Hardware Acceleration for ZK Cryptography w/ Leo Fan (Cysic)

Transcript

[Music] hello and welcome back to the strange waterer podcast thank you for joining us for today's conversation there is a lot to love about zero knowledge cryptography on this show we really only focus on using ZK as a tool to compress computation and to enable decentralized Computing networks but we could easily spend just as much time on zk's implication for privacy or digital Identity or any of the Undiscovered applications that we're inevitably going to find for

ZK we already see so much promise and we're in a world where groundbreaking research is released faster than Apple can get out a new I iPhone but behind all the optimism and the entrepreneurial energy behind ZK today there's a more frustrating side of the technology the concepts are impenetrable and the tooling is hard to find and even harder to use and what few Primitives that do exist are a nightmare to integrate and really that's just the start again it can be frustrating but I guess that's always the case when you're deploying research hot off the

forge but we are here today to focus on one particular pain Point performance even the best ZK systems right now take so long to run that they're practically useless in real world applications just like the internet would be functionally useless and largely ignored if we could only transfer one or two sentences in like an hour we would be in trouble if the performance of ZK systems was limited to what we had today today's guest is Leo fan founder of cisc and a big reason why the ZK Community is so confident that we will overcome the performance problem that we

are racing to build Empires here cisc is working on creating Hardware specifically etched with the ZK Primitives that will power our trustless future this episode is a comprehensive look at what it means to write applications into silicon and metal including an overview of how chips are made and Reflections on how chip manufacturing factors into geopolitics I won't spoil the numbers for you here but suffice it to say that once you hear Leo tell us how dramatically they can already accelerate ZK proofs you'll begin to understand just how it feels to be at the beginning of zk's Moors law

Frenzy one more thing before we begin please do not take Financial advice from this or any podcast ethereum and ZK will change the world one day but you can easily lose all of your money between now and then all right let's go Leo thank you so much for joining us for the strange water podcast thanks another for have be here yeah of course so um before we like dive into ZK and get into Hardware um I'm a huge believer that the most important part of every

conversation are the people involved so um with that is you know background can you give us a little bit about who you are how you found uh ZK how you found hardware and um yeah just tell us a little bit about yourself thanks yeah um my name is Nea I'm a CO co-founder of Cy and before Cy I got my PhD from Cornell so like uh yeah from from Cornell in cryptography so I'm more like a software like uh algorithm guy and before Corel like I did sometime I yeah I I indeed spend some time in China doing some Bitcoin mining and yeah so after Corel like I spent some time in um a blockchain company called arent so what

was at AR and like I help with Aran stoof so Aran stoof is essentially as proof can be used as a course train bridge between AR a AR so uh when when we finish the PC for that part it was April uh 2022 I I found the post generation time for for the uh for the zkp proof there is like um quite n like I try a lot on the software and on the algorithms but it didn't improve or not on the proof generation time so I think maybe we can design some like specializ Hardware what we did for the Bitcoin mining

before so like we designed some specializ how for that to make sure the proof generation time is like uh acceptable yeah but I do not yeah so at that time I didn't know too much about the market so I talk to my friend in some like uh ZK um comps and then they like promised me like there will be like a huge demand for your Hardware if you like um succeeding in doing that so so that's like why like we started CC so cc is a um uh is a hardware design company like for zkp like we end to design the some specialized Hard like the AIC for to accelerate the proof generation for

zp and right now we are in the phase of using fpgas and also gpus to to do the work yeah so we'll unpack and like really dig into um what it means to be using fpgas and you know the the the road map map to as6 and then what that'll mean for the industry but while we're still in your background here can you uh tell us the story of like how you found Bitcoin and why uh you decided that this was something interesting and not something to run away from right actually kind of um like yeah so like I uh yeah before Corel like I spent some time in China also like doing some research on cryptography uh yeah from like I happen to know a a friend of um a

friend in that in Japan and his name is actually okamato so like when I first read the Bitcoin white paper the name is lamato I think maybe he used a su to to to write that paper so I read it it was uh sometime in uh 2011 I think 201 or 2011 so it's it was like a while back yeah so that's why like I know this paper I think yeah so like we we actually we try to solve like the cryptographers they Tred to design some like ecat long ago but the previous poach is quite maybe boring I would say and uh this like SOS they he come up

with a very nice idea I'm not sure if it's he or sure what I call it yeah so but but they but they came up with a very nice idea to um to design this like peer-to-peer like payment Network I thought like that was very very interesting and and it has um some relationship with my research and I and I I happen to know someone uh in China that time was like doing some Bitcoin mining so like I saw like Jo join then like it was like 2011 or 2012 at that time yeah so I think like today in 2023 it's really really clear that cryptography and everything that we're

working on is like it's so new and we need to like kind of get get it right but once we get it right it's going to have implications on all sorts of Industries and all sorts of um computer science problems I think one of the like more subtle but more cool things about the the Bitcoin white paper moment is like that was the moment where maybe not you as a real cryptography researcher but us on the layer above that we realized that oh my God for the first time that I've ever realized there's another use case for crypto phography other than just encrypting messages like Tighter and Tighter and Tighter and so um I guess like my question for you is in that moment when you know your friend

who you think might be Satoshi is uh like exposing you to this and you're getting more and more interested what percentage would you say uh attracted you of like the ideological the decentralized censorship like those kinds of things and what part of it like was truly about the tech and implementing cryptography in a way that provides like brand new systems uh yeah so I I would say the like it's a technique uh in the paper like first draw my attention uh yeah so like previous like people all Design This like eash system they are cized system and if people want to decentralize it there's always a way to circumvent um uh to to then to to get it

synchronized uh yeah so like the at least in the very beginning Where was the Bitcoin white paper I found is like fascinating due to the techniques to help decentralize this system yeah it was like quite interesting to see that idea and yeah and also like maybe after sometime like a realize maybe like the desynchronization will also be very helpful for us like to to get around the censorship and all the other things yeah so I will say it was a technique that draw my attention then I found maybe the SOS or the um the idea like in the blockchain community somehow online also with maybe some part of me so like yeah

yeah no no that makes sense and um again while we're still in the background section of this conversation so you found yourself at algorand and I think that um you might disagree with this feel free to I think that many of us in this industry uh really believe that kind of the point of this industry is to build out a the and then build um you know from like this core shared backbone settlement layer and I think something that a huge thing that we saw in the last cycle are these alt l1s which for me uh I don't really see a a role for in the longterm like story of uh you know cryptocurrency and like digital property

but the cool thing about all L ones is they provide a um playground and like a entirely different Innovation space for like blockchain related stuff that doesn't screw with you know our perfect our like Precious ethereum so with that kind of as context like can you talk a little bit about um you found yourself at alaran if there's an interesting story of how you found it let us know but um what did you think that alaran was doing that was really cool and really interesting that um you think is worth like that you took with you in the work that you're um kind of applying now uh yeah so I I would like to talk first a little bit about the Ico summer back in 2016 2017 yeah so like I was somehow

also involved in in in in that uh period of time uh yeah so like all this like n ones as you mentioned they try to come up with a better throughput like a higher throughput better efficiency to kill ethereum but yeah as you can see like I would say like maybe um yeah maybe some of them will have like higher performance all over etherum but like in the end it's all about the community and like the community they they can support and they will come up with like different steps on to like to enhance it and also itself got like involved um like from the very beginning so like when I was at Cornel like actually uh we we met like yeah so yeah

so we we as a research group like met with had it quite often um back in 20 2015 16 there is like um uh cryptocurrency boot camp at at um Cornell and yeah we we somehow like interact with Ric at that time and as you can see like some of the coronal professors also come up to come come out to De develop some like NE ones try to beat Ean on the on the the efficiency side and also like the arand is also um um founded by silver M who who is a a two two award and also MIT Professor um yeah so like actually like I I think

like we we actually we uh the professors like they they bring a lot of the research from the Academia to the industry um yeah this like um like like for instance like uh you can get like uh higher throughput using some of the um the techniques like proposed by my advisor ini and you also you can use like maybe like VF to next connect the blog producer like um so this idea was like by like um m i I guess like maybe uh the um the professors like provides like more formul defined and a more reg proving

idea to to the world maybe some of them are not very efficient or like the when when the professor they implement the these ideas in in the in the community World they are not like very good at that but I say the input provided by the professors are quite important to the to the community yeah yeah that makes a lot of sense so I I it is so clear if you're working in uh proofs or ZK proofs at all that um like algorand or Beyond like these things are really cool but it's really really hard to um like get them done right and there's this all these pieces around the tooling

and the Frameworks and just like the difficult pieces around getting like Advanced Moon maath into your computer but then um as is very clear like there's a whole aspect of performance so we'll we'll move into this piece now but um just like as a preface when you're running these proofs and you're seeing that they're running really slow how much background do you personally have in hardware and how much do you see in this moment that like Hardware is a real answer here um or is that become part of the solution as you you know like socialize and communicate more and more with other people that this is the problem and we need to come up with the solution uh yeah so only speaking like I

don't have much background in hardare engineering like the thing like I all the like perspective I have for this algorithm is to say like first to how parel this algorithm can be and the second part is like what is the interaction between the processor and the memory during the competition so like if I can figure out like these two parts then I can say maybe um also um like very at hog uh uh sing about the hardware friendliness of this algorithm weing so that's uh so that's my like first impression on ZK I want to take a look at the ZK algorithm like I know

like there are a lot of algorithm so I first to focus on this like major components like the multiscale multiplication numers already transform and the POS M tree I I saw like um I guess like maybe the majority of them can be like very per perable like on on the processors and Al like some of them have like quite weird mem access patterns but like we can always using some like Shuffle or other techniqus like to to improve on that and if we can if we can like run this like um algorithms in massively power Parable mode then we can borrow the idea like we

develop for for the Bitcoin mining like recording the Bitcoin mining machine it it doesn't have it doesn't have just the one chip it actually have like more than 100 chips in the mining machine to accelerate that and we can do almost the same same also like we need to deal with the flexibility and also the uh yeah so like the they skin like they are not stable yet so like we need to handle the pro PR programmable part also as well got it so um last question before I ask you just what is cysc and can you describe me what you're trying to do but just I think a necessary piece of background is can you uh help me and help the audience understand um let's

just say the spectrum of CPU to fgpa fpga to Asic and why on that Spectrum um like basically what do those different terms mean and why as you move closer to Asic do you get more efficiency but what trade-offs are you making As you move over there okay yeah so um I would say like we we will treat the CPU as um as a maybe as a like Swiss army knife is like very powerful is hand on out of like functionalities and the fpj um fpj is a less powerful with army knife it's but it doesn't have that

functionality which means it can be like more focus on doing like maybe it's just something and you have more ownership on on this fpj you can program it to do something and you do not do not need to waste any resource on the fpj to do the other functionality that you don't care about okay so that's the fpg and the EIC is basically a hammer it's a very powerful machine but it can be used to do only very limited things it's like yeah which means the um the clock frequency of the a will be like much higher than fpj but it doesn't have too much flexibility inside it can do one thing or like maybe one or two things

like very very well but it cannot be expand to do a lot of things and so is it fair to think of um if a CPU is the Swiss army knife it's general purpose it can do anything that any computer can do and a Asic is a hammer it is very specific it can only do one thing is it fair to think of a fpga as like it's a programmable chip that virtually can act like an Asic but the cool part is you have the flexibility to change it a little bit so you get the flexibility um and a little bit of the performance but that's why it sits like right in the middle so first of all is that correct uh yes yes that's correct yeah it's right in the middle and uh you can

basically program the chip to do something like un like and you don't need to risk the resource on the chip to like yeah since like in the in the CPU it always have some uh it always reserve some resource to do other things uh and you can only account for maybe um um minority of of of the of of the of the CP to CPU chips to do your competition yeah and and again I think this is so critical to what cysc is building so I'm just going to hammer it home the idea here is that like a CPU because it's general purpose has to spend a lot of overhead and a lot of nonuseful uh computation setting up the computation that wants to be done right so like let's say I want to do one plus

two the CPU first has to say okay we're going to do addition okay there's going to between two digits okay okay okay all right now we're going to do one plus two whereas an Asic has none of this um generalizability all it knows how to do is take an input take a second input and do the plus and if you ever want to do multiplication or division on that chip you're screwed like it's in the Silicon that it can only do the addition but the good news of that is you have zero overhead and it's that loss of overhead which allows um as6 to like perform basically the same operations but like orders of magnitude faster than a CPU is that correct yes that's correct yeah so

like in a you basically like hardwire all the circuit on on on the chip and you cannot do like do do something like you haven't plan on on CHP so that's some like limitation on the on the a but on the good side it's like very performant and it's basically can wrong on wrong things like in massive Parable way to improve the efficiency and and also since like um and also it's kind of quite cheap and like yeah compare with like designing or like tap taping out a CPU so like you you can control all the resources on the on the EIC chip like like how how how many PS like you want

to put on the on EIC and how how much like memory you want to have on that so like you can control the budget and like you just planer something like for yourself and you just pay for oh yeah yeah you just you just pay for that yeah and I definitely want to spend some time talking with you about the manufacturing process I mean we live in 2023 like all anyone on a global scale is talking about is Chip manufacturing so we'll get there in a minute but um I think we've got enough background to like Leo can you tell us like what is cisc what is the goal and um yeah like what are you trying to do with uh for ZK technology okay yeah so like uh for cic our end

goal is to design some Hardware that can that can generate the zkp proof in real time so which means like we want to generate the zkp proof for all the major uh proof system and for all the major uh circuits such as like the thek event circuits the ZK ml the ZK gpt2 or like some Al po har in in less than one second so that's our goal and we also want to achieve this goal in a cost effective way yeah we don't want to like spend maybe a half million to to Pro to just uh produce just one machine which is capable to do that we want to actually we want to like reduce the the

the the cost for for the machine to be like comparable with the current the mainstream GPU card the 49 or maybe the 1580 J car that will be relased next year so that's something that we want to achieve the best performance uh in a cost effective way for for doing the zkp pro generation yeah so that's the goal for C6 cool and then um is is the goal for cysc to be creating chips that are um like bought by maybe not consumers but like for example I'm someone who's an ethereum home Staker and so I'm willing to dedicate like home home resources to uh something like are cys6 chips aiming to be in that like kind of

dedicated hobbyist Market or are we looking at like you know the the same as like the Nvidia high-end chips like they really only belong in data centers and like those kinds of um chips yeah so like uh so that's something like we want to achieve we want our our machine can be bought actually by like a normal individual not by this like Enterprise or like the kind end one yeah so so then like we we just want to uh at least replace the majority of the GPU cards that is used in The thep Proving currently with our machine like we since like U when people are talking about this like chips or like the mining machine they have a impression that they

could be like very expensive but actually that's not maybe that's correct for now but with fic ship we want to change the impression we want to see if you use our machine to do the proofing like actually know you can get your money bike in in like maybe in you know like uh accept acceptable period of time uh yeah so that's something like we want to achieve cool and I think um we'll we can talk a little bit later about what you just said is very interesting because uh that's circling back the actual Hardware you're creating with some of like the interesting crypto economic systems that are being built on the blockchain side so let's Circle back

to that but um I think the really cool thing about like what you're building is like the there's a lot of problems with ZK technology today we've talked about right there's the implementation details and then there's the the actual performance and today performance is like so rough that you know we're in the dialup era of like ZK technology right the but like let's f Focus for a minute on that actual application development um and and how that marries up with Hardware so the ZK very very complicated let's take one example like plunk plunk is really like this giant protocol um and

in it there's like a bunch of different things that could be considered ZK there's a bunch of different thing that's just I don't know overhead or um Transformations let's say and then like within plon there's a commitment scheme we'll just use kg commitments for a second within kzg commitment there's some things that are considered overhead there's some there's some things that might change in the future there's some things that won't and we can keep going all the way down the stock until we get to like the most basic Primitives things like um mm uh multi Matrix multiplications or ffts or like the the core mathematical Primitives um and I guess just to cut to the chase like my question for you is like when you're looking at holistic Z

systems how do you think about designing Hardware um that is like really applicable and can really accelerate these systems but doesn't tie your users and yourself as a company to like technology that is so quickly changing and like might be outdated within I don't know like a week right yeah so as you as you said like they if you like dive it in deeper us say like you start with prank and then you go down do the like polinomial commitment skin and then if you choose like maybe kg then you go down to like MSN and DT and then you go down to like mod multiplication uh yeah so like maybe the

like we try to design something that's like quite General like I mean like general for the dkp skin since they are evolving like so fast uh yeah and we still want to accelerate it and but we want to support the mainstream system and the potential update so like to handle that we first need to come up with um with a vector machine and this Vector machine can basically handle this like vectorized addition multiplication and also modular multiplication and yeah so B So based on this like uh based on this Vector machine we have um I will say is um um uh firmware part so like we call

it Z Isa ZK instruction set architecture so this architecture we basically Define like several category of instructions and yeah so like the instruction there can be like NTT like MSN like for 40 different curves and NTT for for for different ways so like we can basically using the instructions uh in the Z ion to construct I was say like the major Cal MSN TT Pro and all the other things uh yeah so this will comes for um uh the majority of the current pist I I mean like in terms of the proofing time but there's still like if

you want to accelerate thep faster then you need to accelerate every possible component inp yeah so the nonta operation is like very complicated to handle and we are still like working on that so so some potential ideas is to see like we can have some like CPU like by CPU I mean like the arm CPU course on our chip to handle this like ntail operation yeah since like you want to have a very satisfying performance but you do not want your like maybe your pcbs to be plugging into some other other ones like servers since the communication between this like two things might be like the

bottom leg in the whole proof uh proof generation you want to achieve the you want to have all the things like all the competition happen on chip to save some bandwidth there yeah as we also mentioned like in the in our Twitter um like we observe that the the major the major bottom neck actually not of neck the KP component is um is um is a bandwidth yeah so like if we can somehow avoid that then we can also have like better performance so that's some rough idea so for the so like in summary like for the major components the previous major component we have a clear solution but for the

ntail uh ones like if we want to have all the competition happen on chip like with are still exploring the uh solution for that got it so just let me uh repeat it just to make sure I understand it what you're saying is that it is correct to think about the ZK system as deconstructed down to the base mathematical primitive and those are like super super accelerat you've already kind of figured them out what you're starting to realize is that the new or maybe the original bottlenecks of this system are no like they're not really about the actual mathematical Primitives the ffts that you're running anything like that it's turning out that the bottlenecks are in the the latency

and the bandwidth of just like moving all the data to the correct chip or the correct part of the process and what you're finding now is that the opportunity is to build these more integrated chips that contain like an fft Asic portion but also might contain some other portions to accelerate the whole process and not just the advanced math right right right yeah so like so like previously like if you run the kg like uh kg uh plunk uh like say using Halo 2 as a software backend the uh the Ms and NT accounts for 30 to uh accounts for 60 to 70% of the proof generation

time for some like even circuit also for other C circuit but like we like like I will say like like several months ago we uh we we found out like using our fpj system combined with a CPU where like you do all the um the msnt on the fpj system but you do some other components some like previous minor components on the CPU actually the percentage of MSM plus NT in the the circuit shrink to if I record correctly it shrink to like 3 to 4% like like it's done it's down from 60 or 70% percent like to

three or 4% so which means like you if you want to have better performance uh you need to accelerate fluffy uh components in that they generation yeah and so as you're working towards accelerating the as you say the fluffier components are you marching down a path that is creating chips for specific protocols like are you marching down the path of creating a kzg Asic or are these Asic still like interesting Primitives that can be built together into lots of different ZK protocols uh yeah so like the our common design can supp kg fry Stark folding skin uh yeah all these like major components uh yeah so like we

like yeah so like if SE as a company just design hardware for for the kg one or the for the or like more maybe like more general for the for the in curve based one it will be actually like quite high risk for for seic as there like a lot of like players in the in the market are using the five based ones or like some other proof system so like we try to support other proof system as well like in our common defi got it all right so I want to shift the conversation here a little bit towards the process and the pain of manufacturing chips so um I think first can you just give us

a like brief overview of what does the chip design and manufacture process look like like let's start from if I came to you as a company and said I want to accelerate this um this function like how do you go from this is the goal to this is like what a design might look like to we've tested it we know that it works to like it's worth investing in a die and then then once you have a die like how how do you get it into like production and get thousands of um chips okay yeah so like the first thing is like we have a spec so the spec like defines the algorithm the input output and also the algorithm and like then we

write the RTL code for that so basically is how how coding part I'm not very familiar with but like this code like it can be run on like fpg and also on AIC we basically need to design like we try to modernize the design like to to come up with some Miz Parts like say like we can maybe divide the the competition into another of like sub component and for each sub component we name it as a p which is process enement the most fundamental module in in the Chip And so like we can see um and after we finish the Aro coding then we we will do some synthesizing on our

end to see like what frequency we can wrong like the higher frequency the better uh yeah so normal it can be running at maybe 1.5 gahz or or something like that so that's so so then we reach into a stage which we call archo freeze so like we freeze all this like front end Cod and then we will work with um like for for for startups like if you are just maybe have around 20 people you cannot afford to to like to to like to have a back backand te or yeah U to to have a backend group in your team so

you will work with an external designs service team who will convert your code into the real circuit down the board and yeah so the the main goal of this like design servers is is to improve the PPA which is like performance power and area of your chip and also the are so like since like we are basically working on the computation part you also need to like integrate create your computation part with like the iio the external IPS so like you need to go to like synopsis or some other IP vendors to buy this like PCI DDR IP which is very very expensive yeah and after you buy the IP

like you handle your code and also you are working with a backand design service team like to um to get a higher PPA for your chip and yeah so that's the second stage and I end of the stage you will have a gts2 and you send this next gts2 to to some Fab such as tsmc or Samsung to do the chip tape out which will be also very expensive and you need to pay like you need to make the full payment in advance like maybe three or four months uh ahead of time uh to reserve the the capacity to to do the chip T part yeah yeah so after after the

chip tie how is done like which takes also like several months to to to process you you get your it's like since like we are doing a full mask chip tape out so we are going to get a wefer the weer like it has like a lot of like chips on your on it like we just cut the chips in piece and then this coming come to the most um uh uh exciting moments is to say whether you can bring up your your chip like whether your chip is fully functional um so the bring up time takes like high on third time maybe like one month like two sever months yeah yeah if like if you are like very happy you can

bring up your chip then the next stage is to say you need to pack your chip into the the PCB which is the 20 circuit board like you design your PCB and you have another like vendor in the supply chain to manufacture your pcbs and also then you assemble the same thing into a machine and then like like fortunately you can dispatch your machines to to to the buyers so that's so yeah a couple months and many million dollars later yeah that's like um if you are lucky then like yeah got it so to summarize what you said is you start with the spec and then what you guys do at cysc is take the spec design it in software

apply that software to fpgas to make sure that it works and it's actually you know legitimate and then you ship off this design to the backend developers who will like actually take your design and like write out the Silicon write out the wires like what needs to be done the back that plan goes to the chip manufacturer which let's be real is just tcmc right now and then um you know there's there's many subsets but essentially they create the chip for you based on the design it comes back to you you have to test it and assuming the testing goes well then it's time to send it to uh consumers or to customers is that right all right yeah so but yeah so like one more thing like like the thing

like we are doing like after we finish all the RT coding we need to basically like first wrong it like fpg and then on some emulator to make sure uh our designs correct like we also like play a lot with fpj to explore the design space like to to get like yeah but the fpj um yeah so like yeah we we try some experiment on fpj but the fpj like if you want to try the experiment like you also need to satisfy the resource like restriction on on fpj so like it's kind of a headache yeah to to get a like

Optimal Performance and also to make sure your design like it's like correct yeah and so as you as as cisic as a company grows is is the goal to uh vertically integrate more and more until you can do the entire design and then just give it to TCA t smmc or do you really like sitting in the like space that of the front end development of the Chip And then working with the the uh supply chain to make it happen yeah so I guess like in the very in the beginning which is Curr which is right now like we try to um have um have the chip to to go to the market maybe as soon as possible so which means we we

can only handle the uh the designs uh the design part we we actually we use the we use fpj to to work with a lot of like supply chain vendors so for the fpj we also like we we have a custom we have several customized fpg machine and only thing that's not coming from us is a fpj chip which come which comes from AMD and all the other things the pcps the machine the cooling the power delivery all like all like designed by us and also we work with like supply chain extensively to make sure we can find a reliable and Trust trusted uh vendors to to deliver the product like in time so

so I guess like for now we are We R or not on the V um yeah to to do the to do some manufacturing for us yeah but but the goal long term is to bring as much of that inous as possible as you grow in bandwidth and uh resources right yeah so actually that's also something uh ongoing since like we are right now like working with like working with some like PCB vendors and for some uh simulation we actually like for the for the power for the heat simulation we have to do it by ourself since like our design s like so specific that they haven't encountered something like this before yeah so this question a little bit out

there but um I'm very curious about uh manufacturing chips in 2023 because this whole genre is like kind of the molten core of a lot of the geopolitical tensions between us and China right now and so I just uh kind of very broad question to answer this however you want but do you find that the G politics in the news is um actually creating enough noise to affect the way you're able to do business or is this like many things one of those things where there's the news version of the world and then the business version of the world and they kind of operate non- interactively um yeah so I I guess like some maybe some something in between

like you need to be cautious about that like when whenever like we want to work with some vendors outside the us NE we need to check with the Expo control to make sure we are compliant on that part so yeah and as you can see maybe uh if you are using a very Advanced node to do the chip tip out maybe you have to wait a quite quite some time for that so like which means like we need to shift our maybe chip tip out to some like more mature and less competitive notes yeah to to to use those those ones yeah so in this uh same vein uh the United States like within the last few years has has

made a major bet on chip manufacturing there was the chips act which put like tens if not hundreds of billions of dollars directly towards like new chip manufacturing or chip designing or chip anything in the US one is that something that's an opportunity for cysc but two more holistically do you see the us becoming an actual competitive place to do a lot of this chip manufacturing or do you see this as kind of a tilling it windmills and maybe a uh a good idea but like probably not going to make a real dent in how this industry Works um I will say maybe in the first few years um the maybe the the chip manufacturer will

still happening in Taiwan as we can see next the tsmc D2 yeah so like the US like the the Factory in us they need to do a lot of experiments yeah so maybe I'll just use a example here maybe it's not a very appropriate example yeah so like when Huawei uh so Huawei is a Chinese company who who is forbidden to do chip tape out in almost everywhere outside China um yeah so so they use like a Chinese chip design um uh chip chip like tap Factory Cosmic so as Sak is like like previous use a lot of like the mining machines to try their

like process try their techniques to make sure the the thing like they try out they like fully functional fully correct and after that they will have like maybe after after a lot of try out on on this many machines so like hway then come to smake to do the real Chip tape power for for their cell phone so I guess like for the United States yeah so like I I know some information like the chip the chip chip factory in United States is going the a very similar route yeah using many machine uh to try their Pro try their technique yeah maybe after after this next few

years like like the US chip factory should have the real capacity to do to do some very Advanced uh chip tip out got it so you you look at what's going on in the US and say like obviously it's chips short term nothing's going to change but there's actually some like real reasons to believe that there could be like a Renaissance and maybe a real decentralization of Chip manufacturing globally yes yeah yeah I yeah yeah I I believe that yeah I mean listen like the the less uh like important tsmc is like I think the less scary geopolitically the world is so um very much a fan of like new nodes in the chip manufacturing

Network yeah yeah yeah cool all right so um I think we've like covered what we can cover in manufacturing here let's like shift like to the the little bit higher level of the stack so I think we've talked a little bit about like we want to create chips in the endgame that um are are accessible to hobbyists and to home people um so like let's talk a little bit about like the the business you're trying to create right is selling chips is there any part of you that wants to buy a bunch of your own chips and be a like kind of centralized proving farm that um can be accessed like as API anything like that uh yes yes so like we are we will also you use part of our trips to provide the service

to our like Partners right now so like we have a little bit more than n ZK project who want to use our servers but right now due to our maybe short of GPU servers or fpj servers we can only provide service to a handful of that yeah but with our chip maybe sometime next year we can provide the service to all to maybe like to like we account for a minor percentage of the computing power for all the the community got and so when you're the customers that you're working with right now do you have any Benchmark of um let's say that they had the world's best cryptographers but like the the world's most average um Hardware

engineers and like infrastructure people and they just implemented code on AWS in general in in CPUs how how much more performant and what how much better can you do than a kind of a naive implementation in that way uh I guess like it also depends on the circuit like for for instance like for for the ZK gpt2 circuit which is by Daniel Kong um using like our fpj combined with the the the CPU mode as I mentioned before like we run the major component on fpj well or the fluffy component on the CPU we can get a 50 times speed up over their like CPU performance and the CPU is a 64

core CPU so basically if they use a CPU to run the to run the VK circuit it takes more than 9,000 seconds I think and using our approach we can we can reduce the PO generation time time to 180 seconds yeah so that's wow yeah so that's one example and for for the V even circuit used by um by some n TOS like we can easily get a a five times speed up over there um over there like GPU performance yeah uh GP performance is us three to five times faster than the CP performance yeah and so this is um with your current solution that that is the

uh fpga plus the CPU kind of like hack together version right yeah yeah yeah yeah that's right so once your chip is ready and you're able to install it in the system like what do you expect the um performance to be let's say on the ZK evm one yeah so like so I guess the current uh yeah we we basically we want to generate they even circuit in this one second they can even circuit in less than one second we want to achieve uh real time proving uh yeah so we like all the major comp major ZK circuit the ZK evm ZK ZK gpd2 not sure about the K gbd3 or gbd4 but for the gbt2 3K we can definitely generate in real time God and

because and this might be a little basic but the important thing on the ZK evm is that you need to generate the proof quickly enough and post it to ethereum that like the all the state changes that just happened on the L to make it to ethereum and not like 20 minutes later or two hours later or like some inordinate amount of time right like that is the real impact that your technology can provide today just like for the very few use cases of uh ZK that are in practice right right right yeah and other than that we also have we are we have a few uh customers who are designing the ZK Dex uh decentralized exchange yeah so like in that case you I

basically want to want want us to generate the they C they prove as soon as possible since maybe if you wait like more than one seconds then the price will change and yeah and then you do some some of your so like that's another case and there is like the ZK bridge and the ZK Co processor like who also want us to generate the ZK proof in a considerable short of time yeah yeah I mean it's it's super interesting and I think I'm a huge believer that when we get like 10 or 100 or thousand X improvements on efficiency we like Technologies evolve right and the best example is the internet right in right

like we started with arpanet which is like literally sending single characters to each other and then fast forward we have geoc cities and then fast forward we have Instagram and then like right now we are streaming high quality video talking in real time and it is that like order of magnet itude Improvement just in the like bandwidth that has like really transform what the internet is capable of and so with that I I'm so convinced that these order of magnitudes improvements like people like you and cisic are making or just the order of magnitude improvements people like succinct Labs or um anyone that's just helping developers abstract out all of

this like complicated math are going to provide these order of magnitude improve ments that transform the technology but the big question mark is what what does the world look like when we can do ZK um instantly and trivially like have you really spun your head around on like in a world where cisc is like this multi-billion dollar Behemoth and um ZK compute is cheap and easy like what kind of world are we living in yeah so like um like the ZK can basically so like ZK is somehow independent of the blockchain so they can basically provide a way for people to interact um like like it's basically um erase the layer of like trust like you

can use like you don't need to trust anyone you can use like you just need to trust the the math like the like we can use the the proof to communicate without concerning about the trust behind the the things and also the ZK has a lot of like I guess like other than the blockchain word they can be also used in the in the maybe traditional word so the traditional work work we can use ZK to fight netive fake we can use thek to improve the the the like the the TPS of some of of these Banks or some Financial institutes and also we can use the to um to make sure the traffic the network

traffic they are satisfy the compliance using some V met box Tech techn manage so I guess like with um with a powerful Hardware like we can find more applications of the cas that have that we probably like we couldn't imagine before yeah so like for like as you just mention like if like the with the Improvement of bandwidth in the internet we can right now have like Instagram or even like T talk like like what was like invation like um 10 years ago like I also um start application called it's very similar to Tik Tok but due to the

snow bandwidth at that time can only support in Instagram so like that Stu was not a very successful one yeah but like with um with right now the high speed like we can do a lot of things and also I believe with the high speed of the power for the machine we can also explore like more applications of that yeah for sure and I know it's always like really tough to say like well what's going to happen once this technology is ready I mean if anyone had that answer we would not be on this podcast we be founding a company right now but I'm just so convinced that like while the the purpose and the energy around ZK today is about blockchain I'm just looking around at the world

changing and like what as AI gets better and every device out there becomes Internet connected and cars are driverless and Amazon's delivering drones or sorry delivering packages with drones and like the world is becoming more connected more digital more autonomous and I don't understand how any of this is possible without some sort of coordination system and I think blockchain is a huge part of it but it's not the whole part of it I think the whole part of it is cryptography I totally agree with that and also like maybe one more reason to start feic is to say like I'm a cryptographer and I I know that like the cryptographers they design like very

elegant um system like like in theory and this system cannot be deployed in the real world is due to the efficiency and they case just the one example and we also have like of encryption and other things like uh yeah so like we so like maybe one personal reason to start phics to bring this like wonderful crypto crypto system like to the real world maybe in previous then like virus sign customized the hardware then then like buy some like normal Hardware yeah so that's also maybe a personal reason like to starts I think yeah all right Leo thank you so much I mean I could

seriously go on for hours here and I really would love to sit down with you and talk about like what the ZK enabled future looks like but um I think this is a good time to wrap it up so um like before I let you go can you just like let the audience know um where can they find you if they want to learn more about cysc where can they find cisic and like most importantly like if you just want if you like us are understanding that ZK is not only going to change the world it is like the um shelling point of the next like just technology Revolution how do you learn more and how do you get involved in Z yeah I I I think like the Z Community is a very like very friendly community and like

you can always try you always find a lot of like materials to learn about the dayk you can always like type in awesome VK and you will find a lot of surveys like a very long list of all the resource you need to uh to learn to to like to code like in in the K yeah it's a very very very good good community and you can always find a lot of like YouTube videos on online to know about that and if you want to find think you can just try type in next I think c i uh and maybe like our our twwiter handle is like c i under score XYZ so that I remember yeah yeah that should be the

Twitter handle yeah great no no thank you so much um so yeah please everyone go check out cisc I as I was telling Leo um over the last few days like the this cisc is uh so important to what we're like what what needs to happen in order to make ZK like a real use able technology and um whether it's in just like making the performance good enough for Enterprise applications or like I think even more importantly wrestling this magic math into something that's like accessible by people without phds is um what I like to call the Lord's work so I just I really appreciate what you're doing I really appreciate what you're building for us and um man I just I hope to to stay involved and and watch

you conquer the world thanks a lot yeah thanks for for for for inviting me uh to to talk about in uh in the podcast yeah of course well Leo thank you so much and have a good rest of your day yeah you too