Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
Hyperion Tech AMA
Recorded: April 15, 2025 Duration: 1:06:27Space Recording
Short Summary
Hyperion is set to launch groundbreaking features including real-time on-chain AI agents and parallel execution, enhancing the Metis ecosystem's capabilities. The team is actively engaging the community with upcoming news, while exploring innovative funding and yield opportunities through restaking and liquid staking derivatives.
Full Transcription
Thank you. Thank you. Thank you. Hello, hello.
GM, GM.
Amazing.
I hear you loud and clear.
You sounded a bit better after.
No double sound this time.
Yeah, exactly.
Cool. Let's let some more people join in.
How was your day, Pavel?
It was good.
Eventful. I went outside.
That's a first.
Enjoyed the sun. You've got to always be enjoying the sun sun touching the grass or sand
it's always uh yeah very good for your health i'm assuming you're touching more sand than grass
these days and like more waves you are touching yeah yeah i'm um i'm uh undergoing a bit of a move and um
i guess it it uh it definitely has a uh has an impact uh from from i guess, Canada to Portugal.
That's definitely a big shift.
Yeah.
Warmer here than anywhere.
Well, actually, no, there's definitely warmer places,
but compared to Canada, I think a lot of places are warmer.
I'm assuming maybe Canada still has some snow going on.
I know some areas probably have some snow. So it's good that you got out now.
And probably, yeah, the Portugal weather is so much nicer.
But everywhere, I think in a lot of places right now,
there is a little bit of a cold rebound of the weather.
It has been hot for the last month
and then started to get cold again
just at the beginning of April.
In a lot of places,
I was in different countries
last week
and it's all the same,
always very cold.
Yeah, yeah.
It's definitely a little bit of a cold front as well.
So it can't escape it no matter where you are.
Always, always changing.
Yeah.
So enjoying it while I can, but then again, you know,
summer's around the corner.
Hopefully everyone here is also going out, touching grass. I think that's good for you, you know summer's around the corner hopefully everyone here is also going out touching grass
i think that's uh good for you you know rather than uh being being in online and uh
reviewing documentation and all of that like uh like like we are.
Awesome.
Well, I guess ready when you are, Rami,
if you want to kick things off.
Or if we want, we can wait for a few more people.
Give me a second.
I'm switching.
Yeah, I'm just switching devices right now.
Just give me a sec, and then we'll start.
Of course. Thank you. All right, waiting for Rami.
I think we have,
what is it, a hand raised from Metis.
I don't know if that's
intentional
or if you have a question.
But if not, definitely let me know, I guess.
Otherwise, we'll get right into it.
Oh, Rami messaged me saying,
bit of a technical issue. Okay. I guess it's just me.
Looks like he'll be rejoining very shortly.
I guess this spoke too soon regarding the technical issues.
I remember our last AMA,
Rami had the voice of God.
I was hearing doubles
from his side.
Oh, all right. He's asking from his side so oh alright
he's asking
okay
let me see if I can
alright
I guess All right.
I guess, Shada, can you set Rami as co-host?
I think he rejoined just now.
Hello, hello, everyone. Yes, just did it. Remy, can you try again? He says he's unable.
Maybe there's a different option for him.
He requested to speak, but I don't think I can give him access.
I am but a speaker myself.
I just approved it.
Oh, okay.
Right now.
Okay.
Rami?
Rami?
Yes.
Amazing.
Okay, perfect.
Right.
Yeah, just like always.
All right, cool, cool.
Welcome, everyone.
Welcome, welcome.
We have more people joining.
I think we can kick things off.
I think we can kick things off.
We wanted to discuss a little bit about real-time on-chain AI agents, parallel execution, and maybe a little bit about early builder incentives on Hyperion.
We will have some very exciting news in the next few days
So stay tuned to our socials stay tuned to our Twitter. I think like
We are getting closer getting closer and closer to
to trying out the the long-weighted new high-speed hyper
Parallel execution chain finally
amazing by the way yes we will be we'll be taking some questions uh so can uh do we have a chat here uh i think there's a chat uh there should be like a chat
button here cool and also also like uh you can uh you can request to speak and uh if you have a question and of course like we will answer the
question but yeah first few things I'm gonna just wrap up like what we
discussed about last last AMA we we we spoke a little bit about, we answered some questions about how Hyperion and Andromeda
are going to be coexisting,
how this is an expansion
of the MERS ecosystem.
And it's not like a replacement
of Andromeda.
Like, I think some people
misunderstood it for like
a new chain,
then it is a replacement.
But, uh, we, we explained that last time that it's, uh, it's not what, uh, if
some of you were not here last time, uh, we would be happy to go, uh, through
it a little bit more and explain like, if there are still some doubts, but, uh,
yeah, like, uh, Hyperion is going to be an addition
to the Metis ecosystem.
And, uh, like some of the sneak peeks that I can give here are that it's gonna, it's
gonna have a very fast communication layer with Metis itself with Andromeda and like,
uh, the whole Metis ecosystem system because it's going to be built
using the metas sdk which has a novel bridge layer so uh yeah pretty soon i think
uh we'll be able to send our first transactions there and And it would be like from probably Metisipolia.
You will see how lightning fast that is.
Pavel, maybe you want to share a little bit
about what we spoke on AI agents last time
and like what we can cover this time,
what we can like,
what areas that can be explored,
what areas can be achieved using Hyperion
and cannot be achieved using sequential chains?
Yeah, absolutely.
So I guess some of the features that we talked about regarding Hyperion,
you know, what it can do differently from others is, I guess, the first and foremost is
parallelization. So allowing for like the ability to execute multiple different transactions all in one go, essentially.
And then it allows for like a lot of interactions, essentially things specifically tailored for speed.
things specifically tailored for speed.
And with that, you know, we have that SDK coming along.
So the Metis SDK alongside it, it has, it's written in Rust.
So there's like tons of that uh rust is also very you know high performance based on
speed uh centered around this this this high throughputness as well um as well the the ability
to uh i i don't think it'll be released in this current,
let's say when it's the initial release,
but it'll have this ability to vertically scale,
so allowing for tailored deployments
so you can build other, let's say, layer threes
on top of the architecture,
allowing for maybe more custom chains layer threes on top of the, the, the architecture,
allowing for like maybe more custom, custom chains built for,
you know, specific purposes for AI agents.
And then the, the idea is that it's,
it's unified within the Metis ecosystem. So there's a,
there's that bridge, there's account of abstraction and the, the,
the idea is that we want to simplify these cross chain
interactions as well. So, you know, we, we, we want to make it
compatible with specifically a Metis Andromeda.
So this is the, the, the current, you know, Metisis ecosystem that you know, or the Metis chain that you know.
We want to make it super compatible with it.
Again, like the gas token as well is Metis itself.
So like the idea is that it's like an upgraded version
that will allow for more applications
specifically centered around the AI based use cases.
And again, for high throughput and, you know, minimal,
let's say, minimal complexity
surrounding the
interactions between the two.
Yeah, yeah.
It's going to be pretty seamless.
this is
what would make the
Metis X system larger
and larger, to have more chains
that are seamless to use with the metis uh
andromeda uh this is this is more like an expansion like we said and um uh like the the most i i think
one of them like for me one of the most important things here is that uh like a big chunk of the code is being rewritten in Rust or like written in Rust.
And this, I don't know, like what it means essentially if you're like not a developer,
it means that it's going to be like, you know how fast Andromeda is?
It's going to be even much faster and much more scalable because it's a very fast programming
language.
And yeah, these are some of the highlights.
Again, this allows for unique use cases that needed some adaption.
We needed to adapt to the AI agents,
we needed to adapt to GameFi,
so that's why the expansion comes, yeah.
Amazing. amazing so i guess um i have i guess a few few of the questions that have prepared but um
but i guess do you want to do you want to go off of those
yeah sure let's do it awesome so So I guess the first question is the creation, the overall, like why Hyperion?
Why was that the focus? Why do we need a new L22 for this why can't we just build this using like
andromeda what we currently have that is a good question and the like the the reason why the reason
why is that like we've seen we've seen earlier this this year and like at some point in last year uh sometimes we're hitting a scalability limit
uh with andromeda alone so this has called for an expansion of the x system and um
what uh what this means for users uh of course like the expansion uh here
Of course, the expansion here, we are trying to make it come with the most seamless experience
so that if you're using Hyperion, it will feel like you're using Metis Andromeda.
You'll be able to access Metis Andromeda apps seamlessly.
able to access medicine andromeda apps seamlessly,
you'll be able to interact with, for example, Hercules
from Hyperion.
If someone will be allowing these abilities,
there will be such features.
But of course, someone needs to build on top
of this interoperability layer.
For example, if Hercules wants to deploy on Hyperion,
it wouldn't mean that they're going to fracture the liquidity.
They can just deploy a bridge that bridges the messages from Hyperion into Andromeda and like there is going to be more there is going to
be an SDK to build this easily in a way that so a Hercules user would be able to use Hercules from
Hyperion even though liquidity is on Andromeda so this is what I mean by the expansion. This is what I mean. Like, we'll be able to fit more users.
We'll be able to fit more use cases on the MetisX system,
keeping it one, keeping the whole MetisX system one,
while almost categorizing the chains based on, like,
decentralization, based on scalability uh fast processing and and all that yeah i um
we're actually like very happy to answer any uh question related to the expansion uh so uh if if
anyone has like some follow-up on this question please raise your hand and uh we can answer
up on this question please raise your hand and uh we can answer
yeah yeah and i guess like just to continue off of that that point like well well if again like
anytime if anyone has a question uh related about anything that we're talking about please feel free
to raise your hand request to speak we can uh we can get you in. But I guess like the enabling new use cases part.
So I know, I know the, like, you mentioned how like, you know, that direct link between
Andromeda and Hyperion will exist, like you can make interactions between that. But I guess like,
interactions between that.
But I guess like on the side of Hyperion,
one of the one of one of the key like purposes,
like why we're going in this direction
is to create the ability to support more of these like
AI based applications.
So we want to do like model inferences
um integrating like uh ai co-processors and uh and pre-compiles directly uh part of the blockchain
itself so right now like whenever you use like an AI agent, you have to, uh, essentially you execute everything off chain. And as a result,
it's really hard to verify that execution when everything is off chain.
Now there are, there are ways that you can bring this execution on chain,
um, you know, using like ZK proofs, uh,
using like trusted execution environments.
So there's like really, really smart, like ways that you can bring it on top.
But those have like huge like performance issues, right?
So the, I guess the core of it is how can we create a way to have native AI applications
to be built on top of a blockchain. And that's,
I guess, our goal with all this is to enable the ability to execute these AI operations,
inferences, and anything related to that native on the blockchain without,
um, needing to have these, like, let's say, well, I guess, I guess like in the future,
you know, we will have like additional, let's say ways to bring, bring, uh, this,
this on chain, but having like native ability to, to run this AI inferences on-chain.
And as a result, yeah, it's verified and on-chain
without needing to do any ZK or TEE style executions.
Yeah, yeah.
And also allowing, I mean,
Andromeda users are going to be automatically
Metis ecosystem users since they hold the gas token.
So they will have access also to these AI agents,
AI dApps, decentralized AI uh and all that so uh we we are we are bringing some uh new use cases here
we have a request to speak from Austin so Austin I will give you the the mic I just allowed you in
uh yeah if you have a question, go ahead.
Unmute yourself.
Sir?
Hey, good morning.
Sorry, it's morning for me.
You know, I just had a quick question.
Y'all had mentioned that the gas token for the Hyperion is going to be METIS. I was wondering if y'all had considered the possibility
of actually making that a liquid stake derivative
of the METIS token
instead of just the standard token itself.
There is going to be,
this is like some alpha that I'm going to say right now,
but there's going to be some restaking opportunities
on the rest of the METIS ecosystem.
So currently there is there is liquid staking on.
On Andromeda, right?
But having the METIS token as the as the gas token for the whole METIS system, this will allow restaking over over whole METASX system.
So there will be opportunities for this if someone builds it.
So this is also a great opportunity for builders who want to build restaking or liquid staking.
This is one of the use cases I guess
that can be also
built it's not
going to be native I don't think it's going to be native
like it's not going to be
natively supported but
this will
be this will be like
I think encouraged
to build
yeah yeah like the definitely like looking like, looking at these, like, liquid,
liquid restaking tokens, like, those types of models.
But I guess also with the launch of Hyperion, there will be native paymasters.
So in this case, the, like, we want the ability for people to come in and onboard into the Metis ecosystem to be seamless.
The idea that the user experience is in order to get onboarded into crypto, you need to hold on to a set token in order to interact with the chain itself.
So we wanna make the experience of onboarding new users
very simple and straightforward.
So the idea is we want to, again,
like you still need to have like a sort of gas token
in order to facilitate transactions. But the the ability to facilitate the ease of onboarding for new users using these like
Paymaster based systems will be prevalent. And I can imagine that with the integration of like LST
And I can imagine that with the integration of like LSP mechanisms, you can have like, let's say you are an adapt and you want to help, let's say, onboard users into your specific application.
You might have like a set limit on like, you know, what, how much you want to, you know, users to come in and onboard.
Obviously, there has to be some amount that you have to pay for
to get that user to be onboarded into the ecosystem.
So having some amount of Metis stored,
you can potentially see ways that you can restake
that amount of Metis and then earn yield off of it
and then have essentially the ability to pay for users
using the yield earned off of that restaked Metis.
So the goal is to help facilitate the onboarding uh experience within
the ecosystem and uh we have this set by by you know having having the the pay masters and also
the ability to earn yield off of off of off of the Meta through these restaking functionality.
Hopefully that makes sense,
but let me know, I can maybe like try to explain it further.
No, it does, it makes perfect sense.
But I just wanted to get some clarification
on the Paymaster services that you were talking about.
So this will essentially, obviously, all the transactions
are going to be conducted in Metas, correct? But the pay master services essentially leave the
end user feeling like they're paying for the gas transaction in whatever, say, USDC, stuff like
that, correct? Yeah, yeah. So it's gas abstracted, but also the DAP has the ability to pay for gas on behalf of the user. I guess like we, I created Chorus. We the idea of making it available
for other applications is the whole ecosystem-wide initiative
for this Paymaster.
But the way that, let's say you're using a Chorus,
whenever you make a submission,
that's an on-chain interaction.
And so actually the campaign creator
would be able to fund the wallet in which acts as a paymaster.
So the idea is that you don't have to be in the Metis ecosystem to start using Metis-based
applications.
And in this way, making these on-chain-based interactions,
you don't have to hold any form of token
to be onboarded into the ecosystem.
But of course, we've seen Arbitrum have,
let's say, native USDC being able to be used
as payments across the ecosystem. or sorry, base, base,
sorry, base is the one that, that, that, that did that. But again, we're, we're looking more on like,
how can we do this without requiring any form of token as well to, to facilitate that onboarding?
to facilitate that onboarding.
Great, great.
I'm sorry, I know I keep asking questions
and this isn't all about me, but-
No, no, go ahead, go ahead.
I just wanted to get some further details
about the interoperability,
going back to Ethereum between the ecosystem.
Yeah, I guess, uh these the ecosystem yeah i guess um what what what part of the interoperability i guess like the the settlement is still done on ethereum so uh in this way like like um it's still this this
layer two um i guess what what part of that interoperability would you like?
I guess, okay, so I guess for one example is,
where exactly is the data availability?
Oh, yeah.
Yeah, yeah.
So, oh, sorry, Rami, you can answer that.
So the data availability is, at least for MERS and Ronda,
is going to be pretty soon.
I don't know if you read the latest upcoming upgrade,
which is going to come quite soon.
The data availability layer is going to be moved from MemoLabs
availability layer is going to be moved from memo labs to uh Did I cut out?
to...
No, I think he did
because I was thinking the same thing for myself.
Yeah, I think he's not frimy.
Oh, no.
Okay.
Getting into the connection issue yeah yeah uh i think uh definitely you know rami
rami travels a lot so definitely the internet is is a little bit there but but i think i i think
i'm gonna go off of what what rami was saying is is that we uh he recently posted a proposal
what Rami was saying is that he recently posted a proposal on the Metis governance forum.
And essentially that proposal was to bring data availability to Ethereum layer one. So currently
Currently right now, Metis has the data availability off-chain.
So with the introduction of,
there was a testnet upgrade for,
there was a testnet upgrade that essentially had the ability
to make fraud proofs
to then execute on chain,
essentially slashing actions directly on layer one.
And so with that, I believe we would be technically a stage
zero rollup by
L2Beats standards
but I can see about
providing
the forum post
that Rami was
just getting into mentioning
but the idea is that
this upgrade
took a little bit of time
to implement for Andromeda because Andromeda is
let's say it uses an older version of the Optimism Virtual Machine. So it needed to have these like specific upgrades tailored to the
the tailored to the OVM part. And after that release on testnet, that release will happen on mainnet after that testing period.
And the Hyperion will likely inherit those changes as well.
So data availability will be on the L1, or at least that's the plan, is to have data availability be on the L1, or at least that's the plan
is to have data availability be on the L1
so that you can have the safety and security
of being that L2, like a true L2 in that case.
I'll do in that case.
Robbie, are you, are you, are you back?
Robby, are you back?
Oh no, I can't hear you.
I don't think I can hear Rami.
Yeah.
I can't hear you either.
Yeah.
Let me, uh, I'll, I'll link the link the actual upgrade or like the proposed changes. But hopefully,
everything that I've said is accurate to that degree. And hopefully that answered your question.
And hopefully that answered your question.
Austin.
Yeah, it does.
This has been very insightful.
You know, I, you know, a few months back, you know, a lot of the Ethereum Foundation and, you know, Justin and himself, you know, they were hosting those those YouTube videos about, you know, based roll ups and stuff.
and stuff. And I was sitting there listening to them talk about decentralized sequences.
And I was sitting there listening to them talk about decentralized sequencers.
And I was thinking to myself, like, oh, my gosh, you know, Metis Andromeda has been doing this for
ages. And I also thought to myself, wait a minute, where is Metis? Why aren't they a part of these
conversations? Yeah, yeah, I've actually, um, uh, recently, I recently, I do podcasts as well, doing like the House of ZK.
And we actually hosted Justin with that. I'm part of a separate project as well, like ZKM.
I guess doing less work on that front. But definitely definitely like, I think, you know, the, the,
the, the progress, um, that we're making, um, on the meta side, I think really tightly correlates
to what the vision for Ethereum is. And, and, and the goal of it is to have this decentralized and trustless way to make sure to have everything that makes us a true roll-up in terms of that
direction.
And yes, decentralized sequencing is part of that.
I know a lot of roll-ups out there don't have decentralized sequencing is part of that. I know like a lot of rollups out there
don't have decentralized sequencing.
And that is due to, you know,
the sequencer revenue coming in from a single sequencer
is much larger than it is for having like multiple
sequencers interact.
But at the same time, you have that single point of a failure.
And I think like, we don't really have the,
like if a sequencer shuts down, let's say for any L2,
there's no, let's say ability for regular users to come in and do anything,
really. It's like, if you're highly technical, if you're research oriented, and you know exactly
how to, you know, what contract interactions to do or everything, then yes, it is 100% possible to make those on-chain interactions to,
let's say, swap some tokens.
But we want to make sure that the way that the users are interacting with the network doesn't deviate from the, like, from, you know, the worst case scenario, in this case, like, the sequencer going down.
If, in the case of Metis, again, if one sequencer goes down, we have the ability to come in and have that sequencer be replaced by a different sequencer.
And as a result, you don't have to do any of the, let's say, those interactions,
like research-focused interactions for, I wouldn't say research-focused,
for uh i wouldn't say research focus but like uh you have to specifically know how to um
how to move your money out um as as as those utilities are not let's say easily accessible by
regular users and that's something that that we want to provide the ability
where anybody that comes in and builds inside the Meta's ecosystem,
they have that permanence in which even if the sequencer goes down,
even if something happens,
everyone still has the ability to interact with the network and, you know, not need specialized knowledge in order to make those interactions.
I guess, yeah, and we're seeing this model be adopted in like for the decentralized sequencers inside of, for example, the prover architecture in like ZK proofs,
where we have like networks of different people, networks of individual provers being essentially,
provers being essentially, there's multiple, you only need technically one prover to generate a
proof to like facilitate the on-chain execution. And technically you can be a prover, but it requires a lot of hardware resources that, like, for example, 256 gigs of RAM at the minimum in order to generate a proof.
And that's something that not a lot of users have, both the technical capabilities as well as the resources in order to do that, to do that on-chain based computation.
So there's, there's, we have like these prover networks that are coming into play, which
is essentially, which is essentially a form of like a, a redundancy to make it so that
anybody can come in and generate a proof without requiring highly specialized technical knowledge in order to set up a prover to then be able to execute that.
So our goal with all of this is to help reduce the level of complexity for the person coming into the blockchain-based ecosystem
and definitely going towards the AI approach because I think that's what most people going
into and now programming or starting to learn programming are now going into AI and having all of these aspects,
all of the complexities of blockchain abstracted to this degree where now you don't have to trust
the network. You can just deploy and interact as you would and still have those safety guarantees
in place. Sorry if I went on a long rant,
but hopefully, you know,
that whole like idea,
like the general idea of these,
like why we're doing this and why we're following in this direction
and actually why we're deviating away
from, you know,
what the industry standard is,
which is again,
like the centralized sequencer model is, you know, the industry standard is which is again like this centralized sequencer
model is you know because we we want to achieve that that's uh the simplicity and still have that
that same level of uh let's say trustlessness and abstract that complexity away from
the individual user or project looking to build on the ecosystem.
Thank you, Pavel, for the great insights. Good news is Ramy is back. I can hear Rami.
Rami, I have you as a speaker okay he's uh he's reconnecting sorry guys about about the the difficulties
rami definitely has a lot a say. Okay, he's back.
Okay.
I still can't hear you if you're speaking.
Remy, we see you, your status as a speaker.
Yeah, like we see that you're unmuted.
But, yeah. Yeah. yeah
I uh
I can maybe keep ranting
about uh about these things
but but yeah
oh yeah Austin if you have any
more questions by the way
feel free.
Yes, I do. I just didn't want to, you know, hug up the space.
While you were talking, it actually brought up two more questions or two more.
Well, one question, one thought. And, you know, we were talking about AI, you know, essentially automating, you know, all of these processes, you know, not only within the local ecosystem, like on a singular blockchain, but, you know, even when we think further out on like a bigger level and, you know,
AI essentially automating, you know, transactions from one L2 back to Ethereum and on to another L2,
I think that's one of the biggest value propositions behind the ZKs is that, you know,
there's a lot less trust required and stuff like that.
And, you know, my concern about that whole AI automated process is the fact that, you
know, how do we make sure that these AI agents aren't, you know, routing through applications
or protocols that we may not necessarily trust if we were doing it manually.
And I think that ZK really eliminates a lot of those concerns inherently.
Absolutely.
Like the idea of right now, AI agents and blockchain working together is it's a very,
let's say, highly researched subject.
Even ZKML is highly in the research stage.
What parts can be optimized to provide a similar user experience
to how you interact with AI right now.
In this way, if you type in to ChatGPT or anything,
you get an immediate response.
And you get an immediate response
because you have tons of graphics cards essentially processing the tokens, which is essentially the response that you get from the AI.
Now, who runs those GPUs?
Like, how do you make sure that they're not, responses and altered? How do you how do you do that? So there's multiple different ways that you can, you can make sure that that that that pieces is secure. The one way that you can do it is that you can wrap that into like a ZK based application. So you can make sure that the program that is executed in this way is executed
and guaranteed that that execution was done in this way.
And you can do that with ZK, but that involves a lot of overhead.
that allows like a lot of processing needs to occur.
That allows like a lot of processing needs to occur.
And it, that requires,
that basically translates into latency.
So I need to wait for X amount of minutes
to get my response.
Now, I don't know about you,
but I don't wanna wait like even a minute
to receive a response from like an agent.
So, so, so then you have like, um, trusted execution environments,
which is essentially like, okay, now I have, um, the agents, which then,
uh, which then does a, a certain interaction and it's run in a trusted hardware environment.
And that makes it fast because it's only one machine and it has like this secure enclave,
which basically guarantees that the person that has the machine
can't actually look inside and see what's happening to the program
and can't alter whatever the program is doing.
And so as a result, you basically have that hardware guarantee
that the AI agent is executing in that way.
But then, uh, that, that, but then like, how do we get that into more of the
blockchain based ecosystem?
Uh, and like, how do we enable native operations?
And so that's what, that's what we're trying to do within the Hyperion is create these like native precompiles for the,
a precompile is essentially just a bit of code
or it's like a smart contract
that is natively deployed onto the, onto the, onto Hyperion or any blockchain in that case, but it's, it's
natively deployed. And then you can use that in order to verify that the execution was indeed
correct. Of course, that involves like, if you're running everything through like the blockchain,
Of course that involves like, if you're running everything through like the blockchain,
there is like some costs, right?
Like the storage costs and everything.
But again, the idea is that if you're running,
let's say this verifiable models,
you can have multiple options to get the result that you want.
So you can run it through ZK and potentially get the latency.
You can run it through the, let's say, the trusted execution environment
and get the, let's say, that hardware guarantee.
Or you can run it through this native precompile,
which then you get the on-chain based guarantee,
but with an increased cost as a result
of running the applications on-chain.
So I guess like, hopefully that breaks down
like what we're trying to aim for within the within verifying these these types of like on
chain execution or sorry verifying ai um uh inferences using this this this on-chain
verification method so like relatively higher cost but like like, but the intent is that you ensure that the execution is done correctly without any type of manipulation happening.
Hopefully that answers that question in that way.
And like why we chose that specific model for Hyperion.
Thank you, I appreciate that.
And I have one last question,
I promise I'll get out of your else's hair.
As far as bridging back and forth
between Hyperion to Ethereum.
You know, most ZKs, you know, boast being or having a faster finality than, you know, an optimistic.
So I was just wondering, what does y'all's window look like?
What will it look like at launch? And then what do y'all foresee that window shrinking to in the future?
Very good question.
Rami, by the way, can you, are you, are you, are you there?
Can you, okay.
He's, he's still connecting.
But I saw I'll, I'll take my, my, uh, try it, try it this question.
So, uh, within the optimistic rollups, there are, there is a seven day waiting period.
The reason why there is a seven day waiting period, um, actually it's, it's technically
in a way like arbitrarily set, like technically can be one day but of course the
the reason why is because you have the the um the problem with the um if the sequencer ever
does something malicious um you don't have a way to guarantee that the execution that the sequencer had done is malicious.
So you essentially have to wait for someone to come in and say,
hey, the sequencer executed this code wrong, and here I can prove it.
And then they do an on-chain execution.
And then as a result, on-chain execution on the L1.
And as a result, like if the sequencer did something bad,
like for example, try to like change the balances
of someone that didn't have their,
let's say they tried to maliciously
give themselves a hundred million dollars. Someone would come in and say, no to maliciously give themselves $100 million.
You know, someone would come in and say, no, no, no, no, no.
That is a slashable offense. And as a result, they would make an on-chain action, which then would slash the sequencer,
get them removed from the sequencer pool, and then have the next sequencer in place to replace that.
So in a ZK,
because ZK essentially allows for you to generate a proof
that the computation was correctly executed.
So like all the transactions were executed in this specific order.
And that proof guarantees that the result is correct.
And so you don't really you don't need to have any form of like this,
this type of like challenge period or anything like that.
So as soon as the proof is submitted to the, uh, L1, then, uh, you're free to withdraw or you're
free to like make any type of interactions with, uh, the money that, that, that, well, whatever that the sequencer batch
was sent by the sequencer in like a ZK rollup.
Now, it does take a while for that ZK proof to be generated.
So in that case, you have to wait for the ZK proof
to be generated in order for you to withdraw.
So there is some latency.
Now, it will still be an optimistic based roll up in the case of Hyperion.
But the intent is that we can actually introduce ZK as a finality gadget.
So like we can essentially say, remain optimistic,
but use ZK as a means to say,
hey, I've executed all of these transactions.
And as a result, here's the ZK proof.
So in that way, you can add ZK as a means to achieve that finality faster.
So you don't have to wait for that seven day period.
In terms of launch, in terms of like, you know, when will this be implemented?
be implemented? I don't know. I know Optimism is working with Succinct in order to make this like
I don't know.
an OP Succinct to enable these fast withdrawals, but even they don't have this implemented
for their ecosystem. We're working with actually like the ZKM, which is a ZKVM that essentially makes it so
that all of the like, like we can execute all of the interactions of, of each transaction within
the ZKVM, essentially proving that the transactions are correct and enabling
that fast withdrawal mechanism. Again, like all this complexity, let me simplify it and say,
in terms of the release, it will be an optimistic based rollup.
So it will have a seven day withdrawal time.
But the idea is that it's not limited to that.
We will have the ability to add on these components of ZK
to then be able for faster withdrawals.
Hopefully that answers your question, but there will still be, let's say the seven day withdrawal period for Hyperion as well.
Back to Ethereum.
Hopefully that answers your question.
Thank you so much, Laval.
Great.
I just wanted to remind about the space is going to be end very soon because we just
started like two, three minutes late.
It's just the last minute of space. And just to remind everyone that Hyperion Spaces
is happening weekly. So if there are more questions, feel free to share with us in the chat.
Always, all social media channels are open to the questions. And also, please keep joining Hyperions
because there are great news coming.
I'll give the mic to Pavel Rami for the wrap-up session.
Thank you. Thank you so much, Shayda.
And sorry, everyone, I had problems.
My internet is very unstable here.
So I hope I didn't miss some important questions.
But I'll be happy to answer more and more over the next few sessions.
But yeah, right now, for this week, I'm in an unpredictable region, we can say.
That's why I'm having some connectivity issues. I hope you
heard what I just said. Yes we did. Okay great. I was trying to join and like
become a speaker for like 10 minutes every time it like it shows me requested
but Shada told me on telegram that I am a speaker and my mic is open, but you cannot hear me.
It's impossible, these issues.
Yeah, thank you so much also, Pavel, for taking over, handling the questions.
And Sheda for hosting also.
And everyone who joined, of course.
Thank you for your valuable, valuable questions.
And keep them coming.
We are, if you're in our community chats,
you will have our community chats,
Norbert and David answer the questions
and relay them over to us.
And your questions would be great topics of discussion,
and they would be really entertaining for the whole community.
So keep them coming.
And thank you, everyone.
Awesome.
I guess see you next week.
See you guys.
Bye. Take care. Bye. I guess, see you next week. See you guys.
Take care.
