ZK HACK mini - State of ZK Panel Discussion

Transcript

hey everybody welcome let's check our audio can i hear everyone hello hello hi everyone testing hey hey is that what you're supposed to say i don't know something like that yeah testing three one two three all right so i'm really excited to have the three of you here with me for the wrap-up of zk hack mini number three this is our third session and i feel like what we haven't really done in zk hack is kind of do a look into the larger zk pick what's going on in oh did we lose to roone i guess so oh well i'll continue with my

intro um we don't always get a sense for like upcoming research or what people are talking about what topics is are people thinking about and i was hoping to use this uh session to do that i do also want to welcome people in the chat to put comments questions because we actually with this format we can bring up your questions onto the screen which is kind of a cool little feature we wanted to try out and i do think it would be fun for this to be interactive so but maybe to kick off why don't we do quick intros to everyone here with me so well kobe i think people know who you are but maybe you can give a little bit of a background on what what you do and what you're up up to uh yeah sure so i think some people know

me already here from what i'm doing with zk hack with all the puzzles um but yeah the other things that uh i'm doing in life right now is that i'm a partner in geometry which is a research-based vc fund um and also um cryptography itself cool pratice uh yeah hi everyone uh so we are i'm pathetic i recently wrapped up my phd at berkeley and now i'm working at elio as a cryptographer um working on our

implementation of sexy as well as our proof system nice and tyrone hey i'm taroun um i uh work on d5 stuff uh at gauntlet and then also uh invested robot ventures but have kind of been a zk circler i wouldn't call myself a uh expert but try to figure out which parts of d5 can actually work within ck land cool all right as a kickoff question i wondered um what so i i shared a few of these lead up to our panel but i was curious to ask each of you what do you think the most surprising thing has been in the zk

space this could be an innovation a trend or a new finding and maybe we can cap that as like last year david's asking why is kobe not anymore but yeah i don't know if any of you want to jump in on that um companies which are sort of trying to deploy more sort of beyond just like a simple transactions

to actually support computation i think in the past year we started seeing two three four projects sort of released and i think maybe the other thing is a side of the progress on some of the zika avm efforts that i guess took me less of course i was not expecting i guess they're still set up in the prototype phase but i i'm seeing a lot more sort of announcement on people actually saying oh we can prove an execution off uh solution smart contract uh zero knowledge yeah i think uh i guess oh yeah oh god sorry thanks yeah i think um i think uh one thing maybe maybe not

surprising but impressive that i've been seeing is that um there are a lot more people getting in and trying to to do zk across very very different kind of project and people that you know have not known anything about cryptography and are able to create pretty complex and impressive things and that's that's also due to a lot of other efforts that support them but that's really cool and i think maybe maybe something surprising i don't know if it's last year but at least it was surprising to me that i i was expecting that things like universal snarks would be slower than

than previous ones and but we eventually got to know that long can look up and all these kind of new constructions were eventually extremely expressive and allowed to do very very interesting things very efficiently so i think i think that's maybe something you're about to say something too yeah i think the the main thing to me is uh a people really focusing on hardware acceleration and realizing that that's sort of like a bottleneck that um you know can actually be both designed around but also sort of

you know better compilation tools better sort of like circuit reduction tools things like that like kind of seemingly boring things right they're not like the coolest math but they're kind of like the things that get you like 20 increase here 20 increase here 20 increase here and like you get this compounding of all of these things to make these things a lot faster and i think a lot of that engineering work is starting you know some of it kind of on the pure software virtualmin machine side like on the epm side but also a lot of it i think on the hardware acceleration side we've just seen way more teams companies a lot of people who probably didn't know what their knowledge proof was two years ago but

who are really smart are now all of a sudden kind of like paying attention so thinking about that i mean one of my one of the surprises i i i would put here is actually this this blurring of the lines between starks and snarks and the using of the techniques kind of going back and forth and maybe that's always been there but i've for me at least in the last year it's become more pronounced and um yeah i think the idea like with wonky two and actually the work that jordy was doing on cke vmware you're basically using both technique to something but that was pretty cool so that would be a surprise i haven't my next question actually what have the

biggest disappointments been i think if i if i would put like a disappointment is that we don't yet have a wide ecosystem of general purpose snarks deploy that can talk to each other um i think the closest thing to it right now would be what's happening on ethereum with you know pre-compiles from a few years ago that allows you to verify some very specific kind of snarks um and that's changing like leo for example launching is game changer in that mina

and that's really that's that's going to create a really interesting ecosystem in my opinion and um in um yeah in general i i think all this all the the lack of let's say convenient tooling support that accessible to the people that are not like super strong engineers or cryptographers um that's something that i would want to be less disappointed in later on um

yeah i think i'll sort of i agree with the tooling aspect i think yeah we've sort of come a long way in sort of the technical the proof systems and so on right but um anything like sort of efforts are underway it's not an easy problem right because it's combining aspects of like uh you know human computer interaction and ux and so on um so you can get something essentially usable but yeah i think uh just the lack of great user-facing tooling but yeah i think i think there's good progress being made on a lot of teams um

i g i guess in my from from my own and this might not be like as exciting as hey we need better tooling like or hey we have leo or you know something that is supposed to replace sircom it's kind of amazing that circum has survived and still gets used as long as it has but i think it's surprising we haven't actually seen that many like very clear translations of applications from um normal layer one land right now to zk land like we're kind of like dark force is awesome but i feel like we've kind of we you know there's like this whole thing that's been promised for a long time of like private x and

we just haven't seen that many actually live i guess and part of that's tooling so maybe it's a second order effect but yeah actually another question go ahead oh sorry go ahead go ahead i was going to go on a different topic so you should totally keep going okay and i actually think that part of it is tooling but also it's just sort of difficult to actually construct these private protocols right um because okay it's one thing to know z case routes and the properties they provide but then you have to actually integrate into your application and figure out what is a privacy guarantee you want to go for what is what can you actually achieve with your system um and unfortunately i think you still need some sort of crypto background um to be

able to do that reasonably yeah i think i think that's completely right and i would even add that sometimes you know we think of snarks as like you know the thing that will solve everything but usually it's a component in a larger system and you need to combine different cryptographic techniques or even just composed with other protocols and economic guarantees and all that in order to make something happen and sometimes teams focus you know just on the snark aspect and that's not enough to create this translation that everyone is mentioning yeah and i mean we saw the same thing happen i feel like in machine learning right where like everyone was like

saying like oh like neural nets will solve everything but if you look at the things that actually work in production that scale or like kind of state of the art all of them are like very hybrid models they have some features of different things and you have to understand how the different models work together to actually make the conglomeration i suspect that will be true here but hopefully a little less because the surface area is smaller but rhea remains to be seen there was a question from the audience alex asked do you think it's really the tooling or the different programming paradigm i i would say it's a combination of both because i think you know one of the things that we've seen that is extremely

impressive in my opinion is that the the fact that we have proper zika zkvms operating today and so things like um like cairo and minon and all these kind of things these allow you to actually do go back to to the previous model where you you're able to write if statements and things that you forgot that you can while you were doing snarks for the last few years and like the this going back to this traditional programming paradigm i think it did unlock something for people like they were able to create interesting things like physics engines on top of

on on top of these zika vms so i think it's a combination of both because the tooling is still needed to make this extra efficient things and even engineers struggle to get into that but the different programming pattern definitely is is a large contributor to it yeah yeah also um yeah what henry mentioned in the in this chat is it sort of it's a different state uh model also in the sense that um you have to like yourself in normal execution you just you know you have the computation and you have some input state and you execute it on chain and you get the output state also on chain

but um in all of these zika applications you're just pushing the proof of execution on chain so this makes things like you know concurrent access to the same state difficult um yeah so overall there's a couple of different uh factors and beyond just the two things and i guess one thing maybe add to that is that the side channel effect of that is quite high where maybe i have a proof that proves something about some particular some property some function query on some state but then later in the same calculation i need to reveal some fraction some portion of the state but knowing that this property is true plus also knowing some tiny fraction depending on how how you do the

revelation you may cause a side channel attack and that side channel attack is not a pro fault of the zkp that's like a fault of your data access pattern in your piece of code and you know it's very easy to shoot yourself in the foot i think what what oh there's one more question here are zkps that use trusted setups obsolete that was actually a question i had as well as art can we say that trusted setups are almost gone where do we think we actually stand on that front i think uh we still don't have okay beyond maybe i guess it's like sort of the existing solutions that all lie

on the spectrum either you get like transparent setup with small proof size but linear verification or you get transparent setup with succinct verification and larger proof size in the order of like you know tens to hundreds of kilobytes with the hash based stuff or you get trusted setup and succinct proof size and succinct verification uh but then you have the trusted setup so i think right now we sort of we don't have a solution which i guess meets all three of the requirements of like transparent setup and succinct very tiny proof size and 16 verification so i think as long as we are in that state then i would say setups are not yet obsolete and that's funny i remember kobe you and

like we worked together on the plumo trusted setup behind the scenes and i remember like some of the work you were doing there was like quite breakthrough and i was like this is awesome we should publish and you're like yes except trusted setups might disappear it's like breakthrough for now but will it to be honest i i think i think for reasons that pratice mentioned not not for a long time and also it really depends on the use case so for things that that you're talking about scalability and rollups and you can have high latency and one coordinator that combines everything then it makes sense to use something

like a stark that you can just run the same vm and again and again it makes sense but if you're talking about things that are user facing then it becomes more complex to to argue that because users need to generate their proofs and send them over and put them somewhere and you need them to be both really fast and sometimes really small it really depends on the use case and i think like one tweet that i tweeted yesterday is that you can still have hidden inflation bugs even in a transparent setup so it doesn't really solve all your problems it solves

part of your problems and i think universal setups sit in a very nice middle ground that can satisfy a lot of people for for a long time what is the kind of research or piece of work that has come out recently that really pushed the boundaries or would you consider like a breakthrough that sort of strikes you as as something like in the last year or so curious what yeah if you felt like there was some sort of paradigm shift at some point i i would say

zika evm is like one really like extensive project that is taking a huge task that i wouldn't have imagined possible a few years ago wow i i wouldn't have imagined that people would actually take it on and that's that's really impressive and and one other thing that i'm just like very curious about recently is like this topic of collaborative starts and i think that's a really interesting topics where different poovers have different pieces of the witness they want to keep private and then create collaborate the greatest snark oh wow i think this combination of mpc and

stars could could be interesting but i i don't know exactly how yet yeah no i think definitely oh go ahead sorry i'm sorry i actually just want to ask you a question about that um is is the is the like sharding piece like this secret splitting like the npc splitting type thing here going on is that circuit dependent right now like does how how you do splitting depend on the particular calculation no okay um so i think the yeah so i think i i also think that sort of the collaborative stuff is going to be exciting especially for

you know going back to the um you know private x application design stuff um you know where when you have data split across multiple parties then you really do want to have this um sort of collaborative proven kind of model um i think the initial solutions we have right now are somewhat um uh hamstrung by the fact that you do have to have i mean so they work for like whatever um snark you wanted they can work for plonk they can work for crop 16 etc etc uh because what ends up happening is that you have this linear home office zone which makes things pretty cheap for the prover the problem that really pops up is actually producing the witness in a manner of the secret shed across the parties because you have to go from say

like a mercury path to the variable assignments that correspond to the circuit right and that that step requires standard npc and that is expensive so for some applications you actually already have naturally secret shared witness data um but yeah i think i'm very excited for to see like where the progress goes on that front in terms of having the collaborative uh proving various models for particular problems and so on are there any sort of i feel like we all have a bit of a lag eh sorry about that go ahead turn um

i think uh the you know this might be like sort of a more recent thing unclear if it will find a use case and it might also just be because i'm the sucker for vdfs whenever i see them but i really love the short-lived zkap thing of like giving access control and lifetimes um to proofs it's not like a very complicated idea um i i think it's just more the idea of like hey like there's kind of like a uh you know a way of saying like you know in the same way that you can like revoke off keys in like a normal system you can like revoke sort of provability and i i

really think that that's that's kind of kind of cool idea and actually seems like it might be like a more direct application so i want to bring up a question from the audience from thor who's part of the zk hat community um this is about large companies working in zk for a long time and we rarely hear about them why is that i can tell you from my side of some private sk teams to come on the show often there is multiple hurdles to communication in those companies so even though there's really interesting stuff happening they have they seem to have like sort of

trepidation sharing things until they're exactly perfect in the way that they want so that might be part of it i don't know if kobe or pratich or tarou if you have any other ideas why we're not hearing as much from what work they're doing do you have anything there's a lot there's a lot of uh churn i would say that's that's certainly it's like not zero amount of churn from big companies and i mean the first one listed in the question meta has probably had the highest churn uh but you know it doesn't seem like people either people stay at these places forever or they like leave very quickly and so if

that distribution gives a weird set of outcomes you know i think i think maybe one other factor is that um all these zk systems have been successful and you know that have been deployed at least you know things like zcash and things of that sort they're they require um a complete change in in how things work and it it's hard to to think that you know visa or ui will change how big parts of them work and just in in order to make this compatible with zika systems even even

in terms of file formats or like things that they could read inside the snark or whatever that could work um and it requires an ecosystem to to to start and i i i think it's easier for a startup to do something like that um to make such a bad list um i think that's a really good point about the community the fact is to engage in this world like this part of the industry the way you do communities so different and i don't know if these companies are always ready for that i do want to point out though from that list that there is ey and ey actually does do nightfall which

is like there is a zk project that's like a little bit more championed within their org so we should give them a little bit of credit there that's very true yeah yeah yeah i think the other thing is that um unlike sort of other like most efforts in the blockchain space there's not really okay the properties provided by the zoom knowledge groups are not really sort of that integral to uh critical to the work that these companies do right um in the blockchain space you the point is decentralization so you don't have a single party that is responsible for running the service and um so you need to have strong integrity and privacy guarantees

for like meta or visa you don't really have a big push for you just trust meta to like facebook to run anymore so there's not i feel like there's less it's less critical to their path or whatever to their mission there's another question here that was just pulled up is there anything missing in macy for it to get more adoption especially for qv use cases qv quadratic voting ah okay um

i wonder kobe do you know this um i can give my view um i haven't been involved in some time but i can give my view of where i saw it last um so i think messi is has been deployed in things like clear fund in order to do quadratic funding which is a really nice use case for it um and i think that i would put two things in the front in order to see it what more widely deployed and one is a technical problem which is that currently there is like messi provides you with collusion

resistance between voters um and that works as long as the coordinator in macy does not collude with others and something that would be really useful for this kind of system is that the coordinator itself would get would become decentralized because currently it's a single point of failure for collusion and so that's like a technical thing that i would really want to see improved and might increase confidence in it and and the other is more of a social aspect um which is that it has two phases i guess like one is

that you i think that sometimes this collusion resistance property is not widely understood sometimes people think that private voting is enough but it's not really because you need to make sure that you cannot be bribed and that's an interesting property that um you sometimes miss when you try to do private voting and messi provides you with that but it's maybe not widely understood and the other thing is that you need good civil resistance um because when you do things like quadratic voting or quadratic funding then

you you reward more distinct votes from different entities um and if you do not have good civil resistance you you're gonna be just paying uh in inappropriately um waiting this inappropriately in some sense so more confidence in some civil resistance identity is also needed kind of on that level um what are i kind of want to explore some sort of unsolved problems i don't know

if you have a list of things i mean we mentioned things we we had um it's not necessarily it's like this one was specific to macy but are there remaining unsolved problems in ck that you think we could maybe even point some of the people in the audience towards things that they like research projects things you'd love to see explored i mean applications i've i i'm i'm maybe i'm the hot person harping on applications but i feel like we still have a lot more research on applications and proving

they're secure and guaranteeing like even boring things like hey we don't reuse the same state twice uh when we do use this particular mechanism or like the probability of ever doing that is low enough stuff like that um and uh that's not like a zk problem that's just like how do i you use you analyze an algorithm that uses zk and compare it to the non zk version proving like complexity differences for those yeah i think um but now i'm saying yeah okay maybe touching on that point basically going back to the state model point um

that henry uh brought up uh i think as sort of designing like so far i guess application of the design is a sort of solidity model where you just have a blob of state and you execute the contract on top of that um and you're sort of minor or like part of a block producer sequences uh executions of the same contract in the same block or something like that right i think sort of yeah going back and then i guess i would say in terms of applications the biggest challenges of trying to figure out how you go from that model of like you know everybody touching the same single state to designing applications for sort of the state can be sharded into independent

components and people just touch and different transactions can just interact with one independent part um yeah i think i think yeah for example in the m1 case right now you update the same uh reserves or you track the same reserves after the airmen right maybe there's a way to split it across multiple reserves i don't know to give you some sort of similar efficiency benefits but also then allow it to be to have multiple concurrent emm transactions in the same block they don't conflict yeah i totally agree with all the points and

i think that i would maybe reiterate that we're missing um we're missing good high-level languages especially with the most expressive types of zika systems like because we have already good high-level languages for maybe r1cs but not for blankish organizations um and maybe also in that respect just the fact that um it's super hard to argue security for this kind of systems or not not argue about examine security so some kind of you know wider agreement on what what's

been deemed to be secure and be able to reuse it and so some some kind of not everyone creates everything from scratch all the time and then doing the same mistakes over and over can you give us like on that high level language uh topic actually can you just tell like what what are the high level languages that do exist so far so there is leo that practitioners really well and there are there are like circum and there is noir and there's today things like cairo um i probably missed

like i probably is quite a few but do you feel like that's not enough or they're not developed enough like what's the what's the issue with that because you just listed a couple so that they exist yeah so i i think like that uh some of them try to hide or not try so some of them try to give you a very direct view of how your secret circuit looks like so things like circum are very low level they you literally like write arrows that looks like you're connecting wires which makes a lot of sense because you're writing low level code it's like writing c and so that's very good when you're confident but for example it has no types so it's that problem um

but it's what you see is what you get like you don't get extra constraints when you write circle which is really good because for some people and some other languages like leo try to give you more capabilities for example like conditionals which require creating a lot of extra constraints just to support that but they allow you to write that and but still i think what's missing is a way to create these extra efficient versions of primitives like hashes and brain checks and all these um and be able to use them in higher level languages

i think that the work in progress that i've seen in noir is really maybe a nice point in the trade of space where you still have to modify the language every time you need to add such a thing but you can support this so maybe maybe it's a good trade though so that's kind of the different levels that they see cool i want to bring back this point that made which is improving performance of zkb and wasm is this something that any of you are thinking about or looking at not seriously enough

do you think we should be do do you think this is going to come from because there are some projects like is solano awesome like which tea which which no what oh i know pogba i mean near near polkadot near and polkadot are probably your biggest yeah so i'm guessing i'm calling slightly slightly weird uh if you ever try writing cosmos it's not exactly like writing normal but uh i think that there aren't there are teams on each of those who have been working on proving systems uh but uh empirically they have taken longer than people who haven't so

i'm not sure i'm not sure why that is but yeah it does sound like though if if you want to it's almost like that those those two networks and the work they do on zk or the teams building zk and there might be that might be a good place to start looking for this innovation if if it's happening um there's a question from arielle it's a pretty long one uh i think the i'll bring it up here but it's the idea of full zk evm full evm because it is there is these different types of evm um let's see his opinion is that for full evm not toy including pairings all hashes we need

another order of magnitude improvement to get snarks to get reasonable performance is this correct i don't know if i said that right arielle i hope i did it justice any thoughts on that yeah i think uh yeah right now i would sort of agree with ariel on that perspective um but also sort of the i think the context of deployment of these ek avms is uh for rollups where sort of you pick and choose which applications you want to support um and so you don't have uh you might not even have a contract which has you know pairings or a particular kind of unsupported hash so right now it's not so much of a

concern but yeah i think to actually get the full compatibility we need uh quite a bit more i think i think that efforts are very very early in the space yeah i i would i would agree with that and that you don't really have to support everything but you do need to support a very large subset of the vm op codes and pre-compiles in order to to claim that you have something interesting and that can like you know people can migrate to because otherwise you you will just not fit a lot of other contracts um [Music] and yeah i think that the work that i've

been seeing also tries to you know play with different kinds of modularity that you can use and like you said with the work that jordy is doing for example and the work that lithium foundation and school and all these other teams they're trying to look at how to um how to use different proving systems with different performance characteristics and in combination so that helps a bit and [Music] but also i've also seen discussions on different storage models which don't really affect

the evm up codes but affect other other performance characteristics of the system so that might also help i sort of want to start bringing us back to the question of applications um so i wanna like sort of the so far i mean we talked a little bit about the private privacy choke and transfer maybe but i mean private d5 is something that has definitely been talked about this year there's been talk of like private nfts there's the privacy and gaming but that's all privacy it doesn't

necessarily like zk applications ek use cases could be wider than that are there some interesting ones floating around that that have emerged for you this year do you see sort of like maybe a category coming up or has it been still sort of the same old because i mean to me it was like ck version one was privacy zk version two scaling and roll ups then there's been sort of these smaller cases but yeah i'm just curious if there's like i do know that private defy has been a topic at least that's come up over and over again we've seen teams pitch that to varying degrees of certitude whether they could deliver but like yeah there's definitely been

some interest there but that's just one use case maybe i mean maybe we can start with that but i kind of want to spend some time on this use case thing and i do want to open it to the audience as well if anyone has ideas here we can bring bring up your comments and stuff actually like there's one here the semaphore use case for zkid this is one that i'm seeing here actually maybe we can just pull it up here yeah i think also the park 0x park folks they're doing uh they've been showcasing some interesting applications like the stealth drop

um yeah i think earlier we'll see a few more interesting applications as sort of the tooling matures and people you know start actually building on top of these protocols um yeah in terms of private if i think there's uh i guess penumbra i think is an exciting project um i feel like these books we're going to do like something like threshold decryption i don't know where they are on that front but yeah i think there's going to be yeah a lot of uh more people get experience in ck and um and sort of defy knowledge or

building knowledge becomes more common then i think we'll see more interaction between the two communities and the better design of uh yeah because you really need sort of a co-design of both the t5 and the and the zk stuff to make to get like an efficient protocol yeah i mean uh to that point i think uh auctions still are probably the easiest in my mind use case um you know like i think all the public blockchain and defect primitives that exist now probably are not just a drop-in compatible in any way shape or form i mean obviously amms are the easiest example of that but even the lending

protocols are not there's like a lot of extra state that ends up kind of needing to be revealed in order to do the interest rate calculations um and so you know it's going to have to look different if you want to hide blind the you know kind of like the borrowers from you know identifying the borrower information from like the actual calculation that computes the interest rate at each time um that being said auctions are definitely the easiest um because people have studied blind star where star means like different um different sort of like mechanisms for how you decide the winning price star bid auctions and um that's been studied by economists forever and no

one's ever implemented them because you know in practice you just trust this like you trust google or whatever when you use their adduction but yeah in general i think uh is that my phone talking too much about zero knowledge [Laughter] the board the borg is intersecting um but auctions are actually i think the facilities applications yeah auctions yeah that's that's that's the cia coming coming through um but yeah auctions are i think the easiest thing and for some reason no one's done them and i don't really get why hmm

well what does that mean tyrion what's the auction case that you're talking about i mean even just an nft auction right like i think think about that nft auction is is like the easiest case of like hey there's tons of people front-running people in the mempool and no there's no sealed bid auction right um all of the bids end up being public at to some level some time scale and then you can start front-running people and you see this all the time right um and you know just designing a zk nft auction even though like it's going to be hard to get market share i'm not trying i'm not saying this from a like you know hey this is going gonna be successful product on day one type thing thing that's more from this the point of view of like hey no one's

actually even bothered implementing like the proof of concept of this um partially because everyone in nft land just it's like you know they don't really uh but i i i do think like you know if you actually build these auctions there's tons of applications the first one is like the flashbots auction that's not really a separate option if i run enough relayers i can actually guess certain transaction um bundles like hitting time with like some appreciable probability uh you know other auctions are just like you know there's tons of auctions in d5 all over that could benefit and auctions have very minimal state um and you don't need to choreograph the

state as much and so you know i'm just surprised we haven't seen anyone do it and i think that that's like the lowest hanging would something and probably maybe not here but like would something like macy be useful in auctions or is this would you need to create something like macy specifically for auctions is it a complete new it's a new build it depends like it depends on the auction mechanism so so collusion resists economic collusion resistance is actually dependent on like how you clear the auction like given all the bids like what's the final price and what's the final allocation um because most auctions in crypto minus nft auctions are actually

combinatorial auctions of like people are bidding on subsets they're not bidding on like a single item right so it's like oh and it's block space i'm bidding on subsets of slots in a block right um and so in these combinatorial auctions they're actually like you know there's there's a sort of notion of like what kind of economic collusion resistance resistances over like different subsets kobe for touch were you about to say something yeah i think you also now yeah so i'm just um you also i guess you and there's been some

cryptography for options uh on sort of the npc literature so i suspect we might need to do some sort of specialized mpc but also it depends on the specific kind of uh option mechanism um yeah uh and if you sort of a committed review thing for your bits still i think depending on the specific properties you might want to still have some sort of npc which provides so i think this also goes back to copy's point earlier so that snarks are just one aspect of sort of this private protocol or application design you need to use not just snarks but potentially other tools also

[Music] are there any have any like on this topic of use cases have any non-blockchain ck use cases come to the forefront for you have you seen them have you heard about them does anyone care it's kind of sad like i feel like there's like to me there's this chance that it gets used and all these i mean cloudflare used zk there's like other ways that it could be used outside the block this is another this is another flavor of the earlier question right about the big company you think so yeah

they are using it but we don't know maybe i i think they don't have an incentive to use again they're going back like if i saw out so something to aws right and aws screws up the execution then i like having a proof of the fact that they didn't execute it correctly is not really going to benefit me i just want uh i'll just make my contract and go to like azure or like gcp or something gc whatever uh um right so there's no like in a lot of use kit and i guess traditional tech use cases it doesn't make sense i think one area which i think it could these books could be

useful is um i know folks have heard of transparency law such as certificate transparency binary transparency key transparency all these things and then the idea really is you you know you want you have almost like a blockchain like a ledger of uh something where you put proof inclusion of certificates or keys or whatever right and right and the idea is that people can check if their certificate or their key has been changed or um deleted or without their consent um and they can check it because it's basically like a merkle tree over there over this list um but if you want to check it over a period you have to check you know every time slot this multiple path you could use snarks for example

that to compress that uh that verification time so i think that's one application area where it does make sense snacks are probably too slow right now to handle the throughput of these of these systems um like they do like you know millions of billions of updates maybe billions of updates per thing i want to say i don't know actually but it's like much higher than blockchains um yeah but i think that that's one area if you see progress then it's it's out of the incentive line and uh um yeah it would make sense to apply it there yeah i would echo this kind of blockchain like databases or let's say cryptographically designed databases

that you can later use to to to prove things the inclusion or just properties about elements inside it are things that could be useful if if they are deployed there's a lot of reputation game going on where people still trust a lot of these entities so i don't think that they feel a pressure to do such such extra things um for example i don't know you know you can imagine file coin that gives you a proof of return ability you can do that for your amazon storage and then make sure that you didn't lose your data um but it's a lot of overhead and people just trust them now and

they get compensated if something gets lost so you know some kind you some kind you get you have some kind of slashing i get happening in real time real life um but um i think it's also like like i said previously it's also it requires a big change in how things are represented and how things work um and reputation is just easier for them for for existing use cases i will say that i i thought we would see a bit more movement in identity even beyond blockchain space

um that you know maybe we would see some more adoption into the integration of zero knowledge proofs and identity um that might happen so so i feel like i was gonna move off the topic of use cases but before i did that i actually thought we should maybe visit one of them that has come up a number of times this year and that's the private amms and whether or not zk is the right tool for it drew and i know you have strong opinions here patricia i'm wondering if you're

thinking about this at all over at alio or you're not focused on that sort of application side yet because i know that their zk would be at the core so yeah is zk the right tool for a private amm think it's part of the tooling but uh yeah you need other other crypto also i mean you'll be thinking about sort of a private airman's as part of a larger you know problem of how do you do support how do you support applications where multiple users interact with the same state um so that's what that's how we've been thinking about it earlier um and sort of yeah you could use tools like npc i

think there's this excellent post by radei on sort of various other tools you can use beyond just snarks to develop privacy preserving applications so things like threshold homomorphic encryption um npc all these kinds of techniques uh threshold decryption all this stuff sort of um i think has a role to play in this in this space um but it's a difficult problem to solve there's also some impossibility results like uh what darren um worked on uh the androids like ways to work around it with like again it's differential privacy like what the room worked on stuff like batch uh auctions which

provide some notion of uh privacy but yeah so i think it's like research should be done um and we shouldn't just focus on stocks as the only tool um we have to and there's like a lot of stuff in crypto that we can use and we should use it um i i think like you know what you know penumbra is doing with implementing batches and kind of like allowing people to express strategies but also have like the actual execution ordering being private is like a good hybrid of say like something that's event based where you're

blinding every event like an order book and an amm and i think like end of the day the hybrid designs are probably going to be superior to like hey we just are trying to make it only with snarks and you know you're kind of then handcuffed handicapping yourself um but one thing that's actually kind of interesting is there's a lot of bridges like cross chain bridges that are amms on both sides um so synapse i guess is the simplest example of this um but one reason for this and and and you know the big wormhole 300 million dollar hack is related to this is that let's say i'm going on a bridge between ethereum and solana

well i have to have sort of like some notion of a wrapped version of ethereum on solana that represents how much eath i have for solana accounting and on the other hand i have solana a wrapped version of solana on eth that represents the solana that i've minted on the other side right and the accounting is that they should like offset each other well how do you offset them when their prices are fluctuating well one way of doing that is putting them in an amm on both sides and something like the wormhole attack um zk stuff's actually quite useful for that because one of their problems there was like i mean it's a very low level infrastructure problem of like why you were someone found a bug where you could

double mint and you could basically you know you locked a hundred dies on one side you minted two hundred eighth on the wrapped salon side and those are supposed to be equal but somehow you were able to get double that's like effectively the the simplest version of the attack i think like zk techniques for like solidifying bridge synthetics and things like that and a ms that that's like a totally uh unexplored but like area that we're going to see things happen as like these bridge attacks keep getting larger and larger and like lost value wow hey i think

we lost you anna yeah is he no i'm back yeah you're back okay good okay so i was wondering where does the ck in that case happen is it in the bridge or is it in the amms on either side i think it's like similar to like the like client use case right you're proving to another client like that the state is correct on the other side and it can verify it um and the state here might just be something like did we mince this many synthetic tokens on the other side right it's actually like very it's just a light client use case except bi-directional i would say like both sides are proving to each other i'm gonna bring up alex's spin

his point here it sounds like they're thinking on this um okay i wanna shift gears and talk a little bit about actually funding in ck i feel like and the reason i want to talk about is there's just been some crazy uh announcements recently of large fund raises around zero knowledge tech mostly so far i would say focused on roll-ups or scaling solutions but i i do want to kind of bring this up and i wonder like if and it's funny because i think everyone at this table probably either like has invested in or has been invested in

so we yeah but i'm curious what everyone makes of it is the amount of funding coming into zk a good or is it is it going to the right places is it going to solve the right problems some of the numbers i've been hearing recently are kind of freaking me out because they're so big on like networks that are only like half launched so yeah curious what you make of it i think one thing to remember is number go up is very good for attracting talent okay the hardware all the hardware people i feel like wouldn't have been that interested if zk networks didn't get higher headline numbers if if we're

going to be honest okay brings talent in yeah i think we're gonna go go ahead but yeah saying i think you know i'm kind of biased in this front but i think also the uh i think those are existing fundraisers have not by and large gone to i guess shady or sketch teams i think across the board it's people who have demonstrated in the past and capacity to uh work on zka and defer i guess solid ck products so i feel like it's not as risky as maybe something i see or stuff that happened before yeah

i'm gonna fill it up yeah very nice good um so yeah i would i would agree with those things like attracting talent and the fact that credible teams got the funding that's very good um i think it also allows these teams to do to work on pushing the boundaries in ways that they wouldn't have been able to um if they had to work in a more scrappy way um but it is a risk for them in some sense right because if they don't figure things out fast enough or if they don't time it right with the

market then the fact that they raised all this money and there is no one to accept it or that the valuation is now way too high it might hurt them so in the short term i think it's good and i think that some teams will greatly benefit out of it um [Music] so in general i would think it's a plus and [Music] and i would say that i also what they see that i like which is like different than vc funding it's the fact that we have a lot of funding for you know getting people into this space and all these grants for example like xerox

park and these are these are very beneficial efforts to to get talented people up to speed on what's happening and how to contribute and and you educate and make sure that people work on things that make sense and not just the fleshy things so i think it's uh i really like the aspect that we have a lot of grand funding going around i think that's really useful and that sometimes comes through that like they'll be the big fundraise and then those teams start to give out grants or just are more generous in their funding of education in the space i mean you've seen this number of times

exactly and it helps that the money went to these credible teams that know where they would so that's good speaking of actually education in that case there was a question asked in the zk hack discord about can you hear me okay by the way i know i see my own okay um it was basically about what web3 cryptography should be taught or learned by students in preparation to be kind of contributing so this is almost going back to maybe an undergrad or a master's level like what what tracks should people be taking in order to really be able to build

because also and this this kind of makes sense like just think of the amount of time that goes into the learning at school and at what point like where will it be at by the time they're complete it's almost for four or five years in the future what does it look like what do they need to know yeah i think uh existing uh there's not that many courses which actually teach any of the advanced crypto at maybe like an undergrad level uh that you would need uh so like and e yeah so like you have generally you have like a computer security course where you learn things like

publicly encryption and modes of like something like cbc ecb all that stuff symmetric key block cipher modes um and maybe signing and that's it and you might have at the undergrad level a theory crypto course where you cover definitions and proofs but really the kind of stuff that you need for in the blockchain space is not just knowing the theory uh but actually knowing how you actually apply it right and how you construct systems that use things like uh mpc and cps marks and there's no course i would say right now that actually does this combination much to my frustration yeah

well actually maybe a slightly a slightly different question do you think this course will be created as like an online first course or do you think it will have to come from university because like i could see it going both ways actually for this um i think uh i think it can go both both ways yeah i mean i think probably even if it does come from a university perspective i expect the material to be like online and accessible like the assignments and the lectures recorded all that stuff um like we've been seeing that for some of the uh more theoretical ck courses like um ones by alessandro uh kiesa are sort of it's basically open to

anybody on the internet to sign up um on the piazza and like attend the lectures and so on i mean i don't wanna i don't wanna tutor own horn but i know that zk hack won the original had been used as a boot camp who's in the audience there so that was really promising i i definitely at least envisioned some sort of learning path that's very kind of asynchronous and non-planned out not particularly like academic that could be obtained through a lot of the material that's being developed right now but it's still very chaotic and i also know like it's almost like we do see a pair like

parallel courses like a lot of different projects are doing a learning center of their own kind of bringing oh yay christchurch got a job too that's so cool what a great path um anyway sorry i'm gonna i'll put that one up yeah uh but yeah the uh i do wonder if like if there's not sort of an overlap of education happening right now so there's like a lot of teams who've raised a lot of funds and start a learning center of their own they're kind of doing the same type of education over and over and over again but it doesn't always go to that next level like that early stage entry point seems to be somewhat well covered the like high levels what is it where's waldo

alice and bob that kind of stuff seems like there's a lot of video content around that but not that middle area as much and uh yeah i know we're trying to sort of solve for this with zika hack and and we have some plans around that as well yeah um i think and like one more word about it is that um i think students need to have good good grasp of fundamentals so that they'll be able to understand new things and you know new construction and especially you know maybe being an engineering heart they need to understand what people promise them like paper is sometimes you know talk very abstractly about oh yeah you have a random oracle it's very nice

like what is it how do i use it in practice right and it's very dangerous if you don't understand what you're being promised or what you have to promise when you use the system um so knowing that touchy points is is important and i really believe that learning through doing and like learning through applications is a really strong motivator in to get to understand internals of systems and especially through failures which is again what we do in zika which is like yeah totally um actually yeah we are getting we're

starting to come to the like we're not quite at the end of the session but we are going to be prepping for the job fair coming up right after this um i i know tanya's talking about it a little bit in the gr in the group here i have i had one topic that i wanted to touch on which i think is really quite relevant and then this maybe is our last one i do want to actually welcome anyone to add extra questions or thoughts in the chat as well but this is around like the use of zk privacy tech in the world at this very moment like given some of the things happening i know this has come up in conversation where we were like is zk tech going to be used as a force of good to get funding potentially into

conflict zones where funding's very much needed but privacy is needed or will zk tech be used to avoid sanctions or i don't know like there's lots of different ways that this can be used and i'm wondering if you've been thinking about that um also if anyone in the audience um wait can yeah is it just anna or is it me i think it's just anna did you hear the rest of my question though no oh really okay so i'll start that again then um the question here is

uh do we see zk the shifting gear looking at the state of the world do we see zk as a tool for potentially helping i don't know in conflict zones to get private funding into the hands of people or do we actually see this primate like also being used for sanctions i mean i guess it could be used for both but are are you thinking about it are there ways that we can i don't know in our industry help to direct to direct the development in one direction or another or if maybe we can't since it's all free open source i don't know i guess i just want to say i think the the proper analogy here is

this is like the 1920s when we're discovering alpha decay and like all of these like sort of properties of like heavy atoms and you know i i don't think the point is to like while you're doing that a fundamental thing to start thinking about like oh in 20 years as a nuclear bomb because like it actually will just ruin your ability to uh kind of like isolate you you know like you're not if especially if the applications aren't ready let's suppose this whole industry it turns out nothing works like no real application can ever get built worst case scenario then you know it's kind of worse to have i guess started by assuming like this whole world as ending thing is going to happen and

and you know i think once the tools are there then you have to think about the stuff i think it can cloud your decision making it's almost like it's so unusable at this point we don't need to worry kind of agree and i kind of disagree i think it's valuable to think of um the tech you're building in the sense of you know what changes you want to see in the world and what

changes it could have in the world right so it's i think it's important to think of that not necessarily to like you know self-censor but at least have that in um you know the back of your mind as this is a possible application i will say though that i don't think uh bypassing sanctions is a negative use case for zero knowledge stuff because i mean if you're a rich like oligarchy or whatever who is using this to money launder you're gonna get caught because of other means anyway and all these sanctions end up doing generally just harming you know the the average you know person in russia who has nothing to do with with any energy or like in iran or venezuela whatever right

so i think it i think what we will see um is uh it will be used for all kinds of use cases but i think as alex is saying in the in the chat i think somebody else also has mentioned earlier is that you can i guess what the this privacy tech gives you is sort of selective uh disclosure right so you can say you can enforce even with tk that you know if your transaction is above say ten thousand dollars then revoke anonymity or something or like encrypt the sender receiver and amount to uh some kyc authority for example so you can have average case privacy while still maybe

allowing for disclosure in exceptional cases or in the case of a warrant or whatever and this sort of goes into like a very sort of long-standing debate in the crypto community about building impact doors into your taxes like a lot of a lot of lot of discussion that's gone on in the space i think personally i think that we should be trying to maximize privacy given that the rest of the world is very hard it's working very hard to minimize privacy so i think carving out this niche where we do have a private environment and people are able to talk freely and without um every last sort of fear of censorship from states from other people etc

i think is valuable cool all right well i don't really have any more questions i don't know if there's any more questions in the chat i know there's a bit of a delay anyway but i i guess we should potentially wrap up um i don't know if any of you have anything else you want to share yeah alex's point about the contrarian part i i do feel like that there are definitely teams in zk land who certainly believe that quite strongly that it will improve compliance not worse than compliance

i guess espresso espresso i think exclusively stated this to some extent so yeah i think alex also had this question about uh how do you describe a polynomial commitment scheme to a friend yeah well we can actually put so the the reason this comes up um this is actually this is the zk summit filter question um so yeah i think maybe we can share the link again i actually have sort of our wrap-up slides where i do mention that um i hope to see a lot of you in person in amsterdam in just over a month so hopefully you can make it over i know traveling right now is can be treacherous i know a lot of people feel

i feel like a lot of people went to eat denver and left being like i'm never traveling again but please don't give up on all of us over here in europe maybe you head over um but yeah one of the questions is how to describe a polynomial commitment scheme to a non-technical friend okay now go see if any of you can do it i forgot my answer when i submitted the application um and trying to remember it but other people should go i'm doing the same [Laughter] well we're gonna try to find we're gonna try to find a way to actually showcase some of these answers at the event um

because every time i do this there's so many funny answers and i can't actually share them all so we're gonna try to find something to like show it off but i think what we should do right now is wrap up this panel and um i think some people have already shared the link to that that's awesome and i think yeah we should also wrap up and head over to gather where we're going to be doing our job fair uh somebody's asking is there a live stream for the zk summit i believe there will be on the main stage maybe not the whole event but we're gonna try to do something live um yeah so i hope to see you all there i think we're gonna we're gonna wrap this up now and uh we're gonna share our links in the chat so see you in a bit

see you soon ciao thanks everyone for the panel