# Pushing Blockchain Limits with Hardware Accelerated Provable Computation - John Lilic

- Channel: [ETH Belgrade Community](https://streameth.org/eth-belgrade-community)
- Date: 2024-10-07
- Duration: 35:05
- Watch: https://streameth.org/watch/yt-8MJuRl0e0mY
- YouTube: https://www.youtube.com/watch?v=8MJuRl0e0mY

## Description

Pushing Blockchain Limits with Hardware Accelerated Provable Computation - John Lilic

## Transcript

Doan um yeah last night we had a uh a dinner like a dinner party and I was trying to explain to my friends who have never been to Serbia what to expect in other words there's going to be like a beautiful restaurant and uh a massive stage with Extremely Loud Music the whole night uh while you're eating dinner uh but uh I think we had a great time and it's nice to see so many people in Belgrade uh okay so this morning I'm just going to talk a bit about Telos um and some of the stuff we're doing on the ZK side of things which uh might be interesting for some of the ZK Folks at this event um yeah so I guess kind of what is Telos um I've been uh doing a bunch of talks this week talking to lots of people and I have this little kind of story little little stick to describe Telos if you guys remember uh before Netflix we had VHS and sometimes there would be these like Hollywood be movies that would go straight to VHS and then later they would become like a cult following um and so Telos in some ways is kind of like that uh so this thing launched in uh 2018 uh it was actually kind of like a fork of EOS if you guys remember EOS EOS was that big massive project that raised $4 and a. half billion dollars um and uh some people in the community felt like they weren't um honoring their commitment to to invest in the ecosystem U maybe some of that $4.5 billion dollar wasn't uh allocated properly into the community and so the community decided to Fork away from that and uh take a organic approach uh and so they did um in 2018 in the midst of IC om Mania they did a uh share drop um or a fair drop as they called it at the time which is basically airdropping the entire Supply so there was never an Ico there was never a token sale there was never a big money event and instead uh TS had to sort of somehow figure out how to exist and survive uh and along the way and and I also I also like to think of Telos as the grinder chain it's just everyone keeps grinding every single day working hard just grinding away uh and so in 2021 they uh actually built an evm um and this is where uh Telos first entered into ethereum lore and uh history actually this is where it came on my radar um Sentinel which was a very wellknown and respected audit uh smart contract auditing firm um discovered a critical bug that actually could have brought down ethereum itself um in the in the process of fuzz testing the TBM uh of course the uh various clients were notified um the bug was fixed and uh and then this massive story came out later like wow this would have been pretty catastrophic um the TVM actually is a very performant evm it does lots and lots of transactions per second and over the years uh we've been onboarding uh a bunch of different projects maybe not the biggest in ethereum like the AES of the world but uh lots of interesting young talented teams uh that maybe weren't getting tremendous support from Big networks so Telos in some ways uh has been kind of like the minor leagues of baseball on boarding these young projects that uh over the years also keep grinding and now many of them have some great traction um yeah and then uh and then in late 2023 one of the founders kind of rage quit and the project was having a tough time and that's when I thought wow this is a great opportunity uh particularly because I've been thinking about ZK for a number of years uh so at the start of this year I joined as the uh executive director of Telos um I'm a volunteer I don't get paid um I actually sponsor personally all of the Telos uh uh events that we've been part participating in this year because I think the uh Hardware accelerated um let let me restate that the hardware software codesign approach to building robust ZK systems is very important and it's a message that uh uh we feel like we should try and start getting out and building uh building a meme um earlier this year we also launched uh as a first part of our ZK Journey something called snark Tor uh there's a wonderful gentleman I don't know if he's here yet today but he'll be here his name is Alberto gar he's our lead lead ZK researcher and he uh co-wrote the snark Tor white paper and it's a decentralized recursive proof protocol uh we're betting very big on recursive proofs so these are uh proofs that set of underlying proofs are uh sort of valid um andan you can aggregate these things and maybe in somehow in some way achieve uh tremendous performance gains um so this is just a little bit about our evm this was a bit of a meme that we had for many years um um so the evm on Telos so the base layer of Telos is delegated proof of stake it runs something called Antelope which actually is a very good uh uh VM execution environment um and Telos the Telos evm is emulated on top of this layer zero uh and the trade-off is um you don't have necessarily the same type of compatibility with ethereum so deploying has a bit of overhead for de Vel opers um just like with anything there's trade-offs uh but if you are willing to do a bit of extra work and we're happy to support you then you get these like tremendous performance uh characteristics um okay so leading into the uh ZK talk um I guess trying to understand what it is that uh prevents Mass adoption okay and uh Alberto has a great example and maybe some of you heard me uh in private discussions give this example but uh you know it's like you would never like you come to Belgrade for the weekend and you rent a car or you want to rent a vehicle um you would never rent that vehicle using a blockchain system if it meant that your driver's license would be onchain forever um and so when you think about that in the context of many many other use cases uh that you know the internet the rest of the internet is servicing today which necessarily require data protection um you can see how it's quite a big hurdle um and so um thinking about ZK enabling those use cases um by allowing for proofs that some attributes about your identity are real in this case um imagine you could generate a proof that you have a valid license say you're 18 years old um all the various characteristics you would need to prove by handing over your driver's license to the car rental agency well instead of giving them your driver's license you simply generate a proof and share that proof with them uh and that can go onchain and so in this way you can still achieve the same objective of proving that you're um of of a Val that you have a valid driver's license um and you're qualified to rent a vehicle without disclosing and having your license on chain forever moreover um aside from let's say realizing what I just described I would I would I would say is kind of the Holy Grail of the blockchain which is to say self- Sovereign identity and reputation systems so realizing self- Sovereign identity reputation systems means that you can establish your identity on chain you can have a canonical Identity or a canonical Branch with Persona constructs that describe you in different contexts right so maybe you're a gamer and your gamer Persona might have attributes that correspond to your username uh which games you play how many points you have Etc uh but your government uh or kyc Persona will would need to have very different attributes such as your name your birthday um in America your social security number Etc uh these attributes can then collect attestations from others so you could imagine going into a notary office and getting an attestation that indeed uh your passport number is what you claim it is or that your birthday is on this particular day and so now you have your Chronicle Branch Persona constructs attributes oh excuse me attributes and attest against those attributes okay and of course you have the private key to all of this so you have custody over your identity ZK we think will enable uh the user to then selectively disclose and granular authenticate so you'll be able to rather than put all of these identity attributes on chain uh you'll be able to put proofs and more specifically recursive proofs um that these attributes about you are true and valid and that they have corresponding attestations from others okay this this is this is sort of the notion of self- sovereign identity with persistent portable transitive uh reputation as well and so you can imagine a future where rather than logging in with your Apple ID your Google ID and your Facebook ID um you log in with your private key um beyond that uh ZK also has a very important role um in terms of scalability so massive throughput uh performance characteristics that we've all been kind of chasing all these years of course data protection and uh and then on the compliance side there's lots and lots of regulations which protect user data um and of course we can think about some of the classic examples in America we had Equifax um I don't know how many Americans are at this event in this room right now but probably some elements of your credit score are available on the dark web right now um or you look at many other countries as well it's not just system or endemic to the United States uh for example in the Philippines uh the commission on elections so it's called KC where all the voter records are held uh was hacked a couple years ago I I can't remember the exact number but I think it was 62 million uh voter records are now available for you to buy online if you wanted to um so you know very very strong there's very strong reasons to have these systems in place to protect user data uh particularly if we're talking about this concept of identity um Okay so so what are some of the current uh issues and challenges that we Face first of all I just want to say for all the ZK researchers all the people that are working on this technology you are all my heroes okay I have a tremendous amount of respect for all the different teams who are taking multiple approaches this stuff is very very hard um it's not hard it's very hard um and so consequently um there are lots in lots of trade-offs um and some of those trade-offs obviously um relate to proofs and the fact that generating proofs and posting proofs on chain is expensive and slow and so um what we're seeing today is tremendous Innovation um almost like every single day there's some breakthrough some announcement um I think the risk zero team is here as far as I understand uh and you can see for example some of the incredible work they're doing and many others at teos we've sort of been very fortunate to interact with many of these teams I spent some time recently in Hong Kong with the scroll team obviously they've taken of a a very strong approach focused on Halo 2 um and interestingly enough um they have uh from the start kind of taken this Hardware software co-design approach which we'll talk about in a second uh which I think which which we think is critical uh to uh working through a lot of these issues and realizing these uh these performance characteristics that we're all chasing okay so um I guess a lot of you or some of you might know me as kind of like U an ethereum guy guy I was uh one of the first consensus employees I spent six years there um I was kind of the consensus dude for many years um later I uh I joined uh polygon interestingly enough um that project was kind of to a large extent incubated right here in Belgrade uh Maho is uh one of my best friends I was here with him 2017 18 19 when he was just grinding away thinking about how to build uh an ethereum scaling solution with multip uh Solutions baked into one um and of course along the way um ethereum has changed quite a bit uh we had a big big switch towards uh away from proof of work and back to and and towards proof of stake um and for many years we talked about how uh A6 are really bad uh for all these reasons which I personally feel like many of those reasons have been invalidated um there's lots and lots of great examples of uh clean clean mining happening right now of course the very famous miners in Texas who are uh in addition to using clean power providing ancillary services to the grid um there's great projects like the sustainable Bitcoin protocol and many others who are encouraging green mining or clean mining um and now recently we see um ethereum uh royalty I guess uh key people who are very well respected uh talking about as6 again okay so the first thing that really caught my eye was uh Justin Drake in February of this year um I should have brought uh I I I left it in my hotel room but I have this chip with me actually um I went to shenzen and I found the Asic manufacturer it's on my Twitter I've got a bunch of photos but uh he tweeted uh this uh this photo um and uh and then in April um I was at the uh hash key uh Bitcoin uh blockchain conference there I guess it wasn't a Bitcoin conference but the hash key conference uh I actually gave a talk on uh recursive proofs uh and shortly after my talk uh vitalic did his wonderful keynote and uh he basically uh it was a big shock to us at tell is because we've been working on this stuff quietly and he basically described our entire strategy on stage and we'd never even talk to him about it um and of course uh very similar to some of the things Justin has been saying and then of course axial um they're sort of the first as far as I know uh they have actually taped their first uh ZK proving Asic and from what I understand they're going into mass production fairly soon so there is a big shift in The Meta in the narrative across ethereum itself um where perhaps we're starting to realize that actually uh snark proving as6 may be uh on their way um so this is a snark tour this is the white paper you could uh take a look and um Alberto along with a couple other really great iog uh researchers demitro and Roman uh they put this together and it basically describes our approach to uh scaling snark verifications and essentially our bet on recursive proofs okay so Telos keeps grinding away grinding away we've been um sort of trying to establish this meme as a type of venture Production Studio anchored in uh in our L L1 um wherein in addition to that we sort of innovate and build complimentary products that service our ecosystem and our token holders and our community of course the evm uh in 2021 uh snark tour earlier this year and uh just uh some number of weeks ago at token 2049 in Dubai we announced we're going to be launching a hardware accelerated zkm L2 um and uh so this is my personal opinion I feel very strongly about this one of the reasons I uh I kind of dove deep into Telos and have been committing so much of my time is because I think um ZK again is very very very hard I think it's extremely difficult to have super strong assumptions about how this is all going to go rather it should be treated like a uh lab like an experiment um the reason for that is because the pace of innovation is happening so quickly and there are still quite a number of uh research dependencies in order to realize some of these big Visions uh and so I think you need to be very agile um Telos interestingly doesn't have any like big VCS it doesn't I mean the whole token float is out it's been around for many years it had this kind of like airdrop the whole Supply thing it doesn't have these like very strong vested interests that can influence um the researchers and influence the uh Innovation side of things and so um I thought you know let's experiment let's just try different things see how it goes and maybe in one moment all of these different complimentary pieces that so many of these wonderful spectacular ZK research teams are working on will'll reach a point where you can kind of put the puzzle put the pieces of the puzzle together and somehow you achieve this like tremendous performance um and so what's really really important as a kind of foundational base layer to this approach is understanding that Hardware software Cod design is fundamental it's very very important you have to think about these things in conjunction from the start um you can't just build compelling software and then kind of like uh accelerate it um it's very hard to do um moreover we think that today while ax and others are doing a great job uh trying to develop uh CK proving as6 and and and by the way I think there's going to be an fpga interim step uh before we get to uh as6 you know some functionalities you can accelerate but um the entire ZK system on a board okay so a true system on chip uh design is still some ways away and I think I feel very strongly that you have to have the space um and the opportunity to experiment you have to be able to fail very fast and fail very cheap um okay so that's just one comment uh there a bit about snark Tor um some of the stuff I mentioned earlier um so basically Hardware software co-designed to maximize proof generation performances um taking this approach like I said experimenting trying lots of different things seeing what works what doesn't work eventually you'll get there um so for example um the extent to which and this is I think going to be I think there's going to be some exciting news around this to come but the extent to which you can generate proofs on your mobile uh faster than you can on your computer for example um these types of breakthroughs are coming and they're going to fundamentally change everything um and taking a hardware software codesign approach enables you to really kind of be in a position where you can um leverage some of these uh Innovations uh this is just a shot uh uh an image of our new team so we've uh been very fortunate to on board and partner with some wonderful people so uh um in particular ponos technology so uh many of these guys are probably well known here in Belgrade I guess tellos now between our partner Network we've got like 11 phds um very very dedicated uh these guys have been focused on ZK for years um atka uh I think uh I think William uh is here actually so these are our very good friends from Paris uh one of the uh preeminent accelerators incubators very very smart Guys these guys incubated Moro Moro is probably uh probably one of the best known def5 projects in the world right now they incubated Mangrove and and some others and then Jerome Dei who's actually the founder of SEC he's one of my best friends he's joined us uh he was with me at consensus many years um he's been working on a project called kth dealing with account abstraction he's joined us as a uh uh evm technical adviser uh and then finally digital mob um for some of you who might remember back in the day consensus uh we used to have consensus Romania so these guys are in Bucharest it's funny because I think Bucharest and what happened in Romania in the early days of ethereum is kind of like the vibe it's kind of like Belgrade is almost like Bucharest five six years ago uh so these guys have built some of the biggest projects in ethereum very very good Dev team and uh we're very grateful to be partnered with them as well um just a bit more about ponos uh Slobodan I think might be here or he had to go back to Switzerland but I think he came back um uh wonderful guy very good friend of mine lives in Lugano has a seat at the University there and many of his teammates as well all these guys do is ZK research that's literally all they do um and uh we're very very grateful to be uh working with them um just a just a little little quote from sladon um I guess I guess to add to this I've been talking about this with others in private conversation but I I think um you know this AI blockchain kind of buzzword that's really hard to Define and everybody has lots of different opinions on it and there's a whole bunch of uh AI tokens that have pumped and who knows if any of them are credible um but I think the real AI blockchain story is once we so it's like if you're an intelligent system and an intelligent agent um you're going to necessarily demand a fully provable system okay okay that's like table stakes in other words once you understand what that means it's like you can't unsee it okay uh so I believe and we believe that AI on the blockchain will be a massive massive moment that uh you know right now it's kind of a grind not much going on but once these Technologies and once the infrastructure reaches a point um of maturity um it'll be kind of like suddenly all at once and the reason for that is because you can imagine these intelligent systems and intellig agents um performing much faster at much bigger scale much smarter doing much better in the markets um but they will necessarily demand fully approvable infrastructure um uh just some uh another quote from Alberto um so I guess my hot spicy take is that uh obviously the theum L2 ecosystem has exploded uh it has definitely proven to be um an effect Ive uh ecosystem as far as scaling ethereum however I think that um fundamentally uh it's a little bit flawed in so far as uh tremendous amount of subsidies and these subsidies generally Drive user metrics and a lot of times that user met you know that user behavior is kind of parasitic a little bit like a locust um until the subsidies dry up and then the metrics sort of go away um I feel very strongly that uh the correct approach which is much more difficult that's why I say you know Telos is kind of a grinder chain it's all about grinding through you build a compelling product and because your product uh has performance characteristics that appeal to users then users will use your product because it's a good product um and so I think uh in the context of l2s particularly a lot of the ZK stuff that's coming um we should be really focused on building a great product that uh fills a gap rather than trying to sort of subsidize activity um because I don't think that's very sustainable uh this is just a quote from uh Jesse our CTO uh Jesse's been around uh since since the beginning um he's uh he's definitely a grinder um all he does is think about the wasm um and interestingly enough tellos already the evm has rust and wasum native to it I mean there's some interesting characteristics there if you're curious you can delve into it and uh you know Jesse is also very very excited about our our partnership and uh thinking about how uh how we can provide value to ethereum um in particular with snar tour okay so uh just a quick snapshot um yeah we started uh R&D and started working on snar tour earlier this year uh we released the white paper uh I'm sorry last year in 2023 uh we had some workshops in September uh in January we published the white paper um then we partnered uh with uh ponos and we had a great Workshop in Milan in uh March of 2024 um by the way Milan is wonderful if you haven't been highly recommend checking it out and there's a tremendous amount of talent actually somehow concentrated in Milan very very nice people um and then uh well this is a little bit old but this is the slide from Dubai token 2049 where we announced first made this announcement um and then sort of going forward uh we are also a headline sponsor at FCC next July so hope to see many of you in Brussels we will actually be uh demoing our snar tour proof of concept um I think like I said I think sladon might be coming back today he's around uh you can catch him at our booth and uh we can talk a bit more about exactly what that means um so we're going to be releasing some more technical documentation some more white papers coming soon um and then and then towards the later part of this year the early part of 2025 we're going to release the ZK a uh evm and um I mentioned earlier the Telos token had this like fair drop so because there's an evm you can sort of cut through everything and use the same token across these networks which we think is a little bit unique and so we really want to focus on servicing our community and uh and making sure that t those token holders have the opportunity to use all of the things that we use uh all of the all the products that we build um cool so that's uh that's my Spiel I hope I didn't go over time um feel free to follow us uh on Twitter uh I'll be around if you have any questions um but uh yeah thank you very much thank you are we doing questions or sorry okay uh got a few hands here uh yeah there's a few I guess I don't know if if you've got a mic don't be shy guys do we have any questions this is literally lifechanging Earth shattering technology do we have any questions yes right there do we have a microphone you can have mine if we don't here we go MC stage Runner yes sir yeah John thanks for for a nice talk so as a as a k Builder myself I I have a question uh um currently in ethereum uh there is a track to build the K chain so we we build on top of the K stack so we have U architecture of having in layer one then uh ZK layer 2 with the scal into ZK layer frees but given that uh the throughput and SC scalability is not uh is not even is not that high even compared to the metric that tell us already form and my ask is that as as I see from your road map you gravit towards the same uh the same architecture so you have TS as layer one then you scale to the KMS and probably you will also have another another layer of scaling so uh my question is um is it still possible for the native ethereum to achieve such performance that you guys can achieve with your technology given that you have a layer two as as a layer one as a as a TS itself yeah thanks yeah I mean uh so uh it's kind of like the trade-off is always between security and convenience right so more security less convenience um and likewise uh with performance somewhat similar so for example I'll admit it I use dgen all the time so I'm in the L3 and I'm I'm just having a lot of fun um and obviously it's not not exactly a decentralized or very very secure network uh but the performance characteristics are very fast um I think that uh yeah so sharding is very very hard is very hard um we initially sort of in the beginning of ethereum were saying maybe 1,24 shards and then we said 64 shards and then I think in 2018 ballic posted on his blog like uh you know kind of like one Shard maybe one Shard so um the the answer to your question is I don't know uh it's very very hard to scale this stuff um I think lots of the current trade-offs that we're seeing um have pros and cons obviously uh it's very famous now I mean I don't even think it's a spicy take it's just it's just the truth lanaa just shut down their uh Network right in support of as they claim um user protection so uh the answer is I don't know I really don't know um if you want me to get really spicy about it I think that uh I think that you know when you look at things like Alo when you look at things like moned when you look at some of the uh competing uh monsters that are coming I think uh I think ethereum is about to have more competition than ever and we have to get very very serious about this stuff and uh in in very simple terms that means uh uh supporting the ZK researchers tremendously increasing the support for ZK researchers um moving away from billions of dollars of capital going into subsidies and incentives to drive user dgen Behavior so we can pump our vanity metrics and focusing on um providing grants and making sure that uh we support our researchers as much as possible and then hopefully somehow it works out um that's like my honest answer yeah we have time for one more question yes hold on oh we have a mic I don't have to run to you there we go perfect hello hello um so you probably already explained this uh in your talk but uh if you could go back a little bit to this uh Hardware software code design uh importance of this process that you mentioned uh I'm curious to hear you probably said this in contrast to how things developed with Bitcoin uh in the beginning is it being mined like empty blocks being mined with block with laptops compared to what we have today on Bitcoin so how does this uh Hardware software codesign work in terms of which organizations are most focused on what is there parts of Telos team working on hardware and software or uh Telos is collaborating with parties who focus on Hardware or you know just dive deeper a bit on this yep uh that's a great question so um so the importance of Hardware software code design speaks to um the difficult ulty in accelerating this stuff uh by simply plugging in a machine okay um it's much more complicated than that uh first of all I think you have to uh carefully think through your use cases to try to understand what it is that you uh want to offer to the market because understanding the use case helps inform design from the start um once you have that then you can kind of begin to do some Benchmark testing and frankly I think uh uh so there's definitely stuff I want to be able to say right now but I can't because I'm under NDA but uh what I can say is that uh you know you can you can you can probably put a lot of pieces of this puzzle together yourself when you see somebody like axial who's basically a startup by the way that was incubated by scroll and uh scroll was originally a hardware company uh and then they actually couldn't find a product Market fit or any customers and so they pivoted and became this software company but from the start they've kind of like taken the two Hand inand uh kind of like into their design thinking um but um I think that the ZK proving markets are coming uh the ZK Summit recently in Athens I tweeted this it's kind of like in my opinion the most important conference that that happened all year and that will happen uh this this year um definitely check it out the Z prize as well is another great uh source of Truth and information um and so I feel very strongly that uh in the near future you will see major major um announcements from Hardware manufacturers that uh they're betting big on the ZK proving markets um I personally feel that ethereum uh has a bit of a well I don't know if I should say this but I'll just say it um a bit of a problem because you know we have uh uh there's uh there's no more Alpha in ethereum okay the whole thing is basically a bunch of rehypothecation okay we uh take some stuff we stake it we restake it we borrow we do all these different things um and and actually as it turns out our our execution environment is kind of not well suited for ZK and so um as this narrative begins to grow um I think we're going to have to very seriously consider um what potential future hybrid proof of work useful proof of work proof of stake consensus could look like frankly I'm also very concerned about the regulatory side of things I'm very concerned that uh there's going to be an escalation um and so uh I definitely think that we need to have more serious discussions about useful proof of work um and I'm not the only one by the way this isn't just me saying this this is like there's major major um whether it's Hardware manufacturers Bitcoin miners I mean even if you still look at the E classic and eow and these things it's still like a six seven billion dollar market cap between those things um and so uh yeah all that to say I think the race to compete in the ZK proving markets will be spectacular I think it's going to be the source of new Alpha I think it's going to drive growth tremendously and I think that we will see um major major participants uh getting very serious about competing in that uh in that market to come uh so for me it's very exciting um and uh yeah I think everything's going to accelerate tremendously very soon okay please give it up one more time for Mr John lilos