Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
Metis Hyperion AMA with Tom Ngo, Daniel Kwak, & Natalia Ameline
Recorded: March 20, 2025 Duration: 1:15:24Space Recording
Short Summary
Metis Hyperion is set to launch as a groundbreaking layer two solution designed for AI applications, addressing critical challenges in blockchain technology while fostering strategic partnerships and community engagement.
Full Transcription
you Yn ystod y cyfle hwn, mae'r cyfle hwn yn ystod y cyfle hwn. Mae'r cyfle hwn yn ystod y cyfle hwn.
Mae'r cyfle hwn yn ystod y cyfle hwn.
Mae'r cyfle hwn yn ystod y cyfle hwn.
Mae'r cyfle hwn yn ystod y cyfle hwn.
Mae'r cyfle hwn yn ystod y cyfle hwn.
Mae'r cyfle hwn yn ystod y cyfle hwn.
Mae'r cyfle hwn yn ystod y cyfle hwn. Mae'r cyfle hwn yn ystod y cyfle, mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle.
Mae'r cyfle yn ystod y cyfle. Mae'r cyfle yn ystod y cyfle. GM folks, this is Anthony.
We'll get started in about a minute as our guests join. So just hang back for one bit. Okay. Okay, looks like our guests are in, so let's get started.
We have a very special space today.
We are going to be talking about Metis Hyperion, and I know folks have had a lot of questions.
A lot of the Metis community have been wondering what is Hyperion
all about and what it's going to bring to our ecosystem. So today we're having a space to talk
all about that. Just some quick introductions before we get started. I am Anthony Robles,
your host and your guide through today's conversation. And we are going to be, like I said,
covering Hyperion and the convergence of blockchain and artificial intelligence within the Metis ecosystem.
This is obviously one of the most interesting emerging tech trends that is happening right now in our world.
And so we are excited to present Hyperion today and really dive into it. I am with three brilliant minds from the Metis team
who are leading this initiative on Hyperion,
which is a layer two solution specifically designed
to overcome the limitations that are keeping AI
and blockchain from reaching their full potential.
So let's meet our guests who are leading this
and hear about the journey to create the foundation
for this next generation of decentralized AI applications.
So if y'all can please introduce yourselves
and your role with Metis and a interesting hobby
outside of crypto that you have just so we can get
to know you a little bit more. So let's start with Natalia.
Hello. Thank you, Anthony. Hello, everybody. My name is Natalia. I am a co-founder of CryptoChecks,
which is an educational nonprofit and I am part of the Metites Foundation and I am Decentralization Coordinator at the Mites Foundation.
Awesome. Thank you, Natalia. And what's one fun hobby you have outside of crypto?
Cats and tennis.
Nice. I love cats myself. I've never played tennis though, so you gotta teach me one of these days.
For sure.
Let's go with Tom. Same thing for you. Introduce yourself, your role and a fun hobby you have outside of Web3.
Hey, it's good to be back in the space with community. My name is Tom Ngo, CEO of MetisL2.
And one of the exciting hobbies that people may not know
is I'm a big enthusiast in the health side
and the new, I would say, sector
that people are called biohacking,
but more on progressive health.
A big advocate for that,
been doing that for most of my life
since middle school, I would say.
So that's one interesting thing that I like to do is optimize not just my work,
but my body, my mind. So it's a I would say lifelong hobby and commitment.
Love it. That's awesome. And I know I know you're fit. I know you really take care of yourself.
I was hitting the gym. So same here. I tried, I tried to be on it. Not as much
as you though. Not that dedicated or disciplined. So let's pass it to Daniel. Hey, what's up Daniel?
Not sure if you're speaking. I cannot hear you.
Hey, sorry. No, sorry. Sorry. Yeah. Hi, Anthony. Hi, everyone. I am the newest one of the
newer additions to the team. I am the CCO, the chief commercial officer at Metis L2.
It's exciting to be here. And the introduction is long overdue.
I joined back in November, October, I think.
And I'm overlooking the operations and marketing side of things.
I'll be connecting with the community and all the listeners here quite a bit more going forward. I'll be a lot more involved in not just the marketing initiatives,
but also with the community outreach and engagement activities.
So I look forward to connecting with you a lot further
and building a very strong relationship with not just Metis,
but on a personal level with all of the contributors here.
Also, in terms of hobby, a year ago, I
would have said I'm very jealous of Tom
for having a very interesting hobby.
But I have one of the most boring,
but one of the most fulfilling and actually quite fun hobbies.
I'm a new dad.
I have a 10-month-old baby.
And that takes up all of my time outside of work and I am very excited about it.
Amazing. That's great. As you should be. Thanks for the introduction. Your hello world moment to the MED-US community.
Thanks. Yeah. So thank you. Thank you all for introducing yourselves. So let's get started. Let's get right into it.
So let's get started.
Let's get right into it.
We have about an hour and we want to make sure that we get through it all because right
after this ends, Metis Foundation is having a space as well.
So folks, we'll send the link after.
You should find it on our page.
So let's get into it.
What is Metis Hyperion and what sets it apart from existing layer two solutions?
I'm going to hand this one over to Tom.
Yeah, Metis Hyperion is a high performance
AI targeted infrastructure.
And many people are probably asking why.
I usually use this example so everybody can understand
because they see the market as being very overcrowded,
another new chain, another layer two or another L1 and I think from
a user perspective and retail perspective, they don't really care but they want to understand how
it's going to be used and what's going to be used for. If you think about Metis being one of the
first L2s, we have a legacy chain that has been pretty robust with some innovative approaches with
our decentralized sequencer.
But with time that has changed and improvement in technology and demand, we see that it needs
to be improved or we need to provide a solution that caters to today's demand, which is AI.
And if you think about blockchain as a 3G wireless network and you compare it with this
now, it's like 5G LTE and moving on to 6G.
And here at Metis, we see the need and demand. Hence, we've been working and really fine
tuning what in our approach on AI. So we came up with a network that we feel extremely confident
about the modular setup, parallel execution, paired up with our decentralized sequencing
to really help improve and provide the industry with what it needs today and utilize the efficiency and actually the use case for blockchain.
Yeah, definitely. So that's really fascinating. And Hyperion essentially creates a specialized layer two environment where AI applications can thrive
in a decentralized manner.
Anthony, I'm sorry.
If I may just add one more bit to it.
I know this question is like,
what sets us apart from other layer twos?
I would look beyond how we categorize
even our infrastructure nowadays in this industry.
It's not just a layer two, it's an infrastructure architect to really provide beyond a layer two, right?
I think this is beyond what we're comparing it to,
but that's the only thing I wanted to kind of highlight.
But please continue, Anthony.
Definitely, definitely.
I think that's an important distinction that you made.
You know, it's more than a layer two.
It's a
pretty much comprehensive infrastructure for the convergence of AI and blockchain. So,
yeah, thanks for that distinction. I think it's really important to highlight those things when
people are thinking about what's being built. So, Natalia, I want to pass this one to you. Can you give us a high level overview of why Hyperion was created and the specific problems
it aims to solve?
For sure.
So before we get into that though, I just would like to ask to get a quick reset on
the key points of metaserogenesis, which we recently announced,
if you followed us on the social media. So we talked about aerogenesis as a
rebirth of Mites in the era of AI. So what do we know about AI today? Not much,
right? It's a big black box. AI is growing exponentially. We know that.
We see the news popping up about AI every single day.
We know it's exploding.
But one of the key issues that we see
is that AI is growing in this sort of centralized bubble,
where several tech giants own the data,
your data, by the way, their own models, their
own algorithms, and all the power that comes with this.
So with that in mind, we see blockchain technology and AI converging.
We see them as very much complementary and synergetic technologies, where blockchain
tech has a great potential to elevate AI, bring it on the next level and help it address
some of the key issues such as trust, verifiability, equal access and so on and so forth.
However, in its current state, Ethereum ecosystem
is not really ready to meet the demands imposed by AI,
including speed, computational demands,
data availability layers that are
capable of handling these enormous volumes of data that come with AI.
The bottom line is we really need a new foundational layer
capable of handling this AI demands.
Our goal is to set up this foundation and to empower
Web3 with the tools it needs to
thrive in this new age, which is upon us.
Today, when we scout the blockchain space,
the space is scarce.
And the new AI-focused layer ones
are also a bit scary for the projects to launch on,
because it's a big commitment.
If the network fails, the project fails,
and there's no one else to go.
And we clearly need more dedicated solutions.
And so with that, here comes Hyperion,
which is our new Instant L2,
which is built to handle these demands that I talked about,
with a high throughput, low latency, and it's a modulus decay.
Yeah, and I'd also like to add, if I may, that when you're looking at kind of how the
AI space exploded on blockchain, and how everybody basically just jumped on the bandwagon. If you look at how it was kind of handled initially,
yes, of course, it makes sense for the low hanging fruit,
for the first to market type of projects
to kind of take center stage, totally makes sense.
And for layer ones and twos, completely makes sense
to support that industry through grants
and hackathons and whatnot. But that was kind of the extent of it for quite a while. And
still that takes up majority of the different networks that are saying that they're going
after AI. What we found was a gap in the industry where we saw that the needs of AI wasn't really truly being met by any of the
incumbents. So Hyperion is there to kind of provide the new home for AI, right? It's a custom built
home. It is targeting not just AI because it's going to be the next popular thing, but also
because, and this goes back to an Italian's letter to the community where
it ties back into providing real-world value finding actual exciting use cases and generating value for
not just the
the blockchain industry, but you know the users and the businesses and
The processes that are involved itself. So we're not just focusing on, hey,
this is the next cool thing.
This is the new NFT.
This is the new GameFi hype.
But we're actually trying to build something
that generates and drives value for an industry that also
generates and drives value out in the real world.
I see.
So you're definitely,
we're really addressing a gap in the market
and within the industry,
especially as these two technologies
are beginning to converge and are emerging
and are really gonna eat up the world
in the next few years.
And most blockchain infrastructure
wasn't built with AI's unique requirements in mind. So, I know it must have, you know, a lot of work to identify,
you know, the specific pain points around throughput and latency and tooling and the
different things that were mentioned that are holding AI back. So, yeah, thanks for chiming
in on that one, Daniel. And this one, next question I have is for Tom.
How does Hyperion fit into the broader Metis ecosystem alongside Andromeda? This has been a
question that has been coming or surfacing from our community. So just really trying to highlight
or describe what Hyperions fit in the broader Metis ecosystem with
our current Altzoo Andromeda? Yeah I think this is an interesting and a very
important question in terms of like expansion you know depending on the
point of view and where you sit in your interaction with the blockchain either
you're a developer, you're a retail kind of enthusiast or you're you might be an
investment institution
that maybe see this from all different point of views.
But as an innovative organization,
we have to expand and provide real utility for this space.
And I think where the industry is going, it demands more.
And we see it.
So the two networks will work in parallel.
Metis and Andromeda has the expansion
and I'll go back to my example of,
3G coverage you get all around the country.
You have accessibility, it's known,
and you can access it through many different channels.
Hyperion is the high speed and not everywhere in the country
you have 5G or LTE, but it
provides a very unique use case and demand, fulfills the demand that we need.
Eventually, everywhere in the world will have this speed.
So we're taking this initiative and launching Hyperion to facilitate that gap, which supports
the AI side, but beyond that, like Daniel shared, trading is still very limited on chain
with a little bit of setbacks that I would say is underperforming maybe centralized exchanges.
But with Hyperion, this will provide the solution because the demand of AI is such
high and it requires a lot of, I would say, advanced development. It can provide solutions for gaming, DeFi, trading.
And we see this as an evolution of what we need
and really transform the industry to be on traditional,
maybe peer-to-peer or maybe some DeFi aspect protocols.
So this opens up the door for many developers
and developers that saw limitations of blockchain
before.
Now we have a solution for them to come on chain and not have to worry about that limitation,
especially on the gaming side.
But with AI now, I think there's going to be hyper development in this space that will
find new use cases with Hyperion.
And we're super excited with that.
The thing about, you know, one of the questions I'm pretty sure the community asked is, will
one get rid of, you know, the other?
Actually, no, they will be in parallel.
And all this under the Metis token to support the growth of our ecosystem.
Yeah, Daniel, I see you went off mute.
Yeah, apparently I like to talk. So I apologize for jumping in two questions in a row. But
that's a fantastic explanation on kind of Hyperion and how Hyperion itself will act
alongside Andromeda. But to touch on Andromeda itself too, right?
We really can't forget Andromeda
because that is going to be kind of the general purpose,
all use case web three network
that Metis is providing to capture the rest of the market.
We've built something fantastic at Metis over here.
We have the decentralized sequencers. We will be reaching
stage zero and beyond soon. And we have a thriving community that is growing and that's
very excited about some of the potential possibilities that Metis is facing in the near future. We
really can't let any of that go. And that alone has significant value in and of itself and we really want andromeda to thrive alongside
Andromeda, right? So
Andromeda with higher decentralization with a longer track record with with a proven track records and with already
an existing ecosystem
That lives and breathed by itself
That's kind of you know of the standard go-to
for a lot of these applications and projects
that want to launch on a good L2.
And then Hyperion is going to serve as almost like
the halo, right?
It's the thing that draws attention because it is,
and let's face it, a sexy industry.
It's an AI, it's what everybody's talking about nowadays,
and high performance is always something interesting
to talk about.
So we are going to capture a completely whole new market
that we probably would not have been able to attack
very effectively if we just had Andromeda itself,
or if we had to upgrade Andromeda
to meet those high performance needs, right? That would have either taken too long, or that would
have been significant changes to the Andromeda network itself. So with Hyperion, we're capturing
a completely new market. And together, they can function fantastically as a, I don't want to say
codependent, because that has negative connotations, but there's
going to be a lot of synergies between them to create and drive value for the Metis ecosystem.
Indeed. And you know, I think it's a smart approach and it's similar to other even layer
2s, how they structure their own ecosystems with different chains for different purposes.
But I think what's unique about this Metis' approach
is I've never seen one purpose built for AI.
And that's a layer two purpose built for AI
along with not just a layer two like Tom said
but a whole infrastructure of tooling and SDKs
and other things that are going to be great for AI developers that are trying to merge blockchain
and create decentralized AI dApps. So I just want to in quick interjection, if you look at the top
of the screen, you can see the scoreboard or the jumbotron is what I like to call it.
If you scroll through there, you'll see some posts
that describe Hyperion.
And if you click on them, some of them have links to blogs.
So you can see some of our initial documentation
that we've released on Hyperion.
So you can continue learning a little bit more
beyond this space.
But I just wanted to let
you all know about that. And for those who are joining and are participating in the Metis Voyage
the Karma Quest, the secret code word is HYPE. H-Y-P-E. So that is your code word. So I know some of y'all are participating in that.
So moving on to our next question that we have here.
This one is for you right back at you, Daniel.
Hyperion boasts one to two gigagas per second.
So what does this mean in practical terms for devs and users?
Actually let me jump on this one, Anthony, if you don't mind of course. Of course. Yeah. So gigabytes, very scary words, means billing, right?
So Hyperion, both a huge number of giga gas per second.
So giga gas, maybe let's start from there, is essentially a term that is used to describe the amount of computational work
right in the blockchain space and specifically how much of this computational work blockchain
can handle every second measured in gas. So with such a high number for Hyperion, what does it mean in real world? It essentially will translate into unprecedented scalability,
ultra-low latency, and also of course, very high throughput.
That benefits both developers and users.
Practically, it allows applications,
AI applications that we talked about,
high frequency trading applications for DeFi,
real time on-chain gaming for the gaming,
for any type of application,
it allows to operate without fearing limitations
that traditional blockchains impose.
So essentially, for users it means almost Web2
like a speed experience which you know removes the lag and congestion and of course the high fees
from blockchain interactions. For developers you know they'll experience essentially they're no
longer limited by some of the Ethereum constraints,
and with enabling of high-speed scalable AI-driven applications.
And also not to forget, we always kind of like thinking about things as of what happens today,
what happens tomorrow, but we really need to future proof ourselves, right? And I view Hyperion as a kind of future proofing us, making sure that our network is capable
of supporting DeFi gaming AI and high performance enterprise solutions even later down the road.
Amazing. So thanks for explaining that. And that's pretty impressive. And thank you for
helping, you know, all the listeners here, including myself, understand what this means
in, you know, real terms. And, you know, and what kind of, you know, I know there's going
to be a lot of AI operations that are going to become possible at scale that simply couldn't exist maybe on Andromeda
and on other chains as well. So going back to this being a purpose-built chain that is
going to be able to effectively target the AI market and the market of builders as well. So yeah, thanks for that explanation.
Going to our next question and this one I think we're gonna stick with you
Natalia. Can you explain what MetisVM is and how it optimizes execution for AI
powered applications?
applications? Apparently I like to talk too, not just Daniel, right? Too neurophobic.
So MITIS virtual machine, MITIS VM, right, it is a high-performance virtual
machine that we built and it was developed to enhance the integration of artificial intelligence within a decentralized framework.
While it is compatible with Ethereum virtual machine, which is important, Metis VM also
introduces some advanced features which are tailored for AI applications specifically.
which are tailored for AI applications specifically.
And this design facilitates seamless migration
of existing EVM DApps that require high speed,
almost instant execution to meet this platform
and offering reduced gas fees at that.
So if we dive a little bit deeper into EVM,
one of the innovation that Mites Virtual Machine brings
is a separation of computing and storage functions,
which helps to optimize transactions,
throughput, and efficiency.
Mites Virtual Machine architecture supports
speculative and parallel execution,
which significantly improve throughput and transaction speed,
and is making it particularly suitable for the AI protocols.
But not just AI,
also for high-performance decentralized finance application and
gaming application as I touched on in my previous answer.
It is very important, it's actually crucial.
It's also utilizes dynamic code optimization,
which makes smart contract execution more cost-effective,
and it improves scalability as well.
On the other hand, we're also trying to optimize for data handling.
So Amitis virtual machine supports
the state-aware caching,
which is optimization technique
where frequently accessed or computationally expensive data
is stored temporarily in a cache.
And this approach improves the efficiency
of smart contract execution.
And how does it do it? It basically reduces
the redundant computations and redundant storage access.
Essentially, by bridging the gap between blockchain tech and AI,
Metis Virtual Machine aims to create
an open and scalable decentralized infrastructure that we need, right?
Because in the beginning, I talked
about how a lot of AI applications are centralized.
I think it's crucial that we kind of move it
from that centralized silo and make it more decentralized.
And we are creating the infrastructure to do that.
And so this foundation essentially
will ensure that AI remains transparent,
verifiable, and aligned with the human goals, which is very important.
We don't want one day wake up in some AI Holocaust, right?
And it's aligned with the core values of Web3 as well.
And for me, the core value of Web3 is, of course, decentralization.
Indeed. Thank you for that thorough breakdown. And MetisVM is very intriguing.
And, you know, a lot of people might be familiar with EVM,
but MetisVM is, you know, a virtual machine that's differing,
specifically designed for AI workloads and
AI developers so they can create truly decentralized AI dApps and truly merge blockchain
technology with the emergence of artificial intelligence. Thank you for that thorough breakdown.
The next question, Daniel, for you, can you explain how the parallel execution in Hyperion
differs from traditional sequential transaction processing?
Yeah, sorry. So parallel execution in Hyperion, it basically means having multiple queues of
transactions, right? Execute them concurrently at the same time while ensuring mutual exclusion,
right? You want to make sure that between these transactions there's no conflicts occurring.
All of these are basically above board and done correctly, so to say.
That there will be no, what do you call that, when After a transaction is submitted,
when someone submits a...
Sorry, these days my English is actually getting worse,
I feel.
I can think of the Korean word.
But objections, right?
So there's no objections on chain afterwards.
There's no entity saying this was a fraudulent transaction
or a computation
done on chain.
It's similar to a lot of other chains.
But for us specifically, it's the amount
of parallel executions that you can get
is we're aiming to be some of the industry leading.
And at the very least, the throughput itself
should be some of the most, I would say, exciting
in the industry as a popular large layer, as an incumbent also.
It's also very important for us to be kind of at the forefront rather than chasing some
of the newer chains that are boasting some very high numbers.
Also means in order to scale up, we
can just increase the number of transactions
executed at the same time.
And with the requirements of AI and these high throughput
requiring projects, you can basically
provide them a bit
more freedom in how they design their architecture,
how they design their programs.
And they're able to execute more than just
on a linear fashion.
They're able to get a lot of things done at the same time.
So that's also very exciting for us. In terms of looking at how it's done on a technical basis, I can't give you exact details
because I am a bit more on the marketing and business side.
But overall, the parallel execution is targeting primarily at AI.
So it might actually end up being slightly different
from how it's oftentimes done or structured in the industry.
Because for example, if you're looking at how AI is trained,
you need a lot of data fed into a model
and that in turn changes the model.
But the amount of training that does need to occur
for a model to be fully trained, especially for things like an LLM
or even a social agent, the more robust and the better quality
agent that you're looking for, the more training
you should actually give it.
And to facilitate that, we could have a whole whack
load of parallel transactions done.
And those transactions could be representing
these individual pieces of training
that went into the model.
So if we're trying to facilitate something like that,
then of course we do need this very high performance based,
very flexible parallel execution for Hyperion.
Yeah, definitely the parallel execution approach. It makes
perfect sense for, you know, AI workloads, which
can be computationally intensive heavy workloads.
And, you know, AI operations will definitely benefit dramatically from
this parallelization capability that's being introduced through Hyperion.
Our next question is sticking to the tech of Hyperion.
Natalia already brought it up about the Metis database, the DB, which is effectively
which is effectively going to make it or improve blockchain storage
and how this is going to improve AI workloads.
And so this question for Tom,
if you wanted to go a little bit more into the role that MetaDB plays in improving blockchain
storage and why it's crucial for AI applications.
Yeah, so in terms of how to really explain this in simple terms, I think it's easy to
make simple things complex and complex
things simple. So I'll use the analogy of a computer, right? You have speed, but with speed
itself, so you have a really fast processor. You also need to pair it with something to make it work
efficiently. A high speed processor or a line of data doesn't mean it's always efficient or it is
great use case. You have a car that goes 150 miles per hour but doesn't have any safety features, doesn't
have all the maybe onboard computer information to really keep it stable or usable.
What MetisDB and how it's built into the Metis SDK and the tech stack, you know, that's one
unique proposition that Metis is providing is we're building all this to work tangibly, efficiently, well together for developers, reducing development
costs and time.
So Metis DB is needed in a way where there's many data availability layers, solutions out
there, but what we built is a DB layer that really enhances and expand memory and certain processes of the DA layer.
It will help enhance the blockchain state management.
And it's definitely critical for AI.
With speed, you also need a little bit of bandwidth and efficiency.
And this is where MetisDB really stands out compared to, I would say, kind of the industry,
also builders and projects in that area.
Another good way to really understand what MetisDB is unique is, if you think about RAM and ROM for a
computer, one is to access quick data. And, you know, not all RAM and ROM are the same because we
have evolved in the speed, we have SSD and HDD for storage.
But if you think about MetisDB, we're expanding this to be extremely cost-efficient,
and it works extremely well with our tech stack that we're building out for the infrastructure.
So that's what I would say the best and most simplest way for our audience here to understand in a simple breakdown.
Fascinating. Thanks for breaking that down. And yeah, you know, storage is
often overlooked in blockchain discussions, but it's absolutely
critical, especially for AI and what it's bringing to blockchain
and the necessities or the requirements
and things that developers need.
This is one of the especially important tools
that builders are gonna need
as they build decentralized applications
on a blockchain network like Hyperion.
So moving forward, Daniel,
how does Hyperion enable on-chain AI inference
and machine learning models?
So it all basically goes back into
just being insanely high performance.
The inference and machine learning models,
they all require a large amount of computation, right?
Especially the machine learning,
that's basically just training, you know, a version of AI.
You can think of it that way at least.
For AI inference also,
you don't necessarily need massively high performance
chain to do it, but at the same time,
depending on how many inferences you're getting,
so how many results from the model
that you're getting at a single given time,
it could change what you are able and what you're not
able to put on chain.
So for example, let's say, depending
on how the model is structured, if you ask it
for a mock-up of a data set, right?
Say you ask a weather model, for example, I don't know,
to give you a mock-up of a year's worth of weather data
for a specific list of countries,
then it's going to have to come up with a lot of fake data to support that.
Depending on how it's structured, it could be that, for example, each city is its own
inference, each month is its own inference, or each type of weather. So whether it's cloudy
or not, whether it's raining or not, the temperature, the humidity, all of those things could be
its own separate inference.
It's probably not going to ever be structured that way because that just makes it insanely
complicated.
But that's just an example to make it easy to understand why something may be broken
into more than just one inference for one result that you're asking for.
But at the end of the day, it's probably going to be optimized into having one inference be one data
point on chain.
Regardless, though, if you're able to provide a very high
throughput and low latency network,
you're able to provide all these things on a near instant basis.
Especially for machine learning, you're
able to train it very, very quickly.
Otherwise, you're going to have to wait for how many days,
months to actually properly train something
if you're on a much slower network.
Finality times, transaction times,
how much data you could push through at any given point,
all of these things will come into play
when you're deciding what to actually put on chain and what not to put on chain.
Yeah, and a on-chain, you know, AI inferences, something that projects currently, you know,
struggle to deliver. And so I think this is a really important inclusion
into what is being developed through Hyperion
and what's being offered for builders
who are choosing to come to Metis
and really build on a purpose-built chain
for what they're trying to deliver to the world themselves.
And so, Tom, oh, go ahead.
Sorry, to add to that point, to be fair to the current AI
projects and agents, it wasn't and isn't really
necessary to put a lot of that inference on chain
for those types of projects.
For a lot of the projects that have launched already,
it's not really going to apply to them.
Because let's face it, is it really important to put,
for example, a tweet on chain if you have a social agent?
So if you're looking at a bunch of these use cases
and the type of projects that have already launched,
not the ones that are building, because a lot of them have really interesting use cases
and architectures.
But if you're looking at the ones that have already launched
and the ones that didn't put a lot on chain,
it makes sense for them to not put a lot on chain,
at the very least in how they designed it.
What we're looking for is that new possibilities,
it's that new potential that we can bring
by allowing the freedom to put a lot of these things on chain.
Thanks for that clarity and indeed there is a distinction on what you want to put on chain
and what not, you know, and you want to make the most efficient decisions.
And so moving forward on the next question for Tom, what are some real world AI applications
that could benefit from Hyperion's high performance architecture and all the things we just learned
about?
Yeah, I think Daniel kind of touched on this just lightly.
For me, what I see that what Hyperion can really
bring to this industry, more on the decentralized side,
is basically support in the trading sector.
A lot of trade fire has been looking
into bringing assets on chain.
But due to limitation and speed, there is always a risk.
With having a high throughput chain that can provide that robust,
there's more peace of mind because in traditional finance trading,
timing is everything because the market moves and how do you hold a certain oracle price and which oracle do you use?
So in terms of that, I think the real first use case that we can see in our industry would be the DeFi sector.
From DeFi lending to liquidation side,
because speed in a market scare or a market downturn,
you'll have quick action that could liquidate a position.
And if it's not liquidated fast enough,
then the protocol eats the loss or somebody takes the loss.
So that will help ease it
and bring more traditional finance product
and also tools and audience to this space.
So I think from a real world application,
I think that's one of the strongest and most aligned use case
between blockchain and our chain as well, and AIS itself.
Beyond that, there's a lot of people that utilize optimization
for yield farming and all this effects with the speed itself as well.
High frequency trading and such.
On the social side, you can have a lot of social activities, but it doesn't require a lot of speed.
So from my understanding and what we see is we really see the advantage for Hyperion for the gaming and the trading side. Gaming is still an
evolving area and how much data and how much of the game is on chain is still questionable and
still in discovery mode as we can see with the landscape of the different games in the market.
But from Hyperion, I think it will definitely really benefit the AI sector and trading sector,
especially the AI, since it requires a lot of data
throughput and also even right now when you use you know chat gbt or some of those protocols there
is some lag and some time but if you take it on chain and what you put on chain and what data you
pull on chain as well that could be expanded and really add time to really what you're looking to do.
Daniel gave a really good example earlier in our conversation.
Not sure if you want to share that on the data side or not.
Yeah, sure.
This was actually probably just a few hours ago.
But this is a bit more of a bird's eye view than some specific examples of real world use cases.
So what we're envisioning with Hyperion is kind of comparable, or at least we're taking
Ethereum as kind of a model in how the results might play out.
If you look at Ethereum and its launch of what a smart contract is,
it's completely revolutionized what distributed computing can be, right?
It's completely revolutionized what an immutable ledger could be.
It brought immense amount of use cases, completely new types of businesses,
business models, new business architectures, a new way of defining what value is and how to capture
that value and extract it for the users, how to democratize the value gained from distributed
computing.
It created an immense amount of new use cases where we just didn't even think that this
would be applicable ever. Right?
I mean, who would have thought that, say, crypto punks would have any real world value,
right? But here we are. So that was all possible because now you're able to put a process,
a function, the execution of a program on distributed computing and make it transparent,
immutable, so on and so forth. We're envisioning a similar type of revolution in AI and the AI
industry with the launch of Hyperion. So currently, and we've touched on some of these topics previously. A lot of AI processes is really just not on chain at all.
A lot of the times, it's just the result
of what the AI has done that is actually on chain.
So if you look at deep AI, for example, the trades,
the staking, the leveraging, that type of stuff
is mostly on chain.
Sometimes the payment of the usage of stuff is mostly on chain. Sometimes the payment of the usage of these agents
is on chain.
But really, it's very limited in where
the application of blockchain technology
actually goes into the whole process behind an AI.
AI is basically born from the idea of automating something, having something intelligent enough
to execute on a specific process.
It's based on a data set that trains a model, and that model does work to create the result,
and that result is the one that actually does the work.
So anywhere along this process, we could be putting a lot more on chain.
Let's be honest, the data is never going to be actually put on train, the training data anyways.
If you look at, for example, meta and how much data they use to train their model, it's petabytes, right?
That's never going to be on chain.
But what we can do is we can break it down into different categories or different subsets of data. For example, for a large
language model, you could have historical events, you could have physics, you could have philosophy,
governance, famous books, you could have all of those in separate categories, you could hash it,
you could verify it, and the training of the model with that subset of data could also be verified on chain
through zero knowledge proof and then the changes made to the model can be tracked and verified on
chain and then through that you could also track the evolution of the model on chain. Let's face
it for something like social agent that's not really applicable in terms of is this valuable.
But if you're looking at something
with a bit more of a real world application,
so let's say governance.
Let's say we create an AI agent to help us govern
the governance of Metis or something
that has to do with public goods, or anything
where the users, the end users, and the welfare of the end
users are very important in that whole process.
It could provide immense value in those areas.
And that's just a very simple low-hanging fruit
in terms of concept. It's conceptually a low-hanging fruit in terms of concept.
It's conceptually a low-hanging fruit.
To build it, I'm sure it's immensely complex.
But conceptually, it's something that you could easily
apply if you are given the freedom to put anything
you want on chain.
And that's just the start of it, right?
Similar to how, say, staking or an escrow or a social
blockchain application, something like that could, or trading, sorry, something like those
things.
Those are conceptually the low hanging fruit of what smart contracts can do.
And the other applications are, you know,
still being discovered.
And I would say that's the same thing for AI and blockchain.
Definitely.
We are truly in a moment of discovery,
which makes it very exciting.
This is probably maybe the most exciting time
in tech history because of these emerging technologies
of blockchain, artificial intelligence, and the global accessibility of digital technologies
now which continues to grow.
And I'm excited to see what compelling use cases develop on Hyperion and just with AI
in general.
And one of the things that you all were talking about
right now, I think that is really exciting
that people don't really think about
is how we're creating entirely new value systems
that is fundamentally re-imagining what's possible
with the world and how we trade information,
how we trade data and how we can make it empower users
and builders who are the ones developing this data
or giving this data to right now to big tech
and things like that.
And they're the ones creating these AI models
and then selling it back to us.
So I'm excited to see all that emerge and see how all this is developing.
And I'm personally very thrilled that Metis is taking this leap or taking this
path forward in this, in this way with artificial intelligence and blockchain.
And, you know, I can't wait to see what kind of deep AI use cases develop and end in all the other sectors
as well.
So going on to, we're getting near the end.
We have about eight minutes left.
I think we have like two or three questions left, but just a quick time check for us.
But this one for Natalia, we haven't heard from you in a few minutes.
How do you see Hyperion shaping the future of AI-driven blockchain applications?
I feel like, you know, Tom and Daniel in the recent couple of points answered a lot of it
in terms of reflection on our vision,
I just maybe will summarize it a little, right?
So just what makes Hyperion a next generation platform
for AI, right?
So essentially we're addressing it with architecture,
with its high performance architecture,
with the real timetime transaction processing,
and really optimizing infrastructure for AI applications.
So essentially, we're aiming to address blockchain's traditional issues such as latency, scalability,
and some computational bottlenecks. And by doing that, we will enable a new wave of use cases
and decentralized AI applications that follow from these new cases.
And that will span across multiple industries.
Our vision is really to make Hyperion to become a backbone of decentralized AI.
But this is our vision.
This is our approach.
But it's up to you, our community,
our supporters, to our developers to jump on that and really build it out and show what the new world with decentralized AI can look like.
And hopefully we can provide the platform that you deserve to do that.
Yes, yes, yes. And there will be competitive advantages offered for developers that come to Hyperion who want
to build specific purpose-built applications, whether it's deep AI or AI-Fi or social AI-Fi
or whatever else emerges in these coming times that is going to be exciting.
And there's things we haven't even thought of yet that are going to be coming and Hyperion will be built for you to build
that.
So, Daniel, back to you.
What's next for Metis Hyperion in terms of roadmap and adoption strategies?
This is very simple and very exciting because this is probably one of the first times
I've seen a tech team actually be on track.
We have the testnet coming up probably around late April,
might be pushed a little bit into mid-May,
but we're on target for late April.
We have the mainnet coming up in probably around August.
And those two things together is,
I'm very surprised that they're able to build something
so complex as quickly as they did.
With the testnet and the mainnets,
we'll have a lot of campaigns for everyone to take part in
from users to builders,
whoever you are will probably have a place for you to take part in, from users to builders, wherever you are will probably have a place for you
to partake in and drive some of the excitement
and the adoption for Hyperion itself.
And of course, with Laz.ai.
By the way, I don't think I mentioned this previously.
I am also the head of marketing for Laz.ai.
The reason is because we are very, very closely incubating
Las AI and we absolutely see the value in Las AI's
contribution to the AI industry itself, right?
They're solving a very core problem to the AI industry
and that is the data alignment problem.
I'll be super quick about it,
but the data alignment problem is basically
how the data itself aligns
to the intent of the model when you're trading that model.
So if you have good data, the better results and more aligned
to human values and the intent of the creator, the model
and agent will be.
Obviously, the more robust and more complex the model becomes,
the more difficult it is to get the data to fully align.
And of course, as the data set gets larger and larger, immensely and exponentially more
difficult to actually get this aligned.
Regardless, we see the value that Las.ai is providing.
And like we've been saying, Regenesis is a two pronged approach. We're
providing for AI's needs through Hyperion and we're solving their problems through LAS
AI. And in doing so, we're sharing what team members and expertise we can between the two
teams. So I am also the head of marketing for LAS AI. Having that said, when we're talking about roadmap
and milestones for Hyperion,
we do need to talk about Laz.ai as well
because Laz.ai is also launching their testnet
and their mainnet this year.
So we will try and see how much collaboration we can do
because for example, if we have a hackathon
that combines both Hyperion and Aleth, the
agent framework created by Laz.ai, that creates for a lot of exciting opportunities, right?
And there's a lot of value add and synergies between the two companies, and we're excited to
see those progress. That actually leads into the adoption strategies as well.
progress. That actually leads into the adoption strategies as well. I can't give you specific
tactics right now, but overall the strategy and the points that we want to hit is that because
Hyperion itself will be kind of the go-to home for AI, right? Regardless of how big we are and how big our competitors are,
it's pretty much going to be irrefutable
that Hyperion is one of the top homes for AI
if we deliver on what we're saying we're going to deliver.
Add to that the flexibility and the immense amount of use
cases that could come out of LAS AI.
The adoption strategies just kind of play out by itself.
Immediately, you can think of an immense number of use cases
and adoption strategies concerning AI agents.
We have these things called the IDAOs in LAS AI,
which redefines the scope of what a DAO is.
But essentially, it boils down into people coming together to manage
and own and democratize AI assets.
So that community can become immensely thriving the more we connect to different partners.
And that translates directly into value add for Hyper, because Laz.ai is initially going to use Hyperion
as the network to settle transactions,
to push data onto Ethereum, and confirmation, sorry,
onto Ethereum.
And the interplay between these two companies
is something that I'm very, very excited to see coming up.
For sure.
Thanks for sharing about LAS AI, which is important.
And I know we'll probably have spaces
that go a little bit deeper into the relationship between MEDIS
and LAS AI and why it's important.
But thank you for giving that overview.
LAS AI is really important, especially for a problem that
is not, you really talked about a lot, which
is the data alignment issue
and creating models that are bias resistant, context aware,
especially if you're creating localized AI models
or you wanna create personalized AI agents or models,
it's important with the data and the context
that you provide it.
And even ethics, ethics change
between communities and individuals.
So even that, you know, you want to make sure that it
pinpoints exactly what you're trying to build.
So thanks for bringing that up.
Uh, now we're going to get into our final community questions.
Uh, we are a minute over time.
I just want to remind folks after this space, there will be a space being held by Metis Foundation,
I just want to remind folks after this space, uh, we, there will be a space
which is about to DeFi space, and the link is on the jumbotron up there. If you go on there,
you'll find it. But let's finish it off. What are the next big milestones after Testnet,
and which DeFi protocols plan to launch on Hyperion if there's any at the moment?
All right, so I'll be super quick about it
because we are over time and we wanna make sure
that the foundation space gets all the loves
that it deserves.
But overall, big milestones is basically going to be
the campaigns that we're going to launch. We're very excited
to see what kind of use cases, what kind of adoption we can see. We're super excited to
actually stress test the Hyperion network because we're touting it to be high performance.
We're saying that is the core value, one of the core values that it's actually providing
the AI industry and how we're attacking the AI
vertical. So we're very excited to see those things come up. The case studies and the analysis
of those, which we will publish in due time, that's also going to be super exciting. And
the launch of LazeAI. And I'm very excited to see the first IDAUs pop up, the first AI assets being managed,
and the first agents actually starting to gain popularity
through LAS AI.
Those are going to be super, super exciting.
In terms of tech, that is incredibly difficult
for me to talk about.
Overall, I would say after Testnet, obviously Mainnet.
Mainnet is coming up this year.
That is also very, very exciting,
not just because it's basically the first real start
of Hyperion, but also because that's
the culmination of basically all of Metis for how many months
now over a year? Well, close to a year if you think
about the initial conception and then the architecture and all that. So super exciting.
The milestones will be more defined and will be a lot more communicative going forward about these
things. And I'm excited to bring you any
announcements, any exciting news regarding our progress for Regenesis. In terms of which DeFi
protocols plan to launch on Hyperion, that specifically, I would say instead of telling
you which protocols we're speaking to currently, I would like to kind of give a bit of an insight on what kind
of DeFi protocols we're looking at. So if you're looking at regular DeFi protocols,
the Hyperion network itself with its strengths doesn't directly speak to regular DeFi protocols.
So for example, let's say we're talking about a DEX. Yes, absolutely. A DEX, a usage of DEX can be improved
by having a high performance network.
If you're, for example, trading on Uniswap,
yeah, sometimes the actual transaction takes a while.
There's slippage sometimes, you can get sniped.
And the experience will absolutely, yes,
be improved by having a faster network.
But that's not actually very crucial for the functioning
of Uniswap.
Because a lot of people, they don't really
care that much that you get a 0.2% slippage in your trade.
They don't really care that much that you
had to wait an extra 15 seconds to actually get
the token in your wallet.
So what are we actually looking at?
This kind of goes back into what Tom was saying.
And I'm just realizing now that actually all three of us have, I think, high finance experience.
Yes, Tom was, I think he has experience in trading and in Italia has investment banking.
I was from equity research.
So if you're looking at the different types of traditional finance companies that are
available and if you're looking at the DeFi applications of such, then one immediate thing
that comes to mind is obviously high frequency trading.
We're not seeing much high frequency trading on blockchain
because there's just not many networks that actually
has high frequency, that also has the liquidity
or the access to all these different tokens
available to make high frequency trading make sense.
High frequency trading is actually
a massive, massive industry in the traditional finance world,
they trade a ridiculous, immense amount of value every day,
every second.
To give you a bit of an idea on how important
they find the speed of their network,
there was a fund that spent, I think, tens of millions of dollars to move a bit closer
to the exchange and to lay fiber optic cables so that they can improve their internet speeds
by, I think, a few hundred milliseconds overall.
A few hundred milliseconds is actually quite a long time if you're waiting, but I think
for per trade, they were looking at something like, you know, in the tens, maybe even less milliseconds of improvement
by spending this tens of millions of dollars, right? So that's how much the high frequency
trading world values the speed of the network. So obviously with a very high performance network
such as Hyperion, we could be home
to high-frequency trading.
If you look at, for example, the Metis SDK
and its capabilities, we'll put out some blog pieces
and some exciting AMAs about this as well.
But if you look at the Metis SDK,
it has the capability to pull in liquidity from elsewhere.
That's EVM compatible.
So basically, what we're saying is
that it can be an all-to-one source of liquidity.
That means pulling this liquidity
from all these other layer 1s or layer 2s that
are EVM compatible, sorry, within the Ethereum ecosystem.
And we can bring it all to be interacted with
on the Hyperion network.
So that immediately gives you access to this much, much more
massive market.
And with the speed at which you can execute these trades,
it completely opens the door.
It might not be on the
millisecond type
Speeds that you're looking at in the traditional world, but it does provide something massive. That's not really
That Common in the blockchain world. Yes, there's absolutely there's very very fast networks
I understand that but you know at the very least for Hyperion, that's one thing that we can look at.
And then of course, there's going to need to be some very basic types of DeFi protocols,
but some of the foundational stuff.
And we're still kind of in the final stages of making sure that that whole DeFi ecosystem on Hyperion
makes sense and that we're getting the key pieces covered
for Mainnet.
For Testnet, we're not sure who's going to launch yet.
We're still speaking with them.
And because Testnet is, again, primarily for testing,
we're making sure that for Mainnet, at the very least,
we have all of them on board.
Amazing. What an awesome space we had. And that's what happens thing to know about the possibility of that becoming
part of Hyperion, or at least it's possible with the technology and high frequency trading
not being a thing in blockchain or Web3 at the moment because of the different issues,
whether it's liquidity or technical capabilities, whatever it may be.
So that's really interesting to know.
Thanks for sharing that.
This has been a really great space.
We did go a bit over time,
but thank you everyone for joining us.
I wanna especially thank Natalia, Tom, Daniel
for sharing all the insight you did.
I have a lot more questions,
especially just based off of the last things you said, Daniel,
but I know we'll have more time and more spaces like this
to cover those types of things.
So definitely turn your notifications on for MEDIS.
So you can always be an alert and know
when we are sharing information or dropping by
or having times to engage with the community.
Maybe next time we'll open up for some actual questions
for folks to come on stage.
I did see a couple of requests to come on stage. We didn't have time for that today. We had a lot to cover. But again, just thank
you everyone. Thank you to our guests and we will see you all next time. Go check out
the Metis Foundation space right now. See you.
Also quick last words. So I know that's Metis overall in terms of community engagement, in my eyes, we could do better.
We can always do better. Anybody can do better. But what I'm wanting is going forward, we'll have a lot more of these on a regular basis.
And because we'll have it a lot more frequently, we'll have a lot more engagement as well. So like Anthony said, we'll be able to invite you to speak.
We'll answer questions through text.
We'll answer questions that we've gathered through Twitter
or wherever.
We're just very excited to be a lot more engaging with you.
The marketing itself will be a lot more varied.
It's going to be in a lot more channels.
It's just going to be a lot more exciting
to be part of the Metis journey through Regenesis.
So I very, very much look forward
to sticking with you through 2025,
whatever that brings in terms of economy
and the crypto space.
But yeah, I'm just super excited to be here
and I very much look forward to connecting
with each and every one of you.
Thank you for your time.
Yes, exciting. Thank you, Daniel. Thank you for sharing that. Definitely going to be
even more engaging with the community and I can't wait. I'm excited. Let's do it.
Thank you all for being here again. Thank you.
