickly. And

4:13

in order to build those type of systems, you

4:15

really need private, dedicated fiber. The

4:18

problem with private dedicated fiber historically is

4:20

it always had to be run

4:22

by one centralized company. If you were

4:24

jump trading, you would run your private

4:26

network. And if you were Citadel, you'd run

4:29

your other private network. And if you're

4:31

Google you on a private network, and none

4:33

of these systems can talk to one

4:35

another. And additionally, they're all

4:37

intentionally built. So only one company

4:39

can access them. So

4:41

at double zero, kind of like

4:43

the breakthrough was we figured out

4:45

a way that you can have

4:47

multiple independent networks run by multiple

4:49

independent contributors appear and interoperate

4:51

as one seamless network. The

4:53

exact same way that the public internet

4:55

does today, I don't have to worry about

4:57

like, are you on Verizon and I'm

4:59

on AT &T, the networks just talk to

5:01

one another. But it's based on

5:03

a fundamentally new generation of technology

5:06

that enables the types of throughput

5:08

and performance that is just impossible

5:10

on the public internet today. So

5:13

is it safe to say for

5:15

folks maybe building in crypto that

5:17

the problem you're solving for is

5:19

we can sort of

5:21

at the software layer

5:23

Make blockchains Run

5:26

super fast super efficiently,

5:28

but there's still this

5:30

reliance on the undergirding

5:32

infrastructure that

5:35

Is maybe not in all

5:37

aspects super high -performative

5:39

Yeah, it's your classic

5:41

like you've got a Ferrari, but

5:43

it's stuck in traffic. The software

5:45

clients that exist today across a

5:47

number of networks are quite fast.

5:51

And really the vision for 00 came out

5:53

of the work that I was doing on

5:55

the Fire Dancer project back at Solana Foundation,

5:57

which was a net new validator client for

5:59

the Solana blockchain. And

6:01

that client can do a

6:03

million, two million transactions per

6:05

second in testing environments. And

6:07

when we say testing environments, There's

6:10

always a question about, like, well, is

6:12

that just simple transfers? And basically what

6:14

you see is when you run, like,

6:16

real -world benchmarks on Fire Dancer, it

6:18

is orders of magnitude faster than

6:20

the Solana network is today. And

6:23

the reason for that is the public internet. The

6:26

internet was never built for high -performance

6:28

distributed systems. It was built for one -to

6:30

-one communication anywhere in the world. And

6:32

it's really good at that, right? I'm

6:34

sitting here in Seoul, and you're in

6:36

the United States, and we're recording on

6:38

Riverside. Everything's working pretty great. There's a

6:40

few hiccups every once in a while,

6:42

but this is fundamentally like a fairly

6:45

low performance, low bandwidth application. If

6:47

you try and say, okay,

6:49

now we want 10 ,000

6:51

people all over the world to be

6:53

able to watch this broadcast. Well, that's

6:56

actually a quite hard problem. And kind of

6:58

the best way to explain this is why

7:01

did it take like 15 years for us to figure

7:03

out how to stream the Super Bowl? Like

7:05

you shouldn't think this would be such

7:07

a hard problem that every time there's a

7:10

really big major sporting event Most people go

7:12

to cable or they go to direct TV

7:14

or they go to some form of

7:16

like legacy broadcast because the internet streaming services

7:18

have Historically been pretty bad of these

7:20

types of things because the internet was never

7:22

built for one -to -many communication It

7:24

was built for one -to -one communication is

7:26

if you think about it like if I'm

7:28

a Validator building a block

7:30

I'm building that block. Let's say here in

7:32

Korea And there's 3 ,000

7:34

nodes out there between RPCs

7:37

and consensus validators. I have

7:39

to one by one send that data out.

7:41

The 3 ,000 different computers all around

7:43

the world. And there's sort

7:45

of fancy ways that networks get around

7:47

this. Solana uses something called Turbine, where

7:49

I send to 200, and they send to

7:51

200 more, and they send to 200

7:54

more. Monet uses something called RaptorCast for this,

7:56

which effectively works the same way. But

7:58

in all of these situations, there's a

8:00

lot of work that has to be done

8:03

by the network to basically propagate data

8:05

out. And on top of that,

8:07

it's a slow, arduous, manual process.

8:10

And so if we can use the

8:12

technologies that are available in private networking

8:14

like multicast, where you just send a

8:16

packet once, and the hardware,

8:19

the switches automatically get that data to

8:21

all the places it needs to go to, we

8:23

can push the bounds of what's possible

8:25

in these high -performance systems. So at breakpoint

8:27

last year, if

8:29

you saw there's a demo of a

8:31

Fire Dancer client running a million

8:34

transactions per second on a bunch of nodes

8:36

on four different continents. And we

8:38

didn't say so at the time because we

8:40

hadn't launched the project yet, but that was

8:42

running on the 00 network. And that

8:44

was a real world example

8:46

of how you can have

8:48

high performance communication across multiple

8:50

continents. And this gets into

8:52

kind of like a dynamic we've seen in blockchain, which

8:54

is the faster a network gets,

8:57

the fewer validators it has. If

9:00

you look at kind of a ratio of the

9:02

transactions per second to the number of validators, Solana

9:04

is the only fast network that

9:07

has over 1 ,000 validators at the moment that

9:09

operates in a single global state. You

9:11

look at Sui and Aptos and they're

9:13

very high -performance chains, but they're running,

9:15

you know, 200 validators, something along those

9:17

lines. You get down to hyperliquid, you're

9:19

talking about 11, pretty much all

9:21

run in the same country. and

9:23

it's also a very permissioned validator system. And

9:26

so one way to think of what we're

9:28

doing at 00 is we're removing the connectivity

9:30

layer as a bottleneck. Suddenly you could have

9:33

high performance distributed systems that have 2000,

9:35

3000 nodes and are also able to

9:37

run in a million transactions per second

9:39

because they're not using the internet to

9:41

communicate anymore. They're using 00. So

9:45

to explain or walk through how you're

9:47

doing that, I think there's, we have

9:49

to sort of engage in a bit

9:51

of a Aristotelian definition

9:53

of terms of which there are

9:55

many to unpack. There's

9:57

the concept here of

9:59

underutilized fiber of dark

10:02

fiber, excess subsea

10:04

cables, and

10:06

a lot of different terms with which

10:08

I'm sure a lot of listeners are

10:10

not familiar. So

10:13

maybe juxtapose

10:15

the landscape in which

10:17

you're operating in with

10:21

the project itself. I think

10:23

most people grok that right, like, you know, the

10:25

world's connected by all these things that, you know,

10:28

go underground and you can use satellites to

10:30

sort of hop over them and all sorts

10:32

of stuff. But how does

10:34

it all work? Yeah. So

10:36

99 .99 % of the world's

10:38

internet traffic moves through fiber. It

10:40

doesn't move through satellites or microwave

10:42

towers or anything like that. Fiber

10:45

is still the gold standard in terms

10:47

of capacity, reliability, and cost. There's

10:49

nothing today that competes with it even

10:51

remotely. And so, you

10:54

know, to most people's houses, they don't have fiber,

10:56

but it's kind of that last mile connection

10:58

where you maybe switch to some other type of

11:00

technology. For the most part, everything else runs

11:02

on fiber. And so, you can

11:04

sort of think about this as like, there

11:06

are public hearing networks, which are the

11:08

internet. And the internet today is largely run

11:11

by 20 companies that And this goes

11:13

back to like historically how the internet was

11:15

built out, but you have this idea

11:17

of a tier one internet service provider all

11:19

the way down to a tier three.

11:21

And a tier three was maybe like your

11:23

local company that like provided you a

11:25

dial up connection and maybe only serviced like

11:27

half of your state or one city or

11:29

a small region in New England or something

11:31

along those lines. And then you

11:33

get up to your tier one providers

11:35

and that was something like a

11:38

Verizon or an NTT in Japan. These

11:40

were these. big, big companies that

11:42

would actually go out and build the

11:44

connectivity and the, you know,

11:46

initially copper and eventually

11:48

fiber that connected like Japan

11:50

to like China and South Korea

11:52

and the cables that run under the ocean

11:54

and connected to the United States. And

11:57

so we get into this kind of like history of

11:59

this system that's kind of on like Russian

12:01

nesting dolls of internet connectivity. And

12:03

every time you transfer

12:05

a border from one company

12:07

to another, you're

12:09

basically dealing with a different system

12:11

of prioritization. And the

12:14

way that the business models

12:16

work nowadays is every internet service

12:18

provider is basically incentivized to

12:20

get your data off of their

12:22

network as quickly as possible. You

12:24

would think that like right now, my

12:27

data going from Seoul to your data going

12:29

to the New York area, it would

12:31

sort of be paying a bunch of little

12:33

fees along the way. But that's not

12:35

actually how it works. We didn't have the

12:37

technology to build anything like that back

12:39

in like the 80s or 90s. And so

12:41

all of the Internet SERPs providers kind

12:44

of have this crazy reconciliation model where they

12:46

all sort of are like, oh, I

12:48

sent you this much data, you sent me

12:50

this much data, and we bill each

12:52

other for the difference. And that's kind of

12:54

like this net metering system of billing. And

12:57

it doesn't incentivize good traffic routing,

12:59

and it doesn't incentivize good performance. it

13:02

incentivizes you trying to maintain like a

13:04

very sort of like 2025 view of

13:06

trade balance like view of the world

13:08

is like I want to be sending

13:10

you about as much bandwidth as you

13:12

send me. And so you'll get these

13:14

like companies that are in fights with

13:16

one another and just like you might

13:18

lose access to like a sports team

13:20

in a market because the sport seems

13:22

fighting with like your cable provider or

13:24

something like that. You have all these

13:26

like changes that happen constantly on the

13:28

internet. So you a very inconsistent speed

13:30

of connectivity. So if I'm in

13:32

Seoul and we just do a ping

13:34

test between my computer and your computer, and

13:36

we run it for 24 hours, you'll

13:38

see a huge amount of variance in how

13:41

much time it actually takes my data to

13:43

get to you and your data to get

13:45

to me because the routing changes constantly. For

13:48

text messages and YouTube and

13:50

even, you know, podcast recordings,

13:52

that's pretty much fine, but

13:54

that's absolutely not acceptable for

13:56

high -performance systems, let alone

13:58

high -performance financial systems. where

14:00

data arriving out of order basically

14:02

slows the entire system down. Or

14:04

if I'm trying to make a

14:06

trade, and I know where

14:08

the market's going to be in 75

14:10

milliseconds, but I don't know where it's

14:12

going to be in 100 milliseconds. And

14:14

there's a 50 % chance that if

14:16

I send my data, it's going to

14:19

take 120 milliseconds to get to you.

14:21

I'm not going to make that trade.

14:23

So it makes the indeterminate on the

14:25

physical networking layer. makes markets less efficient.

14:27

And this is part of why the

14:29

traditional financial space is all dedicated private

14:31

network links. It is faster,

14:33

right? And so they get this

14:35

sort of high frequency trading advantage,

14:37

but it also has significantly higher

14:40

predictability. Like imagine if you were

14:42

like trying to fly like from, you know, one

14:44

side of the world to the other, and you had

14:46

three flights you had to take to get there. But

14:48

each flight was like, we'll arrive sometime in a

14:50

two hour period. We're not gonna

14:53

tell you exactly when we'll take off

14:55

whenever we feel like it will arrive

14:57

whenever we feel like it Well, suddenly

14:59

if you only have three hops, you

15:01

know of six hours of variance you

15:03

have to build in so each of

15:05

your transfer times has to be Five

15:07

hours to account for that potential variance

15:09

And it would just make air travel

15:11

unrealistic and now in the internet you're

15:14

dealing with sometimes 30 or 40 hops

15:16

to get from You know Asia to

15:18

the United States and each one of

15:20

those hops there's delay involved and there's

15:22

a lack of predictability. And

15:24

so we have these great graphs where you

15:26

sort of see the amount of time it

15:28

takes to talk from LA to Singapore over

15:30

the public internet versus over double zero. And

15:33

for double zero, it's just one

15:35

consistent line of time. It always takes

15:37

the same amount of time. The

15:39

public internet, it'll jump up to 250

15:41

milliseconds. It'll jump back down. It'll

15:43

go as high as sometimes over a second.

15:46

And you can see the graph, like, oh, they're

15:48

routing over this cable. And then they started routing

15:50

over this cable. And then they routed over this

15:52

cable. And then they routed over this cable. And

15:54

so the whole system, like, I know we're getting

15:56

very deep into, like, network topology.

15:58

But, like, these systems, they're very

16:01

hard to predict how it

16:03

will actually work. And they're very

16:05

inconsistent performance. No, that

16:07

makes sense. And so

16:09

when you think about the...

16:12

crypto trading experience, if

16:14

we were to break it

16:16

down for someone who

16:18

is swapping or moving coins

16:21

or staking or just

16:23

participating in the crypto economy,

16:26

having a more performative,

16:28

predictable sort of base

16:30

infrastructure for the internet,

16:32

what does that mean

16:34

for sort of the

16:36

day -to -day activities in

16:38

which they're participating? Yeah,

16:41

so most like... users

16:43

like you and me using blockchain

16:45

the type of like time savings that

16:47

double zero provides Doesn't matter to

16:49

them when they're executing a trade, right?

16:51

You're you're going through phantom or

16:53

metamask or something and at that point

16:56

like you have a lot of

16:58

delay in that system Anyway, just from

17:00

how those pieces of software are

17:02

built for the benefit that you would

17:04

see as a user from let's

17:06

say like Solana or sui or apdos

17:08

or you know an L2 adopting

17:10

double zero is not necessarily

17:12

in your direct experience. It's in

17:15

the aggregate performance of the blockchain. So

17:18

what you might see is like, oh,

17:20

Solana goes from about 5 ,000 transactions

17:22

per second to 50 ,000 transactions per second.

17:24

Well, what do you care if you're

17:26

only sending 100 transactions a day as

17:28

a retail user? Well, suddenly you see

17:30

much tighter margins on DeFi. You

17:32

see much lower slippage. You see the

17:34

market becoming much more efficient. You see

17:36

the average fee you have to pay

17:38

for a transaction being much lower because

17:40

there's more capacity. And you

17:42

basically just see all of these

17:44

markets working more efficiently. Traditional

17:46

financial markets are extremely efficient. They

17:49

have very little spread compared to

17:51

crypto. And that's not just because

17:53

there's more volume going through them.

17:55

That's because there's much less risk

17:57

for market makers, right? Crypto is

17:59

this weird thing where like Market

18:01

makers have to spend a lot

18:03

of time actually worrying about getting

18:05

picked off. And it's not

18:07

because they're bad at predicting where the state

18:09

is or something along those lines. It's

18:11

because if you're dealing with a traditional financial

18:13

exchange, let's just say like I'm a

18:15

trading firm based in Chicago and I'm trying

18:18

to trade on the New York Stock

18:20

Exchange, right? I always know my

18:22

server is trying to run data

18:24

from Chicago to New York. That's a

18:26

point to point system. I always

18:28

know how long it's going to take

18:30

because I have dedicated bandwidth. And

18:32

I always know that if I get

18:34

it there in X amount of

18:36

time, it'll be processed in Y amount

18:38

of time. But that's not

18:40

the case in blockchain, right? I could be a

18:42

market maker and I say, oh, I got to

18:44

get a cancel order in because the price has

18:46

moved. Where am I sending that? And

18:48

like, like physically, where in the world

18:50

am I sending that? Well, maybe the

18:53

leader is being built right now in

18:55

New York, right? Maybe it's being built

18:57

in Amsterdam. Maybe it's being built in

18:59

Sydney or Perth. And suddenly, if I

19:01

have to send a transaction from an

19:03

RPC server to get into the blockchain

19:05

to find wherever the leader in the

19:07

world is, it might be going 25

19:09

milliseconds because the leader is being built

19:12

in Chicago. Or could be going all

19:14

over the world. Exactly. side of

19:16

the world, rather. Right. And so

19:18

all of that lack of determinism

19:20

makes it very hard for these, for

19:22

market makers in DeFi to sort

19:24

of provide the most efficient price discovery

19:26

it can with the lowest tolerances. Hmm.

19:29

So does it really just build

19:31

down to the fact that our crypto

19:33

markets don't have sort of the

19:35

same? Determinism of

19:37

the location of where these

19:39

trades are being matched. We

19:42

don't know, for instance, if

19:44

I'm, you know, trying to move

19:46

money on chain, that that's

19:48

getting that's getting routed or it's

19:50

getting sent out to. So

19:52

caucus or what's the other place where

19:55

they have them at? I think it's

19:57

so caucus and. uh,

19:59

Carl Stat, New Jersey. Yeah.

20:03

Yeah. So there's kind of two components

20:05

that go into this. One, our

20:07

blockchains are not deterministic in terms of

20:09

transaction inclusion. Can't do anything

20:11

about that, right? Like, if I'm a block builder,

20:13

I'm a dictator, I get to pick exactly

20:15

what goes in the block. There's some networks that

20:17

are trying to do some stuff to address

20:20

that, but, um, we can get into why I

20:22

don't think that's gonna necessarily work, but the

20:24

connectivity layer is also a really, a really big

20:26

one. And so if you think about it

20:28

like what has to happen. When

20:30

a block is created before a

20:32

price on chain is updated and updated

20:34

is like kind of a hard

20:36

term, but let's say my validator builds

20:38

a block and you are trying

20:41

to trade based on new data that's

20:43

come in. So that block

20:45

first has to get out to all

20:47

of the other validators in the ecosystem

20:49

so they can vote on it and

20:51

accept it. Let's just assume for a

20:53

minute like that almost every block is.

20:55

confirmed because that is what happens in

20:57

blockchain today. So it's a security check,

20:59

but like you can kind of optimistically keep going. That

21:02

data has to get from my

21:04

leader to the RPC server, the RPC

21:06

or server has to process it,

21:09

then be able to give you data

21:11

and then you can trade against

21:13

that data. And so even if you

21:15

are running a single sequencer like

21:17

base or any type of other L2

21:19

and the data is all produced

21:21

in one location, it needs to get

21:23

out to the entire RPC network

21:25

before anyone knows about it, right? It's

21:28

the classic thing of, like, just

21:30

because an event happened in New York

21:32

doesn't mean that Singapore knows about it yet.

21:34

We have to actually move that data

21:36

from New York to Singapore for people in

21:38

Singapore to learn about it to be

21:40

able to trade, right, or do any other

21:42

action that they're going to take. And

21:45

so even in a centralized

21:47

sequencer, 00 offers a much faster

21:49

path of propagation. then

21:51

you get on a traditional internet

21:53

infrastructure system. And you see

21:55

this all the time in exchanges,

21:57

right? The New York Stock Exchange offers

21:59

a high -speed priority fee that you pay

22:01

millions of dollars a year to get

22:03

access to, let alone the fact they

22:06

let you run servers kind of

22:08

bolted onto their systems. But

22:10

speed is the name of the game in

22:12

trading, right? In the traditional financial markets, if

22:14

I give you two options, you can be

22:16

the smartest firm or the fastest firm. Everyone

22:19

out there will pick fastest firm

22:21

because you can be smart and lose

22:23

money. It's very hard to be

22:26

fast and to be the fastest and

22:28

lose money. You're not going to

22:30

have like crazy bets and crazy asymmetric

22:32

upside, but you're going to consistently

22:34

be making more profit than anyone else

22:36

in the system because you're faster.

22:39

And even if you're only taking 10

22:41

basis points, every, you know, trade

22:43

over someone else, the laws of compounding

22:45

mean. it's an incredible advantage over

22:47

everyone else in the system. And

22:50

so this is kind of a

22:52

long way of saying we're kind of

22:54

trying to bring a lot of

22:56

those mechanics to crypto and blockchain. And

22:58

we think it will be a

23:00

real advantage to users because we'll have

23:02

more capacity on these networks. We'll

23:04

have multiple networks running in a million

23:06

transactions per second, both across layer

23:08

ones and layer twos, and we'll also

23:10

have much more efficient markets. and

23:12

it will be a much better experience

23:14

for traders without having to centralize

23:16

the markets, right? Almost every other approach

23:18

today to make crypto more like

23:20

Treadfy, whatever that means, involves centralization, massive

23:22

centralization, far fewer validators, preconformations, all

23:25

of this type of technology that really

23:27

kind of compromises, I would say,

23:29

the point of crypto. And so with

23:31

double zero, the whole thing is

23:33

like, we need thousands of validators all

23:35

over the world. We need you

23:37

to be able to trade against a

23:39

local server, but we also need

23:41

these markets to be much more efficient.

23:43

They can't move as slowly as

23:45

Bitcoin and expect, you know, this technology

23:47

to be adopted around the world.

23:49

You made me think of the movie

23:51

March and Call, right, where the

23:53

CEO says, you know, I didn't get

23:55

here. You can either win by

23:57

being fast, smart, or cheat. And

24:00

I don't cheat. What

24:03

do you think is

24:05

the realistic timeframe for

24:07

rebuilding decentralized Internet using

24:09

or repurposing infrastructure? Is

24:11

this a three -year

24:13

journey or a multi -decade

24:15

long journey? I

24:17

mean the thing about all software and

24:19

hardware projects is they're never done,

24:21

right? And so there will never be

24:23

quote -unquote enough bandwidth on double zero.

24:25

There will always be additional marginal

24:27

cost uses for it. But that being

24:29

said, we actually have a testnet

24:31

up and running at the moment, and

24:33

we'll be in mainnet before the

24:35

end of the year. And

24:38

this is largely in

24:40

part due to a network

24:42

of contributors that are

24:44

incredibly excited about this technology.

24:46

So we have six

24:48

independent network contributors at the

24:50

moment. That includes everything

24:52

from firms like Jump Crypto and

24:54

Galaxy. down to VCs that

24:56

have expertise in building out fiber

24:58

networks from, you know, past experience or

25:00

something like that, and they're willing to

25:02

take a bet that this is going

25:05

to be a technology that'll be well

25:07

-adopted and blockchain. And so

25:09

we're starting out with supporting the Solana

25:11

network because it has the largest validator

25:13

set to performance ratio, and so it's

25:15

the hardest network for us to basically

25:17

support. But we'll be

25:19

expanding to other blockchains from

25:21

there and even non -blockchain systems

25:23

because... data throughput problems apply in

25:25

the AI space. They apply

25:28

to storage networks like Filecoin and

25:30

Rweave and new generations of

25:32

storage networks coming up as well.

25:35

And so there's actually a ton of demand

25:37

for this type of thing outside of

25:39

crypto because if you're the size of Microsoft,

25:41

you can just build your own data

25:43

centers and run fiber between them. But if

25:45

you're not the size of Microsoft and

25:47

you want to be training AI in multiple

25:49

different data centers all around the world,

25:51

The public internet's not fast enough to move

25:53

that data around. It's actually a limiting

25:55

factor. And yes, you need GPUs, but once

25:57

you get the GPUs, they need to

25:59

be able to talk to each other and

26:01

share a lot of data back and

26:03

forth. And this is

26:05

why like, uh, do you know Amazon Snowball?

26:08

Have you heard of this thing? Yeah.

26:10

Yeah. So this is like Amazon will

26:12

basically send you a semi truck full

26:14

of hard drives that you basically back

26:16

up to your data center and plug

26:19

a bunch of cables into it and

26:21

you transfer the data locally. from your

26:23

servers onto the semi truck, then the

26:25

semi truck drives to the Amazon data

26:27

center plugs in and it sucks all

26:29

the data back out. And that is

26:31

actually the fastest way to get data

26:33

into the public cloud because the internet

26:35

is not fast enough today, which is

26:37

like crazy that like, we're basically moving

26:40

data the way we would move like,

26:42

you know, oil or like coal or

26:44

something like that because the public internet

26:46

infrastructure isn't fast enough to support these

26:48

types of things. So

26:50

if you were to think about

26:52

the Internet or rather, how do

26:54

you see the Internet evolving if

26:57

projects like this succeed? Are we

26:59

heading towards a more localized, sovereign,

27:02

proud -owned network framework?

27:04

Yeah. I mean,

27:06

so we're not looking to replace the Internet.

27:09

I think this is kind of like an

27:11

important thing about the project for like all

27:13

general uses. The Internet is incredible. We're huge

27:15

fans of it. Big

27:17

Internet guy. Yeah, big internet

27:19

guy. Uh, huge on the internet.

27:21

Um, but it

27:23

is not designed for the things we're

27:25

using it for today and distributed system

27:27

technology. And the big companies work around

27:29

it, right? They run their own fiber

27:32

cables. Meta is one of the largest

27:34

investors of fiber because they want to

27:36

sell people ads. Like it's, it's crazy

27:38

to think about it, but Facebook came

27:40

in and wired up the Philippines with

27:42

fiber internet access because they want to

27:44

sell people ads. All of

27:46

these tech businesses that we think of

27:48

as software businesses, they get to a

27:50

point where they say, man, we

27:52

really need hardware if we're going to

27:54

actually scale. Our TAM is limited by

27:56

the capacity of the public internet. And

27:59

I think blockchain is at that point

28:01

as well, where our scalability, our performance, our

28:03

TAM, for lack of a better term,

28:05

is very much limited by how fast these

28:07

servers and these validators can talk to

28:09

one another. But you fast forward,

28:11

you know, maybe five or 10 years and

28:13

like if double zero is successful, like

28:16

what is this new internet infrastructure look like?

28:18

Well, first off, we're not routing based

28:20

on IP addresses anymore. We're routing based on

28:22

public key and private key encryption. And

28:24

so when I'm talking to the block, right,

28:26

I'm not like trusting a DNS server

28:28

to say, oh, this is the block and

28:30

the DNS probably is not hacked. There's

28:33

a handshake, a cryptographic handshake. that

28:35

is verifiable without relying on a centralized

28:37

authority that says, yep, this is

28:39

the thing I'm trying to talk to.

28:41

And maybe for a news site

28:43

that's like, not that valuable, but it's

28:46

really valuable for Coinbase or PayPal

28:48

or, you know, Uniswap or Jupyter, right?

28:50

The fact that we can have a

28:52

high degree of confidence that these systems are

28:54

not compromised because we have cryptographic assurances. The

28:57

other thing that I think would

29:00

surprise people is there's actually no

29:02

prioritization layer on the internet today.

29:04

But if I'm like Fortnite, and I

29:07

want to make sure that like the

29:09

kids can always game really fast, I

29:11

can't just go out and say, look, I'm

29:13

on Epic Games. We make billions of

29:15

dollars a year in Fortnite. I want

29:18

to subsidize everyone's Fortnite Internet traffic to

29:20

always make sure that like it's a

29:22

good experience, even if you're playing in

29:24

an area with poor Internet connectivity. You

29:27

can't do that today. It's not

29:29

possible to just go out and

29:31

pay the Internet, for lack of a

29:33

better term, to prioritize Fortnite data.

29:35

Sometimes they make very specific deals with

29:37

specific internet service providers. But

29:39

00, you know, our

29:41

aim is to create a

29:43

prioritization layer for any type of

29:45

high -performance traffic and give them

29:47

an alternative pathway to go

29:49

through. And so if you're

29:51

a creator and you're trying to build

29:54

a network that competes with YouTube

29:56

and isn't owned by Google, you can

29:58

actually go out and build a

30:00

CDN network on top of 00. Because

30:03

right now, you're kind of trapped,

30:05

right? If you're building an alternative to

30:07

YouTube, you're still probably building it

30:09

on big cloud tech infrastructure. And

30:11

so your margins are caught in

30:13

this sort of like, you're buying your

30:15

picks and shovels from the same

30:17

people who you're selling the gold to.

30:19

We've basically created a company town

30:22

model on the internet today where it's

30:24

very hard to actually compete with

30:26

the incumbents because they offer pretty good

30:28

service. And so

30:30

if we're successful, there

30:32

are far more alternatives to

30:34

build and operate internet companies

30:36

on than the three to

30:38

four major cloud providers of

30:40

today. So let's

30:43

walk through maybe

30:45

the technical playbook. How

30:47

does double zero

30:49

identify and thereby access

30:52

underused fibers or

30:54

cables? Are you knocking

30:56

on the door

30:58

of the municipal? municipality

31:01

dealing directly with carriers

31:04

or data center operators, what

31:06

does that look like? So

31:08

today we are not dealing

31:11

with carriers directly, simply

31:13

because carriers are old school, six to

31:15

7 % cash margin businesses, and a

31:17

lot of them are owned by

31:19

PE firms, and they just don't understand

31:21

tokens yet. And also we're not

31:23

in mainnet, so we can't show them

31:25

like, here's about how much money

31:27

you would make if you join double

31:29

zero today. So we're primarily going

31:31

through the existing customers of those groups.

31:33

So if you're someone like a

31:35

jump trading or a Terra switch, Terra

31:37

switch is a very popular bare

31:39

metal provider in the blockchain space. They

31:41

have leased a bunch of fiber

31:43

already. But when you're leasing fiber, you

31:45

kind of, you don't get many

31:47

choices, right? It's sort of like your

31:49

home internet connection. You can get

31:51

a gigabit, you can get 10 gigabytes,

31:53

you can get 100 gigabytes, but

31:55

maybe you only need 40 gigabytes. Well,

31:57

your option is just you have

31:59

to buy a 100. And so what

32:01

00 lets you do is contribute,

32:03

you know, only 10 gigabits, let's say,

32:05

about 100 gigabit line or 100

32:07

gigabits of a 400 gigabit line to

32:09

00. And that in that

32:11

way, you can sort of think

32:13

about it as like an alternative monetization

32:16

channel for unused capacity in your

32:18

systems. We hope, you know, in the

32:20

future that people are getting fiber

32:22

just to contribute to 00 because, you

32:24

know, the economics are there to

32:26

support it. And I think we'll, I

32:28

think we'll get there without kind of, you know, a

32:30

question at some point. But that's kind

32:32

of the way that that system, system

32:34

works today. And so your question about

32:36

like, where do we need to get

32:38

the fiber? Well, the easy answer is

32:40

you just query the blockchains. Yeah,

32:43

just query the blockchains, right? And you

32:45

say, okay, well, blockchains are public databases. Where

32:47

are all the IP addresses of all the

32:49

stake out there? You can match IP addresses to

32:51

data centers and you can kind of figure

32:54

out like a map of the world. From block

32:56

to there's, as it might surprise no one,

32:58

there's a whole bunch of stake in Europe between

33:00

Amsterdam, Frankfurt and London. There's

33:02

a whole bunch of stake in Tokyo.

33:04

There's a whole bunch of stake in

33:06

Singapore. There's some in Hong Kong. You

33:08

have these kind of clusters around the

33:10

world of where a lot of people

33:12

are already running validators. Where they're not

33:14

running validators, largely speaking, are Latin America, Africa,

33:18

like South East Asia, getting into

33:20

Australia. because the internet in the

33:22

Middle East, because the internet infrastructure

33:24

there is not particularly good. And

33:27

so there's two things that we're

33:29

doing. The first is we're bringing

33:31

fiber connectivity to where there already

33:33

is stake. And the second

33:35

one is we're bringing fiber connectivity to

33:37

areas people would like to run

33:39

validators in, but they currently can't because

33:42

the inner infrastructure is too slow.

33:44

So latitude is one of our partners

33:46

in like network connectivity. And

33:48

they're bringing fiber connectivity down to

33:50

South America and Sao Paulo. And

33:52

that's going to make it possible

33:55

for the first time to really

33:57

profitably run Solana validators in South

33:59

America without dealing with a basically

34:01

an economic penalty for doing that.

34:03

Most people who are running validators

34:05

away from the core, this applies

34:07

to Monad, this applies to Sui,

34:09

this applies to Aptos, this applies

34:11

to Solana. They're doing it because

34:13

they have some other reason. They

34:15

either really want to run validators

34:17

in their country, do national pride,

34:20

or a government, or there's some

34:22

other reason other than just pure economics

34:24

that's allowing them to locate there. And so

34:26

we're going to basically expand the footprint

34:28

of where it's possible to run validators in

34:30

the world. And this is really

34:32

important if you if you think that like. An

34:35

L2 is going to have multiple

34:37

sequencers, remove the sequencer around, or on

34:39

Solana, you're going to have multiple

34:41

concurrent block producers, right?

34:45

You need a fast global connectivity layer

34:47

so that data can move back

34:49

and forth quickly, but also you don't

34:51

want someone who is trading in

34:53

South America to be disadvantaged against someone

34:55

trading in New York. These are

34:57

global distributed systems, and so you should

34:59

be able to get parity access.

35:01

pretty much anywhere in the world that

35:03

you're running the systems. Thinking

35:06

about the. The

35:08

token, do do you reckon

35:10

that? Introducing.

35:15

A token into this kind of

35:17

infrastructure project risk, financializing it

35:19

too early. You think about like

35:21

some of the early. Internet

35:24

people, they were all kind of,

35:26

you know, weirdos in their. Mom's

35:29

basement tinkering. Yeah,

35:33

I mean, I think we're

35:35

rather than profit seekers. Yeah.

35:38

Yeah. Yeah. I mean, we haven't had

35:40

an Internet like that since at least

35:42

the mid 90s. Right. And so

35:44

I think make the Internet great again. Yeah.

35:46

Like, you know, it's sort of like like

35:48

you go back to like defense contractors in

35:50

the Cold War and like they built the

35:52

U2 spy plane for like less than 30

35:54

million dollars in like nine months. It was

35:56

I may have some of the numbers there,

35:58

but it was it was incredible. Like. how

36:00

cheap they did that. And then you look

36:03

at like modern projections, like, well, these are

36:05

run as like, very aggressive for profit businesses

36:07

at the moment. And the internet's gone through

36:09

kind of the same type of transformation. And

36:12

so for one, like, I

36:14

don't think we can, one

36:16

of the superpowers

36:18

of crypto going back

36:20

to the original

36:22

Bitcoin is it weaponizes

36:24

capitalism. Very

36:30

ironic, Austin's cutting out for

36:32

me. That's

36:34

the definition of irony. Is

36:41

this why we need layer zero or double

36:43

zero pieces? Maybe

36:48

we're back. Yeah, would you guys? Would

36:50

you guys lose me? Well, where we left

36:52

off was basically. This

36:55

isn't your grandfather's Internet. Yeah.

36:59

Yeah, okay, so yeah, I

37:01

think we're talking about like

37:03

some of the principles of

37:05

this thing so like Yeah,

37:07

the internet There's too much

37:09

of crypto nowadays I sort

37:11

of like shifted from weaponizing

37:14

greed right like the power

37:16

of Bitcoin was like a

37:18

weaponized greed and capitalism to

37:20

produce optimal results Everyone trying

37:22

to screw each other over

37:24

actually produces correct blocks And

37:27

if we think use things like fee

37:29

markets we actually get like good efficient like

37:31

inclusion and discovery in this stuff. And

37:33

too much of crypto started moving to

37:36

almost like goodwill projects or things that

37:38

rely on donations. And so

37:40

we want to bring like hardcore economic

37:42

incentives back to some of these

37:44

systems and especially on the connectivity layer.

37:46

I think if we're able to

37:48

actually do this successfully with double zero.

37:51

Um, we start to build systems where

37:53

providers are no longer incentivized for getting

37:55

data off their networks as quickly as

37:57

possible. They're incentivized for providing the fastest

37:59

connectivity possible. And part of this is

38:01

like, there's a, there's a component of

38:03

double zero we haven't talked about yet,

38:05

which is not going to be live

38:07

at main nets rolling out in 2026,

38:09

but prioritize data where if we have

38:11

five different private fiber pathways you could

38:13

take from, let's say like New York

38:15

to Singapore, all of those are

38:17

going to be completely fast enough for blockchain

38:19

consensus and voting. But if you are running

38:21

an arbitrage or you're a market maker and

38:23

you always want to make sure you're going

38:25

over the fastest connection, you're going to be

38:27

able to pay for prioritized data to make

38:29

sure it's always moving over the fastest connection

38:32

possible. And that is going

38:34

to be a huge driver of performance

38:36

on these networks. And it also

38:38

aligns the economic upside of the network

38:40

providers with the value they're providing

38:42

to customers, which you'd think is part

38:44

of the public internet architecture today,

38:46

but it's absolutely not part of the

38:48

public internet architecture today. Let's

38:51

maybe wrap up with sort of

38:53

like thinking about just building this out.

38:56

Obviously, you had, um, storied,

38:59

um... Does storied carry a negative

39:01

connotation? I think it does, doesn't it?

39:03

It does, doesn't it? Yeah, no,

39:05

that's not the right word. I mean,

39:08

you've built teams, you've built, you

39:10

know, you've built out Solana. Um,

39:13

what is it like sort of now

39:15

being in that co -founder seat and,

39:17

uh, just from a talent and

39:19

competition standpoint? trying to

39:21

find the best low

39:23

latency infrastructure engineers. Obviously

39:26

you're competing with, from a

39:28

talent perspective with, you're not competing

39:30

from a work perspective with

39:32

some of these other chains, but

39:34

from a talent perspective, how

39:36

do you plan to attract and

39:38

retain the type of caliber

39:40

of talent that, you

39:44

know, you're not just optimizing

39:46

for block time here, you're trying

39:48

to build or rethink. physical

39:50

infrastructure. So where are these people?

39:53

Yeah, so a lot of our

39:55

engineers on the Malbec Labs side

39:57

of things are actually fairly new

39:59

to blockchain. They're hardcore network

40:01

engineers that have spent the

40:03

majority of their career sometimes

40:05

working at centralized exchanges and sometimes

40:08

working at high frequency trading

40:10

firms or sometimes working at

40:12

other really latency sensitive internet

40:14

businesses. And so one of

40:16

the interesting things is like networking infrastructure

40:18

is like what makes HFT firms work

40:20

at the end of the day. Like

40:22

everyone's like, oh, it's the quants. And

40:24

like, obviously the quants are important. But

40:26

if you don't have the underlying network,

40:29

the quant strategies can't can't run and

40:31

can't work. But specifically,

40:33

these network engineers are not

40:36

compensated like the quant trading people,

40:38

right? And so we're actually

40:40

not like, for us, the

40:42

network engineers are the most important component

40:44

of our product. For many of

40:46

the firms we're hiring people from, they're

40:48

not seen as the most important

40:51

component. And so we're actually

40:53

like our talent pool is incredibly

40:55

skilled, and there's a little bit

40:57

less competition for them than you

40:59

might expect there to be. Because

41:02

for most industries, the

41:04

connectivity layer is sort of like, it's

41:07

a you need it, but it's

41:09

not considered to be like the

41:11

make or break in whether you're

41:13

going to have a profitable training

41:15

strategy, even though it's probably should

41:17

be. And so that means that

41:19

like we actually can offer other

41:22

things that these these big firms

41:24

like Citadel and jump can't offer,

41:26

which is the ability to publish

41:28

your code, right? All of these

41:30

firms are deeply secretive because they

41:32

have special sauce and proprietary systems.

41:34

We're building open source technology, right?

41:36

Double zero is an open source

41:38

blockchain protocol and an open source

41:40

networking protocol. And so there's a lot

41:42

of people that just they want to talk about the

41:44

work they're doing. They want to not feel like they're working

41:46

for the CIA. And so

41:48

there's a really compelling answer to

41:50

say, like, come build high performance

41:52

distributed systems, talk about the work

41:54

you're doing, publish papers, like

41:57

help this industry move forward by a

41:59

decade or more. And so we've actually

42:01

had a really good experience on the

42:03

hiring process for these high performance network

42:05

engineers. And then on

42:07

the crypto side, like, there's not a lot

42:09

of projects that say you're going to

42:11

work at a very low level on

42:13

like very low level optimization and you're going

42:15

to work on it across multiple projects.

42:17

And so it's a very interesting like academic

42:19

intellectual problem to say like, okay, like

42:21

I got to figure out like how does

42:24

Solana consensus work and where are the

42:26

areas we can work with the Anza team

42:28

to bring these networking technologies to the

42:30

validator client and make it faster. And then

42:32

you get to do the same thing

42:34

with near and you get to do the

42:36

same thing with like IPFS and like

42:38

you get to do the same thing with

42:40

like Mert and Helius and the RPC

42:42

teams to figure out, like, where are the

42:44

optimizations here? And, like, in some

42:46

ways, it's like we're bringing, like, a

42:48

F -35, like, a medieval siege. Like,

42:51

the mismatch between, like, what blockchain

42:53

people think is high performance and what,

42:55

like, people coming from trading firms

42:57

think are high performance are just so

42:59

far that, like, we're

43:01

able to just add a ton

43:03

of value. Like, there's a lot

43:05

of low -hanging fruit. This is, like,

43:07

the... We've done so much work

43:09

on, like, ZK optimization and like

43:12

the software stack on validators and

43:14

blockchains now is pretty well optimized,

43:16

right? It's like pretty impressive what

43:18

you can do over the public

43:20

internet with a network like Solana

43:22

or Aptos or Sui or Monad

43:24

or MegaEath. But the networking side

43:26

has just been so neglected that

43:28

it's kind of green field, right?

43:30

It's like when Jump Crypto came

43:32

and built the Fire Dancer client

43:34

for Solana. There's just a

43:36

bunch of stuff that they

43:38

brought. They brought a perspective that

43:40

didn't exist in blockchain today.

43:42

We always reinvent the wheel in

43:45

crypto, and sometimes that's awesome.

43:47

But we don't have to reinvent

43:49

the wheel when it comes

43:51

to high -performance networking. We just

43:53

have to build it in a

43:55

way that has cryptographic assurances,

43:57

that has multiple independent contributors, and

43:59

is basically sufficiently crypto -economically and

44:01

cryptographically enforced to be a

44:03

trustable layer for this stuff to

44:05

operate on. And

44:07

we'll leave it there. Thanks,

44:10

Frank. Any open roles? Any

44:13

podcasters that you need? We're

44:15

actually hiring for a bunch of roles. If

44:17

you go to 00 .XYZ, this little careers

44:19

tab in the bottom, and and hey, they

44:22

said, create your own role, you

44:24

know. Influencer, who know. Thanks

44:27

for taking the time. Thanks,

44:29

Frank. You're welcome. You

44:32

too. Move your TV.