Anna Rose (00:00:05): Welcome to Zero Knowledge. I'm your host, Anna Rose. In this podcast, we'll be exploring the latest in zero knowledge research and the decentralized web, as well as new paradigms that promise to change the way we interact and transact online. (00:00:27): So we're back. This is our first episode of 2023, and our first, after a three week hiatus. Very nice to be back here. I hope you've had a little bit of a break. So this week's episode, we will be wrapping up our 2022 wrap up with a little catch up since our last episode and we're also doing a look forward into 2023. I'd say like most of the episode is really about ZK use cases. We sort of explore areas where we think there might be some ZK activity. We also check in on kind of where certain use cases are at, yeah and I hope you like it. Before we kick off, I do wanna let you know that there is an upcoming zkSummit9. It's happening on April 4th in Lisbon. I'm very excited to once again invite some of the best researchers and practitioners in this space to present new research, new ideas, implementations. (00:01:20): It should be a really good time. We're also gonna be doing a hackathon around this time, so keep an eye on our Twitter or Telegram channels. For more info, I've added the link to apply for the summit in the show notes. Tickets are always limited, and it's always better if you go through the application process early, you have a better chance of getting a ticket that way. So be sure to sign up soon and hope to see you there. And now, Tanya will share a little bit about this week's sponsors RISC Zero and Anoma. Tanya (00:01:48): Ever wish you could use existing Rust libraries in ZK? A friendly reminder from the team at RISC Zero that you can things like proof of checkmate, blinded proof of password complexity, and even proving things like protobufs are straightforward when you can use existing code. To learn more, check out the RISC Zero video tutorials from the last zkHack at youtube.com/@risczero. That's R I S C Z E R O. You can also follow them on Twitter @risczero to make sure you don't miss their upcoming 1.0 launch and the alpha launch of the Bonsai Network. So thanks again RISC Zero. (00:02:24): Anoma's first fractal instance Namada is launching soon. Namada is a proof of stake L1 for interchain asset agnostic privacy. Namada natively interoperates with fast finality chains via IBC and with Ethereum via trustless two-way bridge. For privacy Namada deploys an upgraded version of the multi-asset shielded pool circuit, otherwise known as MASP, which allows all assets fungible and non fungible to share a common shielded set. This way, transferring a crypto kitty is indistinguishable from transferring ETH, DAI, ATOM, OSMO, NAM (Namada's native asset), or any other asset on Namada. The MASP's circuit's latest update enables shielded set rewards directly in the shielded set, a novel feature that funds privacy as a public good. Visit namada.net for more information and join the community on Discord via discord.gg/namada. So thanks again, Anoma. And now here's our episode. Anna Rose (00:03:19): So I wanna welcome back to the show, my awesome sometimes co-host Tarun, Guillermo and Kobi, welcome. Guillermo (00:03:26): Whatup Tarun (00:03:26): Hey Kobi (00:03:26): Nice to be here Anna Rose (00:03:27): A quick note, Kobi did let us know that he might have to bounce off this one early, but in the meantime, I hope we can cover some really good ground on ZK use cases. Basically, I wanna use this episode to talk about what is exciting in ZK and what might be coming next. This is our look forward part of the wrap up episode that we started last month. So maybe before we jump into the future stuff, let's get a quick recap of what actually happened since we last spoke. So we recorded our last episode on December 12th. I think in it we were even talking about SBF going to jail or not going to jail that day after we recorded he was arrested and then at the same time, Traun, you had mentioned something like, I hope there aren't any more dominoes that can fall before the end of the year. Like there was something else you were kind of worried about for those that last 19 days of the year. I wanted to check in about that. One: what are those dominoes? And two: have they fallen? Tarun (00:04:23): I mean, I guess the last one was the Genesis/DCG stuff, which seems to have erupted like last week. Anna Rose (00:04:30): Yeah. Tarun (00:04:30): Where the Winklevii kind of, or like, basically saying they're gonna sue them and trying to get to DCG to fire its founder. So that didn't blow up, but it's not looking like it won't blow up, if that makes sense. Anna Rose (00:04:46): Yeah. Guillermo (00:04:47): Wow. It's not looking like it won't blow up is what, what a statement. Anna Rose (00:04:51): But I do wonder, like, given that this seems to be primarily in the CeFi or like crypto-in-TradFi realm, does this still have an impact on DeFi? Tarun (00:05:01): Yes and no. I mean, I think a lot of the centralized lenders did actually source liquidity from DeFi or provide it. So there's some sense in which they're still kind of connected, but it probably doesn't really affect it in the long run. I think the bigger thing is just more like net market confidence in like the centralized entities that haven't died, you know, if everyone is dying like dominoes then and no one survives, then it's sort of like, that would be bad from the perspective of like, there's no on-ramps anymore. But yeah, I don't think it would've like destroyed stuff, but it would've obviously caused a lot of havoc especially if they have to liquidate private positions and stuff like that. A little bit like FTX caused kind of this huge panic. Anna Rose (00:05:47): Hmm. Do you think we're still at risk of that then? Yeah. Recording today on January 11th? Are we still in a position where like something like that can happen in the next month ? Or do you feel like somehow the worst is behind us? Tarun (00:06:00): After the last six months? I don't think I ever can assert that type of claim again. Anna Rose (00:06:08): Yeah. Well, maybe this is then a good time to switch gears and start talking a little bit more about what's coming up and specifically what's going on in the ZK space. Over the holidays, Kobi and I went on another podcast. We went on Epicenter, and in that we actually covered a lot of ZK use cases. I can add the links to that in the show notes if people wanna see the whole thing. But actually I wanted to start off this look forward into the new year with sort of a short summary of some of those interesting use cases that we mentioned there. Kobi, I don't know if you want to take it away and, and start this off? Kobi (00:06:45): Yeah, let's talk about some experiments. I think like one, one thing that we are, we're seeing more this year, which we haven't seen before, or I guess last year, is that we see people being able to use the ZK tooling to deploy applications that are user facing and are a bit further away from just research or just proof of concepts. So I think like we had this Cabal application, which was really fun. Like you could have groups that are pseudonymous in Discord and like other locations which is ZK membership, which previously I think the only on-chain use case that we had for it was for mixers and like anonymous messaging, like in Semaphore. But that was a fun one. There was also this cool experiment with recursive SNARKs, which like the first time that was deployed by an independent developer rather than like a company like Mina that was deploying it. But, so you can prove how close you are to Vitalik. Anna Rose (00:07:49): Oh yeah, this is the ethdos project. Kobi (00:07:51): So what's your "Degree-Of-Separation" - Like ,that's the "ethdos". Yeah, so I have a good number I think, but. Anna Rose (00:07:59): Oh, really? Kobi (00:08:00): Yeah, because I know the developer, that's the only reason I have a good number. Anna Rose (00:08:05): Oh okay. Kobi (00:08:05): Yeah. Anna Rose (00:08:06): It's like a score. Kobi (00:08:07): Yeah. Anna Rose (00:08:08): You get a score for how close you are, but what's the, I actually have always been curious what's the actual linking point? Like what's the tool that's linking? Yeah. Is it GitHub, Twitter, ETH addresses? Kobi (00:08:19): I think basically it's just a signature because you get a link from someone that they say like this link shows that you know the person that sent you the link. Anna Rose (00:08:31): Oh. Kobi (00:08:31): And then you generate like a recursive proof that that takes the previous proof that is in depth link that shows okay, this is degree two, and then you can create a new proof of degree three Anna Rose (00:08:42): And it's private. Guillermo (00:08:43): Right. Anna Rose (00:08:43): So you know the degree, but you don't know who you've gone through Kobi (00:08:45): Exactly. You don't know the whole path. Anna Rose (00:08:47): Oh, interesting. This is, it's, what is it? It used to be 5 Degrees from Kevin Bacon. Wasn't that one? Tarun (00:08:53): 6 degrees from Kevin Bacon? Anna Rose (00:08:55): 6. Okay. 6. Tarun (00:08:55): But it's kind of a stupid number that like 6, the 6 was like, not even like the median of the degree distribution Guillermo (00:09:02): Yeah. I don't know actually where they came up with that. I think it Tarun (00:09:05): 6 was just like the movie was called that. And then they're just like, they just like, but it has, it had very little fact, factual basis. Kobi (00:09:11): I think in the ZK space it's more like 3 Guillermo (00:09:14): 2 Kobi (00:09:15): 2. Okay. Fine. Guillermo (00:09:17): I mean, I don't know, Anna Rose (00:09:18): 2 degrees from Tarun (00:09:19): Yeah. I mean it, Kobi (00:09:20): Everyone through Tarun (00:09:21): The diameter of the ZK graph has to be, I would pretty small. Like, I, I feel like the expected diameter of like if I did a random walk is less than two Guillermo (00:09:34): Yeah. Yeah. No, I mean, like, even in math from Erdős, it's like still quite small, right? It's like six or something or seven. Anna Rose (00:09:41): Is that to like anyone on the planet or something? Guillermo (00:09:44): From mathematicians to like this, like very famous, like Tarun (00:09:48): Well, I mean the, the best version of these is the what, what was it? Is the like Natalie Portman, Kevin Bacon, Erdős number or whatever. Anna Rose (00:09:56): Ooh. Tarun (00:09:56): Because Natalie Portman has a finite Erdős number. I think it's like three. Kobi (00:10:01): Oh wow. Tarun (00:10:02): Something quite low. and then basically like, yeah. Guillermo (00:10:06): Nice. Well there's the, yeah, the, the Metallica Bacon Erdős number I think is what it is, or something like that. Tarun (00:10:11): Yes. Guillermo (00:10:12): And it's like if you've ever put out an album, right? And like, you probably have a finite Metallica number because you've worked with someone who's worked with someone who's worked with someone who like, whoa helped worked on the one, some Metallica album. Then there's a finite Bacon number, which means that you start in a movie or were in some piece of media which had someone who had someone who had someone, you know, who like Anna Rose (00:10:33): 6-fold who worked with him Guillermo (00:10:35): Was part of a Bacon movie and then so on and so forth for, for papers as well. Kobi (00:10:40): Yeah. Guillermo (00:10:41): So that's the coolest. Anna Rose (00:10:41): Oh Guillermo (00:10:41): Unfortunately I'm still, I'm still seeking a finite Erdős Bacon number Tarun (00:10:46): Wait, wait, wait, wait. You, you don't have a finite Erdős Bacon number? Guillermo (00:10:50): No, I have a finite Erdős number. Tarun (00:10:52): Do you know your Erdős finite number? Guillermo (00:10:53): I think it's like 3. I think it's actually, I think yours is 3 too. Tarun (00:10:58): Mine is 3. I think we're both, yeah, we're Guillermo (00:11:00): Both 3 for the same reason. Tarun (00:11:01): Yeah. Guillermo (00:11:02): Which is Steven. Tarun (00:11:03): Yeah. So, Anna Rose (00:11:04): Oh, Tarun (00:11:06): Actually I have another route to 3, but yeah, it's Guillermo (00:11:08): Oh, interesting. Oh, that's spicy. Okay. Yeah. Tarun (00:11:12): But it's a funny thing which means like, I guess Guillermo, you gotta go become an actor soon Guillermo (00:11:18): I know. Clearly that's the, that's the obvious next conclusion Tarun (00:11:21): Here's like, very clearly you're in the wrong career right now. You gotta, you gotta quit, start acting and like, figure out how to get close to people who've been in movies with Natalie Portman Guillermo (00:11:32): It would be cool to have the lowest number, but I think the lowest number is pretty low. Actually. It's like four. So, yeah, Tarun (00:11:37): I mean, actually this is a good question though. In crypto, maybe there's like, you know, when social networks first started growing, you know, let's say even MySpace like 2000, mid-2000s there was a lot of stuff about people trying to be like, how many friends can you really have online? And like, people were like, oh, there's like this, like limit of like a hundred or whatever. And then people kept coming with all these different, like half graph theory, half sociological, like observation per like statements of like, oh, this is like how many friends you should have or how many degrees of separation you should be. An interesting question is like, in the crypto world, how does that different, given the lack of identity? Like, does that number go down? Does the number go up? Are we gonna see these kind of like, sociological theories for like how connected certain addresses should be? Anna Rose (00:12:25): There's also, isn't it sort of weird online? Because there's also people who have multiple identities. You might be friends with like three people that are the same person. Tarun (00:12:34): Right. Right, right. That's what I mean. So it's sort of like interesting of like, you know, I guess in Solana, right? Like the one guy who deployed all the contracts that got used in their DeFi bull run I'm sure he has like the highest, you know, degree, you know, highest connectivity of everyone. He's like Anna Rose (00:12:54): Wow. Tarun (00:12:54): If you ran page rank on Solana, you probably just get to him. Or Anato, like with probability one, always Anna Rose (00:13:02): All right, let's continue then let's, yeah, keep the new use cases rolling. Kobi (00:13:06): Yeah and I think another cool one that I've seen that I also had fun participating in was this "heyanon!" thing where you, like, they created these Twitter users, I guess, that allow you to tweet if you were part of some non-group. So one of the groups that they created was the DAO Hack victims, I guess and yeah, I think I was the only one that tweeted from there. See, I'm deanonymizing myself Anna Rose (00:13:41): heyanon! was the project. Kobi (00:13:42): Yeah so like they have multiple Twitter users from different use cases, like 0xparc participants and hack victims and some other ones. So like, I really like the integration with the real world rather than just being this on chain use cases. Anna Rose (00:14:02): Yeah. Kobi (00:14:02): So that's what I found fun about all of these three Anna Rose (00:14:05): Tarun I want to throw to you. Is there any like, upcoming ZK projects, experiments that you think have high potential, something maybe to look out for? Tarun (00:14:15): Yeah, I mean, I think we talked a little bit about it in the summary of like the idea of language models having some form of distributed training or distributed verification. You know, I think we're very far from that, but yeah, the fact that, you know, people have been able to SNARK imagenet and SNARK-like medium size neural nets sort of says, you know, we're in a world where people could get this verifiability and you know, I think if you look at OpenAI, they've like, or DeepMind, they've like hired all these people to like do not ZK based verifiability like Scott Aaronson and stuff for like proving that like some content was made by, you know a model or not. And I think like fundamentally ZK is gonna be the only way to do that. I just don't think that all of these like kind of half-assed watermark solutions are gonna really work. Guillermo (00:15:05): That's an interesting hot take. Kobi (00:15:07): I mean, I do want to see like how people can reduce the trust in training specifically. I mean, it's not as, as like, I don't know, as sexy of a problem I guess. But if we can get to some situation where it's the same trust as the trust is set up where we can have one honest participant we can know on what data a model was trained, I think that would be really cool. Guillermo (00:15:36): Yeah. I guess like, I still am a little confused on what the use cases of like verifiable, you know, imagenet predictions are. I, you know, like I don't know, right. Like, or not predictions, I guess like classification is, Kobi (00:15:47): Yeah. Guillermo (00:15:48): What is, what is the active use case here for like, we have a model we like verify that the output is is indeed from that model, what do we do with that? Tarun (00:15:56): So I think this idea of like, you know, the value that people get from social networks you know, if we try to like decompose that value into value they get from content generated by other humans, content generated by purely algorithmic means and content generated by like a combination of the two. Like let's, you know, roughly speaking bucket into those three, I think like the more scarce that the content that's actually generated by human gets, the more people value that over time versus like the algorithmic content that's going to be everywhere. Guillermo (00:16:30): Sure. Tarun (00:16:30): And being able to prove that you're not algorithmic content and or having the algorithmic content forced to verify itself and prove that it is algorithmic will sort of be this thing that makes the the real content value go up. Right? Like it's, it's like, it's like the vinyl records, right? (00:16:48): Like everyone's like, vinyl records are dead, you know? Guillermo (00:16:51): Right. Tarun (00:16:52): But like, ironically, they make more money for a lot of record labels, make more money from vinyl records than they do from any of their other digital forms of selling stuff because people buy it as merch. And I sort of think content on the internet will have, that will be the vinyl records where, you know, if you can identify that content was or something was made algorithmically and it's sort of like a standard that everyone can verify, then suddenly your single origin human created thing becomes just worth a lot more. Anna Rose (00:17:23): Whoa. Tarun (00:17:24): What, what, what are the other fair trade human content Guillermo (00:17:27): Fair trade, organic, non GMO human content or non computer modified CMO maybe Tarun (00:17:33): I, this is, again, this is all just my opinion. Like this, you know, obviously it's like I don't know, ask some sci-fi author. They probably have a much larger thesis than I do on this Guillermo (00:17:43): I'm not fully sure I understand. Like, how, how one would prove the non providence or the providence of like a fake not being from a computer. Right. Like fundamentally at separation feels actually much more difficult to prove than not. Tarun (00:17:58): Yeah but I would say networks can enforce that any algorithmic content generated had to be, has to post some provenance proof in order for it to use API or something. Right. Kobi (00:18:08): I mean, you could always like have this open source people create different version that doesn't do it. So I don't kno Guillermo (00:18:16): Yeah. Just like repost it. Tarun (00:18:17): Yeah, for sure. I'm definitely not saying that there's a fool proof way of doing it, but I do believe that there will be this like monetary premium. Kobi (00:18:25): Yeah. Yeah. Tarun (00:18:26): In the same way that like, art gets more valuable when someone commits suicide or they die or definitely way more when it's suicide versus just normal death. I kind of think this is gonna end up being the same thing and like somehow, yeah, single origin human content might be inferior, might like, you know, have not be like as precise or might not have like as much, you know, descriptive but for, for some reason it will be worth more and it'll be like art, I just like, it's like hard for me to imagine that not happening. Guillermo (00:18:58): Yeah. Yeah. I think, I believe the general gist, I just don't know if ZK proofs is gonna be the, the way to do it. That's, that's where I divert on it Tarun (00:19:08): The proposals we see a lot of right now are like all these like statistical water marking mechanisms Guillermo (00:19:13): Oh yeah Tarun (00:19:13): And like, and like those things all seem like fine having to run the modified version that doesn't need the proof is one thing. But these like, things that you can actually fool with, like very simple adversarial models seem like not the, you know, we're just back to this whole like the, the, you know, people in ML have this version of what they think like private ML is and like, it all, half the time it always gets attacked. And you know, actually we had an episode with Florian who's a professor at ETH about, about exactly this phenomena, which is like, you know, in machine learning there's kind of much less of the notion of privacy so weak and these, these watermarking things look exactly the same. So I'm quite nebulous on them. I think all of those are just so that big companies can tell, like regulators like, hey, we checked the box, we tried. But I think ZK is the only thing that will actually provide some actual guarantees, not just some sort of like, bullshit, we added noise. Hope it look, you know, doesn't look like a human. Anna Rose (00:20:13): What do you think this year though? Like, what do you think we could see in that way? Like in the ZK ML front, do you think, I mean, I guess we could see new projects who are trying to do it. Do you think we'll be like actually hearing new use cases being proposed? Yeah. I kind of wonder like, at least when I talk to other people, they're like, timeframe wise we might be looking at seven years or something. Like, it might not be a two year, three year type thing. Kobi (00:20:41): I think related to what Tarun was saying, like you, you have this maybe middle ground. So for example, what Worldcoin is doing or what we talked about with Dan on Epicenter or actually in our episode as well where you have signed inputs like either from the orb or from the cameras. So that helps a bit alleviate the worries of adversarial like scenarios. So then you can offload or move all the detection models to run on the client side which is interesting because then you have privacy against the servers. Anna Rose (00:21:21): Interesting. Kobi (00:21:22): And you don't worry as much about adversarial inputs. But so that, that could be an interesting use case that doesn't take too long to deploy. Anna Rose (00:21:32): Maybe actually that timeframe was also talking about, because as I learned through that interview with Florian, the difference between like inference and model training. So like, yeah, everything we're talking about right now I think is the inference side, right? It's using existing models, but then it's the model training, trying to do that in private would probably take a lot longer. Tarun (00:21:56): That is effectively impossible. Kobi (00:21:58): Yeah. Tarun (00:22:00): I'm effectively gonna just assume that's not gonna happen. Guillermo (00:22:03): I actually, there are people, pretty important people in the system that would disagree with you actually. Tarun (00:22:07): True. Guillermo (00:22:07): I mean I agree with you actually, quite a bit. But there was recently a paper that attempted to disagree with you. Tarun (00:22:15): I think like if you jointly optimize some type of like how big your model is versus how much privacy you want type of thing, maybe, but training is also just like a dark art versus inference. Right? Inference is a very well defined algorithm. Like, you know, there's like, it's like almost deterministic. Training involves like, I do batch norm on a random layer. I like suddenly truncate a bunch of other things. You know, there's like so much dark art in that stuff that I hesitate to even, even if you did, were able to prove that you could do it for one model. It'll never scale to the way people, like, build models Kobi (00:22:52): Maybe if you save some seeds and randomness, that could make, we could make some progress, but it's, I know, I'm not sure Anna Rose (00:22:59): Kobi, I know you have to bounce early. Kobi (00:23:01): Yes. Anna Rose (00:23:01): But thanks for those use cases. Kobi (00:23:03): Thank you for having me. Good to see you all. Anna Rose (00:23:06): Yes. And excited for our next episodes Kobi (00:23:09): Yes. Anna Rose (00:23:09): Coming up in the next few weeks. Guillermo (00:23:11): Catch you later, Kobi. Anna Rose (00:23:13): Thanks Kobi. So now we'll continue without Kobi, we're gonna continue with a couple use cases. I'm curious if you, I think both of you could probably speak on this, but I don't think we would've talked about this last year, but it was the ZK bridges, like we did talk about it in 2022, but I know that that really kind of came into the fore in the last few months. Do you see more things in 2023 around ZK bridges? Tarun (00:23:36): I mean, more meaning like the people who are working on it will deliver, will deliver certain things, more in the sense of like new algorithms. Maybe not. I feel like right now it feels ZK bridging feels like the type of thing where it's like we have enough of a design space and now it's about just like optimizing the code to actually run fast, right? It's like, it's like proof of stake hit this same stride in like late 2018 or mid-to-late 2018 where it was like all the ideas are on the table, there's not that much else you're gonna do, it's just like implement them and then you have this like, Guillermo (00:24:09): Make them good. Tarun (00:24:10): Yeah. Couple years of engineering feels more like that than it feels like like there's some big unsolved mechanistic thing that is out there. I don't know. Guillermo, what do you think? Guillermo (00:24:20): I mean there's really, I don't know, like there's certainly some art to it, but like for the most part it's kind of like, you know, we have these consensus things we're following, right? Whatever, it's consensus, rule following and we need to like verify them. I'm sure there's like maybe some set of ways that you can make something slightly smaller, slightly larger. At the end of the day, you're kind of just like doing the thing, right? Like there's not like some like complicated difficult, hard elliptic curves question in their, I mean there might be but, but like it's pretty unlikely, right? Like for the most part it's like we, it's now more an engineering problem than it is a research problem maybe is the, Tarun (00:24:57): Which is good. Which is good in a lot of words, right? Like part of the 2017 boom stupidity was like everyone under their mom, if they had ChatGPT, you know, they wrote things that sounded like they came out of ChatGPT for describing why their proof of stake mechanism is better and they had like no way of like proving why or how or whatever. And I think by mid 2018 it became a lot more clear. There's like, there's only a few way, few things you can really do. Few ways you can, like directions you can like, make some particular large changes in and then it was like the teams that had the most capital just like went down those roads. Anna Rose (00:25:32): Yeah. I wonder, do you see like the ZK VM / ZK EVM in a similar state as bridges? Like a lot of the ideas have been quite fleshed out and now it's about implementation and the sort of who's going to win is still kind of up for grabs. Like who implements fastest? How, like we don't actually have metrics on a lot of these things yet so Guillermo (00:25:54): You know, because the task is so much more like, complicated, large in the space. I mean like large in like this weird, you know, one, we're like kind of doing arbitrary computation when we're kind of not, there are probably a few more tricks to be played, but I do think it looks much more like engineering than research still. Anna Rose (00:25:54): Yeah. Guillermo (00:25:54): I think if there are improvement, like systematic improvements, it'll probably come more from like better tooling for like ZK, you know, in terms of optimization, in terms of what, whatever, any number of things then I think they will probably come from like, you know, really smart like individual optimizations. So I do, I'd still see that still as more of an engineering artsy fartsy part of engineering because it doesn't require some tricks. But, but still engineering than like, you know, know what, what is like pure research of like, ah, let's figure out this like, general thing that has all these like, you know, nebulous questions about it. Tarun (00:26:52): Yeah. I mean, arguably ZK EVM has a bunch of open problems, but they're not like theoretical problems about ZK. They're much more just about like, you know have you, have you ever seen that that meme about like what modern software engineering is? And it's like they take a bunch of plugs from different countries and then they stack them and until they all like interface correctly and they make this monstrosity, it's sort of like that more than it is like, oh, like, you know, how do I make a music, hey mechanism. I kind of feel like while there is a lot of stuff people are doing on ZK mechanisms and I think more of the actual like true research seems to be on the NPC and FHE side. If I'm giving you kind of my stylized view of it. (00:27:39): I actually wonder if we'll see more breakthroughs there than in like actual ZK. Like there just seemed to be some people who were really plugging away at FHE a little more after the last couple years. Because I feel like there was like a little lull and people were like pretty down on it and then moved towards ZK. And I kind of feel like I've met more people who seemed to be suddenly thinking a lot more about FHE - fully homomorphic encryption and so I'm not trying to say anything that's like some crazy thing. Like, oh yeah, we'll solve it by next year. But like if I were to place like lottery ticket style debts on where you would have the research innovation, it's like in that realm Anna Rose (00:28:23): I think in terms of innovation in ZK, what I see happening more and more is like use cases, like for these tools, like what we described earlier, these experiments, it's at the stage where people can actually do that. And so it's true that like, yes, there are, I mean, I think we've actually felt that since 2019. In 2019, there were like breakthroughs coming in and then early 2020 it was like PLONK comes out. And since then there have been a few milestone breakthroughs, but it definitely got, it was like lots of papers being published, lots of small interesting Tarun (00:28:58): Incremental Anna Rose (00:28:59): Incremental add-ons. Yeah but less of that like, you know, paradigm shifting research coming out. Yeah. I think that makes sense. Tarun (00:29:07): That's why my bet is actually that it will come in like these things that people are quite negative on but are like starting to switch a little bit on or so it seems Anna Rose (00:29:16): Like, just in terms of big research breakthroughs Tarun (00:29:19): Yeah. Yeah, yeah, yeah. Yeah. Guillermo (00:29:21): I also think of like, the frontier for FHE is quite a bit more open for these breakthroughs then like ZK like I feel, you know, maybe there's like another inflection point for ZK, but it does feel like we're like very quickly leveling off on like the, the low hanging fruit. Unless someone comes up with some crazy new technique but just like opens up new field that's that that seems, you know, as time goes on more and more unlikely. Anna Rose (00:29:42): I think that's fair. But I wanna explore one use case that actually, I think it might've been Guillermo, I think this might've been the first time you came on the show. I think we've talked a little bit about this. I mean, I think it was the private DEXs. Guillermo (00:29:55): Oh yeah. Anna Rose (00:29:56): Like the idea of a private DEX, what was what you needed for DEX to be called private? Because there's like parts that can be made private and other parts that can't, but then can you really keep certain things private if other information's available? I do wonder, like, just checking back in on that, where are we at with private DEXs? Have we, like, has there been a standard decided on, I mean I don't, there's not that many groups doing it. I still think it's pretty like yeah, wild west-y. Guillermo (00:30:24): I think there's some initial answers for sure, which have interesting, actually, weirdly enough, I guess may maybe these way to describe it is it's a bit of an open space still. Sure, I mean that's the obvious way of describing it, but within that there's like a big spectrum where you trade off fundamentally, right? Like kind of, and I don't know if you recall, I think you're acknowledging in this paper for like inspiring it like there exists this funny like open space of like, you know, privacy can't really, or when we say privacy we mean specifically like one cannot reconstruct individual trades from like, just by like viewing public data and to know that's actually a surprisingly difficult thing to achieve. Right? And that's kind of what we outline there but that also means that there's kind of like always a fundamental trade off between kind of how quickly things get executed and what pricing get executed at like trades, I mean, and like privacy, right? And we, I think we now have some ideas for certain solutions, right? In the sense of like Penumbra is one example. Anna Rose (00:31:34): Yeah. Yeah. Guillermo (00:31:35): Where you, you have a little bit of this trade off between privacy and what, might call economic efficiency or whatever but I don't think the trade-off has been fully explored. For example, Penumbra used a very particular mechanism for how to kind of scramble together trade, so to speak and, and it turns out it has kind of interesting properties. Anna Rose (00:31:56): Yeah. We did an interview, we actually could link to that. We did an interview with Henry where we talked a little bit about that particular model. But I do wonder, like, have you been following that space? Is there any other, like, are there unanswered questions? There are models that maybe have been sort of proposed but not explored, not implemented. Guillermo (00:32:15): There are some, most people see for some reason, like the hot thing within privacy turns out to be like, I think people call it like RFQs or whatever. Is that what it is Tarun? Tarun (00:32:28): Yeah Guillermo (00:32:29): So it's like somehow, like everyone thinks that a private DEX is solved by like making a private like RFQ. Anna Rose (00:32:36): What does that stand for? Tarun (00:32:37): Request for quotes. So it (00:32:39): Basically just means, like I say, I want to trade five Anna Coins for Guillermo Coins and I say, okay, here's, here's how much I want, what's the price? Then you get you, you know, the mechanism goes on queries people who are willing to take the other side of that and then gives you the minimum of their thing. So the places where you could make that private or are that the, you know, and sort of one bad thing that might happen is like, say you want to buy 10% of the Anna Coin supply. Well, if I tell the mechanism that, and then they goes and ask all the people who wanna sell Guillermo Coin, they may collude and say like, wow, Anna's dumb, she's like selling 10% of this. Like, let's collude to give her a horrible price. Right that's like one mechanism so the notions of privacy you need there are, can you submit your trade size privately? (00:33:31): Can it be done in such a way that the people who are then providing, you know, bids to you can get enough information such that they can give a accurate price that makes sense and in some sense you have to have like privacy on both sides, but also some amount of information that's shared. But it's very like you can analyze each of these components separately. Which is why I think people are, you know, people from the privacy lens want to do it. It;s a little bit more like over the counter trading though, than it is like, like training against an AMM. It doesn't feel like continuous trading, right? It's like there's many interactions. Like you have to go give the the amount you want, then it has to go message every one who's willing to be a liquidity provider. Then take those compute some function of that return to you, then you say you execute. Right? Whereas Uniswap, it's like a one click, one message type of thing, right? So this communication complexity aspect of these is, is actually quite important. And I think I haven't seen anyone on the privacy side who's actually focused on the fact that you're reducing the quality of trading by forcing people to have many different rounds of communication. And like what does that cost you? Guillermo (00:34:47): Right for some reason, like that's been, that specific mechanism has been what's been, like, I've seen some amount of exploration actually even that, not that much, but let's say claim exploration. It's actually shockingly difficult to understand, but kind of on the DEX side, there really hasn't been that much more, I haven't really seen that much personally. I don't know exactly why, but I mean it's fine. Anna Rose (00:35:13): I wonder, well, I mean, Tornado happened and it might have maybe made some people a little bit more cautious to explore that right now. Guillermo (00:35:21): Yeah. I don't know. It, it's just kind of a funny thing because it's like it's a cool open problem. I mean, yeah, you know, you maybe if you think that like Penumbra is the end state of, of private decentralized exchanges, then in that case surely like that, you know, go work for Penumbra or something instead of like making your own. But there is an open question of whether there are more interesting mechanisms. I think there are Tarun (00:35:40): I mean the place I think that's in between these two is sort of like mechanisms that involve sort of like many auctions that are involved, right? So like, you know, actually I'm writing this kind of short paper on this right now, but I was actually trying to study this like gradual Dutch auction thing. This like whatever, I mean Art Gobblers is gross, disgusting, whatever the, I could never, never imagine liking the art or game like it's kind of gross. But I think the mechanisms actually kind of interesting if you try to like understand when it's sort of incentive compatible or not the current version that, you know, the one that Art Gobblers uses and sort of like Dan et Al made is technically not incentive compatible, but there's a way to to basically modify it. (00:36:33): And like, I think once the space of things that they're like continuous repeated auctions that you can join and leave whenever you want and there's sort of some like supply, it's sort of like a middle ground between the pure RFQ, which is like lots of back and forth to do anything versus like the Uniswap like hey, there's just like continuous trading. And I think that space is actually a place where you can make things a lot more private. But, but I think you first need to actually like flesh out the design space of that of like, here are mechanisms that could work not privately here's how they work. Like, original exponential gradual Dutch auction is, is like right idea sort of, but like wrong parameterization and wrong model for like the curve. It's sort of like, it's as if I did a Uniswap curve that was linear when Uniswap started and it didn't, didn't really like, wasn't able to take off. (00:37:29): I kind of think that GDA has that problem right now. But yeah, systematically setting that I think will lead to some private routes and like there's some sense in which, you know, the real North Star for private DEX should be that there should be some mechanism where trading liquid tokens and NFTs should be the same. You just like parametrize it differently for both in some way and you can kind of interpolate between the two so that you do such kind of like one type of trading vehicle that lets you do everything. And right now you see like the vespers of that, right? It's like Uniswap bought that aggregator and like for the NFT exchange, just like mainly an aggregation mechanism, but there's some clear thing where there's like a design space like prior to making it private that unifies trading these like non fungible type of things with liquid tokens and then you try to make that thing private. (00:38:22): So that's sort of, if I were to give you the two year vision of this, that would be what I would say is like missing and you know, I think part of it is like people aren't that interested in building those things when there's not much trading volume so like there's certainly that aspect that you can't ignore, but maybe, you know, like Uniswap, like someone will in the bear market be bored and interested and I think like that's more likely the thing to look out for. But again, I think that design space needs to be flushed out. Like I said, I think there is some sense in which the North Star shouldn't just be private DEX or like private RFQ. It should be like, I wanna trade NFTs and to liquid tokens almost in the same way and that thing if it works, should be able to be made private. But that might be a lofty goal, right? Like you have to have like kind of the hotspot and believe that that thing can be built and whatever. Anna Rose (00:39:19): It sounds actually a little bit like Anoma's pitch, but actually I want to know in the larger DeFi ecosystem, are there any other places where ZK is being incorporated even like maybe on the MEV front? Like is there any sort of place that you do see some sort of incorporation that we haven't seen yet? Or you could see? Guillermo (00:39:39): I don't know actually I don't, I haven't seen any. So usually what happens specifically with an MEV right is like ZK is gonna used for two things that's used for privacy and it's sort of proving the correct execution of something. In the case of MEV, there's often relatively succinct ways of proving the correct execution of or the correct solution or without actually like requiring the full machinery of ZK. So for the most part within MEV the, like real so to speak, like, you know, use of ZK is actually more as a privacy tool. Not always, perhaps. I mean there might be exceptions to this rule, but for the most part it's used as a privacy tool rather than as like a succinct proof tool, a correct execution tool or whatever you wanna call it. Right which is, it is distinguished from ZK almost everywhere else where it's used as both roughly equal, if not actually more in the succinct side than the privacy side. Anna Rose (00:40:37): So I mean one place where it has already been at least proposed is in the voting part. So maybe it won't be in the kind of protocol level, like the software level of the actual mechanism, but rather in the voting or governance side of things with voting using something like CLR fund or Macy, this private voting, but also this idea of like anonymous multi-SIGs. You know, you heard about Constitution DAO too? Guillermo (00:41:02): I heard about it. I actually have no idea what it's you know, new and improved version is Anna Rose (00:41:07): Well I don't actually, I haven't tracked it so I don't know where it's at. But I do know that like Dan had done work on creating private DAOs, the idea here would be like specifically in the case of an auction, if you are bidding against other people and you can see exactly how much they, this is like maybe in a real world auction. Guillermo (00:41:25): Yep. Anna Rose (00:41:26): And it's really hard to sort of like win your auction if yours is visible, it's kind of puts you at a disadvantage. And this in this way you could actually create sort of a private treasury amount when participating in an auction. Guillermo (00:41:41): Interesting. Anna Rose (00:41:42): And I don't know if this fully connects to DeFi, but I know like voting in general and doing that privately might help but has that been explored at all? Like has there been any kind of major projects that have even like, worried about it or discussed any sort of privacy in the voting? Or do you almost feel like it's antithetical cause like you want reputation on the line when you're in those spaces? Tarun (00:42:05): Yeah, it's sort of both, part of it is like the existing ones want the latter. there were certainly some very bad actors of late 2021, early 2022 voting wise. I mean, the easiest example was Wonderland but there were a ton of other like voting malfeasance types of things. Also the VE model kind of like distorts that and we've seen some crazy things like the thing that happened to Balancer a few months a month ago. So I would say like people focus way more on the voting mechanics / payment side. So far there's a huge disconnect between like, real projects that are live and ZK intern projects like at 0xParc and stuff where there's like tons of these like, hey, how do you do private demoing stuff? and I think part of it is like, it's just not very robust right now (00:42:58): And also the voting mechanisms, like you need some form of credibility that that needs to also exist and I actually, there, I mean credibility in sort of like a technical sense, which is, you know if the entire DAO was completely private and you're voting in, you know, this like whatever, I forget what Phil Daian likes to call it dark DAOs. But I don't really like the, I think like privacy shouldn't be connotated as light or dark, it's just a film thing. But you know, hey, like you'll never know where where the money's gonna be, where it's going, how things are voted on, whatever. So like, can you actually have credibility? Can you have any notion that like people aren't cheating in that system? So yeah. So there's definitely that in the practical side, I think that's what's keeping it from reality. (00:43:43): But I think like if there were some set of very good queries where you could prove some state of all the voters in aggregate that gives you confidence that it wasn't manipulated or like some type of thing, like some aggregate statistic can be monitored that's public. You might get to that. But yeah, there, there just hasn't been that much. I think like there's this huge divide in that like people on the ZK side are like, oh yeah, isn't it obvious? Why wouldn't you use this? But they like completely ignore the sort of like sociopolitical dynamic aspects and so that kind of, I feel like is why they're separated. This constitution DAO thing I think is like probably the only thing that's trying, I haven't really looked at that hard. I saw some people shilling in on Twitter but didn't I tried to take an actual vacation and not read any Guillermo (00:44:32): What's funny about this is I remember you mentioning like Constitution DAO 1.0 was like people, you know, like DAOs learning about adverse election in real auctions. Right and it's like, I actually remember you saying something like privacy tools are probably gonna be more useful for DAOs at the beginning cause they're gonna be expensive to use and, and kind of difficult annoying than they are for like the average Joe or whatever because of this fact. Tarun (00:44:59): Right but the average DAO participant is an average Joe at this point. Right. Guillermo (00:45:04): No, exactly. Tarun (00:45:04): So there's, there's sort of this like very weird thing of like the DAOs are democratic but only democratic for people who understand how to use wallets and block explorer as well. Guillermo (00:45:14): Of course, yeah, for sure Tarun (00:45:15): Right. Which isn't necessarily like a bad thing. I mean anyone can go learn how to do that, but it's not gonna be like everyone. Right Guillermo (00:45:21): I mean I think the other things that are, are kind of like interesting. I mean the thing I'm still focused on and but like, again, this is a little bit like this idea of like how do you do NFT trading and normal trading all in the same way NT, MEV auctions also, like certainly, I still stand by my claim last year that they will be the number one consumer or producer of value using ZK at some point. And when I say ZK, I'm using it in the venture investor sense of MPCs, FHE, ZK but I am purposely ignoring SGX because I don't think long-term SGX will survive in that. Anna Rose (00:46:04): So I think there are a few more categories in the ZK space that are emerging that we should mention here. One is like, just to keep an eye on ZK gaming. Dark Force came out I think in 2019 and obviously blazed a trail, but I know that there's still been a lot of experimentation around this. And I'm hoping that we start to see like another breakthrough with ZK gaming. Another space to keep an eye on would be ZK social. So this is things like Semaphore, private messaging or trying to rebuild any sort of social network with ZK and some sort of privacy and sort of close in proximity to both of those topics would be ZK ID. This has been proposed for a while, but I actually think this year we're gonna see it being used a lot more. (00:46:46): I know that there's a few really key projects like iden3 has been working on this for a really long time. There's also a company called sismo, ZKV is an investor there. Sismo, like, every time I see one of their presentations, I kind of like understand better how powerful this can be like I'm gonna take a step back into like what ID is usually. So if you think about it, like say you're trying to get a passport or you're trying to get some sort of identity card, usually what you have to do is fill out a form with a ton of information. Some of that will be used on the card, some of it is actually used by this third party like governmental body to prove that you are who you say you are but once that's proven, it's sort of like it's made private actually, again, it's not shared further on the card or on the passport. (00:47:33): What's on the passport is just your public facing version and I think when it comes to on chain identity stuff right now, at least without ZKPs, it's extremely transparent. So it's like everything you do that adds reputation. It's sort of visible and say you wanna issue a POAP or you wanna offer like, oh you participated in our airdrop or you did some action, you're still kind of linking the action with the address that's gonna get the reward or certificate or stamp or POAP or token or whatever and so it's extremely public. What these new projects are offering is a way to kind of cut that link so you can actually prove that you've done something. So like build reputation, offer all those documents, have your private data, have it somehow kind of checked but it's not checked by a third party. (00:48:28): I'll use the example of Sismo where what they're actually doing is like, you can prove, for example, that you have a Twitter account that's on a certain list or you have like participated in some sort of protocol, you've done some sort of trade, you've added liquidity, you've taken a loan, you've done some sort of action, and you can then kind of choose to create an attestation. So this you think is something that could like show reputation. It kind of proves your humanness, maybe your participation in a certain community, but you can actually create it from one location but send that attestation or badge or in their case it's like a one of those NFTs that can't be moved. it's soulbound, love that term but yeah, you basically can put that into another address that you control and then use that to sign things. (00:49:23): So it's like you can, yeah, it's a really cool idea and I think with this tool or with even the concept of such a tool, I think there's a lot of very interesting use cases that could be created around IDs, group membership reputation and yeah, I think it, I think it's sort of takes it up a level. I think it's a really cool case for you for ZK. Definitely not something I feel very well versed in and something I think I would love to explore more in future episodes. So are there any other use cases you think we've missed on this episode? Tarun (00:49:59): I mean, I think the the real question is like in this kind of like prolonged bear market, like what the new types of assets will be that are needed in the next year. Because like, I feel like each year seems to be punctuated by certain types of assets that are either created or destroyed. 2018/2019 is like the Uniswap LP share or C Tokens or A Tokens, 2020 was probably like vaults like you're involved, like the notion of like those types of assets Anna Rose (00:50:33): Strategies. Tarun (00:50:34): Yeah, strategies. 2021 was probably just Layer 1s. I, I don't even know what NFT and other types of NFTs, I think NFTs 1155, stuff like that. 2022, who the hell was the lack of claim? The year of destruction? Yeah, the destruction of claims and all and all associated benefits. Anna Rose (00:50:58): It was the stress test Tarun (00:50:59): So 2023. What type of asset do you predict? Like and I know Anna's default prediction is private asset, but I don't know, it's not the year of the year of private assets Anna Rose (00:51:11): No, I think my current is like ZK use cases, but I don't know what the tool there is. Like what's the? Tarun (00:51:16): Yeah, but like one way of rephrasing that question is like what asset is important? Like what will it, because like that will actually end up kind of dictating what the use case that gets adopted is if you work backwards. So it's like, what does that asset look like? Right? Like the 2020 innovation was like, you can have like actively managed strategies. Right? 2019 innovation, right? Was like passive, everything was passive. 2021 was maybe like make your own non-security, security. 2022 was bridges and bridge assets blowing up actually. I actually would say synthetic claims across chains, not being what they were. Right? Like, but the bridge asset in my mind stands out, whether it's synthetic or native. I feel like anytime anyone tells me these, like native bridged assets, they never really are when you read the contracts, because like they have some clawback provision or like, you have to put up some collateral, so you're really holding a portfolio. (00:52:19): And so then it's like it's not really native in the sense of like, I went from ETH to Soul, I put in ETH, I get out Soul, I never have to think and then when I go backwards, I get ETH, I never had like a centralized exchange. You don't think about that right. But in bridges, you still have to spend all this time. I think the native asset bridge is a bit of a LARP so far and at least the code, the contracts don't look like that. Okay. So 2023, let's say, let's say there's a new asset type. Maybe it's coming from ZK. What does it look like? Does it look like an NFT? Does it look like a strategy? Does it look like passive thing? Like what, how would you, what do you think that would look like? Anna Rose (00:52:59): Yeah, I really don't know how to answer this. I'm so not thinking like, I don't have any sense for an asset at all. Can I just, I'm just gonna say the thing I want and the thing I think needs to exist for that to exist. So the thing I'd like to see are evermore ZK use cases, kind of the creation of more black box, easy to use out of the box tools, some sort of platform where people can actually build and deploy ZK things in environments, which are really, really catered to that. I know that there's a few projects aiming for that because really what I wanna see by the end of the year is just a ton of really cool experiments. And I know this does not answer your question, where you're really looking for like an ERC something, something something, some sort of token standard, an asset standard. I don't have that. I don't like, I think it's really hard to predict what that looks like. Tarun (00:53:55): To me this is easier than predicting black swans, which was one of your original suggested topics. Anna Rose (00:54:00): Okay. Fair. Fair. Okay. Well what's yours? Do you have a good one? Tarun (00:54:05): I think mine is like, if there is some data structure that exists that allows you to simultaneously store private data, but like make, provide some summary statistics of it and you can like, own some notion of like a fraction of that where like the the private assets are pooled together, but no individual like can control them in uniformly, but the public statistics are how people like, have an understanding of what's happening and you can own a share of that. So like in, in a world where you had, you could, in this non-existent world where you could have had like a totally private Uniswap LP share, like in theory that meets those criteria, right? Where it's like, right, I have my liquidity, people don't know it's my liquidity, but I also know some public thing. I know the price and I know, you know, the price impact but obviously, you know, that has this impossibility theorem. So, okay, let's now try to say like, what is something like that where you can pull together assets privately, but there's some public transparency around the aggregation of those assets. And you can own fractions or trade fractions, right? And it's not gonna, like I said, it's not gonna look like a DEX LP share. It's going to be something like that and whatever that is. I think the first ZK asset will look like that first like well used one. Anna Rose (00:55:25): Could this sort of play into that concept of ZK ID then where you do attestations privately to things existing? So there you'd be able to create like an attest. I mean, it already exists where they have an attestation badge that can be publicly shared and it reflects something private. I know this is like, I mean this is usually used for events and stuff like that. Guillermo (00:55:46): Yeah. It's similar here I guess is the point but here you have like more data. So one example is they can have an attestation that indeed the balance you have is greater than like a million or whatever or whatever. You know, the amount of token, let's say we're talking about like a private LP share, right? Like you can attest to like individual elements of the share right? Let's say it's like you have like 10 coins in this LP share for some reason or another, right? You can attest to the fact that like you know, the LP share and coin three has like this much without revealing anything else for the, the remaining parts of share. I mean that's one example, right? You could also imagine like if you're collateralizing like a debt position using this thing, right? (00:56:28): You could say like, based on the current price given by whatever exchange, the total worth of the share is X but you actually don't know how how much individual assets within the share are actually worth. That's one example. Of course, Tarun is thinking about a slightly more general data structure than this specific case, that can include maybe other things. But the point is you can prove kind of like economic or non-economic, just like general, this is what I think he means by summary statements. Like statements about the claim right, so you can make claims about like this thing, you don't know what it is. Some like opaque object, but I can guarantee that it's worth is like whatever. Anna Rose (00:57:06): Yeah. Tarun (00:57:06): Not just, not just worth, it can be attestations about the content, right? Anything like any attestation that's like, but my point is like, you know, one interesting thing, if I take the asset of the year of 2020, these are all my of the years and obviously I'm not, everyone else can kind of tell me go fuck off if they want. But you know, if I take the vault, right, the vault sort of was created a necessity because you're at DeFi summer and there was tons of farming and you had to manage these portfolios and it was way easier to manage them in the smart contract than to do it individually in wallets and then it became a standard over time, like the ERC4626 standard and obviously in Solana they have an equivalent. So like I realistically, the ETH and Solana ones are the only like true vault standards. (00:57:56): I think other networks have them, but they're not like extremely standardized of like what it means to be a vault. But the existence of a vault was like one of the best developer tools ever. It made it super easy. And you can argue that like Uniswap V3 learned from that, like the way it stores things and so there's a sense in which maybe in ZK land, a way to actually get adoption is to invert that thing where you actually make a standard like this that's like the token standard of like how to combine assets and like make attestations and then have those attestations be tied to like the minimum, some functions of the composition economically without revealing a lot interest and if that was an easy standard to use, then you could start making these like ZK assets. And then I haven't seen anyone like really try to standard because like ZK people define making standards unsexy. (00:58:48): There's no way of getting around that. Like I'm not that type, I don't like the idea of like writing specs for standards. It sounds boring, right? Like, but in every programming and open source community, there's always like some fraction of the community that is very much like, no, we must like only write like this type of module in this way, in this language and here's a spec and if you don't agree, get out, no poll requests and like there, there, there's like the quote unquote like super Nazis of every open source community and I think like ZK could use a little bit of like someone like that whipping this type of asset into shape. Anna Rose (00:59:27): I want to explore what you just said quickly, this, this idea of what needs to be underneath. Do you think there's any work that's come from like the bridge work, the roll-up work, like basically doing the Tarun (00:59:38): The bridge work, the bridges that have to make this, like the bridges implicitly are making this, and so the question is like can you standardize Anna Rose (00:59:44): proof of stake, yes Tarun (00:59:44): Exactly and so the questions, can you standardize this so that it can be used in a lot of environments? Anna Rose (00:59:51): Interesting. Tarun (00:59:51): Because I feel like that would make it a lot easier for like a DeFi developer who doesn't want to learn anything on ZK, but wants to just be able to like get some like, values from this blob that's kept private. Anna Rose (01:00:01): I like this. Tarun (01:00:02): That's my prediction for asset of this year. Anna Rose (01:00:06): Cool. Guillermo (01:00:06): That's a cool one actually. I mean, it's a little bit general that it feels almost like we are just kind of rehashing what ZK proof is but I think the standard for DeFi developers, Tarun (01:00:15): The way you hold a bridged asset across a bridge, if it was private, not just for succinctness, would basically look like this, right? You would be like, there's some collateral that's private on the remote side, there's some attestations that that collateral without leaking its value is locked correctly and that, you know, you have balances here of this, that's an account tied to that. Guillermo (01:00:36): That's right. Yeah but like, one could imagine making a more general standard that isn't here is like generic blob, here is generic computation that actually is useful and easy to develop for. Anna Rose (01:00:47): We have three minutes. Guillermo, can you give a quick prediction in, in a short Guillermo (01:00:52): Oh man Anna Rose (01:00:52): moment? Guillermo (01:00:52): Can I give a quick prediction for what kind of asset is gonna be like the asset? I've been a little stuck on, you know, kind of EMU lately and a few other things that are similar. But my suspicion is things that are going to happen is, one of the things that's gonna happen is there's so many weird future claims for kind of like inter chain MEB and inter chain arbitrage and these kinds of things like these like weird derivatives for essentially what you're doing is you bid now for an expected future slot at some point. And I think that's actually gonna be like the, the asset that defines 2023 at least Tarun (01:01:34): I mean, that's probably honestly a more realistic prediction in the sense that like the ZK one needs like all the ZK stuff to work. So it might be like a year or so. But yeah, that's a very good one in one other respect, which is that like almost every MEV on every chain, including ETH like Flashbots is moving the direction of having this, there's a word for this that like functional programming people like, and I kind of like the word for this, like there needs to be some version of this, but there's a thing called a thunk and a thunk is a way of like wrapping deferred computation that you can transfer around. So it's like, let's say I have some deferred computation that's like, find the nth Fibonacci number and I give you a little box, Anna that's like, hey, this thing can compute the nth Fibonacci number. (01:02:24): I've computed the first five but like, you know, you have to put in some like gas to get the sixth one out. So the idea of the computation exists, but it hasn't been paid for yet. And it's a very natural way of thinking about these kind of like MEV future type of things. But yeah, these cross chain like futures and promises definitely inevitably will have to somehow be like acidized. I mean the most hyped asset right now is like staking derivatives, but I kind of agree that these like MEV derivatives are like future rights derivatives, like land rights. Those are like clearly going to be, those people already have made them, they just haven't like standardized them maybe, but Guillermo (01:03:06): They haven't gone into a public consciousness in the same way that like any of the other things, well, when I say public, I mean like very esoteric niche in my loving crypto degenerates world Tarun (01:03:19): But actually, do you think, here's a question philosophically, do you think that those will be standardized or do you think there's actually excess value to not standardizing them because like each pair wants Guillermo (01:03:31): Ooh. I think, I mean certainly at some point they'll be standardized. I think you think there will be (01:03:39): A right way of doing like these, like futures but these partial features, whatever with like, you have like additional you know, things that you need to do to like step through the next point of the state machine. That being said, I think currently there will be some internal strife as to, you know, should we not standardize it because that means that there's Tarun (01:03:59): Other people will start stealing our liquidity. Yeah. Exactly. Guillermo (01:04:03): Also we can break it in whatever way we want. Yeah. Like, you know, in the next iteration of someone does something and be like, oh, we added this like new epicycle and like, that's why it's kind of interesting Tarun (01:04:10): Yeah. I mean that's a very good question that I think next year we should revisit is like, did the bear market cause people like it did with L1s in 2018/2019 to all work together a bit, right? Or will it cause people to be like, no, I don't want you to steal my liquidity. I'm making my own standard. Anna Rose (01:04:28): It does both right. I think. But yeah, we should check in. That's good. That's sort of wraps up our look across sort of ideas and use cases for ZK, a little bit of a look forward. How are you feeling about 2023? Are you excited? Guillermo (01:04:41): Hell yeah. Well, hopefully, I don't know. I think I'm excited. Anna Rose (01:04:44): Okay. Tarun? Tarun (01:04:45): It was good. It was good to take a vacation after many years Anna Rose (01:04:50): A rosier your outlook maybe? Tarun (01:04:55): I don't know about rosier, but you know, I will say during my vacation I went to Switzerland or I went to Zurich and don't worry not because I was like starting this stiff tongue or something, but I had relatives there but like everyone in Switzerland is still positive on crypto. Like all of them were just like I was like shocked. I was kind of shocked. But like, obviously there are, you know, Switzerland does the wrong plate in some ways it's a very biased sample. This is a country that like had Bitcoin ATMs on trains in 2015. Right. But they're still Anna Rose (01:05:29): It's only benefited. That's the thing. It's like no minus with Tarun (01:05:33): Yeah. Yeah but I haven't, even then they're still pretty, I was surprised by how positive people were. They were just like, yeah, it's good to have the frauds out early. I was like, okay, (01:05:45): Well, that's one way of looking at it. Anna Rose (01:05:48): Cool. All right. Well thanks to both of you for coming on. Guillermo (01:05:51): Thank you Anna. Tarun (01:05:51): Thanks. Anna Rose (01:05:52): Thanks to Kobi who had to leave early. Thanks to the ZK podcast crew. Tanya, Rachel, Henrik, Adam, and to our listeners, thanks for listening.