Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
𝗖𝗘𝗢𝘀 𝗗𝗲𝗰𝗼𝗱𝗲 𝗔𝗜 𝗖𝗼𝗻𝗳𝗶𝗱𝗲𝗻𝘁𝗶𝗮𝗹: 𝗭𝗞, 𝗧𝗘𝗘, 𝗙𝗛𝗘 | w/ @robviglione, @marvin_tong, @GuyItzh & @NNOPT | @HorizenLabs @PhalaNetwork @FhenixIO @Optalysys
Recorded: Sept. 17, 2025 Duration: 0:40:42Space Recording
Short Summary
In a dynamic panel discussion, leaders from ZK Verify, Fala Network, Phoenix, and Optalysis explored the evolving landscape of confidential computing, highlighting strategic partnerships, innovative project launches, and the growing importance of hybrid solutions in addressing privacy challenges in decentralized systems.
Full Transcription
Thank you. Thank you. Hello, everybody.
Welcome, everyone.
We're just waiting for a few people to join, and then we'll get cracking.
Really excited for this CEO panel discussion on the ZK Verify AMAX base.
Been waiting for this one for a while.
Just giving everyone a... Hey, Rob.
How's my audio coming through?
It's loud and clear. Loud and clear.
That's exactly what I was hoping to hear. Thank you.
Testing, testing, ZKV. Yep, it's all good. Hey Guy, how are you? Hey, good morning everybody.
Great to hear from you. Hey Nick, how are you?
Hey Mark, all good, thank you. Hopefully you can hear me.
I can hear you loud and clear, indeed. And I think we're just still waiting for Marvin to join from FALA.
So we'll just give him a few minutes. We're also having a few people join the session live as well.
So we'll just wait a couple more minutes. Yeah, so it's just Marvin, right? The last speaker?
Yeah, so it's Marvin from Falle, who's also going to be speaking. I'll just ping him now
and see if he can join one sec
cool he just messaged me he's joining right now so we're just about to kick this show off very soon
more people joining
for those joining we've got an action-packed show today.
Leaders in the confidential compute space, CEOs of some of the leading encryption and verification technologies out there,
sharing their thoughts in terms of where the industry's at, where the confidential AI compute market is, where it
might be going. We'll be talking about hybrid solutions. We'll be talking about solutions
where ZK can serve well, where TEEs serve best, where FHE comes in, and a lot more.
Just waiting for our final speaker to join, and we should be up and running any minute now. There we go. Hello Marvin, can you hear me? I think then we're going to kick off.
Hey there. Hey.
It's very clear to me.
Marvin, are you... Marvin...
Sorry, Fernando. Ah, there we go. I think you're connected now. Perfect.
Marvin, can you say something just so we can make sure you're connected?
Cool, cool, cool. Can you hear me?
Loud and clear. Loud and clear.
Awesome.
Testing, testing. Z-K-O-V-T-E-E-F-H-E.
Right, everyone. Welcome to...
This is going to be a fun X-Spaces today.
We've got leaders in the confidential compute space with us, CEOs of some
of the leading verification and encryption technologies in the industry today. We have
the CEO of FALA Network, Marvin Tong, with us. Rob Viglione, who is the CEO of ZK Verify. We also have Guy Itzaki, who's the CEO of Phoenix,
and Dr Nick New, who's the CEO of Optalysis, each representing ZK, F-H-E-T-E-E in this space.
So really, really excited to bring all of these speakers together.
So really, really excited to bring all of these speakers together.
So this came about, really, before I introduce the speakers to say hi.
This conversation came about after ZK Verify formed a partnership with Falla a few months ago,
where we figured out how to verify FallaCloud's TE attestations on-chain for at least 20% cheaper on the proving side and 90% cheaper on the verification side.
And shortly after that partnership was formed, I found a really interesting article titled The State of Confidential AI in 2025, Who Wins and Who Loses?
And it really kind of went into detail about all of the privacy challenges that AI processing
brings in decentralized systems and the importance of technologies like TEE and ZK and FHE
and this advent of hybrid solutions being formed, which is what Faller and ZK Verify
have already now done. And we're looking to kind of explore this more in this session.
If you want to know more about ZK Verify's partnership with Faller Network and how that
works, you can feel free to dig into ZK Verify's blogs. There's a couple on there.
There's a deep dive with a video demo on that.
We also have done another AMA session with FALA's DevRel,
all about that partnership and TEE and ZK integration.
And we've done an episode on FALA's podcast, Blackbox,
about our partnership as well.
So check those out when you can.
This is really about shaping the confidential AI compute market in this session.
And so without further ado, I want to introduce our speakers onto the stage.
Maybe if you could say a few words about yourself and
what you're up to and I'll start with you Marvin. Maybe if you could introduce yourself from
Fallow Network. Yeah thanks for the introduction man and just a quick intro for myself. I'm Marvin, I'm the co-founder and CEO of Fala.
And it's just a pleasure to be here
and thank you for having me.
And yeah, as Matt just mentioned,
we are working with CKVerify on many exciting ideas to make AI workloads
and confidential container compute
with a better security assumption using ZKVerify
and other cryptographic solutions so um you know uh
i see uh guys here and many other guests here they are all uh experts from cryptography
um in general like uh the original idea of father is that uh back to our 2018 right we saw uh um uh papers
like iKaden uh and uh secret uh we understand like uh uh it's a it's an exciting idea right we use te to scale blockchain to produce blocks inside
uh security enclaves uh to you know uh to uh just to upload some uh computation workloads from
on-chain to option so um uh but uh by then there's kind of two trenches.
One trench is like Oasis Network,
using TE for sharding, for separate competition power to make a higher TPS,
a better confidential protection blockchain.
And secret network is, of course, using Oasis,
sorry, Cosmos SDK, right,
to produce blocks in a secret environment within TE.
And because of the blockchain,
the blockchain so we don't need to worry about a consensus uh uh you know uh as a security assumption
so we don't need to worry about consensus,
so uh where far as then begin to working is that we think uh okay uh if we can offload right some
of the workloads from on-chain to off-chain we've seen the encl the enclaves, why don't we build a cloud, right?
We use each, why don't we, let's say,
each workers that using TE is a separate,
you know, state container, right?
They don't need to share state with each other in real time,
but they could use some on-chain mechanism to communicate
the state and heartbeat after a while.
So just a TLDR, we want to build a cloud instead of a chain.
And the reason we have a chain is that we want to make sure
all of the key management of these cloud workers
in a sequence and consensus.
So this is kind of the very, very original idea of Fala.
And then we kick off and we do realize there's a benefit of using TE,
which is using TE can solve the general issue for most of the P2P compute network,
most of the P2P compute network which is cheating.
which is cheating.
Whenever there's incentivized the contributors always have motivation to cheat the system.
For example, Filecoin, Protocol Labs spend a lot of hours in there.
But using the EE is easier to avoid cheating.
So we're going to scale the compute quantity of
the network by leveraging
this strong advantage of using TE.
So yeah, that's the original idea of BALA.
And when it came to 2023,
we finally kind of begin to scale a little bit on the
demand end because the barrier of TE is getting,
you know, much easier with D-Stack framework.
And another story is we begin to
face and onboard more use cases from AI.
I think these two game changer is
helping us to scale a little bit.
Yeah. I'll give them a contest.
Very cool. Yeah, but it comes to the next conversation. Okay. Thanks. skill a little bit yeah so uh yeah uh i'll give more content yeah yeah i'll be coming to a
conversation next conversation so okay thanks awesome thanks for the intro um marvin and
actually you started to dive into um the te space and kind of the merits of of te and and also our
kind of partnership together with ck verify um we're going to jump into a kind of panel debate uh in in one
in a few in a couple of minutes um but first i just wanted to introduce uh our other speakers
um guy maybe if you could just give uh like a one minute intro about uh you um phoenix and uh
yeah anything else you want to share and then we we're going to dive into FHE very shortly after that.
For sure.
Yeah.
Great.
I'm glad to be here.
A little bit about my background.
I spent a big part of my career working on trusted execution environment.
I was managing a team at Intel that was responsible for taking the core TE components and driving
them into all of the OEM platforms.
I think I can say I'm a proud father
of probably 90% of the TEE projects
that are out there today in the blockchain space.
And then later on, I was also managing the team
that was building the hardware acceleration for FHE.
And then left Intel and started Phoenix around 2023.
And really our vision is to address the challenge of lack of confidentiality and lack of data
privacy in the blockchain space.
When we started Phoenix, our first product was actually a Phoenix L2.
And the vision there was that we build an L2, we offer data confidentiality using FHE, which
we'll talk about what it is in a minute.
But very quickly, we understood that what is missing in today's world is not another
L2, but rather the ability to provide confidentiality as a service.
And we've released an FHE coprocessor, which is not a hardware component.
It's a software component that enables developers and applications that are running on any EVM
chain to very easily and very natively start to drive confidentiality into their applications.
So they basically call the coprocessor, send the computation that they want
to be executed confidentiality,
and then the coprocessor executes it and returns a result.
And it provides the security guarantees
of fullyomorphic encryption.
And I'm super happy to be here.
I like these conversations a lot,
given the T nature of it,
as well as having Nick from the FHE hardware component.
So looking forward to it.
Thanks, Guy. Thanks for the intro.
And finally, not finally, Nick, Dr. Nick New from Optalysis.
Maybe if you could introduce yourself for a minute or so.
Sure. Hi, thank you very much, Matt. Pleasure to be here on this panel.
So, to introduce Optalysis a little bit, really, we come from the hardware development end.
So, and in particular, our background and my background going back some 25 years plus now is in optical computing and specifically towards Fourier optical computing, which relates to the ability of to perform Fourier transform based operations using a light based interference process, which we now encompass into silicon into chips we are currently developing and applying those to specific cases
where the processing of transforms
and the movements of data are really the limiting factors.
This brings us to FHE and related methods alongside FHE,
which you might come onto as well,
where we can really take on the biggest challenges
attached to the processing of encrypted data
under FHE methods, which is in related to the speed
at which those processes can be executed
and thus the complexity of the processes.
So anybody that knows about FHE will know about
the large amounts of computation that's required to process data in encrypted form.
And what that means right now is that to run on conventional processes, to CPUs, GPUs, you face a very big slowdown in the speed at which you can operate.
speed at which you can operate.
Methods, those methods are thus restricted to fairly simple methods at the moment under
conventional processes being used.
And that's exactly what we're attacking with our optical approaches.
So think of us as providing the kind of acceleration for FHE methods and implementation implementations
going forward. We work closely with a number of companies in this space including Guy and
Phoenix that we're working really closely with now which is really good and providing
that hardware acceleration from which to build from going forward which then removes the
restrictions in terms of the limitations relating to the speed
and thus complexity of the processing
attached to FHE-based methods going forward.
So we've just launched the first product in this space,
which is a digital PGA-based product
for processing encrypted data on the blockchain.
I might talk about that a little bit later as well.
And that's the start of things to come from our developments
into bringing in the photonics into the boxes going forward.
So really exciting place to be right now.
Lots of challenges, lots of early implementation stages to get through,
but great to see really good momentum in this space
in terms of secure computing,
trusted computing going forward as well.
So that's where Octalysis comes from.
Thank you so much, Nick.
Great introduction there.
And we'll be exploring more on how you accelerate
FHE encryption and how your product works in a bit more detail.
Rob, love to get an introduction from you from ZK Verify.
Sure, Matt. Happy to. Hi, everyone. I'm Rob Viglione, the CEO of Horizon Labs.
We're the core builder for ZK Verify, which is our new dedicated ZK proof verification layer.
which is our new dedicated ZK proof verification layer.
Super excited about it.
I feel like a bit of an imposter here
on a verifiable compute panel
because we're not an AI company, clearly,
but although maybe we will be one day
given what we're doing,
but we do have this really, really awesome
enabling technology for verifiable computation
and AI and many other things.
And we're now working with Fala and we're also working with some other TE solutions that
I think are kind of game changer in the sense of we're able to augment, you know, the kind of a
trust assumption that you have for operating in these verifiable compute systems with cryptography. And to have that cryptography trust assumption verifiable on a public,
easily auditable and everything out in the open type of layer, I think is huge.
And just to give you guys a little bit of a sense, we are the first Web3 verification layer to hit the market.
And we're going to be doing this on mainnet in the coming weeks.
And I promised Matt we won't talk about TGE stuff here for our token today. So that's not the point
of it, but the network's going live and we're already on track. I mean, just with our test net
volume, we're processing something like 50 million proofs annualized or ZK proofs annualized now,
which is about to jump up to about a quarter billion.
If you've been in the ZK space for a while,
and we have as a kind of a ZK builder since 2017,
a quarter billion ZK proofs being processed is massive, actually.
So it puts us hands down by far the biggest layer for processing
this type of cryptography in Web3.
So we're pretty pumped about it.
You combine this with the verifiable compute stuff that we're doing with FALA
and the other things that we're looking to do just in terms of, you know,
like secure enclave, you know, environments and pieces of hardware.
Could not be more excited and really happy to be here on this panel.
Awesome, Rob. Thanks for the introduction.
Well, let's kick off with this panel um discussion
and i think the first question that is on everyone's mind is you know what are the pros and
cons of te versus zk versus fhe um you know i've been hanging out in some forums and Discord channels and, you know, there are members of certain projects that are really pro TEE.
They're against other encryption methods. There are others that are kind of really pro ZK and some that are really pro FHE.
I think all have their merits, right? You know, there are speed trade-offs, there are security trade-offs,
there are privacy limitations for certain encryption methods
and verification methods,
which is why a hybrid market
is starting to emerge.
So I'm going to ask each of you,
each CEO, to provide a,
maybe like a two, three minute answer
in terms of what are the pros
of your encryption method versus some
other encryption methods. And also, if you want to be real and honest as well, you know, what are
maybe some of the disadvantages of your tech stack versus some of the other ones that are
represented here? So I think I'm going to start maybe with Marvin.
Tell me a little bit about maybe in two minutes, what are the kind of pros that you see TEE brings right now to confidential compute?
And how do they compare to ZK and to FHE in your mind?
That's a great question, Matt.
Just to tell you,
I mean, at least me and, like,
follow-up team that we believe in
that TE and PCFE and ZKP
are not competitive technologies
compared with each other.
They are serving for a common goal to our perspective,
which is the God Protocol-level computer network.
So, and God Protocol basically means like a simulator of a well-known, you know, well-acknowledged, well-capacitability computer network that can do everything and in the most ideal world, it means that any computer during this network
can fully encrypted the message and communicate
and compute these common public goods for
the whole network without
exposing the secret data with each other.
We believe that to reach out that goal,
there's no single technology could achieve it.
So there's definitely a merge in
the upcoming days and the future of
these technologies together.
Just to simplify to explain what's different with each other,
I'll try to go quick.
MPC basically means that you need to put the computers in a common protocol that
user secrets shared their private inputs,
and they can send in encryption data to the compute nodes.
After the compute calculator result,
the protocol will give output with these nodes signing.
But the trade-off is that MPC works,
it's not guaranteed for private.
At least the private state is not guaranteed by default.
Whenever the MPC knows the numbers of
computer nodes adding up,
the overlap for the compute trade-off
is also increasing rapidly.
Let's say 10 nodes, probably there's a 1 KB complexity,
the data exchange can approach probably 10 GB.
And when it comes to 1 G nodes,
it reaches about maybe 10 TB.
So the workload trade-off is adding up very quickly
when the MPC competition power added up.
FHE is ideally, I think FHE is ideally
is one of the best technology to address
a longer term challenge.
Like it can enable the competition in secret technology to address a longer-term challenge.
It can enable the competition in secret and
security in encrypted environments without requiring
encryption to verify the integrity of the compute.
It means that a user can encrypt their sensitive data and
run it in a server and server can execute
computations on these separate tags.
The output of the result is still encrypted and
the result can only be decrypted by
the user using their private keys.
So it's a little bit different from traditional E2E encryption
where the computation on encrypted data is not feasible.
This FHE requires this data transfer than MCP. So it doesn't go exactly the same,
you know, computation trade-off
when the computing servers
adding up into this network.
But it could be a little bit slower
compared with traditional MCP,
except in the situations that are using maybe ASIC
or very good hardware design that feasible to the VM compute nodes.
So, yeah, so hardware is definitely a keystone
of how to boost up, I'd say, ratio of action.
And how TE play a part of it is that TE itself,
if you're running a program inside TE, the performance and encryption is good,
but it's not the encryption trust assumption is not as good as the same programs running inside FHE or, you know, JKP node.
or, you know, JKP and alt.
But it kind of has a better balance
between compute performance, self-proving,
and, you know, and verification.
So, yeah, I think that's roughly true.
Thanks, Marvin.
And same question to you, Guy. Guy I mean Marvin talked about um you know
FHE's speed uh and like you know I guess high high computation requirements um maybe you can
kind of share uh you know some of the pros and cons of of FHE and and how you think it's going
to emerge um over the coming years with developments that Phoenix are working on
and also your partner Ophthalasys as well.
We'll talk to Nick as well about that soon.
Yeah, 100%.
Totally agree with what Marvin said.
I think if we look at this landscape of three-letter type of cryptographies,
you have the ones that enable you to do computation,
MPC, FHE,
and then you have verification capabilities like ZK.
And then within that space,
the positioning of FHE is that it's been out there
for quite some time.
And I do think that there has been this concept
of Fhe being
complex and slow however since craig gentry's first association of fhe there's been a lot of progress
around around homomorphic encryption before i start talking about the progress just for those
who've never heard about fhe and what FHE is, it's a cryptography that
enables you to compute on data while the data is encrypted. And the reason that it enables it is
that you encrypt the data in a unique way. You encrypt the data homomorphically, which maintains
the mathematical structure of the data that you encrypt. So you can then compute over that
encrypted data and get an encrypted result, which you can then compute over that encrypted data
and get an encrypted result, which you can later on decrypt with the same key
that you encrypt with it, you encrypted the data.
And it is quite powerful in that regards.
So in order to sort out the performance and complexity,
which are the main two challenges around complexity, I think there's been a lot
of progress. And today, FHE is very easy to use, specifically with us. We're Solidity-based,
so developers don't need to understand cryptography. They don't need to understand FHE.
They write their smart contracts, and they can decide what type of data they want to encrypt
and decrypt. It's EVM compatible. So compilers,
progression around compilers that happened, sort of that part, I would say, for the Web3 space.
And also there is quite a lot of work happening around compilers for Web2 with, I would say,
major leading industry companies building compilers, including Google and others. So still work to be done, but definitely feasible, at least in the Web3 space.
And then around the performance, so FHE performance is being tackled in two ways.
The first one is just improving the cryptography itself.
There are a couple of variants of FHE. We call them FHE schemes.
Some of them are more efficient. Others are less efficient. Some of them are better for machine
learning type of computation. Others are better for general computation. But there is content
progression that is happening around the schemes themselves. That is the first part and the second part which i would say is probably
even more exciting is hardware um and in different than mpc which is very it's communication bound
it requires a lot of communication in order to operate which can't really scale well in the
internet fhe is computation bound which means that the more hardware you
throw on fhe the better performance you get so we're starting to see optimization of fhe schemes
to gpus and then fpgas which um optalysis have recently released and nick will talk about that
and then the aim is to get ASICs which are specific hardware
built for FHE which will improve the performance in orders of magnitude and that would really like
for me uh sort out many of the the complications that we see today um I think I want to be clear
that like I don't think FHE will ever be as fast as T is, or if it will, it will be. And there is still time for that.
So totally agree with the way Marvin mentioned it,
that if we want to get this very fast computation,
encrypted computation, TEs are probably today the better solution.
But for the majority of computation, which is not heavy computation,
FHE is definitely ready and will just get better and better as time progress.
I think that's pretty much the state of the union where we're at with FHE today.
Awesome. Thank you, Guy.
Nick, I'll come to you in a minute.
I just want to go to Rob and get your perspective on this in terms of, like know you've been leading you've been in the zk space since i think what 2016 um
what is your kind of perspective in on zk versus some of these other technologies te fhe um where
does it sit what are the kind of pros and cons from your perspective and what have you what have
you seen over the years as well i mean i, I'll give you definitely my kind of,
my inside point of view here is because we are a cryptography company
at the end of the day.
And I'll say we dabble and we're dabbled in kind of all of these technologies
to some extent.
So we've been in ZK basically since the beginning of ZK and Web3,
you know, started building in 2016. And, you know, I would say like why use ZK, basically since the beginning of ZK and Web3, you know, started building in 2016.
And, you know, I would say like why use ZK is because it has the strongest trust assumptions, I would say, just in terms of you've got the math out there.
Circuits are all publicly or published open source.
You know, they're all audited.
And, you know, as long as the math is good and everything's verifiable in the software, it works.
And it works very precisely to get you the result that you need.
And it's composable today on-chain.
So you can actually plug ZKP-Cruves into any part of kind of like the on-chain Web3 world.
Maybe not all of the audience here are Web3.
Maybe it's not a completely Web3 world. Maybe not all of the audience here are Web3. Maybe it's not a completely Web3 audience.
Maybe it is, but where we sit is we're very much in Web3
of getting proofs and being able to do things on-chain
in a verifiable way.
That's where ZK shines, is where you need transparency
and public verifiability for something.
And ZK Verify is designed to be the best system at doing that,
at least kind of the bottom of that value chain of verifying results
anywhere from any type of cryptographic proof system.
That said, Horizon Labs is actually in process of building
a trusted execution environment ourselves.
And we're doing this for another project that we're the builders on called Horizon,
which is a private execution environment in the base ecosystem on top of Ethereum. So actually,
we do have some experience. We're just starting to build our TE muscles. But maybe the relevant
thing here is TEs and ZK are highly complementary. They do different things. And I think the
combination of them is extremely powerful. And again, like coming back to, we really got into
like, like working with FALA has just exploded, I'm not gonna say piqued our interest, but
exploded our interest and a combination of these two technologies. FHE, we don't really have
muscles in FHE. We do have some
FHE experts in-house, and we did explore, working especially with that Google compiler that was
supposed to come out last year. We actually didn't get into it for specific reasons, just because of
where we sit in the industry. But I will say we're very interested in FHE. I think FHE is one of those
extremely promising technologies. So
I'm really happy to be here with the guys to go deeper into it. But that's kind of like my
rundown. I would say the quick cons, just to keep myself intellectually honest here,
ZK can't do everything. For sure it can't. ZK proofs are still computationally expensive,
depending on the proof system. That's why we're actually as an industry creating specialized proof generation networks where we can actually have hardware that specializes in
generating proofs because again, they're expensive. So you're not seeing DK proofs everywhere for every
type of thing. Also, we can verify known statements. They're not good for arbitrary encrypted
compute. That's why we have some of these other technologies. So there's trade-offs, guys. We live in that world of trade-offs, and it's really,
from a product perspective, about segmenting and understanding what you can do better than
the other things that are out there and honing in on that.
Awesome, Rob. Thanks for sharing that. I want to go to you, Nick, actually,
and talk a little bit about...
