Subscribe
Share
Share
Embed
The Starting Block is The Block's new flagship daily live show — the breaking news, the context behind it, and the people shaping where crypto goes next. Each episode is a 45-minute live broadcast hosted by award-winning journalist and broadcaster Gareth Jenkinson and powered by The Block's newsroom and research desk.
Good morning, good afternoon,
wherever you are joining us from
today, welcome to the starting
block.
It's a Friday, so that's one thing
to be happy about.
And if you look above me in the
price ticker, that's another thing
to happy about, Bitcoin nearly
hit $80,000
for the first time in a while.
And yeah, it's looking green across
the board. So it's been an exciting
24 hours or more
for the overall cryptocurrency
market. And you can definitely feel
a palpable, we are back.
Across crypto Twitter.
So I'm loving that today, feeling
very positive,
been really interesting,
few hours on Twitter as
well, seeing everyone's reactions
and breakdowns of what
is really leading this big
recovery across the markets.
And it's great to see a little bit
of Hopium right before the weekend,
what's there not to be excited
about. If you're joining us for the
first time, welcome to the show.
Today is officially us
wrapping up two months of the
starting block.
Crazy to think more than eight weeks
ago we were in Abu Dhabi sitting
alongside CZ live
launching the show.
If you missed that conversation you
can go onto the Blocks YouTube
channel and find our interview with
CZ and a number of
other big names in the space Charles
Hoskinson, Paolo Arduino,
many many other big name in the
industry that we've had the pleasure
of speaking to on the starting block
in the last few weeks.
So go and check those out on our
YouTube channel If you don't
already, please subscribe.
And if you're watching us on X
today, go ahead and like and retweet
this broadcast so we can get more
people joining us for today's
conversation. A little bit later on
the show, we're gonna be joined by
Costas Chalkias, who's of course one
of the co-founders of Mistin
Labs, a real OG in the
cryptography space as well.
So we're going to have a really
fascinating conversation today
around AI, cryptography
and the general industry.
So really looking forward to getting
him on screen with us
a little bit late on the Show.
But of course.
We always start the show with the
block briefing, which is our
breakdown of top stories
you need to know in the last 24
hours. This is, of course, curated
from news that our
exceptional editorial team is
putting out every day.
So if you don't already, bookmark
theblock.co, our website,
on your browser.
Go and check out our news.
We also have a mobile app if you
want to stay more up to date.
And, of, course, if you follow us
already on X, please hit that Follow
button. And again, like and retweet
this broadcast so we can have more
people joining us today.
But I'm gonna jump straight into the
news so we get to the big
conversation of the day.
And starting off with our first
headline, breaking on the DeFi
front.
Mantra slammed the brakes on its
entire network earlier today.
The real world asset layer one froze
all transactions after spotting
an incident inside its Cosmos
EVM module telling users on
exit everything was paused while the
team investigated.
Here's the good news.
Mantra says it has since contained
that threat, traced the root cause
to just two wallet addresses
and crucially no user
funds have been exploited.
That chain could be up within
a day and we'll be keeping an eye
out from updates from the team.
This is, of course, pending testing.
And for a bit of history here, the
OM token though took another
hit down nearly 10% in
24 hours.
And this is, of course, after.
Mantra collapsed back in
2025, which was widely covered
in the industry.
I also had a conversation with JP
Mullin, the CEO at the time,
in the hours after that incident.
So not good news for Mantra once
again, and we'll be keeping
tabs on that story as it
develops.
Then onto markets, Bitcoin is on a
tear, pushing as high as $27,000,
$5,000 on Friday before
easing back towards $78,000.
Bernstein analysts think this could
be a real momentum shift.
And the thesis here is all about
liquidity.
The U.S. Treasury is doubling its
buybacks of longer data debt.
And Bitcoin has always loved loose
liquidity.
Meanwhile, the ETF story has flipped
spot. Bitcoin funds pulled in
more than one point six billion
dollars this week with BlackRock's
iBit leading the charge and
strategies back in the black on its
eight hundred forty thousand Bitcoin
stack.
Ether is actually outrunning
Bitcoin on this leg
higher. So really, really
interesting day across the
cryptocurrency markets, Bitcoin on
an absolute tear.
We nearly got to 80K and
seeing lots of takes on this really
being driven by that US
Treasury News in the last couple of
days, doubling down
on those buybacks of
debt.
So gonna be interesting to see how
things progress into the weekend,
but right now looking very much
green above my head on screen.
Then to Washington, a rare peek
inside the Oval Office.
President Trump reportedly huddled
with the who's who of crypto on Wall
Street, including Coinbase,
Kraken, Ripple, Robinhood, NASDAQ,
and ICE.
One source told the block's news
team that Trump was, quote, bullish
on getting the Clarity Act done.
Chainlink co-founder Sergey Nazarov,
who was in the room, called a
productive, mostly soliciting
feedback on what it'll take to pass
this bill.
A strategic Bitcoin reserve even
came up, though no next steps.
Have been explained
upon. The first Senate procedural
vote is set for September 15th
for clarity, but the bill is still
tangled in an ethics fight over
Trump's own crypto wealth
and interests.
Then it got spicy over at the CFTC
during an advisory committee meeting
yesterday.
CME Chief Terrence Duffy warned that
some prediction market contracts are
ripe for manipulation,
quote, horrible for our industry.
CFTC chair Michael Selig
cut him off, snapping those
examples, that those examples
were offshore and quote,
fake news.
Then Kalshi's COO
Luana Lopez Lara jumped in asking
if CME had ever
had a manipulation problem.
Duffy fired back here that he has
more people in his regulatory
department than her entire company.
Her reply, maybe you should learn
a little bit about efficiency.
Duffy then came back, maybe you
shouldn't learn about credible
markets.
So the turf war is clearly on
over who polices prediction markets
in the United States.
And I think it's a story we're gonna
be talking about for months and
months to come.
And Kraken wants to be your bank,
just not in America at least.
Parent company Payward is exploring
full banking licenses overseas.
Co-CEO Dave Ripley told the block
at the Wyoming Blockchain Symposium.
He sat down with my colleague,
Calvin Sparks, yesterday.
And we will be producing that video
and having it up on our YouTube
channel. And our X account in
the next 24 hours or so.
So keep your eyes out on that.
The pitch here, trading, banking,
and asset management, all under one
roof at Kraken.
Payward already runs Kraken
Financial, a Wyoming chartered
institution that recently snagged
a limited federal reserve master
account, a first for a crypto
firm, though it still can't lend
fiat or offer FDIC insurance.
And looking further out, executives
say they'd even like to offer
customers mortgages one
day. So.
Clearly seeing the evolution of
companies that started in the crypto
industry moving closer and closer
towards full stream financial
services that big
TradFi companies have been
offering for decades.
And that is it for the Block
Briefing this morning.
If you want to check out more news
relating to the industry, please do
go ahead and check out the Blocks
website that's at the Block Co.
We've got some fantastic journalists
who have their finger on the pulse
every day.
We've also got a couple of
newsletters that you can subscribe
to. On our website.
So I do encourage you to go and
check out the Blocks website if
you're interested in keeping up to
tabs with what's happening in the
industry. But that is it for the
news. We're gonna be moving on to
the hot seat.
It's a great pleasure to welcome
Costas Chalkias, who's one
of the co-founders of Mistin Labs.
We actually met in person a couple
of months ago in Paris at
Proof of Talk.
So first things first, Costas,
welcome back to the show.
It's good to see you again.
How are you doing?
Hey Gareth, good to see you guys,
good to see everybody.
Completely different environment
even since the Paris event.
As we can see, how AI is evolving.
Quantum is coming into the picture,
of course, the last few months.
I think we have a lot of things to
see this year.
Actually, 2026, at least in
my brain, is probably something
of a change in era
of AI and crypto and
general security.
This is something I'm.
I'm very curious to learn more and
how I can even contribute myself.
You know my background, Gareth,
right? I'm in this space forever.
My hacking abilities have been
diminished.
Just because I have a competitor
now.
That's the reality.
Of course, you need to use it
properly. But yeah, here we are.
Let's discuss on what's happening
today. What is the risks?
Ask me very difficult questions.
I'm someone who is investing a lot
of time since the very past
from the insurability of all of the
blockchains.
And now it's probably the time for
everybody to understand the risks
and also what we're doing as
cryptographers to protect your
assets.
Yeah, sure. I think let's just start
there, Kostas.
Obviously, the DeFi sector in
2026 has been in the crosshairs
with a number of big exploits,
but lots of experts have been
telling us that these are not smart
contract exploits.
But if we look more recently at
incidents like the cold card hack,
which is really a look at
their systems and
people using AI tools to
go and vulnerabilities across
the internet, across software
systems.
Really does seem that we're at this
very dangerous point where the
power of the AI tools that we have
are far more
consequential than we may have
realized and everything might
be at risk. And right now it seems
like crypto is in the crosshairs,
but that could be the entire
internet.
What's your general sentiment right
now on the power of these tools and
the risk that they pose to
our industry and the broader
industry? Or do you think that I'm a
little bit too alarmist and maybe
blowing it out of proportion?
You're totally actually on the
point. The reality is a few years
ago, we thought that crypto might be
broken by quantum computer, right?
You remember everybody started in
2022, 23, even Satoshi
at the very beginning mentioned one
day, maybe we have to switch to
something else.
But then we gradually realized we
have a new species in our world.
And this new species is actually
something that is even smarter than
us. I wouldn't say necessarily on
everything, but it definitely scales
more than the human brain.
And the whole idea and what you see,
I believe the
important date for
humanity was actually December 2025.
And obviously, everything adds up
into 2026, where for the
first time we realized with all of
this agent
coordination and a few other things
that you can automate search on
random stuff across the world, you
don't need to be a hacker anymore.
You don't have to be the security
expert.
People are just using it for
fun or some more experts, including
North Korea and a few other
organizations, even
on the dark web.
Now they have a weapon in their
hands and obviously they use it
without even them knowing what it
will discover.
This was not the case in
the far past. I would say that 2023,
2024, even at
Mist and we invented some,
discovered a few algorithms that
could use the early versions of
ChatGPT back then.
We call the project Bagdar.
We found a few situations like
scanning GitHub.
We realized that if you even defy
platforms outside sui, they had
their secret keys in their test
net for some reason.
And it was by far easier now to do
all of this scanning without the
human just opening the files, then
you create some automated scripts.
Now you have a personal assistant
and this is where we are, right?
These in the hands of someone who is
not an expert is already a weapon in
the hand of someone who is already
a white hat hacker
or a black hat hacker.
It's literally a nuclear bomb.
And to those who are telling
me sometimes, do you still
believe quantum is a thing?
We see what's happening with cold
card. We see, what's happened with
all of the random news you
hear every now and then, even
mantra that today's situation,
we don't know if it was a human who
discovered it, right?
Probably someone, even a security
guy, found bugged you
some AI automation testing,
and then obviously they had to
protect against this and so on.
I don't believe any attack anymore
is happening just with human brain.
If you ask my like position,
even if you are one of the best
hackers in the world, you still need
this help because you have an army
of personal assistants.
Now.
The cold card in particular is
probably the perfect example because
what we realized in this situation
and I want our people actually to
focus on this.
There was a bug like
for a few years.
It was already there.
Many people were looking at it but
nobody probably spent some attention
into automating and having a
continuous audit on what's happening
like with the old software.
Cold card is probably the best
example we realized that we cannot
stop auditing.
I think everything has to change.
From typically audit,
SIP, and then you forget into
continuous audits.
We cannot forget this.
Any new model that comes to our
hands, in my opinion, should be a
small crisis for everybody who's
working on the security industry.
And they have to run all of their
new models against the previous
code. In my opinion if the team
behind it actually has done it
before, unless it's something
internal like inside their
attack and so on.
With the new models, they should go
back. To all of the open source
code and this is a debate between,
you know, ledger of the closed
source information against Trezor
and Coldcard and all of these open
source wallets.
What is better these days?
If you ask me for the first time as
a cryptographer, I believe that
obfuscation in some situation might
be a difference because now you
have a security engineer who is
looking at all of the doors one by
one figuring out if you have an
attack, but the attacker, what do
they need, right? One door to be
opened.
And this is the big change, in
my opinion, in 2026.
For the first time, we realized that
if you don't do automation
internally, and I'm talking from the
different side, I don't care about
what the Black Hat hacker will do,
because they will try everything.
We need to automate everything,
not only on the new launches,
but everything in the past.
Gold Card is the perfect example
here.
Every substantial improvement in
frontier models effectively requires
you to do audits.
That's all, right?
When you see anthropic...
Publishing something, figure out if
you have some connection, you can
have a pre-release.
If you don't, the first day you have
to have a small crisis,
a small like war room to figure
out if your affected by, I don't
know, some particular prompt or
skills. This is my personal view and
I believe it will surpass all of the
attacks we will see from the quantum
computers.
AI is actually a bigger threat
to security and
if coins,
blockchains, DeFi platforms do
not have a security in at least one
person who is using AI to
defend. In my opinion, they are
making a big mistake.
Even myself as an investor, one
of the things, because yes, I'm
working on sui, I worked on
Ethereum as an auditor, in
other blockchains in the past, I'm
in the Bitcoin space since ever.
It's one of first question I'm
asking. Do you have a red team?
Do you a security guy who
is actually checking?
Because I want to put some money,
even if it's like 1,000 or 1
million, I need to ensure that you
have some protection against these
AI attacks, because this is the
biggest threat at the moment.
Let's face it.
It really is.
And I mean, I think you need look,
as you've said, no further than
the cold card incident.
And the response from the Bitcoin
community has been amazing.
I'm sure you're well aware of the
Bitcoin Red team that's formed.
And they've gone about using
Frontier models,
Kimi's best model and gone and
looked at all the Bitcoin open
source code, found as many
exploits or vulnerabilities
as possible, disclosed those.
He finds things.
Yeah, they find lots of things.
Which is really scary.
Something triggered at least our
community to actually be more
serious and sensitive
on this. Actually, it's good.
You know, in all of the innovations
that humanity did from the nuclear
energy and everything, we have a
small period where it's not a grace
period necessarily, it's a period
where this is exploited.
And then we need to figure out how
we can defend against this, then
eventually would be a balance.
And in my opinion, we're progressing
as a field
not only substantially,
but in my opinion in a huge way.
It says that we're learning to live
with this threat, right?
We have the wolf, we have the tiger
here, and now we try to create our
own differences.
I give one or two years, right,
one or three years for most of the
teams to realize without
the similar effort that the Bitcoin
community is doing, Sui as
well, we do have a specific rating
here.
You cannot move on, right.
It's very simple.
Don't invest in anything
if you don't have a security team
behind. Because previously it was
just sales, sales, and sales.
Now you realize insurance of
your network is actually more
important, otherwise you cannot
sell.
It's very simple.
Yeah, it makes me think to Zach
XBT's recent tweet, which was
actually eerily on
the money because I think it was
about two weeks before the cold card
incident happened where he was like,
hardware wallets are junk,
you can't trust these things.
And then two weeks later, the
cold-card incident happened.
And he was kind of suggesting that
you're better off using an
iPhone that you don't use anymore
that's a few years old because
of the closed source nature of
the code, with the security.
Would you argue that maybe
that is kind of true?
I have mixed feelings.
In my opinion, all of the world will
get into a multi-sig mode.
You need the hardware.
There is a reason you need hardware
because AI will also attack your
phone, it will also attack older
phones even easier than before.
We need the second factor of
indicator where you need proximity
to the actual victim
in order to steal the password or
whatever we're using for
authentication.
People are not, I think...
The early versions of blockchains
made, in my opinion, an important
mistake.
You need to charge the users
more just because they're using
multi-sig. And if you see what we
did at Sui, we erased that.
We don't penalize users because they
are using more complex cryptographic
authentication.
Because it's as if you're going to
Google and you say you're using
8-digit passwords or 30-digits
passwords where we need slightly
more hash invocations.
Google needs to pay slightly
more on energy.
You need to pay us.
It doesn't make sense.
We have to get into a mode where,
in my opinion, the hardware wallet
will actually be more important than
we believe. I'm against that
comment, by the way.
I believe you need both.
You need the old iPhone, of course,
but you need also to have a second
factor of indication because pure
software, especially on devices
that are old.
Are even more vulnerable to this
AI revolution, right?
This is my personal view as a
cryptographer.
I don't know if I'm giving advices
now that some people cannot do.
Figure out how to use
like wallets or put some pressure
to the communities like including
Bitcoin, Ethereum, Shui, as
ourselves, I'm not putting ourselves
outside the equation.
Guys, build very easy to
use multi-sig wallets with
some hardware support.
This is the only way to, you need
two factor authentication.
Like You cannot rely on anything
anymore.
And imagine myself, like I
live in Dubai, and imagine
everything is under camera
surveillance.
Sometimes you're even using
hardware.
There might be a camera behind you
stealing your password or stealing
your credentials, right?
We also have to be careful by
ourselves. It's like natural
defense. We see the attacks.
It's good that we have these
incidents, in my opinion, not the
end of the world.
Yes, a lot of money has been lost.
At the same time, There is progress
and we proved a lot of things before
something happened.
Imagine if Claude published Fable
or Mythos without telling anybody,
not having a pre-release, and this
came to the hands of North Korea for
one year, then it will be the end of
Web2 and Web3, right?
We have to do it cautiously, but
in my opinion, every individual
should also protect their keys and
ensure that they have some multi-seq
capability, even if it's more
expensive for other blockchains, and
so it will cost free.
Yeah, for sure.
And I think for most people,
it's just a big wake up call
to be very conscious
about the potential
mishaps of, you know,
having one single point of failure,
you know, so if it's a hardware
wallet and your seed phrase like
that, you're
playing a dangerous game.
Yes, you are safe.
Go ahead, Costas.
Gareth, my biggest fear here,
it's not only web3.
Have you heard some news that
someone actually lost the
seed, the mnemonic, because they
had it on Gmail and their
email was hacked?
And obviously, agents now, because
previously human, even if you could
attack an email, you needed some
scripts, you need so many other
things to understand, okay, where
could I find the mnmonic if someone,
I don't know, make some
anagrams on how they refer to
the mnomonic and so on.
Now it's very easy.
If I manage to get into your Gmail
account or even your Facebook
account, I can extract information
about you that I can use it either
for social engineering or literally
find the information inside your
inbox.
So we have to change that mentality,
not because we are talking about
Web3 and you're obviously
like me, an advocate of
this community. We need for
everything, for Web2 as well.
And this is, in my opinion, one
extra plus point I'm
giving personally to the
cryptographers and security
engineers in our space.
Because without us, and
if you see all of the big
innovations in the cryptography
space across the world, forget Web3,
most of them because of the money
that was in the blockchain space,
was because blockchain companies,
blockchain organizations, blockchain
foundations, had some of the
best cryptographers in the world.
And I can tell you for sure what
Mark Zuckerberg did in the Libra
days. You remember Facebook, for
those who don't know, a few years
ago tried to build another
blockchain internally, right?
It was called Libra.
It changed the name afterwards to
Diem.
But before that, Facebook didn't
have the amount of cryptographers.
You see what happened with Cambridge
Analytica and all of this.
Because of blockchain, however,
Facebook needed to hire and transfer
a few people across the globe, like
even myself. I moved from London to
the United States and it created
some of the best teams in the world
for security and cryptography.
All of these teams are now working,
I don't know, either for A16Z, some
VCs who are also investing in
crypto. Some other like companies,
some of them went back to NSA and
NIST and so on, but it started
with blockchain research.
We cannot, in my opinion, say
that blockchain is only like for
transferring money and maybe some
people don't believe in it.
On science side,
and the security of everything, our
AI agents, our emails and
everything, the blockchain
improvements on security
and cryptography was probably
fundamental for us to
defend against the new threat which
is AI.
Not only for our crypto, for
everything, right?
For your social life, for every
thing. And this is one of the
reasons that keep me busy
and engaged with the
sector. Otherwise, some people say,
oh, I will move from web threat to
AI because I don't trust the sector
anymore. I would say,
you will see that AI will definitely
need verified agent
execution very quickly,
vulnerability detection against AI
threats. Where is the red button
that will stop AI from running and
conquering the world, right.
This will happen from
cryptographers. Most of the
cryptographers live in the
blockchain world.
Let's face it, that's the reality,
right? Without this, you wouldn't
have this expert to defend against
AI. And it's one of the reasons that
I want crypto to survive because of
this.
Yeah Yeah, for sure.
And I mean, like you, I look at
people like you guys like Ilya
Polosukhin at NIA, who's, you
know, one of a foundational AI
engineer, who's now, you know,
everything is going back that way.
And exactly conversation we had with
him a few weeks ago was exactly
this, you know, we've solved a lot
of the problems that, you
know, and tackled a lot of the
problem that the AI and
the, like, general sort of
technology sector is dealing with.
So we're like purpose built for
this. I did want to speak to you
just very quickly.
Obviously, Sam Blackshear
announced that he was leaving Mistin
and going to Anthropic, which is a
big deal, right?
But it kind of speaks volumes
of the category
of people that were working at
Mistin Labs that are now going to a
company like Anthropic to help on
defensive security research.
Maybe you can talk to us a little
about that, like Sam's
contribution to Mistin and what
he's going to be doing there if you
do that.
By the way, for those who don't know,
Sam was the reason that I eventually
ended up into Silicon Valley because
he saw me in a conference in
Canada back in 2018,
I think, that I had the
paper with Mike Hahn.
Mike Hahn was the creator of
BitcoinJ in
the Satoshi protocol and the last
one who received an email from
Satoshi, he was my manager.
The reason I'm in the Bitcoin space
forever was probably like
potentially and partially because of
Mike. And then Sam invited me to
Facebook and I have very good
connection with Sam.
I know exactly how he's thinking.
He helped me a lot in my career and
he's probably one of the best
software engineers we have in
the Web3 space.
We still have, I didn't say had, and
I will explain why.
Obviously this is a personal career
decision for Sam, but don't forget
that this opens the door for crypto
companies to speak with real AI
foundational LLM companies.
Sam is still an advisor to Mr.
Nsui and there is a reason because
he's very, very sensitive.
On the security-like
defenses against AI and how we
can use AI to
do all of the security surveillance
and everything.
I don't believe, to tell you the
truth, that this is actually a bad
thing, but it's mostly a good thing,
because even for ourselves, we
see a new door opening into
these companies.
You already know that
Mr. In particular had some good
connections with Google and a few of
the early Gemini-like
progress on vulnerability detection.
Now we have with Anthropic as well.
Sam is fundamentally a security
engineer and researcher.
This is what he did for our
space. He tried to create probably
one of the best languages after
Solidity that was focusing on
security, which is Move, right
on top of Rust.
Obviously, Rust is known for being a
secure language.
And Move is a Web 3 language
in particular, like an evolution of
Rust to make this happen.
In my opinion, the.
Like modern version of Solidity.
I see Solidity like, I know some
people will make fun of me,
but it's like the Cobol and Pascal
of Web3, like
the early programming languages of
the world. And then you need an
evolution. The next evolution that
we have at the moment and which is
the best, in
my opinion, effort is Move.
And some is Move, right?
There is a reason I believe that
Anthropic also hired Sam.
I don't know exactly how he
performed in the interviews.
By the way, He was not also immune
from interviews.
Don't people, people shouldn't
believe that because.
Yeah, yeah, no, no.
He got an interview. He got the real
interview actually.
And Sam is very sensitive, right?
And by the time he realized,
oh, Anthropic cares about me and
they want me to lead the few things
around security for with
AI and all of this stuff.
Obviously I'm still young.
He's younger than me, by the way.
He had to take a decision, but he's
staying in web three, like as I
said, as not only an advisor.
The Move Advocate and all of this,
which is the best language in the
world at the moment for these
things. You never say the goodbye,
right? So in
my opinion, that's a major thing.
I personally switch, obviously,
it hurts me when I have to run some
tasks that some had to do
internally.
Like as a quiddle, right, I have do
a lot of more things
under my table now and my umbrella.
But at the same time, for the...
Like for the first time, since I
remember myself, I do have a door to
Anthropic.
This is generally, if I have it,
imagine there is an ecosystem around
us. I'm also active in other
communities. As you see, I'm working
on quantum, not only for Sui,
but some of the quantum solutions we
had, it's applying to Solana, to
Near, and a few other places.
Imagine if you also have someone who
knows Web3 really well, and now it's
inside the nest of AI,
literally in the bee's nest, in
the hive.
So that's why I believe it's
actually a good outcome, and I have
to actually myself make more
out of this.
Not only accept that Sam opened
the door, let's use the door now,
let's go there.
I think it's excellent, honestly,
and I appreciate you being so candid
and open about talking about what it
actually means because I think a lot
of people are skirting around it.
And for me, it was like, this is
awesome.
Someone who's built one of the
newer blockchain
protocols that we have is
moving to Anthropic,
to one of three or
four AI companies in
the world. Like take
a bow, crypto industry, we can
be pretty proud of how far we've
come. And I wanted to take the
conversation in the direction of
Walrus, because I wasn't even that
aware that Walrus has kind of taken
a big direction change.
You've got Walrus memory now,
which is essentially, as far
as I understand, from a layman's
explanation, a protocol that
allows you to take all of your
history and all the things you've
been doing with one LLM
and take it across to another LLm,
which for me is a big
deal, right? Because someone might
be a chat GPT user for the
last year.
But then they realize that, hey, I
really like Claude. I really what I
can do with Claude code.
And I want to go and do it there.
But I've done everything on chat GPT
up until now. And it's all in the
open AI systems.
How do I pull that across to
Anthropic System?
Can you talk to us about how
this kind of came about
and what it solves for the industry
that we don't have?
This is actually one of my babies
here.
Obviously, having a
storage and data
portability network like Walrus,
this is the reason that we had to
build Walrus because most of the
blockchains can handle short
chunks of data, probably just
transactions.
What do you do when you have to deal
with larger data files and so
on?
And our civilization is becoming
more hungry for data.
We see it like AI, even AI needs
data.
And then we realized that for the
first time, and to tell you the
truth, something that triggered a
lot what we do on Walrus
is Trump's statement,
I'm European.
Back in the days, not a few months
ago, actually, where Europeans
or anybody who is not an
American resident, citizen,
doesn't have access to mythos and
fable.
I was shocked.
I'm not imagining, in my
opinion, because I'm also like,
you know, I'm raised.
At least business-wise and
technically-wise in the United
States, Silicon Valley and some
London.
Now Dubai, of course.
My early career, of course, was in
Europe, but I'm still like my family
is from Europe and so on.
I felt like Europe doesn't have a
chance to survive what's
happening, right? We don't have
anything apart from Mistral.
That's the only effort I could
probably remember.
And this is like miles
away from what's happening in the
West and then to Far East.
Like to China.
Open source versus closed source
model. This is what we have and
there is nothing in between.
So I said, look, geopolitically,
and the fact that I have many, many
friends, I'm also an angel investor
in some situations and I know
some of them are working with Sui as
well on verifiability.
There are three elements that we
need to protect our data against
things that we're predicting
are coming.
One of them is the need for people
because the LLM performance
is changing.
To migrate from one place to the
other, like from Gemini.
Now I realize, oh, the new model
from Claude is better.
I go back to Claude.
And then, oh Gemini is now better
again. Why should I be second
compared to my colleagues and my
competitors in the world?
I go to Gemini and
then, Oh, Tzadzipiti made the new
evolution and then DeepSeq and then
whoever.
We need to port data.
But if you have only one American
company doing that, then eventually
you are a hostage of a particular
nation or a particular situation
like a company.
That eventually has all of the data
from all of the LLMs and then they
can build their own model.
And I know how this thing works,
right? I mean, I used to work at
Facebook three years.
I was the lead cryptographer at
Facebook. It's super important to
use this data for your own
evolution, building your products
and so on. Now, the new trend
is to build new LLM models because
after we reach the state where
the LM models became, I don't know,
you're improving slightly, then you
need data. Data is the king, data is
the new money.
But because it's the new money, it's
also our personalities.
Like you're talking, I guess you,
Gareth, as well, more to
the chat GPT, Claude and
Gemini than your
partner, your friends, even
inside your mind anymore because
you have a personal assistant that
you can ask them any day, like at
the restroom, like when you're
sleeping, when you are even driving.
Now you have voice, like when your
interviewing anything.
So it has everything about you
and actually knows more than
yourself because your memory as a
human is limited.
You might be surprised that in a few
years, the memories that you're
using on your GPTs, on your
Gemini and Claude is going to
be more than you can handle
internally as a Human Being, right?
Which means that we need a way to
decentralize this and not give the
power to a particular entity.
And Walrus was obviously the
obvious solution.
It's decentralized.
In a way that the data is not
controlled by a particular continent
or nation or company,
right? These are the three things.
People are porting now because it's
a new trend, just because the
performance is moving.
Also the cost, now the cost is also
very important.
So these three factors require
someone to be the base,
not addicted to a particular
nation or country,
that you can move your data from one
place to the other. The only thing
we had is decentralization.
The only decentralization that is
working with multisig, with some
quantum resistance now
efforts and so on, it's Walrus,
there was nothing else.
So we took the decision that
Walrus itself will deal with
portability of data and AI
because this is a natural evolution
of how we started and where the
world is now.
Geopolitics, the requirement of
people moving from one place to the
other, LLMs, and the cost.
Sometimes you might live with
like less power, less performance,
but you cannot for the cost.
So you need to migrate to another
LLM that is cheaper than you,
right? The one that you used to
have. So I believe it was natural
evolution. I personally see it as
convergent evolution at the same
time, Darwinism here.
It's a Darwin evolution.
We had to do it.
We had do it, otherwise there
is nothing else at the moment that I
think we are one of the first
who actually take that step in
a decentralized form, not a
particular entity. It's an
organization, a foundation.
Let's see how this will go, but I
have high expectations.
Seats. It's literally part of my
umbrella projects.
You need cryptography here a lot.
You also need verifiability of your
data because you need to verify, oh,
there is nobody injecting
data that you shouldn't want,
because then it affects your brain,
it affects your decisions.
So you need cryptographic.
It's actually one of the best
examples of cryptography being used
from web 3 to protect
against AI now and
AI migration of data.
And data is the key.
Data is the king.
I wanted to just chat
to you and pull on that thread a
little bit more about, you know, I
think it's about a month ago now
when, you know Anthropic announced
that, yeah, they would be limiting
access to Fable and
Methos and it was crazy, right?
No one expected that.
Do you know that I talked to my
wife, Gareth, I talked my wife
the very first, the very next
moment.
Maybe we need to move back to the
United States.
Like I've got one of my children
coming in this is the first time
this has happened on stream
Welcome, welcome.
You got to go downstairs you own
lots of people are watching on the
internet this is my son samuel go
on go downstairs
sorry i've got my wife coming
together no no no i have two
give me a mic
I have to as well, do not take your
time, take your time.
There we go.
I've been waiting two months of
the show and finally my son is
coming to the studio and bomb
me.
This funny video of someone
interviewing and his kid actually
came inside with someone like
grabbing the key with force like
outside the door.
This is this is amazing.
I think it was a famous BBC
interview.
But let me land the question again.
I'll reset very quickly.
Obviously, when they made that
announcement, it made me realize
how centrally controlled
these companies are and how
centrically controlled these
tools are.
And everyone wants to have equitable
access to AI.
And it's in humanity's best interest
that that is the case, whether
or not there are nefarious people
that can use these tools or you
know, white hat versus black hat
kind of, you know approach to this.
And I don't think that there is a
wholesale decentralized
AI alternative.
I know like Tethers QVAC is trying
to solve for this.
I think NIA AI is probably leaning
in that direction.
There's probably a few others,
models of companies that I don't
know about right now.
But how important do you think it is
to have a decentralized alternative
in this space?
In my opinion, it will definitely,
whatever we do, unfortunately,
it will create a societal gap.
I cannot see how we can avoid this.
We're not ready as like as
I would say culturally,
ethically, and a few more people
like there is also this capitalism
and you have to make also money.
Otherwise, it's the same thing with
drugs, right?
Some people are criticizing them,
you know, medicine companies and
you know the pharma situation.
Oh, why do you charge for these
drugs so much?
It shouldn't cost you more than
like one dollar now to produce it.
But you had to invest so many
billions until you come to this
stage where it's quite complicated
what you do. It's the same thing
because the same actually statements
have happened for pharma situation.
I don't know if you remember, right,
Trump also mentioning about, oh,
why do we pay so much for drugs and
like medical things,
right? Why Europeans are paying less
and so on.
It's not different.
Like we see that culturally we're
ending up in these situations.
Where for the very beginning, when
we have a new innovation, there is
also this patting world and
some debates on who is going to
win this war, because eventually
it's money. Let's face it, these
companies are for profit.
Like if an OpenAI started for
non-profit, and then we had all of
these debates, everybody knows.
And then at the other side of the
planet, we're having this more open
source way.
We don't know exactly how
the open source launches are
happening or if the data is the most
important stuff, because.
Even if you have an open source
product, then probably have access
to data and then this data actually
feed your next evolution of
like a product that one day might
not be open source.
So we cannot trust anything.
Because we cannot trust anything and
the world like in a capitalistic
scenario works like this, especially
at the very beginning of any
innovation that humans had, we
have to face it.
We need to gradually build.
It's good at the very beginning to
have this competition and obviously
protect your your IPs.
But we're getting in a more mature
environment, like AI
is running for at least three, four
years now that is so powerful that
can change like humans'
behavior and way of thinking.
We need to start investing in
these things. It will take time.
I'm not, I cannot promise to
everybody that whatever we do, Nia
is doing, some other like
visionaries are trying to do for
decentralization of AI.
It's still not there.
That's the fact.
We are working on whatever
we can to make the transition easier
down the road.
Walrus Memory is doing all of these
things for data portability
between LLMs, which is super
important. I personally use it, I
personally think it's a problem we
have to solve, but it's part
of the problem. It's not only this.
What do you do with all of this
models and the data that these
companies are accumulating, not
how people are storing the data and
they don't tell you or even
injecting information that can drive
opinions?
These things will require
cryptography.
My goal, and I believe this was one
of the reasons that
people don't believe it, but I think
in the back of
Sam's mind was,
let me go to the nest,
figuring out how we're doing all of
this security and privacy for our
data.
Sam is a very sane man.
Sam is someone who cares
about humanity.
I know, he cared about me.
He didn't know about me, right?
He sees like...
I don't know if he's
a sapiosexual or sapio,
you know, all of this.
He likes intelligence, but in
all of his aspects of his life, he
wants to contribute to humanity.
And in my opinion, this is also
something that I believe a few
visionaries on Web3
I taking
this into consideration.
We want to be able to
figure out how we can offer
decentralization, step by
step. It's not something that will
be sold in two years.
We have to start from somewhere
though, right?
There will be a few years until
these companies feel quite safe,
even financially, to be be able
share their ideas and models and
move to the next step.
And the next that might be, I don't
know, biomedical, you know,
Zuckerberg's foundation, by the way,
they're working on AI for
pharma, for finding.
I don't know, biochemical stuff.
We only, at the
moment, focus on text
and probably some optimizations
on trading and other things.
But in my opinion, AI will be far
more than what we have today.
So we're going to solve one of the
problem, eventually going to the
next step, which is protect humanity
against diseases and other stuff.
In the meantime, we're preparing the
ground for decentralization of all
of these data.
It will take time because if you
don't give some time to these
companies, they will close down.
And, you know, there is this debate,
what do you do in this, like,
business world at the same time,
they cannot do it for free,
right? It's quite complicated,
unless it's a national effort.
Yeah. Probably what's happening in
Far East, maybe,
and who knows, right, then this is
not exactly, exactly the case.
Eventually, there's someone behind
it. So we need to be careful
and, in my opinion, prepare the
ground for this. And this is what
I'm personally doing, right.
Starting with this even suite.
You will see suey becoming by far
more agentic and by far more
sensitive into the privacy sector.
There is a reason because we realize
that data, data is now
more important than everything.
And yes, there is trading, there's
DeFi, but along with this, there is
also data that is accumulated and
the blockchain should be, it is the
only weapon we have at the moment as
humanity for decentralization.
There is nothing else.
Yeah, it's it's very
interesting to me and I'm sorry cost
us I know we're probably eating into
maybe some other time this morning
but you know just the thought
of the the
The closed nature of these
businesses the Anthropics the open
AI is the deep seeks etc That
own and control the most powerful
models These models probably
wouldn't have been as powerful if
they as they are if they hadn't
had access to all the people using
Them right because the models learn
from the inputs And if we went
to a decentralized alternative, like
QVAC is kind of positing,
where the user keeps all of that
information and data on their
device, it never leaves their
device. The model itself never gains
access to that information.
Would it have grown
at the exponential rate, the
capabilities of the AI,
as opposed to how Anthropic and
OpenAI and DeepSeq and, you know,
at Kimmy.
All these frontier models have,
because so many people have been
using them. Because as my
understanding goes, it's
the inputs that really matter.
It's the sheer amount of people that
are using them that are helping
these AIs grow in
understanding and capabilities
so fast.
I was, Gareth, in this dilemma
for many months in the last
few couple of years, right?
I don't know, to tell you the truth,
what is better.
There is a problem.
AI is energy hungry.
All of this model, by the time
you're adding decentralization, of
course, as you know, I know it very
well, you have to optimize on
everything because you cannot easily
be in par with the current
non-decentralized architecture.
Everything that is decentralized is
more expensive.
And AI is already more expensive,
right? It requires so much access
to GPUs.
We are not controlling even,
like, you know, why do they have
limits on the context, right, and
the number of tokens?
Because you need the energy.
It's a cost, right.
It's cost. By the time you're
getting into decentralization, this
will be multiplied by something,
right depending on the, you now,
Nakamoto coefficient or whatever we
consider decentralization.
Then eventually you end up into a
situation where you have to balance.
Like how much time do you give to
these companies to run like
a proprietary software,
like closed source and
improve the energy consumption of
these models because they need to
optimize as well compared to going
directly to decentralized, yes,
protect the data of users.
But at the same time having, yes,
more data but slower
like reasoning just
because For decentralization, you
have to pay a penalty, right?
So we need, I
don't know when this bar will be
met, but there is
a race between the energy
consumption, what you can
do to audit these companies.
It's not always so easy because they
already have so much data that this
data might be so private that it can
be a weapon by itself.
So it has to have some control.
We don't want to give all of this
into hackers.
And at the same time, yes, it
can on user's device, but by the
time it's on the user's devices.
Then you rely on
the CPU and the GPU that it's in
some device that is very cheap,
right? It's $1,000 best
case scenario with iPhones.
Then eventually this will slow down
improvement and evolution into green
AI.
So it's a race.
I believe both of them will exist.
I want to see the decentralized
version of the full AI, not only
memory that we're doing here, like
whatever others are doing with
putting AI on user device, at least
the data. But I don't think by
itself and alone We're still
at the state that AI is
very hungry. Imagine that.
I don't remember how much do we get
as humans. Is it 20 milliwatts or
something?
The human brain to think of
something and burn energy for this.
AI is already hundreds
of times more expensive.
We're not there.
We're no faster than biochemical
reasoning, right?
So we need also to
invest into making these things
fast. Unfortunately sacrificing
and creating a grace period for
these companies to work.
At the same time, I cannot imagine
just opening a lane for
these company and not the open
source and the decentralized lane.
We need to do it on parallel.
But there will be, let's be honest,
there will for a while a period
where all of them
will progress on parallel and
eventually, I don't know, by
convergent evolution or by whatever
we believe is, as humanity is,
sane and wise, we're going to make a
pick one day.
This is happening with all of the
technologies. In my opinion, this
eventually, they end
up into a small
cycle of people who have this power.
But there is also a new
do-it-yourself evolution on
the site. Until the do-
it-yourselve, imagine Europe is
doing everything for green energy
and all of this stuff.
But you will never see a Google or
Facebook in Europe because this goes
against regulation.
Europe is very sensitive with Mika
and GDPR on this.
So we need, actually, I believe
that different countries or
continents will follow different
paths,
just because of how their society
is built.
From full capitalism to very
green energy,
focused and sensitive to, I don't
know, to socialism and a few other
things. Unfortunately, it will end
up there because AI will
affect society.
It's affecting society already.
We don't even know what will happen
with unemployment, right, where I
was talking, I'm at.
Like I'm talking with some like
French ministers in various
countries of course if you if
you know cryptography you can also
help like on protection
of like governmental
data infrastructures and so on.
Nobody knows what will happen.
Like we have this interview now
about the effect of
closed source decentralized open
source. Now we even have three
lanes.
You don't know because in different
jurisdiction different countries one
of them is more profitable or
a better suite, better fit for this
particular continent, but not for
the other.
So eventually, I
believe it will happen in all of the
parallels.
And then based on where you live or
where you want to live, this is what
you're going to do as human.
And we're going see innovations in
all of the three directions.
It's three directions, I'm telling
you. It's not easy.
Not an easy peak, not in this
transitional grace period
where we don't know exactly which
one is better.
Yeah, no, we don't.
We don't cost us and
I think we're gonna have to book you
on the show again in the next
few weeks or months to
continue this conversation.
Otherwise, we'll be talking for an
hour and a half or two hours.
I really appreciate you taking the
time to chat to us.
And yeah, I was really enjoying the
conversation and I'm loathe to end
it. But I know that I would take
another hour of your time.
So we will leave it there and
we will commit to having another
conversation sometime soon.
Yeah, and really appreciate
your insights and everything that
you and this select special
group of people are doing for both
the cryptocurrency and blockchain
industry and its convergence
with AI. Although I think AI is
converging with everything, which is
the crazy thing, to be honest, I'm
not even sure like I'm doing a
crypto show anymore because every
day I'm talking so much about AI.
Not even myself.
Am I doing just crypto?
You believe now?
You see, Walrus Memory is one of my
kids, and there is also quantum
protection and so on.
Next time, I'm going to discuss
something very, very interesting,
pretty much in the same lines of
conversation, why AI
is our ticket to open quantum.
But I leave it here as a
teaser, and I will explain why
humans cannot go to the quantum
revolution, but maybe humans and AI
can go.
Beautiful, definitely.
I'll bookmark that.
Thank you so much, Costas.
And thank you for joining us on the
starting block today.
We will be back next week.
But I do want to make a public
service announcement. My wife is
expecting a third baby anytime soon.
So it could be that we take a
two-week break from next week, but
you can expect us back on Monday.
If not, you'll know where exactly I
am. Thank you, so much for joining.
If you want to catch up on any of
the content or interviews, you can
find that on our YouTube page or
our ex account. To go ahead and
check that out. Like, follow,
retweet, subscribe on whatever
platform you're joining us on today.
We'll see you next week. Have a
fantastic weekend.