Anna Rose: We're all here. We're almost all here. There is one person missing from the roster of co-hosts. Tarun Chitra: We have to shame him publicly. Anna Rose: Well, I don't know. For those of you who came here to see Kobi in person, he is not here. Actually, he remains the elusive Kobi. We're sad that he's actually not here with us today. He will be hopefully here on a panel like this in the future, but I am lucky to have three of my regular co-hosts with me. I guess if you listen to the show... Tarun Chitra: Wait. How many people actually listen to the podcast regularly? Raise your hand. Okay, cool. Just checking. Good background. Anna Rose: What was that about? Tarun Chitra: Like 60, 70%. Anna Rose: It was pretty good, but why don't we do little intros if people aren't, maybe they know your voice and they don't actually know what you look like. Okay, let's start with you Nico. Nico Mohnblatt: Yeah, so I'm Nico. I do research in cryptography, zero knowledge work at Geometry with he who is not here today. He is real. He is real. Yeah, I think that's sort of it. Anna Rose: Cool. Nico Mohnblatt: He who shall not be named. Guillermo Angeris: Hi, I am Guillermo. I am a glorified paper editor at Bain Capital Crypto. I guess I write papers with Tarun and Alex who is sitting right there and yeah, I guess that's kind of it. I don't really have too much to say. Anna Rose: That's all you do? Guillermo Angeris: Yeah, kind of. Pretty much my free time. Anna Rose: You can play piano. Guillermo Angeris: Oh, I play piano. Yeah, yeah, Anna Rose: That's right. You can bake pizza. You do? Guillermo Angeris: Oh, that's a good point. Yeah. Yeah. If you ever follow my Twitter. I have great pizza tips. If you want a sourdough recipe or a sourdough starter, happy to mail you one or something. Tarun Chitra: Mailing a sourdough starter? Guillermo Angeris: Oh yeah, of course. Nico Mohnblatt: For zk11. Guillermo Angeris: Really? We're not Nico Mohnblatt: As a starter. Guillermo Angeris: You got to do what? Anna Rose: Okay, Tarun Chitra: Look, I don't know. I don't know anything about cooking. Anna Rose: Okay, you're up Tarun Chitra: I'm Tarun, founder of Gauntlet, also investor Robot Ventures and have been on this podcast for a long time covering all the non-ZK things and convincing Anna, there're there is interesting things in MEV Land and DeFi. Anna Rose: Yeah, actually, so I mean I think all of you, but especially you Tarun, you've brought on so many guests that I would never have had on that became either people who really influenced my thinking friends. I mean, I think Guillermo was a guest that you brought on and then became a co-host, so that's pretty cool. I think we have a few questions. We were brainstorming what we wanted to talk about. As I mentioned, please remember that this should be very interactive. I just want to find out where is the mic? Where are the mics? You have the mics. Okay, so yeah, if you want to ask a question, just put up your hand and wait for a mic because we want to capture it on the audio stream. A theme that I feel has come up actually repeatedly today. I think you hinted at it. It was really just the ZK use case question mark, will it happen? Where does it come from? What does it look like? I felt after zkHack Lisbon, I was so many applications, so many ideas, but since then I actually haven't, I think I was asked recently, what's the new ZK idea use case that you've seen and I was a bit like, oh, I don't know if I've been bombarded by them since then. And so yeah, I think that's maybe a starting point. ZK use cases, where are we at? Are they coming? Have you heard about them? Do you think they're real? Guillermo Angeris: I think Tarun has a good answer to this, at least a partial answer to this. Tarun Chitra: I think the ZK co-processor kind of boom that started with Axiom last year was sort of hinting at this idea that people want something else out of ZK than just rollups compression validating an L one because at some level that's just not that interesting. If I pick a random person on the street in London and I go, oh my God, we can fucking compress this blockchain a hundred times faster. They're just going to call the police. And so the real question becomes something that convinces people has to have something interesting and needs to, maybe it needs to reveal something interesting about something that already exists. Maybe it needs to reveal, it needs to have some sort of monetary premium in some ways. If I think about object X with a proof and object X without a proof, is there a reason that the thing should be worth more with a proof? That's one avenue of things I've seen. But yeah, I would say I'm wholeheartedly generally disappointed both in the community and people working on this stuff because everyone, every time they start working application, they pivot into working on proving systems. Nico Mohnblatt: I'm a bit confused with the co-processor thing. What do we mean? Is it just executing something off chain and giving the proof on chain that it happened and is this not what we've been doing so far anyway? Tarun Chitra: I think it's more this idea that there's a lack of synchrony in a rollup proofing system. You kind of have to proofs running synchronously continuously over time. But in these asynchronous use cases, it might be like, oh hey, when certain state changes happen, I do a calculation on historical state compute some number, send it to a contract, but again, it still feels like it's infrastructure. Everyone working stuff is like, oh, but it's trying to be less than a rollup and you can kind of see people inching at it. But I kind of think the most interesting thing about going to ZK hackathons, or going to say ETHGlobal hackathons is every ZK hackathon seems to only think about applications that people who understand what a finite field is, want to use, which is not really a great start. Guillermo Angeris: I guess it feels like you're aiming towards what are non-blockchain, but potentially blockchain adjacent applications of zk, right? It feels like every application of ZK is, and in some sense ZK feels very tailor made for blockchains more broadly, but just I want to make sure that I'm understand the question, which is are there non-blockchain applications of ZK that clearly demonstrates some value that is obvious to just your average person on the street? Anna Rose: But I think it could be connected to blockchain. I think it's the end user part that is still potentially missing. Guillermo Angeris: I mean, there's a whole question of are blockchains even connected to the end user in a reasonable way? Tarun Chitra: So a great statistic is I think Flashbots are putting out a blog on this next week is that, and I'm going to, for those of you who cringe when you hear some security vulnerability sounding things that people love using, get ready... but something like 15% of volume to MEV auctions and MEV RPCs and private mempools is currently coming from Telegram bots. These Telegram bots, the first versions of them were people sent you a seed, sent you their seed phrase in Telegram and then you would trade on their behalf. They wrote a little script that would read the seed phrase, then start trading. Then of course they got pilloried for that, and then it became MPC wallet where it was like, "oh, you need two out of two shares to do a withdraw. One share can do certain types of trades", still tons of vulnerabilities, but 15% volume. These didn't exist three months ago. This should just tell you that this consumer wants convenience and wants to not think, and people in ZK application land don't seem to think at that. Anna Rose: Wow. You just on your comment earlier, you did just dis this audience. I thought that was pretty brave Tarun Chitra: Sorry, sorry, sorry, sorry. Anna Rose: Didn't make a note of that. Tarun Chitra: I'm not trying to... Anna Rose: Disappointed... all of you. It's great though. Nico Mohnblatt: So have we answered the question? Anna Rose: Well, we've definitely started the conversation about use cases. I think that focus on the end user, you were trying to get clarification if we meant non-blockchain / off-blockchain, I don't know. Do we have any stats on 'what is the product using ZK that has actually had the most impressions?' And I think I know which one it might be and I don't like it. Uh oh... Tarun Chitra: It is your least favorite application. Is Anna Rose: It my least favorite? That might be the one that, do you know what we're talking about? Nico Mohnblatt: I don't know. What is this? Oh, can we say it? Tarun Chitra: Audience guess. Guess what Anna's least favorite application is that generates Zero Knowledge Proofs? Anna Rose: I don't say it. I don't know if that picked up on the mic actually... we could just leave it as this vague unknown. I didn't say it! Nico Mohnblatt: Actually. Okay. I'll try to rephrase or relate. An argument that was given to me just maybe an hour ago about applications is that it's maybe okay that we don't know what they are, they will probably just emerge out of these things being available and out of having people with a different mindset, people who aren't thinking about finite fields to show up and see like, "oh, I can do X and Y thing" and start doing it. Anna Rose: Yeah. Actually to defend... Tarun, I think your disappointment is fair for this moment, but just picture in the next few months there are going to be a number of platforms coming out with way better tools so people can start experimenting. So far you've had to be so in the weeds to use a lot of this stuff and I mean, to be honest, we don't know how easy, fully easy it'll be right away, but there's definitely strong, a strong push by a number of teams to create those sandboxes or playgrounds or things where you can start to deploy and play and build something fast. Guillermo Angeris: I mean, the contrarian take isn't that just cope, isn't it just like we're all, we've built a bunch of tools and if we build it, they will come, but will they? Anna Rose: Well, I think if they exist then we will see, right? How do you judge when it's super hard to build stuff still? Guillermo Angeris: But the question is, it is kind of asking us to predict the future in a way. But you could imagine the future goes something I don't think it will, which is why I think we're sitting here, but could go something like 'everyone builds a bunch of really sick tools that do all great things very quickly, but then everyone just ends up using JavaScript and not caring about the ZK part of it.' It's like, ah, whatever users at the end of the day don't care about that. I don't know. I don't really have a question on, well, Nico Mohnblatt: We also have that in ZK world, we have a lot of tools that are built to adapt regular code into ZK. If you think of things like RISC Zero, the whole point is forget about the ZK parts, just compile to RISC-V and we'll do the rest for you so we can bring these people in who just want to think in JavaScript. Guillermo Angeris: Yeah, no, I agree with that. I guess you could imagine that for a very long time, the entire purpose of building this ZK thing is kind of lost on someone. It's easy. Let's say it's very easy. It's actually just one line change from running JavaScript and they're like, ah, but it takes 15 minutes to run this on an AWS server somewhere in the thing. Anna Rose: But it almost sounds then the characteristics that ZK brings need to be the thing that you're building for. So then is it privacy again? Nico Mohnblatt: Or the thing that we need to broadcast, we need to make clear what it is that we are unlocking for people? Guillermo Angeris: I think that's a perfectly reasonable thing to say too. I mean it's like look, maybe if not only we make this available, but actually we have to go touch grass every once in a while, which is a fine thing to ask for. I think if you do that, maybe you will get some real adoption. I don't know. I mean, again, my obvious opinion on this is I think it will in the future, ZK will actually be a thing that people actually use very transparently, but the opposite case is reasonable I think. Anna Rose: I think we have a question from the audience. Guest 1: I'm wondering who are the end users in this hypothetical scenario right now? Because in my imagination, cryptography or not just in my imagination, sorry, in my 'worldview', I think, cryptography is something that the end users not concerned about. It is just there and it works. For example, if you use Signal, the average user doesn't care that it's end-to-end encrypted, but it's still there. So are we targeting developers here or are we targeting or targeting the average person on London Streets? Anna Rose: I mean, I guess for that it's like there needs to be urgency for what is being offered and I think there's these moments where something happens in the world where people realize they have no privacy and that matters because something some just horrible story, anecdote, narrative pops up and for a little while everyone caress and they kind of scrambled to get on Signal instead of whatever they were on or on. I think we saw Tarun Chitra: Mastodon? Anna Rose: Yeah, I mean I think there's been these mini spikes of people caring, but it hasn't seemingly been sustained yet, but I think it's still coming. Nico Mohnblatt: So is the conclusion we should fund more Black Mirror episodes? Anna Rose: I think it's unfortunately, I think it's like real world things have to happen and then people seem to care Tarun Chitra: Or you find things that are other benefits of ZK that are not strictly privacy, right? There are a bunch of other benefits you get from verifiable compute and those things should be the things that you push, not necessarily privacy. I am not saying that privacy isn't the main one, but I do think privacy is the one where we'll be waiting a long time or basically infinite variance in the time you have to wait. Guillermo Angeris: I think that another thing that happened, which is how we got privacy, security, cryptography everywhere, is it became essentially indistinguishable from not cryptography. You write a little 's' in your web browser, 'https', and alright, congratulations. You got a nice little green bar that says you're chilling. But at the end of the day, if you ask someone to describe the difference between 'https' and 'http', it's like okay, good luck. So there's also some notion where it becomes so easy and so transparent that it's kind of obvious there's no reason to do anything else because why the hell would you have something that's strictly worse? I dunno, this is another part, Nico Mohnblatt: But we're not there yet right now doing it in ZK. Even if you add a little s and you have ZK it's strictly slower. And by a large amount. Guillermo Angeris: Yes, exactly. Anna Rose: Okay, I think we had a question over here. I'm going to, oh, you have the mic. Perfect. Guest 2: So for a long time governments have been looking for ways to make online voting happen. Why isn't ZK the perfect solution for this? It's verifiable after the fact. I can go and use whatever I got returned after voting. I can use that to verify, oh, actually my vote got tallied in the total. Why isn't that a good use case? Anna Rose: I think it is a good use case. Nico Mohnblatt: It's a great use case and there have been experiments. I remember in France the last Parliament elections that we had French people living abroad could vote online and it did use some form of zero knowledge proof. It wasn't fully ZK and fully bulletproof. There was a lot of trust in the government, so do this correctly, but there was an element of ZK. Guillermo Angeris: Sorry. What part of the voting problem does ZK solve? Actually, I actually have no idea? Nico Mohnblatt: So prove that you're a registered voter, that you haven't voted yet, that you're above a certain age that you're voting for one of the eligible candidates without revealing who you're voting for, who you are, et cetera. Anna Rose: Yeah, I would actually to continue on that idea, I've also heard about not quite projects in the works or even they're sort of thinking. There's been basically inquiries by governments to learn more about that. I mean that's even come through some of the things that we've been doing. I've heard about that. So I think your point is good. I think the timing is not there yet. Also, I think tooling for a long time was not there and I think trust in it is not quite there. I think there needs to be sort of great tools that can be well understood that are super well vetted that people feel comfortable with somewhere out in the world. Then you do a project just to try it out with some government somewhere and then that's kind of proven and then that's how it would happen. I actually think we are kind of at stage between stage one and stage two of that just not yet. Nico Mohnblatt: A few comments. It's like online payments at first. People are like, don't pay online, don't put your credit card information ever. And today we're fine just writing in our card details in a website and just throwing it off. So there's been some kind of mentality change of yes, I can trust that if I put my things here, Https we're all good. The other thing is there are still challenges that aren't resolved and one of the big ones is civil resistance. How do I make sure that they aren't the same user? Making a bunch of bots essentially and voting and ZK alone doesn't solve. And I think we're still trying to figure out the best way to do it. Tarun Chitra: I also would say, I think an online internet native institution would be the first real consumer. Governments are just inevitably going to have too high a bar and they have to cater to the least common denominator to be able to use the system. And that places huge UX constraints on the developers of these systems. To the point about the Https and the lock. The lock took a long time and it took a lot of standardization, took a lot of browser changes. I kind of think ZK just hasn't figured out what its lock is. It doesn't have a way of proving to the user that without this proof, you get the wrong answer. Right? There's no UX for it. I mean, using a ZK mobile wallet makes me want to just commit like a horrible experience. Anna Rose: I think there's actually a question right here Guest 3: Regarding jumping back on the privacy aspect of ZK. I think the focus is too much on the consumer, which might doesn't care in the end about privacy, even though you might want it. So you might want to shift the perspective to the businesses. They do have the incentive to not have to store all this data because of the regulation, especially in Europe with the GDPR. So if you give them the opportunity to not having to store all those confidential data and just have single proofs for those, especially for example, proof of identity, then they may be pushed financially speaking to actually integrate it, which is good for the user and good for them. Anna Rose: That means though that we're going to have ZK SaaS products. Oh dear. I mean not, oh dear. It's a thing. Whatever Nico Mohnblatt: ZKYC, right? ZKYC. So you sign up to some exchange rather than having to send all your personal information and they have to respect GDPR and save these things correctly. That's really good use case. Anna Rose: Although, I mean for businesses it could be all sorts of things. I think you can also, you don't want to share, you want to pay on chain, you don't want to share the incomes of everyone in your team to each other, whatever, or your suppliers or there's always been these ideas of how to do that. Guest 3: That also applies to it. For example, if you want to prove that as a business, you have certain financial abilities, but you don't want to share the entire bank statement or everything, you can just make a proof as well. Anna Rose: It is crazy, there are no businesses or sorry there's very few, I'm sorry to go back to SaaS businesses and I know they're very untrendy. Nico Mohnblatt: So much 'SaaS' Anna Rose: But there aren't very, I think there's a couple projects I've seen in sort of around the ZK space who hint at doing something like that, but it's just very untrendy right now in our ecosystem it seems. But maybe that's the way that happens. You almost need a bridge. It might be like living on some network deep cryptography. You have some SaaS business that interfaces to an existing businesses. They don't need to know what's happening under the hood and the middle. It's this middleware, Ugh. Tarun Chitra: I will, Anna Rose: It's not why I got into this industry Tarun Chitra: Will. Anna Rose: It's a needed thing. Tarun Chitra: I will say from the perspective of businesses that have to cooperate with each other, share some information, but not all their information. For some reason, almost all the products that do that use MPC. I don't think there's that many that try to even consider ZK. Anna Rose: But are they there? Do those businesses, the full businesses? Tarun Chitra: Yeah, it's more like Google and Apple have an MPC thing between and CloudFlare, right? It's actually between data centers and it's not a very consumer use case of MPC, but it is actually quite a big use case between businesses that don't want to share all their information. We want to share some targeted thing, especially for incident response and stuff, but it's extremely niche, right? It's like these tech companies building their own little private MPC stuff between them. But yeah, there's no ZK in that. Anna Rose: ....I think Mary had her hand up first... Nico Mohnblatt: Front running... Tarun Chitra: After the ultimate rug pull... Guest 4: Sorry, this is actually going back to the online voting thing. There are actually some technical concerns with that which would make me really not want to recommend recommending that use case. A big one being that if you have people's votes encrypted and public, then there does not exist an encryption scheme that is secure today that we know will be secure 10 years from now that just does not exist. Tarun Chitra: Yeah, forward secrecy is definitely a hard problem, but there are expiring ZKP's. So yeah, Anna Rose: I mean that topic, we did this episode recently, and actually I just had a conversation before about this, the post quantum cryptography and then the cryptography for quantum and yeah, that's, that's the crazy thing about even all that we're building, if it gets broken eventually, then all those past secrets could be revealed, which is kind of terrifying. And I know that maybe wasn't what you meant, but Nico Mohnblatt: Yeah, with zero knowledge, it's not the case. The privacy guarantee is your knowledge is perfect, but with encryption it is the problem. And that's why encrypted votes is a problem. Encrypted data about your genome. If you've done any of those 23 and me tests, I dunno if that cipher text is public anywhere. Someone can keep it for a while and try to hack at it later and they'll find out. Anna Rose: Yeah. So I guess you weren't mentioning that though. You were just saying there are no systems today Nico Mohnblatt: Same problem Anna Rose: That you'd feel comfortable with. I think you have the next question. Guest 5: Okay, so you're talking about new ZK use cases and exciting things, but maybe it makes sense to kind of focus on the old ones because is everybody is like, well, ZK is for privacy, kind of like let's forget about sexiness and all this stuff because it's not exactly ZK, but whatever. Back to privacy. So we are working hard on creating new proven systems. It's like new proving system every three months or something like that. And it's fine, but maybe it makes sense to work on abstracting it away from the user so that we don't have to look for new ZK use cases, but we can develop a stable, I dunno, structure for it so that users don't know that they use ZK for privacy or whatever. They just use something and we just work on making this something accessible. And then behind that there is a ZK and I guess for that it requires to work on the old use cases. Anna Rose: Do you mean standards? Guest 5: Not exactly standards, but I guess more thinking about how to improve what we do now. The project that we have now that provide privacy instead of creating a new project that does something like, I don't know, private voting, private something. It's like you have chairs and you don't come up with new exciting way to use chairs every week. You just make it as comfortable as possible and then people just sit in it and they're like, Anna Rose: Isn't that kind of happening? Don't you feel like some of those teams that have built a lot of stuff, they're like they're waiting and hoping that people will start building on top of them, so it's abstracted away from them? Guest 5: Yeah, I mean I agree. It just kind of felt like people feel a little bit upset about lack of new exciting ZK use cases. Guillermo Angeris: I mean, in a sense that's what zkVM's are doing. They're simply abstracting the underlying ZK. They just say, look, we give you a guarantee, you compile to X language, whatever it is, or you write X language and we'll interpret it for you. And you don't ever have to think about ZK except in so far as you do something like whatever proof verify and then that returns true or whatever. So I think people are doing that. I mean I think there's still plenty of unfortunate technical details that do have to go into some of it. I mean with zero being one that is trying to avoid those as much as possible. But many other VMs certainly are. Some are watching specific, some are not. But yeah, I think it's starting. Certainly it's recognized that this is a huge pain point. Why the hell should I have to give a crap about the underlying field of whatever proving system might give a crap about and when in reality I should just be able to write as in old fart like C or something and have that correctly execute and compile with whatever memory. Tarun Chitra: Yeah, I mean another way of thinking about it is that there just hasn't been the tensor flow moment, even though now everyone shits on TensorFlow. If you think about it, in 2012, that was like 2013 when it came out. It was like that was the thing that enabled a lot of people to go try training models who had no training or anything. They just followed some guides and they were able to bullshit their way through something and then it seemed to work and then it kind of spun out and grew. And then of course people made better and more easy frameworks. Literally, PyTorch is extraordinarily easy and there's a sense in which no one seems to care that much about doing that in ZK land, not me, not no one, but it definitely feels like everyone goes for let's have this amazing functional language and make someone have to think so much about storage structures. And I just don't think any of these people ever fucking talked to a JavaScript developer. Anna Rose: I want to grab back the mic, the theme because we had a few questions and actually you sort of just hinted at one of them, which is that definition of ZK, which I think is kind of controversial. I feel like I have a slightly controversial take on this. So a few months ago, I guess it's now, a few months ago, Justin Thaler at a16z published an article with these misconceptions about ZK. And one of them was very much about the word ZK, the way it's being used. We just hinted at it with ZK zero knowledge for privacy versus sort of parts of SNARKs for succinctness. It's actually not ZK anymore, but it still falls under the umbrella. So I just want to, since I have a room full of ZK people and Succinct people, what do you make of that? Should we still use ZK as the umbrella term or not? Guillermo Angeris: I think there's a type system and at the highest level when someone says ZK, and it depends on the context in which it's measured, right? It's like an anthropological singleton. Tarun Chitra: Singleton Guillermo Angeris: So this is a singleton and then there's a partial order. But the point is when someone says ZK, and it's in this context, it's in good company and polite company, you wouldn't want to say this, but in this company we're all a bit weird. So it's fine. It refers to everything. It's fine. You say ZK, Succinct proofs are all the same Succinct arguments, whatever you want all ZK. The second that you step out into the real world and you say ZK to someone you've lost, you've automatically lost. That's when you put on your nicest whatever, fancy dress or whatever, and then you go to someone, you're like, okay. We talk about Succinct proofs, we talk about Succinct arguments, and when we talk about zero knowledge, we really do mean zero knowledge. Everything else just parses differently in a different context. Anna Rose: It's just wait. So when you go out in the fancy dress, you have to use ZK properly? Guillermo Angeris: Have to use ZK properly. That's correct. Anna Rose: That's so interesting Nico Mohnblatt: Those are the people who might get confused. If you say ZK to someone who can't make the difference and they think they're using something private when they're not, there's a problem. Whereas probably in this room, if I say ZK to people, they'll know to check, oh, is this privacy or is just scalability? Anna Rose: It's interesting. Tarun Chitra: I think maybe a somewhat more pragmatic thing that maybe people in this room might appreciate more though, is that ZK as a shelling point for a term that investors could glom onto was extremely successful. Exactly. True. It was very successful for fundraising. And every single person in this room who raised any type of capital, owes it to all of the people who basically branded everything as one thing and then sold it as a bucket to limited partners and other people as a thing to fund. And fundamentally, that actually funded a lot of this stuff. So you can kind of bash the non-cognizant who get confused, but half of you don't have jobs if this isn't the case Anna Rose: So on board with that. Nico Mohnblatt: No, I agree. But it's more like protection and avoiding the foot gun of calling something private when it's not. Tarun Chitra: I'm just saying it's the type of thing that you can't go backwards on. Look, people turn this meme and the meme is like, what? Nico Mohnblatt: ZK, ZK rollups, right? You start going into, Guillermo Angeris: So sorry. I guess my main point was I would not deem talking to investors, polite company. Your entire argument is predicated on this. Tarun Chitra: Okay, fine. I like that you're dissing yourself there. That's part Guillermo Angeris: As well. But you know.. Anna Rose: I'll add to this. So I think you highlighted the fundraising part, but I also think that catchy short terms like MEV, like ZK also capture mindshare even when they're kind of peripheral, maybe not exactly. And so I, as someone who's named a lot of my company's products, newsletter identity with that first letter Z, those first letter ZK, I'm so guilty of it. You all know the Zero Knowledge podcast when it started wasn't about zero knowledge. We had zero knowledge. It was about scalability. No, it wasn't even about that. But yeah, my sense is the comradery, it is a bit of a branding thing at this moment or so far. It's also seemed like a good center point for attracting talent and people who want to learn and get excited. And there's a community and that's special. And I really think Tarun Chitra: You're a fan of the vibes, the ZK vibe. ZK stands for ZK vibe. Anna Rose: I think it's very obvious that I'm a fan of the vibes. Nico Mohnblatt: So in the same way we have the https, if the privacy equivalent of this s is ZK, then we have a problem. If we have another way of signifying privacy, then I'm happy to call everything ZK. Anna Rose: I mean, I think that point, and actually what you were saying is out in the world, one should be accurate, especially if the properties of ZK are being very, very much communicated and then they're not being delivered. I think that would be very dangerous. I think that's what a lot of the arguments against calling everything ZK say. But I think what might happen instead is that ZK as a term just becomes kind of, it's not immediately. I think that's actually happened. I don't know that ZK is associated with only privacy anymore. So it as a term does not signify private. Tarun Chitra: For the record, I blame the Rollups. Yeah, they literally caused this. Wait, what problem? Blame the Rollups. Basically were like, Hey, we're generating zero knowledge proofs and they call themselves zk. Yeah, yeah, sorry, sorry. They generated SNARKs, but then they basically started branding themselves as ZK, and then they raised a ton of money doing that, and they've convinced everyone who buys tokens that ZK means Anna Rose: Actually Tarun. I think it went the other way. I think a lot of those teams were at zkOne, zkTwo, they were ZK companies. And then they focused on this actually, and there's a lot of examples of this fair. They started to focus on the scalable because it was less scary and more, there was more potential upside. Sorry. Tarun Chitra: May be right. It's true. It goes back to fundraising Anna Rose: A little bit. But the names remained. The names remained. And actually a lot of those teams were incredible contributors to early libraries. They have the legitimacy in a way to be in, they were in the space, they were there early, but they just shifted. And I think that's where that weird legacy comes from. Nico Mohnblatt: There was also some teams playing on the ambiguity where the website would advertise zk-SNARK, zk-SNARK or zkSTARK zkSTARK. But then the actual system that they're building and that they get people to use ignores the ZK part. So I don't know. Anna Rose: But yeah, I mean since then I think there's been teams. Nico Mohnblatt: Yeah, a lot of people have rolled back. Anna Rose: Some of the older ones, they were originally zero knowledge, zero zero knowledge, the ZK ZK rollups. Guillermo Angeris: I guess, yeah, maybe the highest level of way of thinking about it's ZK is just like a term of endearment. That's maybe the best Tarun Chitra: Way of Mr. Mrs. Doctor Nico Mohnblatt: And ZK. Yeah. What's happening in the ML space? Do we have a lot of people branding things? ML, which isn't ML? Yes. So how do they solve it? Tarun Chitra: There's a ton of AI companies that have people in the Philippines doing all the work. Anna Rose: Are we looking to ML to solve our problems? Nico Mohnblatt: No I'm just trying to see if this problem was somewhere else. Tarun Chitra: So I will say one interesting kind of observation about this is people who are doing layer ones or pure new chains, they actually seem to respect the privacy aspect of the moniker. But people who are doing rollups don't give a shit. And I know some of you will take offense to that, but fundamentally, rollups only really care about this compression aspect with SNARK. Anna Rose: Except for Aztec. Tarun Chitra: Okay, fine. Most roll-ups, they really don't care because it's also fundamentally not, they have to do so much more work to be private versus an L1 where they can build it in. So there was also this kind of fork philosophically between the two of them that I think caused this insane. But then they also both need to raise money at the same time, so then they sold the same term. Nico Mohnblatt: Back to the fundraising. Tarun Chitra: I think it does bow down to that. Anna Rose: I wanted to continue on. There was a topic about ZK crossovers. So you kind of mentioned ZKML. I'm curious about what your thoughts are. You've mentioned a number of times like MPC, there is a ZK MPC crossover. There's ZKFHE, there's ZK MEV, there's ZK DeFi. I guess that's ZK MEV. Tarun Chitra: Zk DeFi is more legit than ZK MEV at this point in time. Anna Rose: Okay. What are those two things? Tarun Chitra: Given that everyone just wants to use SGX, but I mean this is just an honest truth. Anna Rose: I got MEV-TE. Tarun Chitra: Yeah, Anna Rose: Different crossover episode there. Tarun Chitra: I feel like you got to make some swag for this, Anna Rose: All the crossovers. Tarun Chitra: The crossovers. It's like a shoe drop Anna shoe drop. Anna Rose: What do you think that might happen one day? So into that, have you guys seen our swag game? It's getting good. It's really like socks. Socks, I dunno, books. What was the cool thing that we did? What's the cool what? Tea bags. Tea bags. This time. Yeah, we're in London. I dunno if you guys saw that. Those little squares are tea bags for ZK10. But yeah, I think going back to just that crossover, ZK crossover episodes, which ones do you think are legit and which ones do you think are kind of like, Nico Mohnblatt: I wouldn't say there's legit and non legit, but some of them are ZK as a thin wrapper over something else, but we're just doing this thing that we used to do in a ZK circuit and other ones where we are just changing the things that we're doing. So ZK and PC and the ability to generate proofs as a committee, that's new and that's very interesting. ZKML is very interesting in its promise and what comes out of it. But in terms of the research, it is just like, can I run a big ZK circuit? Anna Rose: Is ZKML real? Nico Mohnblatt: Yeah, yeah, absolutely. And it's a wonderful engineering effort. I don't know if it's a cryptographic effort which ZK came out sort of hints at maybe, or maybe that's me reading too much into it. And again, reading ZK and getting angry. Anna Rose: Well, I mean you're saying it's like a wrapper, do you mean the work is not in the ZK part? Nico Mohnblatt: The work is in building circuits and adapters for ML to be compatible with ZK proof systems. And that's super hard. The work isn't in the ZK itself. Anna Rose: Which side are they adapting though? Are they adapting because kind of heard both adapting the ZK side or adapting the ML side, but I mean I don't know what's got more potential. I feel like ML must be way more developed. You've got to adapt ZK, right? I'm assuming. Nico Mohnblatt: I don't know. So far it's not like I've seen drastic changes to proof systems for ML. Anna Rose: It's okay. So they are using existing... Nico Mohnblatt: Some incredible uses of existing arithmatization, like super smart ways of doing it. But it's not like someone came up with a proof system specifically for ML. Tarun Chitra: Guillermo and I were actually just having this conversation earlier, which is that in fact it's actually a travesty. No one is making specific proof systems for ML because things like folding schemes are perfect for matrix multiplication and doing optimizing for these operations instead of optimizing for a VM. It feels like you're kind of like there is simple calculation, shoving it into a big circuit, big VM just so your general purpose thing can execute it. And then generating a small proof versus special purpose proving. And in some ways ML is in general, statistical calculations are just way better when there's tons of special purpose stuff done. Guillermo Angeris: I think that's right. Yeah. Fundamentally there's, when you build a processor, the entire point of the processor is that it can execute anything. But this is not true of ML. ML isn't executing anything. It's executing a very specific set of instructions like 500 million times. So fundamentally it's we're doing all this wasted effort, reducing one thing to another and then doing the other reduction and whatever weird up until we get to the end. But we don't really need to do all of that. It just advises to get good. Anna Rose: The comment you just made about the folding though is the work unfolding leading to proving systems that would be better suited for ZKML. Nico Mohnblatt: So there is this one repo called Zeta, which uses folding to sort of compute different layers of the ML, whatever's going on there. I'm not familiar with that, but they essentially, yeah, but they use folding to sort of scrunch that down into one layer and then prove just that. So yeah, there are more integrated use users of ZK tools. Anna Rose: Interesting. Nico Mohnblatt: And there's more to be done because these proof systems for generic circuits, you pay an overhead by making it generic. We don't need that. Tarun Chitra: And there's a lot of known optimizations in the ML side that a lot of the calculations are actually very repeated patterns. Most of the models have computation that's repeated multiple times. Even something like a boosted tree or just forest to boosted trees. Those are very regular calculations. They're not extremely irregular worst case behavior calculations. And in some sense it's almost a travesty that almost a lot of the work in ZKML has definitely focused on getting out the initial prototypes, which have been take a really simple calculation but then put it into this very complicated machine. And that translation layer actually adds a lot of overhead. But I think I suspect by next year we'll see some very good special purpose ML. Anna Rose: I do want to remind the audience that if you have any comments or things you want to add, just remember to put your hand up. We'll call on you. We'll bring a mic over. I want to kind of keep going with those crossovers though. So we've covered, you say ZK MEV. Tarun Chitra: Doesn't exist. No, no, no. I was at an MEV conference giving a talk and I made this point that was almost all the mitigation mechanisms of different things people were talking about were pretty useless unless you had FHE and they were like FHE, no we're fine with just using SGX for everything. So it's just like I don't think people in MEV land, they only care about performance. They don't care about verifiability in a strong manner. So it's culturally distinct and don't care about it. Nico Mohnblatt: That's a good way of putting it, it's anthropologically different. Tarun Chitra: Completely. The value system is like orthogonal. Anna Rose: Okay. ZK, FHE. Yeah. So last year, I mean I think at the last summit we had Ravital speak on ZK and FHE, and actually I had had her on the show right around that time. She introduced me to the idea that there's such thing as what is lattice based ZKP's, which you kind of need if you're working with FHE, similar thing, right? Where you're adjusting the ZK proving system, the kinds of maths underneath so that they match to one of these crossovers. And I mean, she's very pragmatic and she's like, it's not ready. It's going to take some time. But it made me interested. There was not a ton of work at the time. Nico Mohnblatt: Do, what do we mean by ZK FHE I give you a proof that I ran something correctly and that's something happens to be unencrypted data. Anna Rose: Yeah, I wouldn't be able to define it. I think she defined it well on the episode, but it was definitely using them in tandem. Maybe you just said it correctly. I'm not entirely sure. I just remember her talking about altering the math of on the ZK side or yeah, the type of... Nico Mohnblatt: What I just described is again, just a thin wrapper of ZK around what FHE is already doing. So I take encrypted data, I do some calculations, I give it back to you. You can decrypt if I now prove that I did my calculation correctly, I'm just doing just ZK as usual. Anna Rose: By the way, if somebody here, we're talking to an audience of experts on this thing. I see someone going like this. Do you want to say something? Do you want to put your hand up so I can find you Nico Mohnblatt: Aware of FHE stuff? Anna Rose: Yeah, I think we just, Guillermo Angeris: Yeah, define it for us. I still actually have no idea what this is. So Guest 6: I think the definition's correct, but it's hotter than that. Of course. Okay. Yeah, because it's not just a thin wrapper. I think that, I mean there's probably cryptographers in the room that can go into more details, but the math don't really align. So you can't just use existing proof systems and hope to prove an FHE circuit. Guillermo Angeris: Sorry, here's the dumb question. So essentially what it is is we can think of kind of FHE, you can do very few calculations and then essentially you have some sort of oracle call that says, I need someone else to perform a computation, and you provide me a proof that the computation was done correctly. And then so is that correct? Guest 6: Yeah that would be this kind of things, but I guess I'm not too sure what Ravital is talking about, but I guess the low hanging fruit in a way is that you do need some form of proofs to make some FHE use cases work. Because if you can, as you were saying, if you have an oracle and you can ask a decryption of an FHE cipher text, which could be like the result of something that has been computing with FHE, you can spam this oracle with correctly formed cipher text, but that are random in a way, you can very easily extract the secret key of the FHE scheme. So you do need some kind of proof that you knew the original plain text that was encoded to lead to this computation that produced the results. Otherwise you screwed and very quickly screwed. Guillermo Angeris: Interesting. Okay. Guest 6: And Tarun, I disagree. The MEV people care about FHE. Tarun Chitra: Oh, no, no, no. I think there are people who think they do, but you go, if you talk to them and ask them what they want out of these systems and why they're engineering them a certain way, they give you a completely different answer than their philosophical answer. Guest 6: Define MEV people then. Because I think what you mean is that the MEV searchers, they only care about the speed. Tarun Chitra: No, no. I literally am talking about people designing the auctions, right? The people building the auctions, they all are like, we like FHE, but it's actually too annoying for us to use. So we're going to do this SGX thing and then hope that one day we just cross over and it works perfectly well. Guest 6: I care about FHE Anna Rose: And you work in that? Guest 6: And I'm with flash bots. It's a pragmatic thing. It's like we care about making things available for end users and whether we like it or not, it's true that SGX is probably the most pragmatic choice today, but we would love to have FHE and probably down the line it probably means some kind of FHE plus ZK if we could. But of course it's like way, way Anna Rose: Triple threat Guest 6: In the future. So we'll see. Anna Rose: All the acronyms, MEV ZK FHE, no one will know what we're talking about at all. Very good. Guillermo Angeris: Maybe what Tarun might be getting at as well is you could probably create systems now, which FHE is very limited in what it can do. And I guess the point is instead of creating complicated systems, which you can then just stick into a TE and then hope FHE gets good enough that you just pour it over. You could imagine creating fairly, you have to go pretty deep into the guts of these systems and construct things that can be delegated very small tasks that could be delegated to FHE, just enough to give you privacy, but it requires rethinking what auctions or things like that. Tarun Chitra: Doesn't look the same AST changes. Guest 6: No, absolutely. I think that's the right way of doing it. It's like either you want to read something quick and then you wrap the whole thing in the TE or you manage to carve out part of the protocol part of the use cases and you make it happen into an FHE circuit. Totally. Yes. Guillermo Angeris: Yeah. Okay. Anna Rose: Nice. Okay, the last one that I had on my list, I know we have zkMPC, but I don't know if we have, do we have any MPC experts here? There was a project.. Nico Mohnblatt: There were a few Anna Rose: There was a few projects over the years. I actually don't know that combo, but I almost wonder is that a similar thing where it's like a wrapper? Nico Mohnblatt: No, that's more can people who have parts of a secret generate a ZK proof together in a way that prove something about that secret that we all have or that we put together, but without ever revealing our shares of the secret? Anna Rose: Okay, so that's more in the weeds or more integrated actually as a architecture, Guillermo Angeris: I feel like this is not ZK Operation Nico Mohnblatt: It's almost the other way Guillermo Angeris: It should be MPC ZK Nico Mohnblatt: Exactly. In that case, it's almost like the MPC. That's a wrapper around ZK. You're doing MPC over a ZK proof. Tarun Chitra: But ZK as a linguistic operator seems to only prefer precedents. Nico Mohnblatt: There may be MPC ZK is a better way of thinking. Anna Rose: Is this Oh yeah. Wait, wait, you have to wait. This is, we need to get your audio. Guest 7: Is this secret shared between all of the users, all of the.... Nico Mohnblatt: So there are different models. Yeah. One of them is if I have a secret but I don't have the capacity to prove anything about it, I'll secret share it. I'll do shares of it, give one share to some cloud service, another share to another cloud service. And together they can sort of make a proof or I think each one of them makes their own proof. And at the end, I combine the two proofs into one real proof about the secret I had. That's one model. The other model is I know part of a secret. Guillermo knows the other part. Tarun knows the other parts, but we don't want to reveal to each other what parts we know. But yet the three of us want to prove something to Anna about the aggregate secret that we know. How do we do that? Guest 7: Interesting. And do you have some kind of threshold cryptography plus ZK? So we all know a shared secret, and if a bunch of us gather together and create Nico Mohnblatt: So fact, check me on this one, but I think the first model I discussed of secret shares, one person knows the secret produce secret shares and distributes them out. I think you can do a two out of five thing and with only two people producing proofs for you, that's enough, but you're also reducing your trust budgets. Where now if only two of these people collude, they'll find your full secret. Guest 7: Do you know any names that I can search about this? Right? Nico Mohnblatt: Eos, it's a paper. I know Pratyush Mishra is on it. I don't know who, I forgot who else is on it. Anna Rose: This is Eos, not to be mistaken for Brock Pierce's for EOS the blockchain. When I saw that, I was like odd naming, Nico Mohnblatt: Which was presented at ZK Summit a while back, I think seven in Amsterdam. Anna Rose: That's when I was like, what is this? There was a hand right there. I just want to make sure you get a chance to say something. Guest 8: To go back to MPC and ZK. I think we have it actually both ways because we have, on one hand, as you say, you want to produce a proof and the witness is going to be private from different people. And so they have to collaborate in creating this proof. I think I heard that is a paper from Dan Boneh about this, but you also have MPC in the head, which is we're going to use MPC techniques to produce the ZK proof. And so it's reversed. Nico Mohnblatt: ZK as sub routines for MPC protocols? There's a lot of crossover. Guest 8: So the course of us are both ways. Anna Rose: Both ways. But just to make sure I heard it was MPC in the head. Yeah. And that was Nigel Smart, I think. Was he part of that? He works on that. I had him on the show at the time. I think he was doing that. By the way, we say Tarun Chitra: That for most things, Anna Rose: Any topic Tarun Chitra: Will address in cryptography. Did Nigel Smart Anna Rose: Ever look at did Nigel Smart.. Tarun Chitra: At this point, what haven't you had? Is there something that's the big gaping hole after so many years that you're like, Anna Rose: Ah, 200 and... Tarun Chitra: Thing I'm missing? Anna Rose: I'm sure a lot come on. Tarun Chitra: But is there something that comes to mind that's like, oh, this is the thing that's like. Anna Rose: Pretty much most of these crossover ones I have not talked about. So this is very interesting. I want to explore them just so you know any, and this is also to the audience potentially listening later. Anything we're mentioning here, we will have show notes. We'll be able to put links to the previous episodes and everything. So we'll try to dig those up. Zk DeFi, it's not really the same kind of acronym. You said there's a lot. Tarun Chitra: Yeah, I think there is stuff. I think the main thing you have to remember is that the ZK is not a panacea. So there was, I think 2020, 2021, there were just so many of these teams that are like, oh, we're doing ZK Uniswap. And they did it mindlessly without thinking. And then we wrote a bunch of papers that show when that doesn't work. Guillermo Angeris: Anna's acknowledged on them actually, because you were the reason why we wrote the paper. Yeah, Anna Rose: So nice. Tarun Chitra: We've evolved since then. I think people have kind of realized, oh, I can't just throw ZK and hope that people won't pay attention because people do care about the privacy aspects in DeFi quite a bit actually. And I think some of the things are auctions. So designs for particular types of MEV auctions that use ZK, that's maybe closer to that type of stuff. But I think there's also a lot of stuff on cross chain DeFi for proving properties about collateral on other chain such that you can use it locally. And that is actually sort of like a weaker form of a bridge. So there's definitely a lot of stuff in that vein. I think a lot of it is just slowly being built up right now. I think in layer two world where there's many roll-ups with these kind of somewhat shared sequencing aspects, a lot of the partiality guarantees are coming from ZK, but it's more like evolutionary rather than revolutionary to DeFi. Anna Rose: I just remembered that work. And even that show, we did this show, I think where we talked about it, it was such a sanity check at the time to the kinds of claims being made. There were so many private text, private text and all meant different things to different people. I remember your work really was a little bit of a lightning, like a let's get out of this weird stupor we're in. Let's be realistic. Some of those claims are not possible. I think it shifted people who were in that to take it into account, change direction, be a little bit smarter. I remember that feeling of like, and the teams responded. Guillermo Angeris: Yeah, some teams not so kindly, but yeah, I've got some emails Anna Rose: There were some emails. Guillermo Angeris: I think the cross chain stuff is way, way more of the Anna Rose: And that's now though, right? And this is kind of referring to your talk where it's intents. You start to like... Guillermo Angeris: On that vein, I actually think one of the interesting things that hasn't really happened a lot is we're preoccupied a lot with what's possible. A lot of ZK is like, okay, what can we do with this? And I think one of the things that I haven't seen a lot in the field is actually impossibility results. What things can't, if you only have ZK and you don't have any other guarantees, what can't you do? Right. In some sense there's a bunch of questions. I mean, one of them happens to be privacy indexes is really not achievable with just ZK. You have to actually get smart about how you do a bunch of these things. But more broadly, you could imagine that there's a lot of these crossovers that actually don't make sense. I don't think this is true, but one example is maybe there really isn't a panacea for ZKML and it requires a lot of communication to actually get things done. And that places a bound on how well you're likely to do, how fast you can run a given model or something of the like. So it's kind of interesting I guess that this gets us as, I think there is a place right now for the more negative work that outlines what the boundaries of what we can do are in ZK and then what we actually need to enable thing further applications. I feel like we're getting close to the edge of what's possible, but it's still very fuzzy in both theoretical sense and the practical one. Anna Rose: I think. I mean, I have one last topic that I wanted to cover, which is a little bit about the trajectory of ZK and where we are. The name of this was State of ZK. As a project. So zkSummit1 was in 2018, beginning of early 2018. I think that was around the same time we named the show Zero Knowledge and it went, I remember the first summit was a hundred people, 110 people. The second summit was packed. People were very excited and then it went down again, which is a lot of fun as an event planner. And then it sort of slowly meandered along, got a little bit bigger. And then in the last two years especially, I mean you can see it in the events. The events is like one metric. You don't all see this, but I can see it in the show, the listen, the download count on the show, you see these spikes. But yeah, I have a sense a little bit for at least the metrics that I have access to, but where are we on that ZK scale? Have we just gone through almost a delayed ZK bull market? Just so you know this, as the crypto market crashed, ZK took off. It was very strange. My metrics went up when everyone went down and I was like, we're defying gravity people. This is amazing. But yeah, and I will say this ZK10, we've made it to 10, but I will say currently on our metrics is leveled. We're not getting the big swoops up anymore and I'm totally cool with that. I find those swoops kind of scary. They weird me out a little bit. I like slow and steady. That's my way. But I'm just curious what you guys think about that from where you're sitting. Is ZK on the up. Guillermo Angeris: Coming up? Before we get to that, I actually want to say it is insane and Alex Turner and I were just talking about this, which is you said the first ZK event, the first zkSummit was in March of 2018. We were just talking about the fact that what existed in 2018, Anna Rose: Zcash. Guillermo Angeris: Yeah, right. But okay, is that ZK now? Are we doing thing that existed? The fact that that was the only thing that existed Exactly right. It's Zcash, but Anna Rose: It was Zcash and a lot of research groups and universities at the time. Berkeley, I don't even know if Alessandro had already started his, but there were some Berkeley people. No, he probably had and Dan Boneh's group and I'm sure there were others for sure. Guillermo Angeris: The FRI paper was around then, right? Yeah, 2018. Anna Rose: But it was tiny. And also we weren't on the map, so I was kind of lucky that ZK researchers even learned about it. It was, Guillermo Angeris: Yeah, this is what I'm getting at. I was like, how did you have the big brain vision to be like, ah, yes, ZK is the thing that will be big. Anna Rose: I've told this story. You know this story. Well, I Guillermo Angeris: Don't know, maybe I know this story. Anna Rose: I'm sure you all know this story. So the pod, Guillermo Angeris: Do you know this story? Does anyone know this story. Anna Rose: The podcast for the first. Okay, we're going to do a little history of the podcast for the, you can all listen to it. The first 11 episodes are terrible. But if you go back to the beginning of Zero launch podcast, we had no name. We were this podcast, welcome to, I think that's what Frederick would say, like this podcast, my old co-host. And then around episode 11, I think it's around there. We're like, we have a name. We're the Zero Knowledge podcast because, and I don't know if we said it, but we were thinking it because we have zero knowledge, not the cryptography and ZK, this is the truth, this is how this happened. Non-technical, not math person in this position. And then I had just, so I had a startup before a SaaS based startup by the way, which is where my disdain comes from. But yeah, I had done events. I think we had thrown parties with my old company and I knew that they were just cool. They were fun to do. And I wanted to do an event around the podcast. So something Summit became insert name of show placeholder Zero Knowledge Summit. That is how this event started. Incredible. And we had named it by then, I mean I think I had met Zuko too, and I was trying to get him on the show and I was trying to definitely get zero knowledge people to come. And we did know what it was. We weren't completely pulling it out of the air. I was working at the time in an engineering team. We knew that ZKP's existed, but they were still pretty far away from us. But, and that's how the first summit came to be. And actually what happened was we sent out the invite for it. I think it was still application only at the time. It was kind of the crypto bull market. So it was more just, there's could have been a lot of random people coming who wouldn't have been very technical, but someone passed it over, I think it was Howard Wu. Got it. And he was like, I want to apply to the zkSummit and I'm going to bring eight of my cryptography friends who do ZK. And at the same time, Zuko was like, we have someone on our team who's up for coming Strad, who's here actually today. And then Strad came. So we had those two people. Strad started it off. He was the first talk. Howard was the second as far as I remember. And then there was a few others, but the event was also a lot of just random blockchain who was local. We were in Berlin. But yeah, that's how that started. And then every summit after that became more ZK, less other until, I mean ZK four was all ZK. Oh no, no, actually no, not fully. Maybe like 90% ZK. But it was like getting there. That's the story. Sorry, that was a lot of story time. No, that was, and actually the audience of this show, I have told this story on the show before. Guillermo Angeris: I haven't heard the story. I mean, not on the show at least, Anna Rose: But does that destroy the magic for everyone? Tarun Chitra: Fake it till you make it as a well trodden tradition in this industry? Just look at Tether. Anna Rose: Have I ever really faked it though? I don't think I've ever pretended to be good at this. Nico Mohnblatt: I dunno. Guillermo Angeris: So anyway, I didn't mean to derail too far. Sorry. I'm like, Anna Rose: Where's ZK at? This is the question, Guillermo Angeris: What is the trajectory of zk now? I'm sorry, I interrupted you while Nico Mohnblatt: You were, so the metric I have is research and the pace at which it's coming out. And also I have a very short timeframe that I'm looking at. I joined very recently, but it still feels like it's super fast and it's probably getting faster. So as far as I'm concerned, Lots of speed, lots of energy. And Anna Rose: On that trajectory, Guillermo Angeris: All the papers that were started in the bull market are being finished now. They're being edited as we speak and coming out of the oven. Yeah, I think we've maybe you think is the wrong term. I hope we've hit a bit of an apex. It seems like the more normy channel, so to speak, have stopped raving about ZK and now we're kind of left with more slow, maybe more deliberate work. I wouldn't call research slow or deliberate in any way, but it's the closest thing that we have. So I think it feels like that's kind of continuing on. But I feel like the frenzy of getting in now and doing ZK as quickly as possible, I feel like has slowed and it feels very good in a lot of ways. Anna Rose: Yeah, it sounds like kind of quality. Guillermo Angeris: Yeah. Yeah, exactly. And sometimes Anna Rose: As you say that, that's what I'm thinking is quality. Is there, yeah, Tarun Chitra: I think I agree at both of those points, but I'll give maybe the unfortunate capitals and point, which is that I will say investments in ZK seem to be down quite a bit. So if you're trying to raise the ZK fund, you might have missed the boat. I think six months ago would've been the ideal time. I'm not trying to be the Debbie Downer. Okay. I'm just trying to say the facts. Anna Rose: So I believe we are actually getting the, what's that stick, the cane, the thing that pulls you off stage. Oh, what are those called? Guillermo Angeris: Yeah, yeah. Anna Rose: What are they actually called? The cane. Guillermo Angeris: I think it's a cane. Yeah. When they just pull you out. Anna Rose: The hooked cane that pulls you off the stage in those cartoons from the twenties, whatever that is, we're getting it. I've been told now multiple times we need to wrap it up. So I want to thank you all for listening to this panel for Tarun Chitra: Can we end with each person on stage giving their prediction for what will happen in ZK land in 2024? Anna Rose: Oh, okay. Yeah. Okay. Before I close. Okay, let's do it real quick. We have to be quick though. I am getting the signals. Okay. What's 2024? Tarun Chitra: I predict we will have our first application and it will not look like anything talked about here, but it will probably be used in a way that probably discuss the purity of people working on the proving systems. Anna Rose: Debbie Downer. Tarun Chitra: Because gamblers always find a way. Anna Rose: You think it's going to be gambling. Tarun Chitra: Gambling. I think it's going to be some type of speculatory. Bamboozle. Guillermo Angeris: Yeah. I don't have a real prediction. I mean I think we're just going to see more of the culmination of the current wave of research actually becoming more real. But I dunno if I have a prediction for what that means At an animal level, at the base of your stem level, you, it's kind of like, oh, cool, continues. Nico Mohnblatt: I think we'll probably see client side proofs that are fast enough to not be too annoying, fast enough to be almost usable almost. Anna Rose: Okay. I don't think I even have time to think of one. I'm actually getting signals from people that I think they're holding the door closed so we don't get actually thrown out. Alright, thank you so much to the panelists. Thank you so much to this audience and the questions you just gave us. Thanks for having a Live zkPodcast episode with us. I think this was fun of what it's like. And obviously thank you so much for coming to the zkSummit10. Thank you again to this team for putting it on really. And there's Agni. Say thank you to her. Say thank you to everyone here, but especially Agni. Yeah. And I think I'm supposed to say something. Thank you to the podcast team and to our listeners. Thanks for listening. There we go.